Citation
A systems approach to characterizing farm sustainability

Material Information

Title:
A systems approach to characterizing farm sustainability
Creator:
Hansen, James William, 1960-
Publication Date:
Language:
English
Physical Description:
xix, 265 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Agricultural and Biological Engineering thesis, Ph. D
Agricultural systems ( lcsh )
Dissertations, Academic -- Agricultural and Biological Engineering -- UF
Sustainable agriculture ( lcsh )
City of Gainesville ( local )
Crops ( jstor )
Farms ( jstor )
Corn ( jstor )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1996.
Bibliography:
Includes bibliographical references (leaves 249-264).
General Note:
Typescript.
General Note:
Vita.
Funding:
Electronic resources created as part of a prototype UF Institutional Repository and Faculty Papers project by the University of Florida.
Statement of Responsibility:
by James William Hansen.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
023165313 ( ALEPH )
35010234 ( OCLC )
AKT6148 ( NOTIS )

Downloads

This item has the following downloads:

UF00056221_.pdf

UF00056221_00001.pdf

00006.txt

UF00056221_0099.txt

UF00056221_0109.txt

00265.txt

UF00056221_0160.txt

00199.txt

UF00056221_0126.txt

00206.txt

UF00056221_0004.txt

00026.txt

UF00056221_0204.txt

00047.txt

00080.txt

UF00056221_0259.txt

UF00056221_0209.txt

UF00056221_0211.txt

00058.txt

UF00056221_0251.txt

UF00056221_0135.txt

UF00056221_0240.txt

UF00056221_0167.txt

UF00056221_0157.txt

UF00056221_0087.txt

00105.txt

00060.txt

UF00056221_0001.txt

00054.txt

00092.txt

UF00056221_0112.txt

UF00056221_0016.txt

00282.txt

UF00056221_0194.txt

UF00056221_0104.txt

00233.txt

UF00056221_0266.txt

UF00056221_0124.txt

00280.txt

00051.txt

UF00056221_0248.txt

00269.txt

UF00056221_0012.txt

00177.txt

UF00056221_0230.txt

00231.txt

00263.txt

UF00056221_0214.txt

UF00056221_0083.txt

00252.txt

00055.txt

UF00056221_0264.txt

00061.txt

00153.txt

00162.txt

00137.txt

UF00056221_0189.txt

UF00056221_0181.txt

00205.txt

00253.txt

UF00056221_0059.txt

00183.txt

00067.txt

00142.txt

00181.txt

00237.txt

00037.txt

UF00056221_0196.txt

UF00056221_0165.txt

UF00056221_0253.txt

UF00056221_0156.txt

UF00056221_0108.txt

UF00056221_0134.txt

UF00056221_0198.txt

UF00056221_0018.txt

00262.txt

00033.txt

00215.txt

UF00056221_0014.txt

UF00056221_0195.txt

00100.txt

00224.txt

UF00056221_0144.txt

00096.txt

00145.txt

UF00056221_0162.txt

UF00056221_0231.txt

UF00056221_0100.txt

UF00056221_0278.txt

UF00056221_0118.txt

UF00056221_0286.txt

UF00056221_0102.txt

UF00056221_0009.txt

UF00056221_0115.txt

00108.txt

UF00056221_0206.txt

UF00056221_0260.txt

UF00056221_0044.txt

UF00056221_0150.txt

UF00056221_0023.txt

00174.txt

UF00056221_0187.txt

UF00056221_0279.txt

UF00056221_0017.txt

00062.txt

UF00056221_0052.txt

00002.txt

UF00056221_0011.txt

UF00056221_0030.txt

UF00056221_0038.txt

00112.txt

00146.txt

UF00056221_0178.txt

00243.txt

UF00056221_0051.txt

UF00056221_0034.txt

00076.txt

UF00056221_0056.txt

00057.txt

UF00056221_0075.txt

00148.txt

UF00056221_0280.txt

00182.txt

UF00056221_0243.txt

UF00056221_0037.txt

UF00056221_0006.txt

UF00056221_0033.txt

00158.txt

UF00056221_0245.txt

00087.txt

UF00056221_0185.txt

00066.txt

00186.txt

UF00056221_0184.txt

00073.txt

00075.txt

UF00056221_0269.txt

UF00056221_0049.txt

00267.txt

UF00056221_0072.txt

00279.txt

UF00056221_0219.txt

UF00056221_0171.txt

UF00056221_0235.txt

00194.txt

UF00056221_0122.txt

00007.txt

UF00056221_0154.txt

00127.txt

00235.txt

UF00056221_0094.txt

00027.txt

UF00056221_0035.txt

00063.txt

UF00056221_0110.txt

UF00056221_0228.txt

UF00056221_0273.txt

00270.txt

UF00056221_0224.txt

UF00056221_0142.txt

00114.txt

00221.txt

00091.txt

00071.txt

00120.txt

UF00056221_0143.txt

00059.txt

00223.txt

00136.txt

UF00056221_0086.txt

00259.txt

00284.txt

UF00056221_0164.txt

00150.txt

UF00056221_0173.txt

UF00056221_0114.txt

UF00056221_0146.txt

UF00056221_0105.txt

UF00056221_0237.txt

UF00056221_0161.txt

UF00056221_0242.txt

UF00056221_0021.txt

00042.txt

00012.txt

00201.txt

00156.txt

UF00056221_0199.txt

00125.txt

00023.txt

UF00056221_0257.txt

UF00056221_0101.txt

00167.txt

00039.txt

UF00056221_0067.txt

UF00056221_0180.txt

UF00056221_0119.txt

00218.txt

00122.txt

UF00056221_0159.txt

00258.txt

00163.txt

00255.txt

UF00056221_0121.txt

00256.txt

00133.txt

00210.txt

00072.txt

UF00056221_0093.txt

00081.txt

00020.txt

UF00056221_0175.txt

UF00056221_0136.txt

00274.txt

00038.txt

00268.txt

00213.txt

00250.txt

00188.txt

UF00056221_0244.txt

UF00056221_0077.txt

UF00056221_0163.txt

00179.txt

00193.txt

00151.txt

UF00056221_0255.txt

UF00056221_0281.txt

00101.txt

00011.txt

UF00056221_0058.txt

00238.txt

UF00056221_0078.txt

00277.txt

00190.txt

00285.txt

00160.txt

UF00056221_0074.txt

00034.txt

00010.txt

UF00056221_0129.txt

00083.txt

00157.txt

00143.txt

00024.txt

UF00056221_0090.txt

00110.txt

UF00056221_0130.txt

00093.txt

UF00056221_0047.txt

00117.txt

UF00056221_0207.txt

00247.txt

00234.txt

UF00056221_0145.txt

UF00056221_0221.txt

UF00056221_0246.txt

00152.txt

00184.txt

UF00056221_0193.txt

UF00056221_00001_pdf.txt

00022.txt

00204.txt

00119.txt

UF00056221_0203.txt

00189.txt

00168.txt

UF00056221_0133.txt

UF00056221_0045.txt

UF00056221_0216.txt

00111.txt

UF00056221_0088.txt

00154.txt

UF00056221_0188.txt

UF00056221_0210.txt

UF00056221_0082.txt

00248.txt

00207.txt

00019.txt

UF00056221_0252.txt

00203.txt

UF00056221_0026.txt

00251.txt

00126.txt

00135.txt

UF00056221_0282.txt

UF00056221_0028.txt

00283.txt

UF00056221_0208.txt

00172.txt

00191.txt

00170.txt

00220.txt

UF00056221_0233.txt

00246.txt

UF00056221_0120.txt

00169.txt

UF00056221_0140.txt

00070.txt

00032.txt

UF00056221_0015.txt

UF00056221_0222.txt

UF00056221_0043.txt

00138.txt

00068.txt

UF00056221_0002.txt

UF00056221_0107.txt

00241.txt

UF00056221_0073.txt

UF00056221_0113.txt

UF00056221_0013.txt

00107.txt

00217.txt

UF00056221_0138.txt

UF00056221_0019.txt

UF00056221_0274.txt

UF00056221_0141.txt

00128.txt

00140.txt

00212.txt

00064.txt

UF00056221_0168.txt

UF00056221_0238.txt

00008.txt

00035.txt

00095.txt

00200.txt

UF00056221_0125.txt

UF00056221_0096.txt

00264.txt

00271.txt

00090.txt

00196.txt

UF00056221_0076.txt

00016.txt

00222.txt

00116.txt

UF00056221_0149.txt

UF00056221_0024.txt

UF00056221_0200.txt

00118.txt

00005.txt

UF00056221_0081.txt

UF00056221_0218.txt

00103.txt

UF00056221_0048.txt

00208.txt

00166.txt

UF00056221_0065.txt

00197.txt

00017.txt

00139.txt

UF00056221_0283.txt

00178.txt

00097.txt

00050.txt

00121.txt

UF00056221_0029.txt

00085.txt

00195.txt

UF00056221_0091.txt

00018.txt

00227.txt

UF00056221_0201.txt

00098.txt

UF00056221_0179.txt

UF00056221_0153.txt

00209.txt

UF00056221_0131.txt

00113.txt

00052.txt

UF00056221_0137.txt

UF00056221_0095.txt

UF00056221_0155.txt

00144.txt

UF00056221_0217.txt

UF00056221_0212.txt

00084.txt

UF00056221_0270.txt

UF00056221_0041.txt

UF00056221_0241.txt

00069.txt

00245.txt

00134.txt

UF00056221_0071.txt

00239.txt

UF00056221_0147.txt

00004.txt

UF00056221_0055.txt

UF00056221_0057.txt

00088.txt

UF00056221_0172.txt

00187.txt

UF00056221_0117.txt

UF00056221_0169.txt

00240.txt

UF00056221_0020.txt

00286.txt

00029.txt

00257.txt

UF00056221_0227.txt

00175.txt

00226.txt

00272.txt

00074.txt

UF00056221_0128.txt

UF00056221_0022.txt

UF00056221_0220.txt

00254.txt

UF00056221_0182.txt

UF00056221_0127.txt

UF00056221_0070.txt

00249.txt

00132.txt

UF00056221_0250.txt

UF00056221_0007.txt

UF00056221_0229.txt

UF00056221_0111.txt

UF00056221_0277.txt

UF00056221_0116.txt

00077.txt

UF00056221_0276.txt

00219.txt

UF00056221_0272.txt

00041.txt

UF00056221_0275.txt

00236.txt

00053.txt

UF00056221_0062.txt

UF00056221_0226.txt

00164.txt

00198.txt

UF00056221_0268.txt

UF00056221_0064.txt

UF00056221_0151.txt

00229.txt

UF00056221_0265.txt

00104.txt

00185.txt

UF00056221_0098.txt

00115.txt

00078.txt

00149.txt

00141.txt

UF00056221_0050.txt

00131.txt

UF00056221_0249.txt

UF00056221_0148.txt

UF00056221_0042.txt

00021.txt

UF00056221_0003.txt

00028.txt

UF00056221_0176.txt

UF00056221_0036.txt

00216.txt

00275.txt

UF00056221_0089.txt

UF00056221_0254.txt

00031.txt

00009.txt

UF00056221_0262.txt

UF00056221_0215.txt

00230.txt

00276.txt

00281.txt

00046.txt

UF00056221_0106.txt

UF00056221_0040.txt

UF00056221_0005.txt

UF00056221_0174.txt

00278.txt

UF00056221_0068.txt

UF00056221_0170.txt

UF00056221_0186.txt

00266.txt

00147.txt

UF00056221_0092.txt

00044.txt

00013.txt

UF00056221_0285.txt

UF00056221_0205.txt

00228.txt

UF00056221_0232.txt

UF00056221_0234.txt

00001.txt

00109.txt

00225.txt

00099.txt

UF00056221_0039.txt

00102.txt

UF00056221_0008.txt

UF00056221_0031.txt

00180.txt

UF00056221_0247.txt

UF00056221_0158.txt

00040.txt

UF00056221_0080.txt

UF00056221_0223.txt

UF00056221_0284.txt

UF00056221_0132.txt

00129.txt

UF00056221_0053.txt

00094.txt

00159.txt

00014.txt

00086.txt

00242.txt

UF00056221_0079.txt

00232.txt

00130.txt

00049.txt

UF00056221_0123.txt

UF00056221_0192.txt

00079.txt

UF00056221_0213.txt

UF00056221_0190.txt

00048.txt

00165.txt

UF00056221_0256.txt

UF00056221_0183.txt

UF00056221__pdf.txt

UF00056221_0191.txt

UF00056221_0197.txt

UF00056221_0202.txt

00211.txt

00123.txt

UF00056221_0027.txt

UF00056221_0084.txt

00065.txt

00261.txt

00106.txt

UF00056221_0025.txt

00214.txt

UF00056221_0063.txt

UF00056221_0085.txt

00015.txt

UF00056221_0267.txt

UF00056221_0069.txt

UF00056221_0236.txt

UF00056221_0103.txt

00056.txt

UF00056221_0271.txt

00192.txt

00045.txt

00161.txt

00171.txt

00176.txt

UF00056221_0046.txt

00173.txt

00202.txt

UF00056221_0139.txt

UF00056221_0010.txt

UF00056221_0239.txt

UF00056221_0177.txt

00030.txt

UF00056221_0225.txt

UF00056221_0263.txt

UF00056221_0097.txt

00244.txt

UF00056221_0258.txt

00089.txt

UF00056221_0060.txt

00082.txt

UF00056221_0061.txt

UF00056221_0166.txt

00155.txt

UF00056221_0054.txt

00273.txt

00036.txt

00124.txt

UF00056221_0032.txt

00260.txt

UF00056221_0066.txt

00043.txt

UF00056221_0152.txt

00025.txt

UF00056221_0261.txt

00003.txt


Full Text










A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY


By
JAMES WILLIAM HANSEN

















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1996


























Copyright 1996

by

James William Hansen














ACKNOWLEDGMENTS


I would like first to express my gratitude to my advisor and mentor, Dr. Jim Jones.

He took a risk by giving me the freedom to pursue an idea, then guided and encouraged

me through the process of making it reality. His competence, integrity and compassion

serve as an example to many. Dr. Robert Caldwell first suggested to me the idea that

became the basis for this study. Dr. Phillip Thornton helped me numerous times to clarify

and express my ideas. He and Drs. Peart, Boggess and Hildebrand provided

encouragement and wise guidance for this research. I owe a debt of gratitude to Dr. Ron

Knapp for his kindness and hospitality during my visits to CIAT, and for many long hours

of hunting down information for this research. His enthusiasm helped me to maintain

mine. Thanks goes also to his staff, especially Jorge and Jorge. Jos6 Domingo deserves

honor and thanks. His successful effort to extract a livelihood for his family and higher

education for his children from a few hectares of hilly land provided the basis for much of

this study.

I am grateful to my wife, Merlie, who endured with me and remained my best

friend through the process. In her love I found strength as I worked and a refuge as I

rested. Thanks also to friends at North Central Baptist Church and others who supported

me in prayer. Ultimate thanks goes to the Lord Jesus who gave me strength to complete

this study and compassion that gives it purpose.















TABLE OF CONTENTS


ACKNOWLEDGMENTS .........


. .. . iii


LIST OF TABLES ....................... ........... ......

LIST OF FIGURES ......... ........ .................... ...

ABSTRACT ............................. ... ...... ........

CHAPTERS

1 INTRODUCTION .......................................


2 IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT?

Introduction ...........................................
Sustainability as an Approach to Agriculture .................
Sustainability as an Alternative Ideology ....................
Sustainability as a Set of Strategies ........................
D discussion ........................................
Sustainability as a Property of Agriculture .....................
Sustainability as an Ability to Satisfy Goals ..................
Sustainability as an Ability to Continue .....................
Approaches to Characterizing Sustainability ....................
Adherence to Prescribed Approaches ......................
M multiple Qualitative Indicators ...........................
Integrated, Quantitative Indicators ...................
Time Trends ....................................
R esilience ............... ...........................
System Sim ulation ....................................
Elements of a Useful Approach for Characterizing Sustainability ....
C conclusions .........................................


. .. .. 4


. viii

.xii

.xviii


. 1











3 A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM
SUSTAINABILITY ....................................... 33

Introduction ................................................. 33
Defining Sustainability ......................................... 34
Quantifying Sustainability .......................... ............. 36
Failure as Violation of State Thresholds ........................ 36
Sustainability Hazard ............................ ............ 41
Simulating Sustainability .............................. ............ 42
An Example: Sustainability of a Simple Time-series Model ............. 44
Diagnosing Constraints to Sustainability ...................... ........ 46
Sensitivity Analysis ......................................... 48
Significance Tests ..................... ....... ................ 49
Sustainability Applied to Farming Systems ........ ............ .52
Selecting a Time Frame .................. .................... 53
Assumptions about Inputs .................................... 54
Farm Failure Criteria ......................................... 55
Determinants of Farm Sustainability ............................. 56
An Example: Sustainability of a Coastal Texas Rice Farm ................. 58
Methods ................................................. 59
Results .................................................... 60
Discussion ........................ ......................... 62

4 AN OBJECT-ORIENTED REPRESENTATION OF A FARMING
SYSTEM ................................................ 65

Introduction .................................................. 65
Overview of the Farming System Simulator ........................... 67
Inputs .............. ............... ....................... 72
Scenario File ............................................. 73
Price File ................ ..... ..... .................... 79
Crop Minimum Data Set .................................... 80
Processes ............ ............................ .. ... ... .......... 80
Overview .............................................. 81
Random Number Sequences .. .................................. 82
Price Generation .......................................... 82
Crop and Ecosystem Processes ............................. ... 85
Event Handling ................................ ............ 86
Resource Accounting ....................................... 88
Illustration: A Fertilizer Application Operation .................. 95
Household Consumption ..................................... 98









Outputs ................. ...... .. .... ............. 99
Resource Status ...................................... ....... 99
Sustainability ............ .... ................. ........ 101
Discussion ............................................... 101


5 CROP SIMULATION FOR CHARACTERIZING SUSTAINABILITY
OF A COLOMBIAN HILLSIDE FARM ....................... 104

Introduction ................. ........... ... .................. 104
A Colombian Hillside Environment ................................ 106
Crop Simulation ................... ................. ......... 116
Approach .......... .............. ....................... 120
Weather Data ........................................... 120
Soil D ata .................................. ...... ...... 125
Simulation Conditions .............. ............. ........ 126
Development and Yield .................................. 127
Response to Environmental Factors ...................... .. ... 128
Results and Discussion ................... .. .... ..... .... 129
Crop Development and Yield .................................. 129
Weather Variability ................. ....................... 136
Response to Nitrogen Dynamics ................................ 139
Response to Soil Erosion ................... .............. 152
Linking Models ............................................. 158
Issues in Linking Crop and Whole-farm Models .......... ... ... 158
Issues in Linking Crop and Erosion Models ....................... 161
Conclusions ............................................... 164

6 DETERMINANTS OF SUSTAINABILITY OF A COLOMBIAN
HILLSIDE FARM .......................................... 167

Introduction ....................... ........................ 167
Approach .................................................. 170
A Colombian Hillside Farm ................................... 170
Sources of Information ..................................... 171
Assumptions ............................................ 176
Scenarios .................................... ........... 180
Simulation and Analysis ...................................... 185
Results .................................................... 186
Base Scenario ................... ......................... 186
Cropping Systems ...................................... 190
Soil Management .......................................... 192
Costs and Prices ................... ....................... 194










R sources .................................
Sources of Risk ...............................
Constraints to Sustainability ......................
D iscu ssion ......................................
Practical Im plications ...........................
L im stations .................................

7 SUMMARY AND CONCLUSIONS ..................

APPENDICES

A OBJECT-ORIENTED PROGRAMMING CONCEPTS ....

B A MINIMUM DATA SET FOR SIMULATING FARM
SUSTAINABILITY ............................

C FARMING SYSTEM SIMULATOR USER'S GUIDE ....

D INPUT FILES USED FOR FARM SIMULATIONS ......

BIBLIOGRAPHY ......................................

BIOGRAPHICAL SKETCH ..............................


196
198
202
202
203
204


. .......... 2 13


235

238














LIST OF TABLES


Table page

2-1 Interpretations of agricultural sustainability ............................ 5

2-2 Contrasting approaches of conventional and sustainable agriculture as
characterized by Hill and MacRae ................................... 9

2-3 Strategies frequently associated with sustainability ...................... 11

2-4 Contingency table for inferring sustainability based on trends of system inputs
and outputs ................................................... 23

2-5 Elements of a useful approach to characterizing sustainability of agricultural
systems .................................................... 28

2-6 Approaches to characterizing agricultural sustainability .................. '31

3-1 Definitions of symbols used ....................................... 37

3-2 Sensitivity of simulated sustainability to system properties ................ 46

3-3 Frequency table for tests of difference between simulated sustainabilities,
independent observations ......................................... 50

3-4 Frequency table for testing differences in sustainabilities: paired observations .. 51

3-5 Relative sensitivity of simulated five-year sustainability of a Texas rice farm
to continuous factors ............................................ 60

3-6 Absolute sensitivity of simulated five-year sustainability of a Texas rice farm
to discrete factors .............................................. 61

4-1 Sections in the farm scenario file ................................... 74










4-2 Description of resource classes ........................ ............. 77

4-3 Format of the operations output file ............................... 87

5-1 Area in each slope class, Domingo farm, Cauca, Colombia ............... 106

5-2 Properties of soil layers, site near on-farm trials, Domingo farm, Cauca,
Colombia ..................................... ............ 112

5-3 Spatially interpolated WGEN coefficients for the Domingo farm, Cauca,
Colombia ........................... .... ................... 125

5-4 Planting information for crop simulation studies ....................... 126

5-5 Observed and predicted bean yields, Domingo and Trujillo farms, Cauca,
Colombia ................................................. 130

5-6 Observed and predicted timing of phenological events for bean and maize,
Domingo farm, Cauca, Colombia, planted March 30, 1994 ............... 130

5-7 Treatment description and observed and predicted yields for the October 14,
1993 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132

5-8 Treatment description and observed and predicted yields for the March 30,
1994 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132

5-9 Observed and predicted maize yields, Domingo and Trujillo farms, Cauca,
Colombia ................................................ 134

5-10 Observed and predicted cassava yields, Domingo and Trujillo farms, Cauca,
C olom bia ........................................... ......... 135

5-11 Mean, standard deviation, skewness and Kolmogorov-Smirnov test statistic
for distributions of crop yield and maturity time simulated with observed and
simulated weather from La Florida, Cauca, Colombia .................. 137

5-12 Mean, standard deviation, skewness, and Kolmogorov-Smirnov test statistic for
distributions of observed and simulated monthly and annual rainfall totals, La
Florida, Popayan, Colombia ...................................... 139

6-1 Hypotheses related to determinants of farm sustainability ................ 169









6-2 Estimates of labor requirements for annual crop production, Cauca,
Colombia .................................................. 173

6-3 Prices of production inputs, Cauca, Colombia, December 1992 ........... 174

6-4 Fitted parameters for deterministic component of production commodity
price time series models ....................................... 175

6-5 Fitted parameters for stochastic component of production commodity price
time series models .......................................... 175

6-6 Adjustments to simulated crop yields and reported prices, Cauca, Colombia 176

6-7 Description of farm scenarios ..................................... 181

6-8 Predicted 15 year sustainability of cropping system scenarios, and McNemar
test statistic for difference from the base scenario ...................... 192

6-9 Soil loss and predicted 15 year sustainability of erosion scenarios. .......... 195

6-10 Relative sensitivity of predicted 15 year sustainability to continuous factors,
and McNemar test statistic for difference from the base scenario. .......... 197

6-11 Predicted nine-year sustainability, McNemar test statistic for difference from
the base scenario, and standard deviation of liquid assets after three years for
sources of risk scenarios .................................... 199

A-1 Object-oriented programming terms ......................... ... 211

B-1 Format of the SCENARIO section of the scenario file .................. 215

B-2 Format of the OUTPUTS section of the scenario file ................... 216

B-3 Format of the ANALYSES section of the scenario file .................. 217

B-4 Format of the RESOURCES section of the scenario file ................. 219

B-5 Format of the LINKAGES section of the scenario file .................. 221

B-6 Format of the OPERATION REQUIREMENTS section of the scenario file .. 223

B-7 A resource block within the OPERATION REQUIREMENTS section ..... 223









B-8 Format of the SCHEDULED OPERATIONS section of the scenario file .... 225

B-9 Format of the LANDSCAPE section of the scenario file ................. 226

B-10 Format of the STRATEGIES section of the scenario file ................ 228

B-11 Format of the ENTERPRISES section of the scenario file ............... 230

B-12 Format of the CONSUMPTION DECISIONS section of the scenario file ... 231

B-13 Format of the CONSTANT section of the price file .................... 232

B-12 Format of the ARMA section of the price file ......................... 233

B-12 Format of the HISTORICAL section of the price file ................... 234

C-1 Options available from FSS menu items ............................ 236















LIST OF FIGURES


Figure page


2-1 Contrasting interpretations of the relationship between chemical input levels
and sustainability ........................................... 13

3-1 Relationship between time, distribution of state and sustainability under
constant mean and high variability, and negative trend and low variability ..... 40

3-2 Estimating sustainability by sampling a small number of simulated
realizations of future system behavior ................................ 43

3-3 Sustainability and hazard of an AR(1) process with constant parameters,
declining expected value, abrupt decrease and increase in mean, decrease
and increase in variance, and decrease and increase in autocorrelation ....... 47

3-4 Simulated sustainability of a Texas rice farm under four scenarios .......... 62

4-1 Parallel ecological, economic and social hierarchies of agricultural systems .... 68

4-2 Object representation of the main components of a farming system .......... 70

4-3 Tree of resource class hierarchy in FSS ............. ................. 75

4-4 Flow of information in FSS from the generation of a field operation to its
effect on resource accounting ...................................... 89

4-5 Pseudocode representation of the USE method of the consumable resource
class ............... ............................. ........... 90

4-6 Pseudocode representation of the SELL method of the consumable resource
class ..................................................... 91

4-7 Pseudocode representation of the USE method of the timed resource class .... 92











4-8 Pseudocode representation of the USE method of the credit resource class .. .. 92

4-9 Pseudocode representation of the UPDATE method of the consumable
resource class ............... ..... .......... ............ 93

4-10 Pseudocode representation of the UPDATE method of the capital resource
class ..................................................... 93

4-11 Pseudocode representation of the UPDATE method of the credit resource
class ..................................................... 94

4-12 Pseudocode representation of the USE method of the cost list class ......... 95

4-13 Forrester representation of variable cost linkages between a consumable
resource and three linked consumable resources ........................ 96

4-14 Flow of information from the generation of a fertilizer application operation
by a crop model to its effect on the operating fund ................ ...... 97

4-15 Example of a resource box plot generated by FSS ...................... 99

4-16 Example of a final resource distribution plot generated by FSS ............ 100

4-17 Example of a sustainability time plot generated by FSS ............... 102

5-1 Location map of the study area, Cauca, Colombia .................... 107

5-2 Land use map of the Domingo farm, Cauca, Colombia ................. 108

5-3 Monthly climate statistics: mean daily solar radiation, mean daily maximum
and minimum temperature, and total rainfall, Domingo farm, Cauca,
Colombia ................................... ................ 110

5-4 Mean monthly rainfall totals at weather station used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation .......... .... .................... ......... 111

5-5 Phosphorus sorption isotherms for four soils with different mineralogy ...... 113

5-6 Decomposition of soil organic C in three allophanic and seven
non-allophanic soils ..................................... 114










5-7 Decomposition of 4C-labeled wheat straw in Lo Aguire sandy loam with
added allophane ............................................... 115

5-8 Locations and elevations of weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation ............. ............. .... ......... 123

5-9 Trend in the ratio of bean yields observed in on-farm trials and predicted
by CROPGRO, Domingo and Trujillo farms, Cauca, Colombia ........... 131

5-10 Simulated and observed bean yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994, October 1994 and
M arch 1995 ............... ..................... ......... 133

5-11 Simulated and observed maize grain yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994 and October 1994 .... 135

5-12 Distribution of simulated yields of October- and March-planted maize and
bean in response to historical and generated weather variability, La Florida,
Popayan, Colombia ............................................ 138

5-13 Simulated grain and biomass yield response of maize and October- and
March-planted bean applied N using the default mineralization factor ....... 140

5-14 Maize response to applied N on several volcanic soils in Narifio, Colombia .. 141

5-15 Effect of N mineralization factor, SLNF, on simulated maize response to
applied N ................... ...................... ........ 142

5-16 Simulated grain and biomass yield response of maize and October- and
March-planted bean to applied N using the adjusted mineralization factor .... 143

5-17 Grain and biomass yields of October- and March-planted bean simulated
by the original and modified versions of CROPGRO .................... 144

5-18 Grain and biomass yields of irrigated and rainfed soybean simulated by the
original and modified versions of CROPGRO ......................... 145

5-19 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Domingo farm, Cauca, Colombia, 1994 ................... 147










5-20 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Kuiaha, Hawaii, 1993 ................................. 148

5-21 Plant and nodule mass at 35 days, and grain yield of container-grown
beans in response to applied N .................................... 150

5-22 Simulated 60 year sequences of maize grain yields at different levels of
applied N, Domingo farm, Cauca, Colombia .......................... 151

5-23 Simulated grain yields of maize and bean in response to soil loss, Domingo
farm, Cauca, Colombia ........... ............................ 153

5-24 Simulated root distributions of October- and March-planted bean in response
to 0 cm, 20 cm, and 50 cm of soil loss, Domingo farm, Cauca, Colombia .... 155

5-25 Relative root distribution factors, F and SRGF after 0 cm, 20 cm and 50 cm
soil loss ........................................ .......... 157

5-26 Simulated grain yields of October- and March-planted bean in response to soil
loss, Domingo farm, Cauca, Colombia ............................. 159

5-27 Pseudocode representation of an algorithm for resolving resource conflicts
among crop enterprises ........................................ 160

5-28 Pseudocode representation of an algorithm for simulating erosion and crop
growth on a complex hillslope ............................... 164

6-1 Historical crop wholesale prices, Cauca, Colombia .................... 172

6-2 Cropping pattern included in farm scenario ........................... 178

6-3 Influence of type of farmer participation in on-farm trials on bean response
to ground and partially acidulated rock phosphate, chicken manure, and
10-30-10 ................ ... .. ................. .. ...... 179

6-4 Box plot of liquid assets, base scenario ........................... 187

6-5 Cumulative distribution of liquid assets after 1, 3, 6, 9, 12 and 15 years, base
scenario .................................................. 188

6-6 Sustainability and hazard time plots of the base scenario .. ............. 189









6-7 Sustainability time plot of annual cropping systems ................... 190

6-8 Sustainability time plot of base and coffee scenarios .................... 191

6-9 Sustainability time plot of base and nitrogen management scenarios ........ 193

6-10 Sustainability time plot of base and soil erosion scenarios ................ 194

6-11 Sustainability time plot of base and price scenarios ..................... 195

6-12 Sustainability time plot of base and household consumption scenarios ...... 196

6-13 Sustainability time plot of base and resource scenarios .................. 198

6-14 Sustainability time plot of base and credit scenarios .................... 199

6-15 Distribution of liquid assets after 1, 3, 6 and 9 years for the source of risk
scenarios ................................................. 200

6-16 Sustainability time plot of source of risk scenarios ................... 201

B-1 Example SCENARIO section ................. ....... ............ 215

B-2 Example OUTPUTS section ..................................... 216

B-3 Example ANALYSES section .................................... 217

B-4 Example item from the RESOURCES section ........................ 218

B-5 Example item from the LINKAGES section ...... ... ........... .... 221

B-6 Example item from the OPERATION REQUIREMENTS section ......... 224

B-7 Example item from the SCHEDULED OPERATIONS section ............ 224

B-8 Example item from the LANDSCAPE section ........................ 226

B-9 Example item from the STRATEGIES section ........................ 229

B-10 Example item from the ENTERPRISES section ....................... 229

B-11 Example item from the CONSUMPTION DECISIONS section ........... 231









D-1 Farm scenario file used to simulate the base scenario ................... 239

D-2 Crop management file used to simulate farm scenarios .................. 242

D-3 Soil profile description file used to simulate farm scenarios ............... 247


xvii














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY

By

James William Hansen

May 1996


Chairperson: Dr. James W. Jones
Major Department: Agricultural and Biological Engineering



The potential usefulness of the concept of sustainability as a criterion for

evaluating and improving agricultural systems has been hindered by inadequacy of

approaches for its characterization. The approach presented here for characterizing farm

sustainability is based on a definition of sustainability as the ability of a system to continue

into the future, expressed by the probability that the system will continue without violating

failure thresholds during a particular future period. Characterization includes

quantification and diagnosis of constraints. Sustainability is quantified by using long-term,

stochastic simulation to sample realizations of the future behavior of a model of the

farming system. Sensitivity analysis provides a tool for testing hypotheses about

constraints to sustainability. An object-oriented farming system simulator developed for


xviii










this purpose can simulate a replicated farm scenario with stochastic inputs of weather and

prices. The farm simulator simulates the balance of farm resources and accounts for

ecological determinants of crop production by calling external crop models. Several crop

simulation models were evaluated for compatibility with a simulation study of the

sustainability of a hillside farm in the Cauca region of Colombia. The crop models were

useful for capturing response to weather variability and N management, but did not

account for important yield determinants, including P deficiency and nematode damage.

Simple modifications improved simulated response to applied N and sensitivity to soil loss.

A study of a hillside farm in Colombia showed the practical value of using simulation to

characterize sustainability. Results identified cropping system, area under cultivation,

consumption requirements, crop prices, and soil erosion as important determinants of

sustainability. The study showed that price variability contributes more than weather

variability to farm risk in this location, and that spatial diversification reduces risk and

improves sustainability. Results suggest that the farmer can enhance farm sustainability by

diversifying and intensifying crop production. Researchers can contribute to sustainability

by identifying promising high-value crops, by testing the proposed cropping systems, and

by quantifying severity and impacts of erosion. Policy makers can address sustainability

constraints caused by price volatility and lack of affordable credit.














CHAPTER 1
INTRODUCTION


The past decade has shown a shift of focus from productivity to sustainability of

agricultural systems. The growing emphasis on sustainability stems from concern about

both threats to, and negative impacts of agriculture, and from the realization that decisions

made now can have unforeseeable impacts on future generations. In spite of the growing

interest in agricultural sustainability, there is no generally accepted definition of what it is.

There is even less agreement about how to achieve it. Little progress has been made

toward developing methods for characterizing sustainability of particular agricultural

systems because of the conceptual problem of agreeing on a definition and the practical

problems that result from the fact that sustainability deals with the future and therefore

cannot be readily observed. Although some have argued that sustainability cannot and

should not be measured, there is growing realization that progress toward improving the

sustainability of agriculture is not possible without the ability to measure it and identify its

constraints.

Although concerns about agricultural sustainability focus on system levels ranging

from field to global, the farm is an appropriate level for dealing with sustainability. It is

appropriate from the standpoint of relevance; it is the farmer who makes the final

decisions about what to produce and what resources and methods will be employed. If










2

agriculture is to meet the needs of society as a whole--producing food and other products

while protecting natural resources--it must first meet the needs of the farmers who

implement and manage it. The farm level is also appropriate from a practical standpoint.

Since human goals are not intrinsic to fields or enterprises, it is difficult to identify what is

to be sustained. On the other hand, characterization of sustainability at system levels

higher than the farm is complicated by the complexity of the systems, the difficulty of

specifying system boundaries, and the often conflicting goals of multiple human actors.

The purpose of this study is to present and demonstrate a simulation-based systems

approach to characterizing sustainability of farming systems. Specific objectives of this

dissertation are as follows:

1. Review existing conceptual and methodological barriers to using the concept of
sustainability for guiding change in agriculture

2. Propose a set of elements necessary for an approach to characterizing
sustainability to provide a useful basis for guiding agriculture.

3. Present the logical and mathematical basis for a definition of sustainability that
applies generally to dynamic, hierarchical, stochastic, purposeful systems, and
relate this definition to farming systems.

4. Describe and illustrate a framework for using system simulation to quantify
sustainability and test hypotheses about its determinants.

5. Describe an object-oriented farming system simulator that was developed for the
purpose of characterizing farm sustainability, and present its data requirements.

6. Evaluate the compatibility of a set of crop models with requirements imposed by
a simulation study of sustainability in a hillside environment in Colombia.

7. Demonstrate and evaluate the approach for characterizing farm sustainability by
identifying determinants of the sustainability of a hillside farm in Colombia.












This dissertation is organized around these objectives. Chapter 2 critically reviews

existing interpretations of sustainability and approaches proposed for its characterization.

A conceptual and mathematical framework for characterizing farm sustainability is

presented in Chapter 3. Chapter 4 presents the farm simulator that was developed to

apply the proposed framework to a particular farm. The simulator uses process-level crop

simulation to characterize the contribution of ecosystem processes to farm production and

sustainability. Chapter 5 describes the physical environment in the region of Colombia

where the approach was applied, and evaluates the suitability of a set of crop models for

characterizing production and sustainability in that environment. Finally, Chapter 6

applies the approach to a particular farm in the lower Andes of southwestern Colombia.














CHAPTER 2
IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT?



Introduction



In literal English usage, sustainability is the ability to "keep in existence; keep up;

maintain or prolong" (Neufeldt, 1988, p. 1349). The variety of meanings acquired by

sustainability as applied to agriculture (Table 2-1) have been classified according to the

issues motivating concern (Douglass, 1984; Weil, 1990), their historical and ideological

roots (Kidd, 1992; Brklacich et al., 1991), and the hierarchical levels of systems

considered (Lowrance et al., 1986).

The distinction between sustainability as a system-describing and as a goal-

prescribing concept (Thompson, 1992) identifies two current schools-of-thought that

differ in their underlying goals. The goal-prescribing concept interprets sustainability as an

ideological or management approach to agriculture. This concept developed in response

to concerns about negative impacts of agriculture, with the underlying goal of motivating

adoption of alternative approaches. The system-describing concept interprets

sustainability either as an ability to fulfill a diverse set of goals or as an ability to continue.

This concept can be related to concerns about impacts of global change on the viability of














CHAPTER 2
IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT?



Introduction



In literal English usage, sustainability is the ability to "keep in existence; keep up;

maintain or prolong" (Neufeldt, 1988, p. 1349). The variety of meanings acquired by

sustainability as applied to agriculture (Table 2-1) have been classified according to the

issues motivating concern (Douglass, 1984; Weil, 1990), their historical and ideological

roots (Kidd, 1992; Brklacich et al., 1991), and the hierarchical levels of systems

considered (Lowrance et al., 1986).

The distinction between sustainability as a system-describing and as a goal-

prescribing concept (Thompson, 1992) identifies two current schools-of-thought that

differ in their underlying goals. The goal-prescribing concept interprets sustainability as an

ideological or management approach to agriculture. This concept developed in response

to concerns about negative impacts of agriculture, with the underlying goal of motivating

adoption of alternative approaches. The system-describing concept interprets

sustainability either as an ability to fulfill a diverse set of goals or as an ability to continue.

This concept can be related to concerns about impacts of global change on the viability of











5

Table 2-1. Interpretations of agricultural sustainability.
Sustainability as an ideology:

". .. a philosophy and system of farming. It has its roots in a set of values that reflect a state of empowerment, of awareness
of ecological and social realities, and of one's ability to take effective action." (MacRae et al., 1990, p. 156)

". .. an approach or a philosophy that integrates land stewardship with agriculture. Land stewardship is the philosophy
that land is managed with respect for use by future generations." (Neher, 1992, p. 54)

"... a philosophy based on human goals and on understanding the long-term impact of our activities on the environment
and on other species. Use of this philosophy guides our application of prior experience and the latest scientific advances to
create integrated, resource-conserving, equitable farming systems." (Francis & Youngberg, 1990, p. 8)

"... farming in the image of Nature and predicated on the spiritual and practical notions and ethical dimensions of
responsible stewardship and sustainable production of wholesome food." (Bidwell, 1986, p. 317)

Sustainability as a set ofstrategies:

"... a management strategy which helps the producers to choose hybrids and varieties, a soil fertility package, a pest
management approach, a tillage system, and a crop rotation to reduce costs of purchased inputs, minimize the impact of the
system on the immediate and the off-farm environment, and provide a sustained level of production and profit from
farming." (Francis, Sander & Martin, 1987, p. 12)

"... a loosely defined term for a range of strategies to cope with several agriculturally related problems causing increased
concern in the U.S. and around the world." (Lockeretz, 1988, p. 174)

Farming systems are sustainable if "they minimize the use of external inputs and maximize the use of internal inputs already
existing on the farm." (Carter, 1989, p. 16)

"... (a) the development of technology and practices that maintain and/or enhance the quality of land and water resources,
and (b) the improvements in plants and animals and the advances in production practices that will facilitate the substitution
of biological technology for chemical technology." (Ruttan, 1988, p. 129)

Sustainability as the ability tofulfill a set of goals:
"A sustainable agriculture is one that, over the long term, enhances environmental quality and the resource base on which
agriculture depends, provides for basic human food and fiber needs, is economically viable, and enhances the quality of life
for farmers and society as a whole." (American Society of Agronomy, 1989, p. 15):

". .. agricultural systems that are environmentally sound, profitable, and productive and that maintain the social fabric of the
rural community." (Keeney, 1989, p. 102)

"... an agrifood sector that over the long term can simultaneously (1) maintain or enhance environmental quality, (2)
provide adequate economic and social rewards to all individuals and firms in the production system, and (3) produce a
sufficient and accessible food supply." (Brklacich, Bryant & Smit, 1991, p. 10)

". an agriculture that can evolve indefinitely toward greater human utility, greater efficiency of resource use, and a
balance with the environment that is favorable both to humans and to most other species." (Harwood, 1990, p. 4)

Sustainability as the ability to continue:

"A system is sustainable over a defined period if outputs do not decrease when inputs are not increased." (Monteith, 1990,
p. 91)
"... .the ability of a system to maintain productivity in spite of a major disturbance, such as is caused by intensive stress or a
large perturbation." (Conway, 1985, p. 35)
"... the maintenance of the net benefits agriculture provides to society for present and future generations." (Gray, 1991, p.
628)

"Agriculture is sustainable when it remains the dominant land use over time and the resource base can continually support
production at levels needed for profitability (cash economy) or survival (subsistence economy)." (Hamblin, 1992, p. 90)










6

agriculture, and to the goal of using sustainability as a criterion for guiding agriculture as it

responds to rapid changes in its physical, social and economic environment.

Although the concept of sustainability has been useful for consolidating concerns

and motivating change, concrete examples of its use as an operational criterion for guiding

efforts to improve agricultural systems are difficult to identify. The objectives of this

chapter are (1) to examine conceptual and methodological barriers to using the concept of

sustainability for guiding change in agriculture, and (2) to propose a set of elements

necessary for an approach to characterizing sustainability to provide a useful criterion for

guiding agriculture.



Sustainability as an Approach to Agriculture



The sustainable agriculture movement evolved from several reform movements in

the U.S.A., Canada and Western Europe that developed in response to concerns about

impacts of agriculture such as depletion of nonrenewable resources, soil degradation,

health and environmental effects of agricultural chemicals, inequity, declining rural

communities, loss of traditional agrarian values, food quality, farm worker safety, decline

in self-sufficiency, and decreasing number and increasing size of farms. These problems

became associated with "conventional agriculture" that was perceived as unsustainable

(Dahlberg, 1991). "Alternative agriculture" is often equated with sustainable agriculture

(O'Connell, 1992; Madden, 1987; Harwood, 1990; Dahlberg, 1991; Bidwell, 1986) and

reflects the goal of promoting alternatives to conventional agriculture. Reviews by












Harwood (1990) and Kidd (1992) trace the historical development of the sustainable, or

alternative agriculture movement.

Differences in values and practices promoted as sustainable have been attributed to

differences in the problems emphasized (Carter, 1989) and to different visions of what

agriculture should be like (Thompson, 1992). "Originally, the advocates of alternative

approaches to agriculture--all united in their critique of industrial agriculture as being

unsustainable--debated among themselves the future direction and shape of agriculture"

(Dahlberg, 1991, p. 337). Some have focused on identifying sustainable alternatives to

existing management practices while others have advocated new philosophical orientations

toward agriculture.


Sustainability as an alternative ideology


MacRae et al. (1990), Neher (1992) and Francis and Youngberg (1990) defined

sustainable agriculture as a philosophy (Table 2-1). Ikerd (1991) described low-input,

sustainable agriculture (LISA) as more a philosophy than a practice. Examining the

concept of conventional agriculture is important since sustainable agriculture is often

described by its contrast with conventional agriculture (Lockeretz, 1988; MacRae et al.,

1989; Hauptli et a., 1990; Dobbs et al., 1991; O'Connell, 1992; Hill and MacRae, 1988).

Conventional agriculture. The concept of conventional agriculture was apparently

developed in order to clarify, and justify alternative approaches to agriculture.

Conventional agriculture is characterized as "capital-intensive, large-scale, highly

mechanized agriculture with monocultures of crops and extensive use of artificial












fertilizers, herbicides and pesticides, with intensive animal husbandry" (Knorr and

Watkins, 1984, p. x) with a paradigm of "strength through exhaustion" (Bidwell, 1986, p.

317). Hill and MacRae (1988) contrasted approaches of conventional and sustainable

agriculture (Table 2-2). Based on a review of actual cropping practices in the U.S.,

Madden (1990) appropriately identified conventional agriculture as a caricature.

Beus and Dunlap (1990) identified centralization, dependence, competition,

domination of nature, specialization, and exploitation as key elements of conventional

agriculture from the writings of six conventional agriculture advocates. Although

"conventional agriculture" was applied to mainstream U.S. agriculture as one side of a

debate between competing paradigms, its description was admittedly a construct for the

purpose of"clarifying opposing positions, facilitating comparisons, and sharpening the

focus of the debate" (p. 597). A survey of agriculturalists in Washington state by the same

authors (Beus and Dunlap, 1991) suggested that the conventional agriculture that they

described does not represent mainstream U.S. agriculture. A random sample of farmers

and three of the four groups of conventional agriculturalists surveyed showed greater

agreement with the alternative than with the conventional agriculture paradigm.

Characterization of conventional agriculture extends to mainstream research and

education institutions where research has been described as too narrow, short-sighted,

biased by interests of agribusiness funding sources, and distorted by the values of scientists

to be able to deal with the issues necessary to achieve sustainability (Bidwell, 1986; Allen

and van Dusen, 1988; MacRae et al., 1989; Dahlberg, 1991; Kirschenmann, 1991; Hill and

MacRae, 1988). Beus and Dunlap (1990) and Dahlberg (1991) expressed concern that














Table 2-2. Contrasting approaches of conventional and sustainable agriculture as
characterized by Hill and MacRae (1988, p. 95).


Conventional agriculture


Symptoms

Reductionist

Eliminate "Enemies"

Narrow focus (neglects side effects, health
& environmental costs ignored)

Instant

Single, simple (magic bullet, single discipline)


Temporary solutions

Unexpected disbenefits (to person & planet)

High power (risk of overkill & errors/ accidents)

Direct "attack"


Imported

Products

Physico-chemical (often unnatural, synthetic)

Technology-intensive

Centralized

Values secondary

Expert, paternalistic (arrogant)

Dependent

Inflexible

Ignores freedom of choice (unjust)

Disempowering

Competitive

Authored


Sustainable agriculture


Causes, prevention

Holistic

Respond to indicators

Broad focus (subcellular to all life to globe, all
costs internalized)

Long time frame (future generations)

Multifaceted, complex (multi- & trans-
disciplinary)

Permanent solutions

Unexpected benefits

Low power (minimal risk)

Indirect, benign approaches (catalytic, multiplier,
synergistic effects)

Local solutions and materials

Processes, services

Bio-ecological (natural)

Knowledge/skill intensive

Decentralized (human scale)

Compatible with higher values

Individual/community responsibility (humble)

Self-maintaining/regulating

Flexible

Respects freedom of choice (just)

Empowering

Co-operative

Anonymous (seeking neither reward or fame)












such institutions threaten to dilute the concept of sustainable agriculture by co-opting it

while ignoring its more important and radical aspects.

Alternative values. Sustainable agriculture has been described as an umbrella term

encompassing several ideological approaches to agriculture (Gips, 1988) including organic

farming, biological agriculture, alternative agriculture, ecological agriculture, low-input

agriculture, biodynamic agriculture, regenerative agriculture, permaculture, and

agroecology (Carter, 1989; MacRae et al., 1989; Bidwell, 1986; O'Connell, 1992;

Kirschenmann, 1991; Dahlberg, 1991).

Beus and Dunlap (1990) listed decentralization, independence, community,

harmony with nature, diversity, and restraint as key values of alternative agriculture.

Social values such as equity, the value of traditional agricultural systems, self-sufficiency,

preservation of agrarian culture, and preference for small, owner-operated farms have

been incorporated into definitions of sustainability (Weil, 1990; Keeney, 1989; Bidwell,

1986; Francis and Youngberg, 1990). The concept of equity is extended to include future

generations (Batie, 1989; Norgaard, 1991). Environmental values associated with

sustainability include mimicry of nature and an "ecocentric" ethic. Hauptli et al. (1990)

described mimicry of nature: ". sustainable agriculture attempts to mimic the key

characteristics of a natural ecosystem .. (p. 143). The ecocentric position--valuing

ecosystems or species without regard to their impact on human welfare--is illustrated by

Douglass (1984) who stated ecology-minded people ". define agricultural sustainability

in biophysical terms, and to allow its measurement to determine desirable population

levels" (p. 5).












Sustainability as a set of strategies


Francis and Youngberg (1990) described sustainable agriculture as a philosophy

that guides the creation of farming systems. Specific management strategies are often

suggested by ideological interpretations of sustainability. The strategies promoted as

sustainable (Table 2-3) are based on the types of problems emphasized and on views of

what would constitute an improvement.


Table 2-3. Strategies frequently associated with sustainability.
Strategy References

Self-sufficiency through preferred use of on-farm or locally a, b, g, d
available "internal" resources to purchased "external" resources.

Reduced use or elimination of soluble or synthetic fertilizers, a, e, f, h, i, d, k

Reduced use or elimination of chemical pesticides, substituting a, c, d, e, f, h, i, j, k
integrated pest management practices.

Increased or improved use of crop rotations for diversification, a, c, d, e, f, h, j
soil fertility, and pest control.

Increased or improved use of manures and other organic a, c, f, h, j, k
materials as soil amendments.

Increased diversity of crop (and animal) species. a, d, g, i

Maintenance of crop or residue cover on the soil. a, d, e

Reduced stocking rates for animals, a, c, d
SLockeretz, 1988
bHarwood, 1990
" MacRae et al. 1990
dNeher, 1992
" Dobbs et al. 1991
'MacRae et al. 1989
SGliessman, 1990
h Edwards, 1990
'Hauptli et al. 1990
SO'Connell, 1992
k Hill & MacRae, 1988










12

The strategy most frequently linked to sustainability is reduction or elimination of

the use of processed chemicals, particularly fertilizers and pesticides (Stinner and House,

1987; Lockeretz, 1988; Carter, 1989; Hauptli et al., 1990; Madden, 1990; Dobbs et al.,

1991). In 1988, the U.S. Department of Agriculture linked sustainability to levels of

inputs by establishing the LISA (low-input, sustainable agriculture) research program

(O'Connell, 1990; Dicks, 1992). Arguments for reducing chemical inputs include limited

supplies of fossil fuels, decreasing commodity prices necessitating reducing input costs, a

need for self-sufficiency, concerns about pollution, and health and safety concerns (Francis

and King, 1988; Carter, 1989; Stinner and House, 1989; Conway and Barbier, 1990;

MacRae et al., 1990; Rodale, 1990).

York (1991) argued that fewer options exist for reducing fertilizer inputs than

pesticide inputs in agricultural systems while maintaining sustainable production. Unlike

pesticides, soil nutrient elements generally have no substitutes and are subjected to harvest

and other losses that must be replaced by weathering or imported from outside the system

if production is to be sustained. The high energy cost unique to N fertilizer production

and the potential for biological fixation suggest a need and potential for seeking

alternatives to synthetic N fertilizers that does not exist for mineral-based nutrients such as

P and K.

The important distinction between production systems that currently employ high

levels of chemical inputs and those that employ low levels (Weil, 1990) is often

overlooked. Zandstra (1994) described sustainability as a function of chemical input levels

(Fig. 2-la). Excessive input levels were said to degrade natural resources through












A

exhaustion


accumulaion


input level


B
/\
/
//




S / \
OT /
inputs control













sustainability


Figure 2.1. Contrasting interpretations of the relationship between chemical input
levels and sustainability by (a) Zandstra (1994) and by (b) Stinner and House (1987).










14

accumulation while inadequate levels degrade resources through exhaustion. This concept

is in sharp contrast to the decreasing relationship between chemical input levels and

sustainability proposed by Stinner and House (1987) (Fig. 2-1b).

Studies in Mali, Benin, Zambia, and Tanzania provide examples of resource

degradation due to inadequate chemical inputs (Budelman and van der Pol, 1992). In each

case, supplies of soil nutrients were exhausted rapidly due to a combination of harvest,

erosion, leaching, denitrification and volatilization, with harvest being the greatest loss.

Nutrient budgets estimated for several crops in southern Mali were always negative for N

and K, and variable but generally better for P. The authors concluded that the only way to

make these cropping systems sustainable is with increased use of fertilizers.


Discussion


Interpreting sustainability as an approach to agriculture has been useful for

motivating change. Sustainability as an ideology has provided as a common banner for

various agricultural reform movements (Gips, 1988; Dahlberg, 1991). Research and

promotion of sustainability interpreted as a set of strategies has become part of policy in

the U.S. in the form of provisions in the 1990 Farm Bill (O'Connell, 1992; Yetley, 1992).

Interpreting sustainability as an approach is not useful for guiding change in

agriculture for several reasons. First, approaches developed in response to problems in

North America and Europe may be inappropriate in regions where circumstances and

problems are different. The alternative agriculture movement has its roots primarily in

regions characterized by high levels of resource consumption, food surpluses, high levels












of chemical inputs, relatively deep, fertile soils, and relatively stable populations. In

contrast, many less developed tropical regions are characterized by lower levels of

resource consumption, frequent or chronic food shortages, lower levels of chemical

inputs, relatively fragile soils, and rapidly growing populations. Attempts to link strategies

to sustainability by definition fail to consider the need to match technologies to specific

environments.

Dicks (1992) argued that interpretations of sustainability in the U.S. have been

shaped by food surpluses. A shift in concern from global food security to environmental

quality in the 1980s (York, 1991) led to the perspective that ". the question is not can

we produce more food, but what are the ecological consequences of doing so?"

(Douglass, 1984, p. 5). However, much of the concern about sustainability in less

developed countries is related to the need to increase productivity to meet future needs of

growing populations (Ruttan, 1988; York, 1988, 1991; Lynam and Herdt, 1989;

Plucknett, 1990). The potential for the desperation imposed by poverty to shorten

people's planning horizons (Ashby, 1985) raises questions about the ecological

consequences of failing to produce more food (Mellor, 1988; Oram, 1988). The

alternative agriculture movement has not adequately addressed the need to feed rapidly

growing populations in order to prevent both human and ecological disaster.

The second problem is that a distorted caricature of conventional agriculture may

cause approaches that may enhance sustainability to be ignored or rejected because of

their association with conventional agricultural institutions. Although the philosophical

roots of the alternative agriculture movement formed outside of the academic community










16

(Rodale, 1990; Bidwell, 1986), most of the practices it promotes as sustainable are largely

products of mainstream research and educational institutions (Francis and Sahs, 1988;

York, 1988).

Third, establishing the contribution of an approach to sustainability through

definition eliminates the perceived need to evaluate approaches that may be poor or

harmful in a particular context. If strategies are identified as sustainable based on their

effect on agricultural systems, and agricultural systems are then judged to be sustainable

based on their implementation of sustainable strategies, then a form of circular logic

results. It is logically impossible to evaluate the contribution of an approach to

sustainability when adherence to that approach has already been used as a criterion for

evaluating sustainability. This circular logic is a fourth reason why interpreting

sustainability as an approach is not useful for guiding change.

Because of the temporal nature of sustainability, errors of either ignoring

approaches that enhance sustainability or promoting approaches that threaten it may not

be obvious when the approaches are implemented. The evaluation needed to recognize

errors and improve approaches is not possible if sustainability is interpreted as a

philosophy or a set of strategies. Thompson (1992) warned that

Our society may collapse because of shortsighted stupidity on the part of
pro-growth, resource exploiting power elites, but the collapse will only be
tragic if it is shortsightedness or ignorance on the part of environmentally
and ethically concerned people that helps bring it about. (p. 19)













Sustainability as a Property of Agriculture



The concept of sustainability as an approach to agriculture evolved in parallel with

the concept of sustainability as a system property. While Dahlberg (1991) argued that

"sustainability" was first used by an emerging alternative agriculture movement to

prescribe a particular set of values, Kidd (1992) countered that the system describing

concept developed earlier but did not use the word "sustainability" until later. As a

property of agriculture, sustainability is interpreted as either the ability to satisfy a diverse

set of goals or an ability to continue through time.


Sustainability as an ability to satisfy goals


A sustainable agricultural system is often defined as one that fulfills a balance of

several goals through time. These goals generally include some expression of maintenance

or enhancement of the natural environment, provision of human food needs, economic

viability, and social welfare (Table 2-1).

Lynam and Herdt (1989) argued that an interpretation of sustainability based on

several qualitative goals fails to provide a criterion useful for guiding agricultural research.

Ifa system is defined as sustainable when it protects the natural environment, provides

adequate food, and maintains producer profitability, then there is no logical way to rank,

for example, the relative importance of commodity price variability and nitrate leaching

into aquifers as determinants of sustainability. Furthermore, the subjectivity of goal













Sustainability as a Property of Agriculture



The concept of sustainability as an approach to agriculture evolved in parallel with

the concept of sustainability as a system property. While Dahlberg (1991) argued that

"sustainability" was first used by an emerging alternative agriculture movement to

prescribe a particular set of values, Kidd (1992) countered that the system describing

concept developed earlier but did not use the word "sustainability" until later. As a

property of agriculture, sustainability is interpreted as either the ability to satisfy a diverse

set of goals or an ability to continue through time.


Sustainability as an ability to satisfy goals


A sustainable agricultural system is often defined as one that fulfills a balance of

several goals through time. These goals generally include some expression of maintenance

or enhancement of the natural environment, provision of human food needs, economic

viability, and social welfare (Table 2-1).

Lynam and Herdt (1989) argued that an interpretation of sustainability based on

several qualitative goals fails to provide a criterion useful for guiding agricultural research.

Ifa system is defined as sustainable when it protects the natural environment, provides

adequate food, and maintains producer profitability, then there is no logical way to rank,

for example, the relative importance of commodity price variability and nitrate leaching

into aquifers as determinants of sustainability. Furthermore, the subjectivity of goal










specification links criteria for determining sustainability to the goals and values of the

analyst or the author of a definition rather than to the agricultural system. At the farm

level and higher, goals belong to the actors within the system and are, therefore,

endogenous. Kidd (1992) argued that it is not helpful to "use sustainability loosely as a

general purpose code word encompassing all of the aspects of agricultural policy that the

authors consider desirable" (p. 24).


Sustainability as an ability to continue


The final concept interprets sustainability as a system's ability to continue through

time. Hildebrand (1990) suggested that sustainability may be interpreted as the length of

time that a system can be maintained. According to Hamblin (1992), sustainability implies

that agriculture remains the dominant land use. Lynam and Herdt (1989) and Jodha

(1990) expressed sustainability in terms of maintaining some level of output. Monteith

(1990) added consideration of the possible confounding interaction of changes in input

and output levels. The definitions of Fox (1991) and Hamblin (1992) emphasized the

continuing ability to meet human needs. Conway (1985), Conway and Barbier (1990) and

Altieri (1987) emphasized the ability to withstand disturbances.

Interpreting sustainability as an ability to continue is consistent with literal English

usage of "sustain" and its derivatives. Its potential usefulness comes from suggesting

criteria for characterizing sustainability, providing a basis for identifying constraints and

evaluating proposed approaches to its improvement. This potential usefulness has been

limited by inadequacy of current approaches for characterizing sustainability.













Approaches to Characterizing Sustainability



Characterization is a prerequisite to applying the concept of sustainability as a

criterion for identifying constraints, focusing research, and evaluating and improving

agricultural policy and practices. The conceptual problem of defining sustainability and

methodological problems imposed by its temporal nature have hindered development of

approaches to characterizing sustainability. Sustainability involves future outcomes that

cannot be observed in the time-frame required for intervention (Lynam and Herdt, 1989;

Harrington, 1992). For this reason, Conway (1994) argued that defining sustainability in

terms of preservation or duration has little practical value.

The variety of approaches reviewed here reflects the different interpretations of

sustainability and methodological difficulties that result from its temporal nature.

Characterization by adherence to prescribed approaches is based on an interpretation of

sustainability as an approach to agriculture. Characterization by multiple qualitative

indicators and attempts to integrate such indicators are consistent with interpreting

sustainability as an ability to satisfy diverse goals. Sustainability as an ability to continue is

usually characterized by time trends or resilience.


Adherence to prescribed approaches


In a study comparing conventional and sustainable farms in South Dakota, Dobbs

et al. (1991) identified farms as sustainable if they reduced chemical inputs relative to













Approaches to Characterizing Sustainability



Characterization is a prerequisite to applying the concept of sustainability as a

criterion for identifying constraints, focusing research, and evaluating and improving

agricultural policy and practices. The conceptual problem of defining sustainability and

methodological problems imposed by its temporal nature have hindered development of

approaches to characterizing sustainability. Sustainability involves future outcomes that

cannot be observed in the time-frame required for intervention (Lynam and Herdt, 1989;

Harrington, 1992). For this reason, Conway (1994) argued that defining sustainability in

terms of preservation or duration has little practical value.

The variety of approaches reviewed here reflects the different interpretations of

sustainability and methodological difficulties that result from its temporal nature.

Characterization by adherence to prescribed approaches is based on an interpretation of

sustainability as an approach to agriculture. Characterization by multiple qualitative

indicators and attempts to integrate such indicators are consistent with interpreting

sustainability as an ability to satisfy diverse goals. Sustainability as an ability to continue is

usually characterized by time trends or resilience.


Adherence to prescribed approaches


In a study comparing conventional and sustainable farms in South Dakota, Dobbs

et al. (1991) identified farms as sustainable if they reduced chemical inputs relative to










20

typical farms, and included rotations, legumes, tillage and cover crops for management of

fertility, erosion and weeds. Sustainable farms were sampled by sending questionnaires to

farmers they ". believed might be using greatly reduced or even zero levels of synthetic

chemicals in their farming operations" (p. 111). Cordray et al. (1993) characterized the

sustainability of farmers in Washington and Oregon based on changes in agricultural

chemical use and adoption of alternative production practices. Taylor et al. (1993)

developed a quantitative index of sustainability based on production practices of

Malaysian cabbage farmers. Practices were assigned values according to their "inherent

sustainability" determined by consensus of the research team, weighted by their expected

contribution to sustainability, then combined into a composite index evaluated for each

farm.

Taylor et al. (1993) were the only authors in the above studies to acknowledge the

necessity of assuming a relationship between farmer practices and future viability if

practices are to serve as a basis for characterizing sustainability. Even if evidence

supporting such a relationship were presented, circular logic prevents using sustainability

measured by adoption of particular practices as a criterion for evaluating and improving

agricultural practices.


Multiple qualitative indicators


Torquebiau (1992) used a set of indicators to characterize sustainability of tropical

agroforestry home gardens. Several system attributes that were believed to influence

sustainability were identified related to the resource base, system performance, and effects












on other systems. Measurable indicators were identified for each system attribute. A

negative change in an individual indicator indicated unsustainability. Jodha (1990)

developed a set of indicators of unsustainability of mountain agriculture in the Himalayan

region consisting of visible changes in natural resources and farming practices that indicate

system degradation, changes in farming practices that compensate for less visible changes

in the environment, and inappropriate development initiatives that may lead to negative

impacts. Neher (1992) described an approach that is being developed to monitor

agroecosystem health at regional and national scales in the U.S. A number of indicators of

environmental quality and agricultural performance are to be measured to provide a

baseline, then monitored to identify changes.

Monitoring sets of qualitative indicators is consistent with interpreting

sustainability as the ability to meet a diverse set of goals and the belief that no single

indicator can exist (Geng et al., 1990; Norgaard, 1991). However, diverse sets of

indicators are difficult to interpret and do not provide mechanisms for diagnosing causes

of unsustainability, or for evaluating effects of proposed interventions. Such indicators do

not facilitate establishing cause-effect relationships between diverse system properties.


Integrated, quantitative indicators


Increasing recognition of the need for quantification has motivated efforts to

combine diverse indicators.of sustainability into integrated, quantitative measures. One

goal of the project described by Neher (1992) for monitoring agroecosystem health is to

combine indicators into an aggregate measure of agricultural sustainability in a manner












that balances productivity, environmental soundness and socioeconomic viability goals.

Lal (1991) proposed a sustainability coefficient as a function of output per unit of input at

optimal per capital productivity or profit, output per unit of decline in the most limiting or

least renewable resource, and the minimum assured output level. Sands and Podmore

(1993) proposed an environmental sustainability index as an aggregation of sub-indices of

soil productivity, ecosystem stability, and potential to degrade the environment. Selection

of components of the sub-indices and the form of aggregation functions were indicated as

important research topics. Stockle et al. (1994) proposed a framework for evaluating

sustainability based on nine system attributes: profitability, productivity, quality of soil,

water, and air, energy efficiency, fish and wildlife habitat, quality of life, and social

acceptance. Production system sustainability is determined by scoring attributes as

weighted functions of quantifiable, long-term constraints, then combining weighted

attribute scores into an integrated measure.

The consistent inability to specify aggregation functions in these studies points to

the weakness of interpreting sustainability as the ability to fulfill diverse sets of goals as a

conceptual foundation for characterization. Diagnosis is limited by need to decide a-priori

the relative importance of different types of constraints to sustainability. Stockle et al.

(1994) acknowledged and defended the subjectivity needed to aggregate diverse system

attributes into an integrated measure of sustainability.












Time trends


Time trend approaches express sustainability in terms of the direction and degree

of measurable changes in system properties through time. Lynam and Herdt (1989)

regarded a system as sustainable if there is a non-negative trend in its output. They

proposed total factor productivity, the total value of system outputs divided by the value

of system inputs, as the output criterion because it accounts for changes in the value of

inputs. Ehui and Spencer (1992) extended total factor productivity to account for changes

in the value of natural resource stocks, particularly soil nutrients. Hedgerow intercropping

in Nigeria was determined to be unsustainable without correction for soil nutrient flows,

but sustainable after accounting for nutrients. Monteith (1990) proposed determining

sustainability from a contingency table of trends of inputs and outputs (Table 2-4). Cereal

production was determined to be sustainable in the Karimnagar district in Andhra Pradesh,

India, based on increasing yields and decreasing land use during 27 years. Decreases in

both land use and yields in the Adilabad district prevented inference about sustainability.



Table 2-4. Contingency table for inferring sustainability based on trends of
system inputs and outputs (Monteith, 1990).
--------------------Inputs-------------------------

Outputs decreasing constant increasing

decreasing indeterminant unsustainable unsustainable
constant sustainable sustainable unsustainable

increasing sustainable sustainable indeterminant












Characterizing sustainability by time trends is appealing because of its simplicity.

The slope of the estimated trend line provides a quantitative index with an intuitive

interpretation as a rate of system deterioration or enhancement. Trends represent an

aggregate response to several determinants of sustainability, eliminating the need to devise

and defend aggregation schemes.

The assumption needed to infer sustainability from trends--that future rates of

system degradation can be approximated by past rates--is often difficult to defend.

Unsustainability can express itself either as a gradual change or as an abrupt collapse

(Conway, 1985; Trenbath et al., 1990). Furthermore, much of the concern about

sustainability comes from recognition that agriculture is being impacted by unprecedented

changes in population pressure, resource demands, market structures, and technology.

Another weakness is the manner in which time-trend approaches interpret temporal

variability. Variability tends to hinder sustainability by driving subsistence farmers to

desperation, leading to environmental degradation that may not recover during normal or

good periods (Mellor, 1988). Price and yield variability have also been shown to increase

the probability of farm failure in the U.S. (Grant et al., 1984; Perry et al., 1986).

However, when characterization is based on time trends, either variability is ignored or it

implicitly enhances sustainability by reducing the probability of identifying a significant

negative trend.

A final criticism is that applications of time trends to sustainability have examined

levels of system performance without considering the levels of needs and goals of the

individuals or segments of society who decide on the fate of those systems.












Resilience


Conway (1985) defined sustainability as resilience: ". the ability of a system to

maintain productivity in spite of a major disturbance" (p. 25). He suggested that

measurement of five system properties are necessary to characterize resilience: inertia,

elasticity, amplitude, hysteresis, and malleability (Conway, 1994). Cramb (1993) based

inferences about the sustainability of two shifting cultivation systems in eastern Malaysia

on both trends and resilience. Pepper production was determined to be sustainable

because production in 1989 recovered to its 1980 level in response to price recovery after

production diminished (in 1985) due to a period of low prices. Rubber production was

considered more sustainable at Batu Linang than at Nanga Tapih, as indicated by recovery

of depressed production in response to price recovery.

Like time trends, resilience can be viewed as an aggregate system response to

determinants of sustainability. However, inability to identify a single measure of resilience

(Conway, 1994) leads to the same problems of interpretation faced when using a diverse

set of indicators to characterize sustainability. Assumptions about the likelihood and

timing of disturbances have been avoided by interpreting sustainability as an intrinsic

property of an agricultural system in isolation from its environment (Conway and Barbier,

1990). However, York (1988) argued that (un)sustainability is not an intrinsic property

but rather a response to changing environmental and socioeconomic circumstances.

Predictions about future sustainability cannot be made in the absence of assumptions about

changes and variability in those higher-level systems that comprise a system's










26

environment. Resilience shares with time trend approaches the criticism that it ignores the

goals of the human actors within agricultural systems.


System simulation


Simulation has been used to characterize the sustainability of crop production in

response to soil dynamics. Singh and Thornton (1992) illustrated the use of long-term

simulation of crop sequences replicated with stochastic inputs of weather data to examine

trends and variability in yields. Lerohl (1991) used the Erosion-Productivity Impact

Calculator (EPIC) (Williams et al., 1984) to study the long-term impact of predicted soil

erosion on productivity of crop rotations on four soil types in Alberta, Canada.

Sustainability was inferred in all soil-rotation combinations because no negative trend in

crop yields could be detected during a simulated 100 year period.

Other studies have used crop simulation models to examine relationships between

production and environmental degradation. Singh and Thornton (1992) illustrated the use

of CERES-Maize (Jones and Kiniry, 1986) to simulate the effects of soil type and rate of

application of N fertilizer on distributions of maize grain yield and NO, leaching into

groundwater from upland fields in Chiang Mai, Thailand. Alocilja and Ritchie (1993) used

CERES-Maize and a multiple goal optimization technique to identify sets of N fertilization

schedules that were optimal in the sense that neither production nor water quality could be

improved without decreasing satisfaction of the other goal.

Several whole-farm simulation studies have looked at the effect of various factors

on farm survivability. For example, Perry et al. (1986) examined effects of production












costs, labor availability, rice grain quality, land tenure, trends and variability of rice and

soybean prices and yields, beginning equity, type of rotations, and participation and terms

of government farm programs on probability of farm survival during a five year period.

Production and environmental processes were not simulated. Although this and similar

studies have not generally used the word "sustainability," the use of farm survival as a

criterion is consistent with an interpretation of sustainability as economic viability

(Madden, 1987; Lockeretz, 1988; Dicks, 1992; Neher, 1992). Survivability addresses

shortcomings of other approaches by integrating levels, trends and variability in system

performance with the needs and goals of farmers.

System simulation is a tool, and does not suggest a particular criterion for

evaluating sustainability. Simulation can be used to examine future impacts of alternative

interventions across the range of expected variability in a manner that is not possible with

empirical observation and experimentation. The value of simulation is limited by

capabilities of, and confidence in simulation models, by availability and reliability of input

data, and by lack of methods for designing and interpreting simulation studies for

characterizing sustainability. So far, there has been little integration of models of crop and

animal production, environmental degradation, economic processes, and farmer decisions.



Elements of a Useful Approach for Characterizing Sustainability



In order for sustainability to be a useful criterion for guiding change in agriculture,

several elements should be incorporated into approaches to its characterization (Table 2-










28

5). First, characterization should be based on a literal interpretation of sustainability (Fox,

1991). Regardless of the merits of goals and ideals frequently incorporated into

definitions of sustainability, if the idea of continuation through time is omitted then those

ideals and goals are something other than sustainability.


Table 2-5. Elements of a useful approach to characterizing sustainability of agricultural
systems.
Element Explanation

Literal Defines sustainability as an ability to continue through time, consistent
with literal English usage.

System- Identifies sustainability as an objective property of a particular
oriented agricultural system whose components, boundaries, and context in
hierarchy are clearly specified.
Quantitative Treats sustainability as a continuous quantity, permitting comparisons of
alternative systems or approaches.

Predictive Deals with the future rather than the past or present.

Stochastic Treats variability as a determinant of sustainability and a component of
predictions.
Diagnostic Uses an integrated measure of sustainability to identify and prioritize
constraints.


Second, characterization should be system-oriented. A literal interpretation

suggests that sustainability is an objective property of an agricultural system. It cannot be

a property of approaches to agriculture if it is to serve as a basis for evaluating and

improving approaches. Lynam and Herdt (1989) argued that sustainability is a relevant

criterion for evaluating technology only when the system is clearly specified, including its

boundaries, components, and context in hierarchy. Sustainability has meaning only in the

context of specific temporal and spatial scales. Fresco and Kroonenberg (1992) cited a












number of examples in which disturbances that threaten sustainability at one spatial and

temporal scale could be seen as natural cycles at broader scales. Both constraints to

sustainability and factors that can be managed for its enhancement depend on the level of

the system (Spencer and Swift, 1992). The objectivity that results from a system-oriented

approach is essential for guiding change, but may work against motivating change because

it may call prescribed approaches into question.

Third, an approach to characterizing sustainability should be quantitative.

Although MacRae et al. (1989) cited quantification as a barrier to sustainability, others see

it as a prerequisite to using sustainability as a criterion for evaluating and improving

agricultural systems (Monteith, 1990; Harrington, 1992). Sustainability is often treated as

a discrete property: "A farm is either sustainable or it's not sustainable. Simply by

definition, you cannot create a system that is half sustainable" (Rodale, 1990, p. 273).

However, comparisons among agricultural systems or alternative approaches are possible

only when sustainability is treated as a continuous quantity.

Fourth, since sustainability deals with future changes, its characterization must be

predictive of the future rather than merely descriptive of the past or present (Harrington,

1992). Sustainability has little meaning after the fact. The deterministic view that ". .. a

farm will either last for a very long period, or it won't" (Rodale, 1990, p. 273) does not

take into account the uncertainty of predictions resulting from the inherent variability of

the farming system's environment. A stochastic approach, the fifth element, recognizes

variability as a determinant of sustainability and appropriately expresses predictions in

terms of probabilities.









30

Finally, characterization of sustainability should be diagnostic. Sustainability is a

useful concept when its characterization focuses research and intervention by identifying

and prioritizing constraints. Diagnosis can be accomplished by testing hypotheses about

constraints based on a measure of sustainability that is both comprehensive and integrated.

Diagnosis is facilitated by use of a single measure of sustainability that combines the range

of possible determinants into a single, integrated measure of system response. An

integrated measure is necessary for comparing, for example, the relative impact of nitrate

leaching into aquifers and product price volatility on sustainability.

Weaknesses of the reviewed approaches for characterizing sustainability can be

related to their failure to incorporate the proposed elements (Table 2-6). Characterization

based on adherence to prescribed approaches fails because it is not founded on a literal

interpretation of sustainability. Lack of integration limits the usefulness of multiple

indicators of sustainability for diagnosing and prioritizing constraints. Integration of

indicators has been difficult because the underlying interpretation of sustainability as an

ability to meet diverse goals is not integrated. A time trend represents an integrated

system response that is potentially useful for diagnosis and can be predictive by

extrapolation, but it is not stochastic in the sense of accounting for variability. An

integrated measure of resilience has not yet been found. The assumptions about future

variability and disturbances necessary for resilience to be predictive and stochastic are

avoided in discussions of its use for characterizing sustainability. Simulated farm

survivability is the only approach reviewed that incorporates all of the elements listed.















Table 2-6. Approaches to characterizing agricultural sustainability. A check (V) indicates the approach incorporated the specified
element. A question mark (?) indicates it addressed some aspect of the element.

Approach Reerence Literal System- Quanti- Predic- Sto- Diag-
oriented tative tive chastic nostic

Reduced use of chemicals relative to other farmers in South Dakota Dobbs et at., 1991

Quantitative index of cabbage farmer practices in Malaysia Taylor et al., 1993 7

Adoption of alternative practices and reduced chemical use on U.S. Pacific Northwest farms Cordray et al., 1993 /

Indicators of resource base, system performance &external effects in agroforestry gardens Torquebiau, 1992 / /

Indicators of visible, masked, & potential degradation in Hymalaian farming systems Jodha, 1990 / 7

Indicators of regional "agroecosystem health" Neher, 1992 ?

Index of assured output, and output per unit input and per unit limiting resource decline Lal, 1991 / /

Index of productivity, stability and degradivity Sands & Podmore, 1993 / /

Subjectively weighted index of constraints to nine system attributes Stockle et al., 1994/ /

Non-negative time trend in output Lynam & Herdt, 1989 / / / 7

Total factor productivity accounting for natu-ral resources in cropping systems in Nigeria Ehui & Spencer, 1992 / / /

Regional trends in inputs and outputs in India Monteith, 1990 / / / 7

Properties representing resilience Conway, 1994 / / ? ?

Production trends and resilience in shifting cultivation in Malaysia Cramb, 1993 / / 7

Simulated trends and variability in crop production and NO; leaching Singh & Thornton, 1992 / / / / /

Simulated crop production trends in response to simulated soil erosion in Canada Lerohl, 1991 / / / /

Simulated survivability of Texas rice farms Perry etal., 1985 / / / / / /













Conclusions



The importance and desirability of agricultural sustainability are generally

recognized. However, its potential as a criterion for guiding agriculture's response to

change has not been realized. Characterization is a prerequisite to using sustainability as a

basis for guiding change. Logical inconsistencies limit the usefulness of characterization

of sustainability interpreted as an ideological or management approach to agriculture.

Interpreting sustainability as an ability to meet diverse goals suggests measuring sets of

system indicators consistent with those goals. However, these measurements have proven

difficult to integrate and interpret in a way that identifies constraints or focuses research.

Literal interpretations of sustainability as an ability to continue into the future suggest

measurable, integrated criteria for its characterization. However, applications of these

criteria--time trends and resilience--have ignored or misinterpreted important aspects of

system behavior. Criteria are needed that relate levels, trends and variability of long-term

system performance to the needs and goals of farmers and of society.

In order for sustainability to be a useful criterion for guiding change in agriculture,

its characterization should be literal, system-oriented, quantitative, predictive, stochastic

and diagnostic. These elements identify weaknesses in existing and proposed approaches,

suggest directions for future development of approaches, and together constitute a

systems approach for characterizing sustainability of agricultural systems. The tools of

system analysis and simulation must be part of approaches that incorporate these elements.














CHAPTER 3
A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM SUSTAINABILITY



Introduction



The potential benefits of applying the concept of agricultural sustainability--

providing feedback about future impacts of current decisions, and focusing research and

intervention by identifying constraints--can only be realized when sustainability of

particular systems is characterized. Characterization includes both quantification and

diagnosis of constraints.

In spite of the tremendous amount of concern about agricultural sustainability,

surprisingly few studies have attempted to characterize the sustainability of specific

agricultural systems. The methods which have been proposed or applied suffer from (a)

conceptual problems associated with interpreting sustainability as an approach rather than

a property of agriculture, and (b) practical difficulties that arise from the fact that

sustainability deals with the future (Chapter 2). Characterization based on management

practices either does not relate to a literal interpretation of sustainability or it leads to

circular logic. Attempts to characterize sustainability based on system response have

generally ignored or misinterpreted important system properties.














CHAPTER 3
A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM SUSTAINABILITY



Introduction



The potential benefits of applying the concept of agricultural sustainability--

providing feedback about future impacts of current decisions, and focusing research and

intervention by identifying constraints--can only be realized when sustainability of

particular systems is characterized. Characterization includes both quantification and

diagnosis of constraints.

In spite of the tremendous amount of concern about agricultural sustainability,

surprisingly few studies have attempted to characterize the sustainability of specific

agricultural systems. The methods which have been proposed or applied suffer from (a)

conceptual problems associated with interpreting sustainability as an approach rather than

a property of agriculture, and (b) practical difficulties that arise from the fact that

sustainability deals with the future (Chapter 2). Characterization based on management

practices either does not relate to a literal interpretation of sustainability or it leads to

circular logic. Attempts to characterize sustainability based on system response have

generally ignored or misinterpreted important system properties.












Since sustainability deals with the future, it cannot be readily observed. Analysis

and simulation of a model system can compensate for the limitations of observation and

experimentation on a real agricultural system. In this chapter, I first present a definition of

sustainability that applies generally to dynamic, hierarchical, stochastic, purposeful

systems. I then describe a framework for using system simulation to quantify sustainability

and test hypotheses about its constraints. A Monte-Carlo study of a simple time-series

model demonstrates how sustainability relates important components of system behavior--

mean, trend, variability and autocorrelation--to threshold goal levels. I then discuss issues

that arise when applying the framework to farming systems. Finally, I illustrate an

application of the framework for characterizing farm sustainability using data from a

previously published farm simulation study.



Defining Sustainability



To sustain is literally "to keep in existence; keep up; maintain or prolong"

(Neufeldt, 1988, p. 1349). Sustainability can therefore be defined as the ability of a

system to continue into thefiuture. Key words of this definition suggest a framework for

quantification.

First, the system addresses the question, "What is to be sustained?" Although the

system of interest in this paper is a farm, the concept of sustainability can be applied to any

system that is dynamic, stochastic, and purposeful. The ability to continue does not apply

to a static system. Furthermore, sustainability has no meaning unless some purpose or










35

threshold condition exists which distinguishes a system that is sustaining from one that has

failed. Finally, sustainability of a deterministic system is binary; in the absence of

uncertainty one can say that a system will either sustain itself or fail during some future

period.

Second, the word continue implies the possibility that a system can fail if some

criteria are met. Failure occurs when a system can no longer fulfill its purpose. Failure

implies irreversibility: a degree of stress from which the system cannot readily recover.

Criteria for failure address the question, "Above what minimum level is the system to be

sustained?"

Third, the future suggests a time period which extends from the present (t = 0) to

some future time (t = 7). For a stochastic system, the future also implies uncertainty;

uncertainty distinguishes the future from the past. The time period addresses the question,

"How long is the system to be sustained?"

Finally, the ability of the system to continue in the future is best expressed as a

probability. The suggestion that "A farm is either sustainable or it's not sustainable ... A

farm will either last for a very long period or it won't" (Rodale, 1990, p. 273) expresses a

common deterministic interpretation of sustainability. However, one cannot determine

with certainty whether a system will continue through some future period. Probability of

continuation provides a measure of sustainability with a zero-to-one range that addresses

the question, "With what degree of certainty will the system sustain itself?" The definition

of sustainability can be restated as the probability that a particular system will not meet

specified criteria for failure during a particular future period.













Quantifying Sustainability



Consider the status of a system, D(t), as a Bernoulli process with state space {0,

1} operating in the period from the present (t = 0) to some future time (t = T) (See symbol

definitions in Table 3-1). Continuation is indicated by D(t) = 1 and failure by D(t) = 0.

The system is initially operating: D(0) = 1. Time of failure, TF, is then a random variable

with a probability density function, f,(t) = P{TF = t), and a cumulative distribution, FrT(t)

= P{Tp t}. The distribution FTF applies to the population of possible time paths, or

realizations, of system behavior. Failure is irreversible such that if D(t) = 0 then D(t+At) =

0 for all At > 0. For the period (0, T], sustainability, S, is defined as,


S(T) = 1 FTr(T). [3-1]


The definition given in Eq. [3-1] is equivalent to the survival function in mortality studies

(eg., Elandt-Johnson and Johnson, 1980) and to reliability in quality control literature (eg.,

Barlow, Proschan and Hunter, 1965).


Failure as violation of state thresholds


Time to failure, T,, is a random variable because the system's state, x(t), behaves

as a stochastic process as it responds to a stochastic environment. Consider a single,

continuous state variable, x(t). At any given time t, x(t) has a probability density function,

fx,,(x), and a cumulative distribution, Fx,(x) that apply to an initial population selected at













Quantifying Sustainability



Consider the status of a system, D(t), as a Bernoulli process with state space {0,

1} operating in the period from the present (t = 0) to some future time (t = T) (See symbol

definitions in Table 3-1). Continuation is indicated by D(t) = 1 and failure by D(t) = 0.

The system is initially operating: D(0) = 1. Time of failure, TF, is then a random variable

with a probability density function, f,(t) = P{TF = t), and a cumulative distribution, FrT(t)

= P{Tp t}. The distribution FTF applies to the population of possible time paths, or

realizations, of system behavior. Failure is irreversible such that if D(t) = 0 then D(t+At) =

0 for all At > 0. For the period (0, T], sustainability, S, is defined as,


S(T) = 1 FTr(T). [3-1]


The definition given in Eq. [3-1] is equivalent to the survival function in mortality studies

(eg., Elandt-Johnson and Johnson, 1980) and to reliability in quality control literature (eg.,

Barlow, Proschan and Hunter, 1965).


Failure as violation of state thresholds


Time to failure, T,, is a random variable because the system's state, x(t), behaves

as a stochastic process as it responds to a stochastic environment. Consider a single,

continuous state variable, x(t). At any given time t, x(t) has a probability density function,

fx,,(x), and a cumulative distribution, Fx,(x) that apply to an initial population selected at












Table 3-1. Description of symbols used.
Symbol Description
t, T Time variable and a particular time
D(t) Status of a system continuingi, 0=failed)
TF Time to system failure

fT, F, Density and cumulative distribution of TF
S(T) Sustainability for the period (0, 7]
h(t) Sustainability hazard probability function
x, x State vector and a particular state variable

x0, xo Failure threshold values for x and x

fx,,, Fx,, Density and cumulative distribution ofx at time t
N Total number of realizations simulated
n(T) Number of realizations continuing at time T
S estimated from a finite number of realizations
SEs Standard error of S

z(t), oz A stationary stochastic process and its standard deviation
a, p Intercept and slope of a deterministic trend

C1 One period lag autocorrelation coefficient
e(t), o0 A white-noise process and its standard deviation

Yo, Y, Base and alternate value of the ith hypothesized determinant of sustainability
9o, ~i Sin response to the base and alternate value of the ith factor
Ri, ri Absolute and relative sensitivity to the ith factor
vab, N Frequency of a and b occurring, and sum of all frequencies
q Correction factor for continuity in G-tests

G1, Gadj Uncorrected and corrected statistic for tests of independent frequencies,
independent observations

Gp, Gp,dj Uncorrected and corrected statistic for tests of independent frequencies,
paired observations










38

time 0. Assume that a threshold value, x0, exists such that the system will fail the first time

x < xo. Ifxo represents a maximum threshold, a sign change is required to allow the

relationship to hold. We can express the relationship between x(t) and TF by considering

the probability that failure occurs during an interval (t, t+At], At > 0:


P{t< TF r t+At} = P{TF t+At I T> t} P{TF> t}, [3-2]


and, since the condition T, < t+At is equivalent to x(t+At) < xo, we can rewrite Eq. [3-2]

as,


P{t< T < t +At) = P{x(t +At) sxo I x(t)>xo (1 -P{TF t}). [3-3]


As At approaches 0, the left side of Eq. [3-3] becomes the probability density of time to

failure:


lim
tlm P{t< T,< t+At = P{T,=t}

= fTF() [3-4]



Since Fx,, is the distribution ofx for the population of realizations that have continued to

time t, Fx,t(xo) expresses the probability that the system violates the threshold xo at t, given

that it has not done so previously. Therefore, taking the limit of the first term on the right

side of Eq. [3-3] gives,


lim
lt 10 P{x(t +At) < xo I x(t) > x0} = Fx,(xo). [3-5]
Atl 0










39

Finally, the last term in Eq. [3-3] is equivalent to sustainability from Eq. [3-1]. By

taking the limit ofEq. [3-3] as At approaches 0 and substituting the simplified terms (Eq.

[3-4] and [3-5]), we obtain,


fTF(t) = Fx,t(xo) S(t). [3-6]


We can now derive an expression of S as a function of only Fx,,. Differentiating Eq. [3-1]

gives,

dS(t) / dt = -fT,(t). [3-7]


Substituting Eq. [3-6] into [3-7] gives the differential equation,


dS(t) / dt = Fxt(x) S(t), [3-8]


which has the solution,


S(T) = exp( fTFx,(Xo) dt [3-9]



at time T. Thus, I have shown that sustainability is determined entirely by the probability

that a system's state falls below a threshold value during some time interval (0, T].

Two examples illustrate how time, thresholds, and the distribution of system state

interact to determine sustainability. In the first example (Fig. 3-la), the variability ofx is

relatively high but its expected value, E[x(t)], remains constant. The system has a 0.15

probability of failing in any particular period. Sustainability declines exponentially with























period, t


period, t


S0.8 O8
0.8
0.6 0.6

c: 0.4 M0
- C' 0.4
I 0.2 cn
n :3 0.2
0.0 I I I I I 0.0 ,
0 1 2 3 4 5 6 0 1 2 3 4 5 6
period, t period, t

Figure 3-1. Relationship between time, distribution of state, and sustainability under (a) constant mean and high variability,
and (b) negative trend and low variability.












time as Eq. [3-9] predicts. In the second example (Fig. 3-1b), variability is lower, but

E[x(t)] decreases linearly with time. Here, the effect of the negative trend in x on

sustainability is not apparent until E[x(t)] approaches x0.

Several generalizations can be applied to the preceding discussion. First, a failure

threshold may exist for more than one system state variable. Second, the thresholds may

be dynamic. The state of a complex system and the corresponding set of minimum

threshold values could then be expressed as vector processes: x(t) and x0(t). The

threshold vector xo(t) bounds the system's state space in one or more dimension. Since

only a few of the state variables may be directly related to the ability of a system to

continue, many members of x(t) may have values of -o for all t. Third, the threshold

vector could be stochastic. The probability of violating system thresholds would then

depend not only on the behavior of x(t) but also on the dispersion of xo(t) and its

correlation with x(t). Finally, thresholds may apply to derived state variables such as a

sum (eg., aggregate wealth) or ratio (eg., a financial ratio) of basic state variables.


Sustainability hazard


The value, Fx,,(xo), is the instantaneous probability of failure applied to the

population of realizations that have continued up to time t. It is referred to as a hazard

function,


h(t) = fTrt) / (1 FT(t))

= f(t) / S) [3-10]














(Barlow et al., 1965). Hazard is a probabilistic expression of the intensity of stress on a

system as a function of time; increasing h indicates an increasing stress or increasing threat

to sustainability. By rearranging and substituting Eq. [3-10] into [3-7], and integrating,

we arrive at,


S(T) = exp h(t) dt. [3-11]




We see that constant hazard results in exponentially declining sustainability. Although S(t)

increases monotonically from an arbitrary starting time (t = 0), the threat to sustainability

expressed by h(t) is independent of starting time, and may increase or decrease.



Simulating Sustainability



Equations [3-1] and [3-9] cannot be used to calculate sustainability in practice

because the distributions of T, and x(t) are generally not known for a real system.

Harrington et al. (1990) expressed the need to artificially construct time paths for current

and alternative strategies to assess sustainability before long-term experiments or

monitoring could be completed. A set of replicated time paths, or realizations, can be

sampled by stochastic simulation of a model of the system.

Consider a set of N realizations simulated for the period (0, T] (Fig 3-2). Although

the environment is sampled randomly, each replicate has the same initial conditions, x(0).









43







S- -* -x2 )W



Z to-) wo
/ .-"^^ 4% -



.. ,xW(t)




xI------------.x--s----






1.0




S(T) ------------------

C


=3
4 4.










4~ 4---------- 4-+








0.0

To TF3 TF5 T

time, t


Figure 3-2. Estimating sustainability by sampling a small number (i.e. 5) of simulated
realizations of future system behavior.
",x()_ '._5t








*1n
Co
c (T












Let n(t) be the number of realizations continuing at time t. Sustainability can then be

estimated by the relative frequency of surviving realizations, or


S(T) = n(T)IN. [3-12]


Because of the uncertainty associated with estimating the probability of success from a

small Bernoulli trial, sustainability estimated from Eq. [3-12] has a standard error

(Snedecor and Cochran, 1980) of


SEg = /(T) (1 S(7)) /N

= n(T)(N- n(T)) N3. [3-13]



An example: sustainability of a simple time-series model


The proposed definition (Eq. [3-1]) integrates several important aspects of system

behavior--means, trends, variability and autocorrelation--and relates them to levels of

goals expressed as system thresholds. I used Monte Carlo simulation of a simple, discrete-

time, univariate time series model to demonstrate the role of these aspects of system

behavior in determining sustainability. The model consisted of a deterministic trend

component,


x(t) = a + pt + z(),












and a first-order autoregressive component,

z(t) = z(t- 1) + e(t),

where a and p are intercepts and slope of a time trend, z(t) is the value of a stationary time

series in period t, 4( is a first-order autoregressive coefficient, I~4 | < 1, and N(0, ao) is

a random shock variable. The standard deviation of the random shock (o() is related to

the asymptotic standard deviation of the generated time series (oz) by,



Gc = _, .



(Pankratz, 1983). I calculated S by Eq. [3-12] from 10,000 replicated realizations, each

simulated for 120 periods.

Table 3-2 summarizes values of the parameters and the results of the sensitivity

analysis. Sustainability, S(120), was 0.567 0.005 ( S.E.) in response to the base

parameters. Decreasing the mean, increasing the goal threshold, or increasing variability

reduced 5(120) by increasing the proportion of the population which fell below x0 at any

time before t=T. Decreasing autocorrelation reduced ( 120) because of its effect on Fx,,.

For discrete time periods, Fx, in any period I is conditioned on the prior status of the

system and, therefore, on the previous value of the state variable. Finally, extending the

time period Tbeyond 120 decreased S(T).

Figure 3-3 shows how declining expected value, and abrupt changes of expected

value, variance and autocorrelation influence the sustainability and hazard time functions













of the time-series model described above. The general pattern of the sustainability time

function remains the same whether hazard is modified by a change in the expected value,

variance or autocorrelation (Fig. 3-3c-h).



Table 3-2. Sensitivity of simulated sustainability to system properties. !120)=0.567 for
the base scenario.


Property

mean

trend

variability

autocorrelation

goal threshold

duration


Base
value

10.0

10.0
0.0

5.0

0.8

3.0

120


------Increased------
value ST)

11.0 0.783

9.0
0.566
+0.0167

5.5 0.401

0.88 0.924

3.3 0.491

132 0.535


-----Decreased------
value S(T)

9.0 0.307

11.0
0.517
-0.0167

4.5 0.774

0.72 0.235

2.7 0.641

108 0.604


Diagnosing Constraints to Sustainability


The potential value of the concept of sustainability lies in its ability to focus

research and intervention by identifying and ranking its constraints. Diagnosing

constraints entails a process of hypothesis formulation and testing using simulation of the

system model. Hypotheses should identify the current (or expected) value of a suspected

constraint, and a specific change that would relax the constraint. Sensitivity analysis then

provides the experimental tool for testing and ranking hypothesized constraints.


Para-
meter

a

a


(3,


xo
T















0.8. 60
12 /
.60.6 10 40
8
0..4 6
4 20
<( 0.2 / P = -0.045 \
.2 o,2= 4.0 B
0.0 I 0 I 0

1.0 25
C D
.0.8 20 30
oI 20

6 15 20

S0.4 I
10 1
0.2 5 5
a = 8.0 a =10.0 /a =10.0 a 8.0
0.0 0 0
1.0 16 16
E 14 F 14
0.8
12 12
.G0.6 10 10
S 10
.0.4 / 16 6

0.2 / N
So =9.0 =7.0 7.0 = 9.0
0.0 0 O0
1.0
r\ ^ -^ H ^ 20 E
,0.8 15

i0 0.6
/ 10
S / 10
C 0.4 -


= 0.87 $ =0.93 =0.87
0.0 0 0
0 40 80 120 0 40 80 120
Time, t Time, t


Figure 3-3. Sustainability and hazard of an AR(1) process with (a) constant parameters,
(b) declining expected value, (c) abrupt decrease and (d) increase in mean, (e) decrease
and (f) increase in variance, and (g) decrease and (h) increase in autocorrelation.
Parameter values were a = 10.0, 13 = 0.0, oa = 8.0, 4j = 0.9, x0 = 4.0 and n = 50,000
except as indicated otherwise.











Sensitivity analysis


Sensitivity analysis is used to quantify the relative importance of hypothesized

constraints to sustainability. It involves changing the value of a factor a small amount in

the direction that would relax the hypothesized constraint relative to a base scenario which

represents existing or expected conditions, then simulating the modified scenario.

Hypothesized constraints can then be ranked based on either absolute,


Ri= Si S [3-14]


or relative sensitivity,


Yi, -o
ri [3-15]



where Y,o is the value of the ith factor in the base scenario, Y, is its adjusted value, and 90

and are sustainability values estimated for the base and alternate scenarios. The

absolute value allows an increase in sustainability to result in ri > 0 regardless of the

direction of change in Yi. Relative sensitivity is interpreted as the percent change in g in

response to a 1% change in Y. Comparisons may be made and ranks assigned among

discrete or among continuous factors. However, absolute sensitivity to discrete factors

cannot be compared with relative sensitivity to continuous factors.












Significance tests


Independent observations. The frequencies (vij) of failure and continuation in the

simulation of the base and alternate scenarios used for a hypothesis test can be represented

by a two-way contingency table arranged as in Table 3-3, where i is the row andj is the

column. Several tests are available for the null hypothesis that frequencies are

independent, equivalent to the null hypothesis that sustainability is the same in the two

scenarios. Sokal and Rohlf (1981) recommended a log likelihood ratio, or G-test, over

the more frequently used X2 test. One can calculate the expected frequencies of

continuation or failure from the observed frequencies based on the null hypothesis of

independence:

j = (Vi1 Vi2) (VIj + j) /N

The GI statistic is then calculated from the observed and expected frequencies:

2 2
G, = 2 E E(vij n(vij /i)), [3-16]
i=1 j=1



and is corrected for continuity (Williams, 1976):


q = 1 + ((N/v1 + N/v2, 1)(N/v1 +N/v*2 1))/6N, [3-17]


Gadj =G q.


[3-18]












The G statistic is distributed approximately X2 with one degree of freedom. If Gldj > X 2

then reject Ho: S(T) = Si().




Table 3-3. 2x2 frequency table for tests of difference between simulated
sustainabilities, independent observations.
Scenario continued failed Total
base Vn V12 V* = V1+V12
alternate V21 V22 V2 = Y21 V22
Total: vi*=V11+v21 v2=V12+V22 N = ij




Paired observations. Different scenarios can be replicated under the same set of

environments by using the same pseudorandom number sequence for sampling stochastic

inputs. The additional information available from such a randomized block design permits

the use of a more powerful test. When data are arranged as in Table 3-4, the row or

column totals then represent the frequency of failure or continuation of the base (row

totals) or alternate (column totals) scenarios. Then vn is the number of replicated

environments in which both scenarios failed, v12 is the number in which the base scenario

failed but the alternate scenario continued, and so forth. The McNemar (1947) test as

adapted by Sokal and Rohlf(1981) uses a G statistic calculated as,












Gp=2 v In 12 V21 + v,, In 2V21 [3-19]



with the correction factor,

q = 1 + 1/2N, [3-20]


applied as in Eq. [3-18] (Williams, 1976). Again, G,adj is compared to X2 with one degree

of freedom. The McNemar test statistic will be undefined if any of the frequencies (Table

3-3) have a value of zero. This can easily occur if, for example, none of the replicates that

continue in a base scenario fail in an alternate scenario (v12 = 0).




Table 3-4. 2x2 frequency table for tests of difference between simulated
sustainabilities, paired observations.
Base --Alternate scenario--
scenario continued failed Total
continued V1n V12 V1* = n 12
failed v2 v22 v2* = v21+v22
Total: v*, = vu+v21 v*2 = v12+ N = v




It is important to keep in mind that statistical inferences based on simulated

sustainability apply to the model system and its environment. Extension to the actual

system depends on the validity of the model and assumptions about the future

environment.













Sustainability Applied to Farming Systems



Although sustainability is an important concern at several levels in the hierarchy of

agricultural systems (Lowrance et al., 1986; Lynam & Herdt, 1989), it is particularly

relevant at the farm level. If agriculture is to meet the needs of society--providing food

and other products while protecting natural resources--it must first meet the needs of the

farmers who implement and manage it.

The framework for characterizing sustainability applies to farming systems because

a farm is a dynamic, stochastic and purposeful system. Furthermore, sustainability is

characterized most easily at the farm level where system goals are more easily specified

and more consistent than at other system levels. For example, human goals are not

intrinsic to fields or enterprises. On the other hand, the emergence of many human actors

at levels higher than the farm leads to multiple and often conflicting goals. The continuing

and failed status would comprise fuzzy, nonexclusive sets at these higher system levels.

Sustainability expressed as a probability of continuation (Eq. [3-1]) must be

applied to some initial population. One could think of a population of farms and attempt

to predict the proportion that will survive through a future period. However, the

appropriate population to consider when characterizing farm sustainability is the

population of possible realizations of future behavior of an individual farm.

A farm's context within hierarchy has implications for selecting an appropriate

time frame for sustainability analysis, identifying failure criteria, making assumptions about










53

the future behavior of system inputs, and hypothesizing constraints. The remainder of this

section examines these issues.


Selecting a time frame


Sustainability has meaning only in the context of a specific time frame. For

example, climate change, soil erosion, or extinction may be seen as irreversible threats to

the sustainability of an ecosystem in a time frame of decades or centuries, but as part of

natural cycles in a time frame of millennia or longer (Fresco and Kroonenberg, 1992).

From another perspective, sustainability can be viewed as a non-increasing function of

time. Considering the extreme cases, all existing agricultural systems can sustain

themselves for an arbitrarily short period. On the other hand, few agricultural systems can

be expected to continue in a recognizable form for tens of millennia. IfFT(t) is a true

probability distribution with a lower bound at t=0 then


lim
1-0o FTF(t) = 1.



Although selecting a time period for analyzing farm sustainability is a subjective

decision, considerations of hierarchy, relevance, and realism suggest a range of about 10

to 15 years. Ecological hierarchy theory states that processes in higher-level systems

operate more slowly than in lower-level systems (Allen and Starr, 1988). The time frame

for analyzing sustainability of a farming system should therefore be longer than the several

months to a few years that are typical of crop and animal production cycles. Relevance












suggests that the time interval should be long enough to allow detection of important

threats to sustainability. For example, three or four years would reveal little about the

impact of soil erosion on sustainability. On the other hand, a study exceeding a century

would not be relevant to the livelihood goals of an individual farm. Lynam and Herdt

(1989) suggested that five to 20 years is a relevant time-frame for analysis of farming

system sustainability. Realism of assumptions about economic, policy, and technological

inputs to the farming system becomes increasingly difficult to defend past about 10 to 15

years.


Assumptions about inputs


The higher-level systems that comprise a farming system's environment exert

control through inputs and through constraints to farmer decisions or farm outputs.

Analysis of farm sustainability requires assumptions about future behavior of inputs and

control mechanisms that are conceptually external to the farming system. The issues to

consider include (a) whether systematic trends or cycles are expected, (b) whether

variability is sufficient to warrant stochastic sampling, and (c) whether a feedback

mechanism exists which allows an input to respond to farm outputs. Although some farm

processes may be inherently stochastic, inputs of weather and prices are usually the major

sources of risk in a farming system. Catastrophic events such as disease epidemics, storms

or wars also represent important stochastic inputs to some farming systems.












Farm failure criteria


Failure criteria denote the minimum level of performance above which a system is

to be sustained. Since farmer livelihood is the primary purpose of most farming systems,

criteria for farm failure can be expressed in terms of minimum levels of livelihood goals.

Hamblin (1992) suggested that agriculture fails to sustain if production falls below the

levels necessary for profitability in a cash economy or survival in a subsistence economy.

In a subsistence economy, a level of poverty or malnutrition from which a farm family

cannot escape without outside intervention might indicate system failure. Lynam and

Herdt (1989) referred to famine as "the ultimate indicator ofunsustainability" (p. 391). In

a cash economy, lenders may impose threshold leverage ratios above which they will force

foreclosure by recalling loans (Perry et al., 1985). Failure could be expressed in several

forms such as farm abandonment, conversion of land to non-agricultural use, the need to

supplement income with off-farm employment, inability to meet critical goals such as

education of children, or major changes in farm enterprises, depending on the analyst's

purpose.

Negative feedback loops between components of a system tend to counteract the

effects of disturbances and stabilize a system, resulting in a stable state or attractor.

Mathematical and empirical evidence suggests that ecosystems can possess multiple stable

states (May, 1977). The region about a stable state in state space is a domain of

attraction, and the boundary between adjacent domains of attraction is a separatix

(Trenbath et al., 1990). If the state of a system is displaced across a separatix, it enters a












new domain of attraction. The ability to return to the original domain of attraction

depends on the relative stability of the two domains, the nature of the separatix between

them, and the existence of disturbances which could displace the system back across the

separatix. Failure of an agricultural system can be viewed as transition from a useful to a

less useful domain of attraction. The state threshold vector x0 forms the separatix between

the domains. Trenbath et al. (1990) used mathematical models to illustrate abrupt

transitions from useful to less useful domains in response to intensification of three

agricultural systems.

In many cases, farm failure criteria may be difficult to determine. However, it may

be possible to obtain meaningful insights into the relative impact of various stresses on a

farming system by assuming particular failure thresholds when those thresholds cannot be

measured.

Since increasingly restrictive failure thresholds increase the probability of system

failure, sustainability may be viewed as a non-increasing function of threshold levels. A

system is less able to continue at a high level than at a low level.


Determinants of farm sustainability


Sustainability is an aggregate response of a system to a range of external factors,

conditioned by internal characteristics of the system. Any factor that influences means,

trends, variability, autocorrelation or goal levels may influence sustainability. A host of

factors influences the balance between income and expenditure that determines the mean

level of farm wealth. Soil degradation, depletion of scarce resources, technological












innovation and trends in prices can affect trends in farm state variables. Variability is

influenced by weather patterns, price volatility and the occurrence of catastrophic events.

Credit availability, market access, and the ability to store agricultural products have

positive effects on autocorrelation of farm wealth. Finally, a household's tolerance to

difficulty, alternative sources of livelihood and lenders' policies can influence goal

thresholds.

Adaptive management generally serves to improve sustainability. Farmers employ

a range of management strategies, such as selling capital assets, reducing input use,

working off-farm, or shifting from cash to subsistence crops, to reduce the risk of failure

during difficult times. Although the proposed framework for characterizing sustainability

can account for adaptive strategies, the farmer decision process may be more difficult to

simulate than biological or economic processes. Simulating a fixed management strategy

may greatly overestimate the probability of farm failure.

Much of the concern about sustainability of agricultural production systems relates

to externalities: non-target outputs with costs (or benefits) which are not born by an

individual farm but by society. A hierarchical perspective suggests that externalities

should have no direct impact on farm sustainability; the impact of an externality emerges

at a higher (eg., regional) system level. However, a feedback mechanism that either

charges the farming system for the externality or constrains the practice that produces it

can easily be incorporated into a simulation analysis. Such a deviation from a strict

hierarchical approach might be justified by assuming that society will attempt to correct

the perceived injustice of a negative externality.










58

A similar issue arises when extending an analysis from a single farm to a group of

similar farms. A group of farmers could adversely affect common grazing land, water

resources, fuelwood, fishing areas, or forest by overuse, whereas a single farmer's impact

might be negligible. Similarly, although an individual farmer is usually assumed to be a

price taker, a group of farmers may alter prices due to their aggregate effect on supply of

a product or demand for an input such as seasonal labor. Incorporating these effects

involves extending the analysis to include some processes (eg., market equilibria) above

the farming system level.



An Example: Sustainability of a Coastal Texas Rice Farm



Whole-farm simulation studies have examined the influence of factors such as

commodity price variability (Grant et al., 1984), farm size and beginning equity level

(Richardson and Condra, 1981), intergenerational estate transfer strategy (Walker et al.,

1979), and land tenure expansion strategy (Held and Helmers, 1981) on probability of

farm survival through various periods. Perry et al. (1986) conducted a more

comprehensive simulation study that examined the impacts of crop rotation, land tenure

arrangement, government programs, costs, labor availability, lenders' policies, interest

rates, and the level and variability of crop yields and prices on rice farms in Texas. A

reinterpretation of the results of this study illustrates the use of long-term, stochastic

simulation to characterize farm sustainability.












Methods


Perry et al. studied a "representative" coastal Texas rice farm rather than an actual

farm. They examined four combinations of rotation and sharecropping arrangement.

Assumptions and initial conditions can be found in Perry et al. (1986).

Perry et al. used a modified version of the FLIPSIM (Firm Level Income Tax and

Farm Policy Simulator) simulation model (Richardson & Nixon, 1985) called RICESIM to

simulate the model farm. RICESIM is primarily a farm accounting model which randomly

samples from probability distributions as a proxy for the biological and ecological

processes involved in crop production.

The study examined two alternate criteria for system failure. First, probability of

survival was based on a threshold leverage ratio (total debt/total equity). Lenders were

assumed to force foreclosure by recalling loans if the leverage ratio exceeded 2.0. The

second criterion, a negative net present value (NPV) of future cash flow, requires a higher

level of minimum system performance to avoid failure. Failure indicated by a negative

NPV means that a secure, non-farm investment would be more profitable than farming.

Probabilities of survival and positive NPV were calculated by Eq. [3-12] from 50

replicates of each five-year scenario.

Perry et al. analyzed sensitivity of the probabilities of survival and positive NPV to

several factors. For our analysis, I selected only the soybean-soybean-rice rotation with

1/2 share of rice and 1/7 share of soybean going to the landowner. I tested those factors

included in the sensitivity analysis that I believed could constrain sustainability (Tables 3-5












and 3-6). To test for differences in sustainabilities, I used a G-test that is based on

independent sampling (Eq. [3-16] to [3-18]) since the study did not provide paired data or

indicate the experimental design.


Results


The values of Sat the end of the five-year scenario presented here were 0.50

0.071 based on leverage ratio and 0.12 0.046 ( S.E.) based on NPV. Figure 3-4 shows

S(t) of all four combinations of crop rotation and tenure arrangement based on the

threshold leverage ratio.




Table 3-5. Relative sensitivity, r (Eq. [3-15]), of simulated five-year sustainability, S5), of
a Texas rice farm to continuous factors. For the base scenario, (5) was 0.50 based on
threshold leverage and 0.12 based on threshold NPV.
Base Alternate Leverage < 2.0 ---- NPV > 0.0 ----
Factor value value S(5) r S(5) r
variable costs 100% 90% 0.82 6.40 ** 0.48 30.00 **
crop share 50% 45% 0.80 6.00 ** 0.46 28.33 **
crop pricest 100% 110% 0.68 3.60 n.s. 0.38 21.67 **
mean rice yield 100% 110% 0.60 2.00 n.s. 0.26 11.67 n.s.
mean soybean yieldt 100% 110% 0.58 1.60 n.s. 0.26 11.67 n.s.
ratoon red rice 25% 15% 0.80 1.50 n.s. 0.32 4.17 n.s.
rice yield variancet 100% 75% 0.54 0.32 n.s. 0.12 0.00 n.s.
soybean yield variancet 100% 75% 0.52 0.16 n.s. 0.10 -0.67 n.s.
tPercent of base value.










61

Table 3-6. Absolute sensitivity, R (Eq. [3-14]), of simulated five-year sustainability, S5),
of a Texas rice farm to discrete factors. For the base scenario, 9(5) was 0.50 based on
threshold leverage and 0.12 based on threshold NPV.
Base Alternate Leverage < 2.0 -- NPV > 0.0 --
Factor value value S(5) R S(5) R
tenure
tene1/2 share 100% owned 1.00 0.50 ** 0.32 0.20 *
arrangement

tenure
r ent 1/2 share 50% owned 1.00 0.50 ** 0.18 0.06 n.s.
arrangement

soybean
sirrion non-irrigated irrigated 0.86 0.36 ** 0.64 0.52 **
irrigation
tenure
a en 1/2 share 1/7 share 0.82 0.32 ** 0.52 0.40 **
arrangement

tenure rent at
1/2 share 0.74 0.24 0.56 0.44 **
arrangement $74/ha

rotation SSR SR 0.72 0.22 0.20 0.08 n.s.

loan interest 2 points
Si variable 0.54 0.04 n.s. 0.22 0.01 n.s.
rates lower

ratoon rice 7% quality no quality 0.50 0.00 n.s. 0.12 0.00 n.s.
quality discount discount


Of the continuous factors tested in the sensitivity analysis, only variable costs and

the rice crop share showed a significant role in constraining sustainability (Table 3-5). The

ranking of continuous factors by relative sensitivity was the same for sustainability based

on the threshold leverage ratio or on positive NPV. Changes in most of the discrete

factors showed a significant improvement in sustainability based on leverage ratios (Table

3-6). However, the ranking of discrete factors was different when sustainability was based

on NPV. These results suggest that reducing variable costs, negotiating more favorable











tenure arrangements, and irrigating the soybean crop would be the most important

strategies for enhancing sustainability of the model farm.



Discussion


By defining sustainability as the ability of a dynamic, stochastic, purposeful system

to continue into the future, I arrived at a useful, quantitative expression of sustainability.


1.0


0.9


>0.8

CU
cr
.%0.7
.6-
c'o

0.6


0.5


0.4
1984


1985 1986 1987 1988


Year

Figure 3-4. Simulated sustainability of a Texas rice farm under four scenarios: a
three year soybean-soybean-rice rotation with a 1/7 (SSR 1/7) and a 1/2 (SSR 1/2)
share arrangement, and a two year soybean-rice rotation with the same two share
arrangements (SR 1/7 and SR 1/2). Data from Perry et al., 1986.












Although sustainability of a real agricultural system cannot be observed because it deals

with the future, it can be estimated from simulation of a system model. Testing

hypotheses about constraints to system sustainability is then straightforward. Applying

this framework to farming systems results in an approach to characterizing sustainability

that is literal, system-oriented, quantitative, predictive, stochastic and diagnostic (Chapter

2). The use of such an approach could provide attempts to improve farm sustainability

with objective feedback.

The requirement for comprehensive and realistic farm simulation tools currently

limits application of the proposed approach. Most existing farm-level simulation models

are not sufficiently comprehensive; they do not integrate models of crop and animal

production, environmental degradation, economic processes, and farmer production and

consumption decisions. A study designed to examine a single constraint to farm

sustainability would be less demanding in its model and data requirements.

One could question the realism of assumptions about the future behavior of inputs

to a farming system that are required for characterizing its sustainability. However, all

approaches to characterizing sustainability involve inferences about the future. If high-

level systems change more slowly than lower-level systems, as ecological hierarchy theory

asserts (Allen & Starr, 1988), then future projections of inputs such as weather, prices,

infrastructure and technology are more defensible than extrapolation of past farming

system behavior.

The study by Perry et al. (1986) illustrates the utility of the framework presented

in this paper for characterizing sustainability. Although I interpreted their study beyond its











64

original purpose, their simulation analysis was comprehensive enough to illustrate how the

probability of continuation integrates the effects of a range of factors, and how sensitivity

analysis can be used to identify and rank those factors that constrain sustainability.














CHAPTER 4
AN OBJECT-ORIENTED REPRESENTATION OF A FARMING SYSTEM



Introduction



Chapter 3 presents a framework for applying simulation to characterize the

sustainability of a farming system. The framework calls for a farm simulation model that is

able to simulate the operation of a farm over an extended period and to replicate that

period with stochastic inputs of the important variables that contribute to farm risk,

particularly prices and weather.

A comprehensive review of existing farm simulation tools is beyond the scope of

this chapter. Fortunately, a few good reviews of farm modeling literature are available.

Klein and Narayanan (1992) provide an excellent summary of the history of farm modeling

efforts. Jones et al. (1995) derived several generalizations from a review of applications

of farm-scale models. Two are particularly relevant to this study. First, ".. there is

usually a new model for each study, with little acknowledgment of the potential for using

the same model for different farms by emphasizing data requirements and collection for

implementation" (p. 5). One of the exceptions is the FLIPSIM family of models

(Richardson & Nixon, 1985), which has been used to study the impacts of farm size

(Richardson & Condra, 1981), price variability (Grant et al., 1984), marketing strategies














CHAPTER 4
AN OBJECT-ORIENTED REPRESENTATION OF A FARMING SYSTEM



Introduction



Chapter 3 presents a framework for applying simulation to characterize the

sustainability of a farming system. The framework calls for a farm simulation model that is

able to simulate the operation of a farm over an extended period and to replicate that

period with stochastic inputs of the important variables that contribute to farm risk,

particularly prices and weather.

A comprehensive review of existing farm simulation tools is beyond the scope of

this chapter. Fortunately, a few good reviews of farm modeling literature are available.

Klein and Narayanan (1992) provide an excellent summary of the history of farm modeling

efforts. Jones et al. (1995) derived several generalizations from a review of applications

of farm-scale models. Two are particularly relevant to this study. First, ".. there is

usually a new model for each study, with little acknowledgment of the potential for using

the same model for different farms by emphasizing data requirements and collection for

implementation" (p. 5). One of the exceptions is the FLIPSIM family of models

(Richardson & Nixon, 1985), which has been used to study the impacts of farm size

(Richardson & Condra, 1981), price variability (Grant et al., 1984), marketing strategies












(von Bailey & Richardson, 1985), tenure arrangements and several other policy and

management factors (Perry et al., 1986) on farm viability. The first generalization

highlights the need for simulation tools and associated data standards that are generic,

flexible and extensible. The second generalization is that "... most farm-scale models

have been developed with a bias toward economics and limited consideration of the

biophysical components" (p. 5). Farm models often represent crop and animal response to

weather variability by sampling from probability distributions that were fitted to historical

or survey data. However, concerns about ecological threats to sustainability such as soil

degradation and climate change are better addressed by the integration of biophysical and

economic models. Recent farm models have incorporated crop simulation to capture the

impact of weather variability on farm risk (Dillon et al., 1989) and to relate global climate

change scenarios to farmer adaptation (Kaiser et al., 1993), for example.

A farm simulation model, the Farming System Simulator (FSS), was developed as

a tool for applying the framework presented in Chapter 3 for characterizing farm

sustainability. Its object-oriented design address the need for a generic simulator that can

represent very different farm types, and whose functionality can be extended relatively

easily. Linkage to process-level crop simulation models addresses the need for balanced

and integrated treatment of the ecological and economic components of a farming system.

The objectives of this chapter are (a) to describe the Farming System Simulator and (b) to

present data requirements for simulating the sustainability of a farming system. The

chapter first presents an overview of the design and functionality of FSS, then describes its

inputs, structure, processes and outputs.













Overview of the Farming System Simulator



The Farming System Simulator (FSS) is an object-oriented, dynamic, stochastic,

discrete-event farm simulator. It is object-oriented in its design, and is implemented using

the object-oriented extensions of Borland Pascal (Borland International, 1992). The

description of FSS that follows cannot be fully understood without some familiarity with

object-oriented programming concepts (Appendix A). FSS is dynamic; it is capable of

simulating the operation of a farming system through many years. It is stochastic in the

sense that it is designed to simulate many replicates of a farm scenario with stochastic

inputs of weather and price data, and to present analyses based on the resulting

distributions. FSS runs external crop simulation models to simulate continuous

physiological and ecosystem processes. However, all farm-level processes (i.e.,

operations, production, management and consumption of resources, and failure) occur in

response to discrete events. A final characteristic ofFSS is that it is primarily a resource

accounting model.

The object-oriented structure of FSS is based on a conceptual model of the

structure and function of a farming system. A farming system integrates ecological,

economic and social components (Fig 4-1). The ecological component of a farming

system--a set of agricultural ecosystems, or agroecosystems--consists of biotic

communities and the landscape that they inhabit, and can be delineated by field boundaries.

The economic component comprises the set of resources that are under the control of the












LOCAL
INDUSTRY?
I


S- I --
---------------------------------------


ecological
systems


economic social
systems systems


Figure 4-1. Parallel ecological, economic and social hierarchies of agricultural systems.


WATERSHED?
I-----;-----I


AGROECOSYSTEM
(LANDSCAPE &
BIOTIC COMMUNITY)


VILLAGE?
I ~ILG1


f












farmer. The social component of a farming system is the farmer or farm household, and

includes goals and decision criteria. In FSS, objects represent each of these components:

fields in the farm landscape, a set of strategies that specify the sequence of crop activities

and their management, a set of enterprises that link management strategies to particular

fields within the landscape, and a set of farm resources (Fig. 4-2). The household is

represented by decision rules for consumption, production and farm failure. The object-

oriented design of FSS provides a flexible means of representing farm resources, possible

interactions among resources, and relationships between operations and the resources that

they use. The Farming System Simulator is a farm model only in a loose sense; the model

structure of a particular farm is specified at run-time by the resource, field, enterprise and

strategy objects that are initialized in response to input data.

The capabilities of FSS reflect its purpose as a tool for characterizing farm

sustainability based on the framework presented in Chapter 3. The first requirement was

the ability to replicate a long-term scenario with constant initial conditions but stochastic

inputs. FSS takes advantage of a stochastic weather generator, WGEN (Richardson,

1985), that has been incorporated into the crop models that are part of the Decision

Support System for Agrotechnology Transfer version 3 (DSSAT, Hoogenboom et al.,

1994). An analogous stochastic price generator is part ofFSS. FSS addresses its second

requirement--an ability to simulate ecological processes of crop production--by calling

external crop simulation models. The crop models simulate weather variability, soil

dynamics and crop growth and development, then return the information that FSS requires

in the form of schedules of field operations. Third, FSS addresses the need to deal with












has a has a
set of has a set of


SResources


has a
set of
\


SProduction
decision criteria


Landscape
has a
set of
\





SGoal
thresholds


Enterprises

has a
set of
/
Fields has a
set of


Production
strategies


Figure 4-2. Object representation of the main components of a farming system.


has a


has a has a
set of set of
1


I Consumption
decision criteria












farmer livelihood by accounting for all farm resources produced, used for production, or

consumed by the farm household. The fourth requirement was the ability to test

conditions for system failure. Failure can be based on insolvency--the inability to cover

fixed costs, obligations or minimum subsistence consumption requirements. Failure can

also occur when an individual or aggregate farm resource violates a user-specified

threshold value. Detailed resource accounting was a prerequisite to the ability to test for

conditions for farm failure. Finally, FSS offers several file and graphical outputs that are

relevant to the analysis of farm sustainability.

The current version of FSS possesses several important limitations. First, it does

not simulate adaptive management; it simulates a fixed, continuously repeating set of

management practices. However, an actual farm operating under stress would normally

employ a range of practices to avoid failure (Chapter 3). Second, FSS does not possess a

mechanism for adjusting crop management for within-season resource constraints. This

limitation is imposed by the need to simulate each crop for an entire season. Chapter 5

discusses the problem and a possible solution. Third, FSS can simulate only crop

production enterprises. No livestock model has yet been adapted for running under FSS.

Fourth, the household resource consumption model (Eq. [4-7]) is simplistic; it does not

consider the impact of risk or anticipated future lifestyle changes on consumption and

savings decisions. The remaining sections of this chapter focus on inputs, processes and

outputs of FSS.













Inputs



Jones et al. (1995) cited a lack of emphasis on data standards as a barrier to

reusing farm models for different farms or applications. FSS input data structures and file

formats were designed to be flexible enough to be able to represent a range of farm types

and to accommodate possible future extensions of FSS. The general organization of data

and scheme for its use are being proposed as a starting point for developing data standards

for enterprise and farm-level systems analyses (Hansen et al., 1995). Detailed

presentation of input data requirements and formats in Appendix B supplements the

discussion in this section.

Afarm scenario is the operation of a farm through a period of time with a given

set of initial conditions and rules for making decisions and scheduling activities. A

scenario may be replicated with stochastic sampling of input variables such as weather and

prices, but with the same initial conditions and decision rules for each replicate. At least

two files--a scenario file and a price file--are required to simulate a farm scenario. The

scenario file contains farm-level information and identifies the other input files. The price

file contains the parameters for models for generating sequences of prices. A scenario that

calls external IBSNAT crop simulation models also requires a minimum data set (MDS)

for each crop consisting of a crop management file, soil file, weather or climate file, and

genetic coefficient file (Tsuji et al., 1994).












The set of input data required to run FSS serves three roles. First, inputs define

the structure of a farm model. The objects that define the model of a particular farm are

initialized from the input files. This is the topic of the next section. Second, inputs specify

initial conditions by specifying initial values of the state variables of objects. Third, inputs

provide the values of dynamic driving variables. The two types of driving variables--prices

and weather--can be represented either by inputs of actual historical series or as stochastic

time-series models.

The structure of the input files relates closely to the object-oriented structure of

FSS. In most cases, a line or section of an input file is an argument for the constructor

method (see Appendix A) that initializes an object. Multiple items within a section of a

file are used to initialize collections of similar objects. For example, each line in the

RESOURCES section of the farm scenario file is used to initialize and insert a resource

object into the collection of farm resources.


Scenario file


The farm scenario file contains the information necessary to initialize a farm model

and simulate a farm scenario. The scenario file also identifies other essential data files: a

crop management file, a soil file and a price file. Table 4-1 describes the sections of a farm

scenario file. Appendix B presents the format of each section.

Simulation control. The first three sections of the scenario file relate to simulation

control. The SCENARIO section identifies the farm scenario and specifies crop

management, price and soil data input files, units of wealth, and options for duration of the












scenario and the number of times it is replicated. The next two sections--OUTPUTS and

ANALYSES--control file and graphical output. These are discussed later in this chapter.



Table 4-1. Sections in the farm scenario file.
Section Purpose
SCENARIO Specifies title, units, other files, and simulation control.
OUTPUTS Controls farm-level file output.
ANALYSES Controls graphic display and file output.
RESOURCES Identifies farm resources.

LINKAGES Defines interrelationships between resources.

OPERATION Specifies types of operations, priorities, a time window, and
REQUIREMENTS timed resources required.
SCHEDULED Schedules operations which are not included in the experiment
OPERATIONS file or returned by crop models.

LANDSCAPE Identifies homogeneous fields and their positions and
characteristics.

LIVESTOCK Notyet developed. When developed, will specify herd and
management information for grazing livestock.
STRATEGIES Specifies sequences of management activities.
ENTERPRISES Links management (i.e., strategies) with landscape (i.e.,
subfields).
PRODUCTION Not yet developed. When developed, will specify a model and
DECISIONS criteria for selecting strategies for each enterprise.

CONSUMPTION Specifies household subsistence and discretionary consumption
DECISIONS of resources.




The remaining sections of the scenario file define both the structure and the initial


conditions of a model of a particular farm.










Resources. The RESOURCES section specifies the particular resources that

constitute a farming system's state variables, and determines their initial characteristics and

supply. A hierarchy of nine resource classes is available in FSS to represent the various

types of farm resources (Table 4-2, Fig. 4-3). Two additional abstract ancestral classes

are used internally as templates for the other classes.


Resource


Simple
resource



Consumable Timed Credit
resource resource resource



Seasonal Capital Activity-linked
resource resource credit resource



Machine
resource


Figure 4-3. Tree of resource class hierarchy in FSS.


Aggregate
resource



Debt
resource




Full Text
80
presented in Appendix B. The models that they initialize for generating daily price values
are discussed in a later section.
Crop minimum data set
FSS obtains detailed information about crop management and the physical
environment from the crop management file, or FILEX (Jones et al., 1994). Thornton et
al. (1994) present relevant guidelines for preparing a crop management file for simulating
sequences. FSS ignores the TREATMENTS section of FILEX; the LANDSCAPE and
STRATEGIES sections of the scenario file serve the function of the TREATMENTS
section.
In addition to the crop management file, FSS requires a soil file, weather files or a
climate file, and genetic coefficient files. These data files are documented in Tsuji et al.
(1994). The climate file contains parameters for stochastic weather generators that are
incorporated into the crop models. Actual weather sequences can be used instead of
generated weather to eliminate weather risk in order to explore the relative importance of
weather and price variability.
Processes
FSS manages a number of processes, including managing random number
sequences, stochastic generation of prices, ecosystem dynamics and crop production,


90
Use and Sell methods. Most operations use or produce some material resource,
represented by a consumable resource object. Examples are water used for an irrigation
application and maize grain produced in a harvest. Most operations call the USE method
of any associated material resource, passing the amount of the resource produced as an
argument. A marketing operation is the exception; it calls the material resources SELL
method. The USE method adjusts supply according to the amount produced (Fig. 4-5). If
the amount produced is negative indicating consumption of the material, and the amount
used is greater than the available supply, then the USE method attempts to cover the
resulting deficit from resources identified in its list of variable costs. If use violates a
maximum storage constraint, then the USE method disposes of the resulting surplus by
selling to resources in its variable cost list.
Algorithm
Execute the method, Update.
Set Supply = Supply + Amount.
If Supply < max(Minimum, 0) then
begin
Set Deficit = mm(Supply Minimum, 0.)
Request VariableCosts to execute UsE(Deficit).
If Deficit < Othen
set Amount = Amount Deficit.
Set Supply = mm(Minimum, 0).
end
otherwise if Maximum is defined and
Supply > Maximum then
begin
Set Surplus = Maximum Supply.
Request VariableCosts to
execute SEhh(Surplus).
Purpose
Make sure linked resources are up-to-date.
Change Supply according to Amount desired.
If the minimum storage constraint is violated...
Deficit represents a negative amount.
Try to meet deficit by using linked resources.
If VariableCosts could not eliminate the deficit...
Reduce Amount used by remaining Deficit.
Supply is exhausted.
The maximum storage constraint is violated.
Dispose of Surplus by selling to linked resources.
Figure 4-5. Pseudocode representation of the USE method of the consumable resource
class.


141
Nario (Muos & Wieczoreck, 1978); maize showed a strong response to N fertilizer in
only three of the six soils shown in Fig. 5-14.
I varied the mineralization rate factor (SLNF) in the soil profile description (Fig. 5-
15). In the absence of observed maize N response data for the study area, I selected a
value of 0.35 for SLNF because it produced a reasonable yield response curve based on
the experience of CIAT scientists (E.B. Knapp, 1994, personal communication). This
Figure 5-14. Maize response to applied N on several volcanic soils in Nario,
Colombia. Data from Muoz & Wieczoreck (1978).


Figure 5-1. Location map of the study area, Cauca, Colombia


TABLE OF CONTENTS
ACKNOWLEDGMENTS iii
LIST OF TABLES viii
LIST OF FIGURES xii
ABSTRACT xviii
CHAPTERS
1 INTRODUCTION 1
2 IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT? 4
Introduction 4
Sustainability as an Approach to Agriculture 6
Sustainability as an Alternative Ideology 7
Sustainability as a Set of Strategies 11
Discussion 14
Sustainability as a Property of Agriculture 17
Sustainability as an Ability to Satisfy Goals 17
Sustainability as an Ability to Continue 18
Approaches to Characterizing Sustainability 19
Adherence to Prescribed Approaches 19
Multiple Qualitative Indicators 20
Integrated, Quantitative Indicators 21
Time Trends 23
Resilience 25
System Simulation 26
Elements of a Useful Approach for Characterizing Sustainability 27
Conclusions 32
IV


185
of liquid assets at the end of nine years, determining the random number seed used to
generate weather for that replicate, then using that random number seed for every
replicate of the no weather risk scenario. The third scenario, no spatial diversification,
was a modification of the base scenario in which all of the cultivated area is in the same
phase of crop rotation (i.e., all in maize then bean then bean then cassava). These
scenarios were compared with the base scenario simulated for nine years.
The scenarios used to examine sources of risk were simulated for only three
rotation cycles (9 years) for two reasons. First, after bean and maize prices were shifted
as described above, only 11 full years of prices were available for all five crops. Second,
risk is easier to interpret if distributions are not severely truncated by a large proportion of
failures. The distribution of liquid assets is the basis for interpreting farm risk through
time for these scenarios.
Simulation and Analysis
The farming system simulator (FSS) described in Chapter 4 was used to simulate
the farm scenarios. FSS simulated 100 replicates of each 15-year scenario. For a given
scenario, the model farming system had the same initial state but different realizations of
weather and prices in each replicate. The same sample of price and weather realizations
was used for all scenarios.
FSS recorded minimum, 25th percentile, median, 75th percentile and maximum
supply of liquid assets among replicates at the end of each simulation year, the value of
each resource each year for each continuing replicate, and the occurrence and timing of


240
OPERATION REQUIREMENTS
@ N
Type
cc
Methd
Pri
Win
Resourcel
BH
Hours
1
PLNT
MZ
PM001
5
7
HIRED
LABOR
A
40.0
2
PLNT
BN
PM001
5
14
HIRED
LABOR
A
96.0
3
PLNT
cs
PM003
8
30
HIRED
LABOR
A
8 0. C'
4
PLNT
TM
PM002
6
21
HIRED
LABOR
A
39.0
5
IRRI
IR005
1
3
JOSE
A
2.0
5
IRRI
IR005
1
3
HIRED
LABOR
A
2.0
6
TILL
TI021
4
21
HIRED
OX
A
24.0
7
TILL
CS
TI023
7
21
HIRED
LABOR
A
60.0
8
FERT
AP002
7
7
HIRED
LABOR
A
2 4.0
9
HARV
MZ
HA001
2
10
HIRED
LABOR
W
20.0
10
HARV
BN
HA001
3
14
HIRED
LABOR
W
60.0
11
HARV
CS
HA001
8
30
HIRED
LABOR
w
30.0
12
HARV
TM
HA001
2
14
HIRED
LABOR
w
40.0
13
HARV
CF
HA001
3
14
HIRED
LABOR
K
40.0
14
MRKT
MZ
0
1
5
JOSE
W
6.0
15
MRKT
BN
0
3
14
JOSE
W
3.0
16
MRKT
CS
0
6
21
JOSE
W
9.0
17
MRKT
TM
0
1
3
JOSE
W
9.0
18
MRKT
CF
0
6
28
JOSE
W
3.0
19
RESD
AP002
8
5
HIRED
LABOR
A
15.0
20
SPRP
TM
1
4
10
JOSE
A
12.0
20
SPRP
TM
1
4
10
HIRED
LABOR
A
12.0
21
CLEA
0
4
21
HIRED
LABOR
A
60.0
22
CHEM
AP006
3
14
JOSE
A
8.0
22
CHEM
AP006
3
14
HIRED
LABOR
A
8.0
23
PRUN
CF
0
8
28
HIRED
LABOR
A
150.0
24
FERT
AP003
8
4
JOSE
A
8.0
24
FERT
AP003
8
4
HIRED
LABOR
A
8.0
25
STAK
TM
0
3
4
HIRED
LABOR
A
72.0
26
PRUN
TM
0
6
4
JOSE
A
15.0
26
PRUN
TM
0
6
4
HIRED
LABOR
A
15.0
27
TILL
TM
TI023
7
21
HIRED
LABOR
A
120.0
28
TILL
MZ
TI023
7
21
HIRED
LABOR
A
120.0
29
TILL
BN
TI023
7
21
HIRED
LABOR
A
120.0
30
PROC
BN
0
5
28
HIRED
LABOR
W
100.0
31
PROC
CF
0
5
45
HIRED
LABOR
W
4 00.0
SCHEDULED OPERATIONS
@ N
Cl Type
CC Methd BT LOp Time Resource
BA
Amount
1
O CLEA
MZ
0
L
9
1
A
1.0
1
O PROC
BN
0
L
10
1
A
1.0
1
O MRKT
MZ
0
L
9
5 HA001
MZ
IB0C
65 R
4 87.5
1
O MRKT
BN
0
L
10 14 HAG 01
BN
IBOO
26 R
112.5
1
O MRKT
CS
Cl
L
11
5 HA01
CS
UCOO
22 S
-1.0
2
O CLEA
MZ
0
L
9
i
A
1.0
2
O PROC
BN
0
L
10
i
A
1.0
O MRKT
MZ
0
T
9
5 HAG01
MZ
IB005 R
487.5
2
O MRKT
BN
0
L
10 14 HA001
BN
IBOO
26 R
112.5
2
O MRKT
CS
0
L
11
5 HA001
CS
UCOO
22 S
-1.0
O
O HARV
CS HA001
A
0 95110 HA001
CS
UC0022 A
10000
LANDSCAPE
@ N
HS PO FL IC ME
ER
Area
Los s
1
J. _
1 0
N
v. 6 0
0.0
2
1 5
3 1 0
N
0.18
0.0
3
2 2
11
i 0
N
3.60
0.0
4
2 3
i:
1 0
N
0.18
o. c*
5
3 2
c
1 0
N
0.60
0.0
6
3 3
6 10
N
0.18
0.0
Figure D-l, continued.


190
season following the years in which maize and bean were harvested from the least
productive fields, during the months (October to December) between planting and the
bean harvest.
Cropping systems
Simulated sustainabilities of the annual cropping system scenarios are shown in
Fig. 6-7 and Table 6-8. The most sustainable cropping system was the two-year maize-
bean-cassava rotation. It was also the most intensive rotation simulated, with very little
Figure 6-7. Sustainability time plot of annual cropping system scenarios.


192
and 2.5 Mg ha'1 yr'1 in the area of the Domingo farm. However, uncertainties about coffee
yields and the variability of results obtained for different crop rotations limit
generalizations about the relative sustainability of coffee and annual crop production.
Table 6-8. Predicted 15 year sustainability ((15)SE$) of cropping system scenarios, and
ifivnvuiui.A^P.adiJ ^ V^Ladi/
Scenario
$15)SE
f^P.adj
^I,adj
coffee @ 2.50 Mg ha'1
0.990.010
... .1
49.0
**
irrigated maize-tomato-bean-cassava (3 yr)
0.990.010
41.9 **
49.0
**
maize-bean-cassava (2 yr)
0.95 0.022
t
32.1
**
coffee @ 2.25 Mg ha'1
0.92 0.027
24.7 **
24.0
**
coffee @ 2.00 Mg ha'1
0.70 0.046
1.0 n.s.
0.8
n.s.
maize-bean-bean-cassava (3 yr)
(base scenario)
0.64 0.048
t
t
cassava monoculture (3 yr)
0.47 0.050
8.6 **
5.8
*
coffee @ 1.75 Mg ha'1
0.30 0.046
26.5 **
23.5
**
maize-bean (1 yr)
0.04 0.020
75.3 **
91.4
**
maize monoculture (1 yr)
0.00 0.000
t
t
T Undefined.
t Comparison does not apply to the base scenario.
Soil management
The impacts of rates of N fertilizer on simulated farm sustainability are shown in
Fig. 6-9. Either increasing or decreasing the amount of N applied to beans and maize
reduced sustainability compared with the base scenario. The reduction of sustainability
with less applied N supports the hypothesis that inadequate N diminishes sustainability by


232
CORRELATION MATRIX, and HISTORICAL. Each item in the CONSTANT, ARMA
and HISTORICAL sections specifies a price series.
A price model initialized from the CONSTANT section of the price file always
returns the same specified value. Table B-13 presents the format of the CONSTANT
section.
Table B-13. Format of the CONSTANT section of the price file.
Description
Header
Formal
Index of price
N
013
Name of price
Descr
1 C 16
Constant value of price
Price
1 R 10
* See footnote, Table B-l.
Stochastic (ARMA) prices require both the ARMA (Table B-14) and
CORRELATION sections. The ARMA section provides the parameters used by Eq. [4-1]
and [4-2] to generate monthly prices. The CORRELATION section contains the cross
correlation matrix, in lower-triangular form, of the residuals of the price models specified
in the ARMA section. It is required to reproduce the cross-correlation structure of the
entire set of ARMA prices. Each cross-correlation coefficient has the format, 0 R 5 2 (see
footnote, Table B-l).


116
allophane is apparently much greater for humified organic matter than for fresh residues
(Monreal etal., 1981).
The characteristic low bulk densities of Udands have been used to explain their
good workability, favorable infiltration and water-holding capacities, aeration and nutrient
availability, and their tendency to lose nutrients easily by leaching (van Wambeke, 1992).
Soil degradation. Soil erosion is a major concern in the Cauca watershed (Ashby,
1985). Annual crops are cultivated on very steep lands, often on slopes near 100%. Rains
are expected to be most erosive in November, before plant canopies have covered recently
tilled soil. However, a high infiltration capacity and the contribution of organic matter to
the formation of stable aggregates in the Ap moderate the erosion hazard somewhat.
The Andisols of Latin America are typically more vulnerable to mass movement
(i.e., landslides and slumping) in which the transport mechanism is gravity, than they are to
water erosion (Sents, 1992). This vulnerability results from their high infiltration capacity
and rheological properties. A mass movement can occur during intense rain if antecedent
moisture is high, the subsoil is less permeable than the surface soil, and the subsoil has low
cohesiveness. Some amorphous clays behave as fluids when they are fully hydrated and
may act as a lubricant between surface and subsoil horizons. The landscape in the Cauca
region shows evidence of a great deal of degradation by mass movement.
Crop simulation
The crop simulation models (Hoogenboom et al., 1994) and support software in
the Decision Support System for Agrotechnology Transfer version 3 (DSSAT3, Tsuji et


221
purchased at the specified price and sold at 65% of that price.
Table B-5. Format of the LINKAGES section of the scenario file.
Variable
Header
Formal
Name of the resource
Resource
OC 15
Name of the linked resource
Source
1 C 15
Type of linkage:
F = fixed cost
V = variable cost
N = numerator (aggregate resource )
D = denominator (aggregate resource)
B = debt {debt resource)
Cl
1 C 1
Name of price which defines the exchange rate with the
linked resource
Price
1 C 15
Price multiplier for purchasing resource
BuyMult
0 R 8
Price multiplier for selling resource
SelMult
0 R 8
T See footnote, Table B-l.
@Resource
Source
Cl
Price
BuyMult
SelMult
HA001 BN IB0026
OPERATING FUND
V
Bean
1.00
0.65
Figure B-5. Example item from the LINKAGES section.
OPERATION REQUIREMENTS section
The OPERATION REQUIREMENTS section specifies how operations use
resources. The OPERATION REQUIREMENTS section contains multiple items. An
item specifies a unique combination of an operation type, crop type and method, and
assigns a priority to the combination. A time window indicates how long completion of
the operation can be delayed before it is abandoned. The remaining columns represent a


O 20 40 60 80
Days after planting
Figure 5-20. Effect of applied N on bean nodule growth (a, b) and cumulative N2 fixed (c, d) observed and
simulated with the original (a, c) and modified (b, d) versions of CROPGRO, Kuiaha, Hawaii, 1993. Data from
Tewari (1995).
oo


Probability Probability
138
Figure 5-12. Distribution of simulated yields of October- and March-planted maize
(a) and bean (b) in response to historical and generated weather variability, La
Florida, Popayan, Colombia.


LOCAL
INDUSTRY?
WATERSHED?
VILLAGE?
1
\
i
1
1
i
1
/
I
ecological economic social
systems systems systems
Figure 4-1. Parallel ecological, economic and social hierarchies of agricultural systems.
o\
00


A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY
By
JAMES WILLIAM HANSEN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1996


93
Update method. The Update methods of consumable (Fig. 4-9) and capital
resources (Fig. 4-10) change supply or value in response to interest or depreciation, and
charge fixed costs of ownership. Credit resources use Update to make scheduled loan
payments. The Update method obtains the total payment required from its collection of
loans, then searches its collection of fixed costs to find a source of funds to complete the
required payment (Fig. 4-11).
Algorithm
Purpose
If the date is later than the last Update then
begin
Calculate A time.
The fraction of a year since the last Update.
Set Supply = Supply exp {Interest A time).
Adjust Supply for interest or depreciation.
Execute the method, Use(0).
Adjust for any new storage constraints.
Execute the method, UpdateFixedCosts.
Charge for any fixed costs of owning Supply.
end.
Figure 4-9. Pseudocode representation of the Update method of the consumable
resource class.
Algorithm
Purpose
If the date is later than the last Update then
begin
Calculate A time.
The fraction of a year since the last Update.
Set Value = Value exp(Interest A time).
Adjust Value for depreciation.
Execute the method, UpdateFixedCosts.
Charge any fixed costs of ownership.
end.
Figure 4-10. Pseudocode representation of the UPDATE method of the capital resource
class.


6-7 Sustainability time plot of annual cropping systems 190
6-8 Sustainability time plot of base and coffee scenarios 191
6-9 Sustainability time plot of base and nitrogen management scenarios 193
6-10 Sustainability time plot of base and soil erosion scenarios 194
6-11 Sustainability time plot of base and price scenarios 195
6-12 Sustainability time plot of base and household consumption scenarios 196
6-13 Sustainability time plot of base and resource scenarios 198
6-14 Sustainability time plot of base and credit scenarios 199
6-15 Distribution of liquid assets after 1, 3, 6 and 9 years for the source of risk
scenarios 200
6-16 Sustainability time plot of source of risk scenarios 201
B-l Example SCENARIO section 215
B-2 Example OUTPUTS section 216
B-3 Example ANALYSES section 217
B-4 Example item from the RESOURCES section 218
B-5 Example item from the LINKAGES section 221
B-6 Example item from the OPERATION REQUIREMENTS section 224
B-7 Example item from the SCHEDULED OPERATIONS section 224
B-8 Example item from the LANDSCAPE section 226
B-9 Example item from the STRATEGIES section 229
B-10 Example item from the ENTERPRISES section 229
B-l 1 Example item from the CONSUMPTION DECISIONS section 231
xvi


Table 6-4. Fitted parameters for deterministic component (Eg. f4-l]) of production commodity price time series models.
Product Trend Seasonal
a
P
Pj.
Pro.
Pm,,
PApr
Milay
Pjm
Pm
M'Aitg
Psep
Poet
f*Nov
Pd
coffee
3.0035
-0.0001
0.0005
-0.0070
-0.0031
0.0008
-0.0019
0.0028
-0.0030
0.0006
0.0002
0.0035
0.0034
0.0035
cassava
2.2553
0.00073
0.0143
0.0056
0.0060
0.0134
0.0112
0.0373
0.0231
0.0091
0.0015
0.0055
0.0117
0.0247
choclo
1.9643
0.00055
0.0754
0.0044
0.0501
0.0634
0.0189
0.0665
0.0461
0.0136
0.0796
0.0695
0.0081
0.0659
tomato
2.5335
-0.0003
0.0075
0.0395
0.0637
0.0711
0.0225
0.0296
0.0273
0.0364
0.0343
0.0118
0.0343
0.0244
bean
3.0079
-9.1E-5
0.0064
0.0106
0.0104
0.0021
0.0075
0.0127
0.0078
0.0021
0.0090
0.0016
0.0052
-0.0003
Table 6-5, Fitted parameters for stochastic component (Eg, [4-2]) of production commodity price time series models,
Product Shock Autoregressive terms Moving average terms Seasonal Cross-correlation coefficients
ot
.
0j
04
$5

e,
02
0j
04
05
s
0s
^"coffee
r
* cassava
^choclo
^tomato
coffee
0.01813
0.069
0.871
0.261
0.271
0.000
0.000
1.034
0.308
0.000
0.000
0.000
62
0.126
1.00
cassava
0.03458
1.014
0.024
0.630
0.664
0.000
0.000
0.043
0.123
0.602
0.173
0.000
0
0.000
0.10
1.00
choclo
0.05115
0.387
0.445
0.000
0.000
0.000
0.000
0.959
0.048
0.017
0.086
0.167
0
0.000
-0.06
-0.13
1.00
tomato
0.07315
0.000
0.000
0.000
0.000
0.000
0.000
0.955
0.555
0.246
0.000
0.000
0
0.000
0.13
0.04
0.00
1.00
bean
0.02098
0.916
0.110
0.109
0.016
0.005
0.229
0.000
0.000
0.000
0.000
0.000
0
0.000
0.32
-0.03
0.11
0.14 1.00


85
or if the observed series was log-transformed,
Sx = io(1o6*-^pa, [4-6]
where yt is the estimate of price at month t, A t is the number of months between the time
the price was observed and the time of the scenario for which the observed price is used as
a proxy, and p is the slope of the trend. Daily prices are linearly interpolated as required.
Crop and ecosystem processes
FSS is designed to run and communicate with external simulation models of the
ecosystem processes involved in agricultural production. It currently works with the
family of crop models in DSSAT3. Ecosystem processes are simulated in order of the
landscape position of fields, which are ordered by hillslope and by position within each
hillslope. The purpose of ordering fields by hillslope position was to facilitate simulation
of processes such as soil erosion for which runoff from one field becomes an input for the
adjacent, down-slope field.
An activity object implements a crop model driver that calls external programs to
simulate each production activity on each field. FSS communicates with IBSNAT crop
models by means of the temporary input file documented in Hoogenboom et al. (1994)
based on the data specified in the crop management file, or FILEX (Jones et al., 1994).
An operations output file (Table 4-3) supplies all of the information that FSS requires
from production process models. It consists of a schedule of field operations and the
material resource balance associated with each operation. The operations output file is


*CCBN940001
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 1A
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
@ SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
76
0.35
1.00
IB001
IB001
IB001
@ SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
20 AP
0.325
0.467
0.575
0.75
25.0
0.45
11.70
16.2
25.8
0.0
0.78
5.3
4.5
3.3
50 B
0.375
0.541
0.622
0.15
5.0
0.48
3.10
75.0
15.0
0.0
0.27
5.4
5.0
0.8
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.10
75.0
15.0
0.0
0.20
5.6
5.1
0.8
*CCBN940002
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO IB
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
22 AP
0.325
0.467
0.575
0.75
25.0
0.45
10.00
16.1
28.3
0.0
0.73
4.8
4.5
2.6
50 B
0.375
0.541
0.622
0.15
5.0
0.48
7.20
75.0
15.0
0.0
0.30
5.0
4.9
0.7
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.50
75.0
15.0
0.0
0.24
5.3
5.0
0.6
CCBN940003
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 1C
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.5
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
23 AP
0.325
0.467
0.575
0.75
25.0
0.45
8.40
18.8
29.7
0.0
0.61
4.9
4.4
2.2
50 B
0.375
0.541
0.622
0.15
5.0
0.48
3.10
75.0
15.0
0.0
0.25
5.0
4.6
1.0
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.10
75.0
15.0
0.0
0.19
5.2
4.7
0.7
CCBN940004
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 2A
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
19 AP
0.325
0.467
0.575
0.75
25.0
0.45
3.90
16.1
29.7
0.0
0.55
5.2
4.7
2.8
50 B
0.375
0.541
0.622
0.15
5.0
0.48
2.80
75.0
15.0
0.0
0.24
5.2
5.1
1.2
200 B
0.375
0.541
0.622
0.10
5.0
0.48
1.70
75.0
15.0
0.0
0.15
5.6
5.2
0.7
Figure D-3. Soil profile data file used for simulating farm scenarios.
247


105
environmental pollution. Jones et al. (1991b) reviewed studies that demonstrate various
capabilities of EPIC (Erosion-Productivity Impact Calculator) such as predicting crop
response to soil erosion, nitrogen inputs, weather variability and climate change scenarios.
They concluded that EPIC is an operational model for evaluating sustainability of cropping
systems because it addresses constraints imposed by productivity, resource conservation,
protection of water quality, and socioeconomic considerations (p. 347). While none of
these studies suggests specific criteria for interpreting simulation results in terms of
sustainability, all agree that crop simulation is a useful tool for studying crop response to
weather risk, and the agricultural and environmental impacts of long-term soil dynamics
that threaten sustainability of crop production systems.
Chapter 3 presents a framework for using long-term, stochastic simulation to
characterize farm sustainability. The Farming System Simulator that was developed as an
experimental tool for characterizing farm sustainability (Chapter 4) uses crop simulation to
capture the effects of agroecological processes and management on agricultural
production. In Chapter 6, the framework is applied to identifying determinants of
sustainability of a hillside farm in the Cauca region of Colombia. The purpose of this
chapter is to evaluate the compatibility of the IBSNAT family of crop models
(Hoogenboom et al., 1994) with requirements imposed by a simulation study of the
sustainability of a hillside farm in the Cauca region of Colombia. Specific objectives are
(a) to describe the physical environment of the study area, (b) to review the status of the
crop models, (c) to evaluate crop model predictions of development and yield in a
Colombian hillside environment in response to potential agroecological determinants of


229
0
N
Ev
Name
1
1
Maize/soybean
doublecrop
0
N
Pe
Ac
Cl
cu
MP
MI
MF
MR
MC
MT
ME
MH
SM
LI
Recover
Name
1
1
1
I
1
2
1
0
1
0
2
0
0
1
0
1.00
Bean
1
1
2
I
2
0
1
0
0
0
0
0
2
3
0
0.00
Fallow
1
1
1
3
I
3
1
1
1
0
0
1
0
0
2
0
0.80
Maize
1
1
4
I
2
0
1
0
0
0
0
0
1
3
0
0.00
Fallow
2
Figure B-9. Example item from the STRATEGIES section.
ENTERPRISES section
The ENTERPRISES section combines strategies with plots within the landscape.
Each section identifies an enterprise that will exist throughout the scenario. When
implemented, the production decision model will allow selection among strategies within
each enterprise. The ENTERPRISES section contains multiple items.
The example in Fig. B-10 links subfields 1 and 2 with strategy 1 to form a
bean-maize doublecrop enterprise.
@ N Name
1 Bean-Maize Doublecrop
@ Subfields
1 2
@ Strategies
1
Figure B-10. Example item from the ENTERPRISES section.


LIST OF FIGURES
Figure
page
2-1 Contrasting interpretations of the relationship between chemical input levels
and sustainability 13
3-1 Relationship between time, distribution of state and sustainability under
constant mean and high variability, and negative trend and low variability 40
3-2 Estimating sustainability by sampling a small number of simulated
realizations of future system behavior 43
3-3 Sustainability and hazard of an AR(1) process with constant parameters,
declining expected value, abrupt decrease and increase in mean, decrease
and increase in variance, and decrease and increase in autocorrelation 47
3-4 Simulated sustainability of a Texas rice farm under four scenarios 62
4-1 Parallel ecological, economic and social hierarchies of agricultural systems .... 68
4-2 Object representation of the main components of a farming system 70
4-3 Tree of resource class hierarchy inFSS 75
4-4 Flow of information in FSS from the generation of a field operation to its
effect on resource accounting 89
4-5 Pseudocode representation of the Use method of the consumable resource
class 90
4-6 Pseudocode representation of the SELL method of the consumable resource
class 91
4-7 Pseudocode representation of the USE method of the timed resource class .... 92
Xll


Table A-l. Object-oriented programming terms.
Term
Explanation
object
The basic structural unit of an object-oriented program.
class
A template for forming an object.
data
An objects set of variables. The set of data defines the state of an
object.
methods
Algorithms that act on an objects data. A method is analogous to a
procedure in procedural programming.
encapsulation
The incorporation of data and methods into a single unit with a single
memory address.
message
The means by which public data and methods are accessed from
outside an object.
instantiation
The operation of creating an object at runtime from a class.
constructor
A method that creates an object in memory and initializes its data.
destructor
A method that disposes of an object from memory.
inheritance
The ability of classes and their derived objects to obtain data and
methods from their ancestral classes.
class hierarchy
A set of classes that descends from a common ancestral class.
polymorphism
The ability of different objects derived from a given class hierarchy to
exhibit different behavior in response to a single message.
collection
An object that implements a dynamically sizeable array of pointers to
objects.
early binding
In early binding, the code that will execute in response to a
procedural call is determined at compile time.
late binding
In late binding, the code that will execute in response to a procedural
call is determined at execution time.
The benefits of polymorphism are best seen in the context of collections. A
collection is an object that implements a dynamically sizeable array of pointers to objects


195
Table 6-9. Soil loss and predicted 15 year sustainability ($15)SEs) of erosion scenarios
and G-test statistic (G^j) for difference from the erosion @ 0 Mg ha'1 yr'1 scenario. The
McNemar test statistic was always undefined.
Scenario
annual
- Soil loss (cm)
total (15)
SE$
Gi,,dj
erosion @ 0 Mg ha'1 yr'1
0.00
0.00
0.677
0.047
t
erosion @ 25 Mg ha'1 yr'1
0.56
8.33
0.60
0.049
1.0 n.s.
erosion @ 50 Mg ha'1 yr'1
1.11
16.67
0.54
0.050
3.5 n.s.
erosion @ 100 Mg ha'1 yr'1
2.22
33.33
0.29
0.045
29.5 **
erosion @ 150 Mg ha'1 yr'1
3.33
50.00
0.16
0.037
56.2 **
1 Comparison does not apply to the erosion @ 0 Mg ha'1 yr'1 scenario.
Figure 6-11. Sustainability time plot of base and price scenarios.


64
original purpose, their simulation analysis was comprehensive enough to illustrate how the
probability of continuation integrates the effects of a range of factors, and how sensitivity
analysis can be used to identify and rank those factors that constrain sustainability.


245
@N
GENERAL
NYERS
NREPS
START
SDATE
RSEED
SHAME
3
GE
10
1
P
93255
2150
Mai se
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
3
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
3
ME
W
S
E
R
S
C
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RES ID
HARVS
3
MA
A
N
D
A
M
@N
OUTPUTS
XCODE
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
3
OU
N
N
A
10
N
N
N
N
N
N
Y
@
AUTOMATIC MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
3
PL
260
305
35
100
15
40
0
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
3
IR
30
50
100
GS001
IR005
0
1.00
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
3
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
3
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
3
HA
0
365
100
0
@N
GENERAL
NYERS
NREPS
START
YRDAY
RSEED
SNAME
4
GE
1
1
S
93255
2150
Fallow 1
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
4
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
4
ME
w
S
E
R
S
c
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RESID
HARVS
4
MA
A
N
N
A
R
@N
OUTPUTS
FNAME
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
4
OU
N
N
N
10
N
N
N
N
N
N
Y
e
AUTOMATIC MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
4
PL
1
366
0
100
15
50
0
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
4
IR
30
50
100
GS001
IR001
10
0.75
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
4
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
4
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
4
HA
0
365
100
0
@N
GENERAL
NYERS
NREPS
START
YRDAY
RSEED
SNAME
5
GE
1
J.
1
£
93255
2150
Fallow 2
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
5
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
5
ME
W
c
E
R
S
r
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RESID
HARVS
5
MA
E.
N
N
A
D
@N
OUTPUTS
FNAME
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
5
OU
N
K
N
10
N
N
N
N
N
N
Y
Figure D-2, continued.


170
predict the sustainability of that farm. Inferences apply not to the real farming system, but
to a model of the system that is forced to follow the management practices defined in each
scenario.
Approach
A Colombian hillside farm
A farm located in the Cabuyal River catchment (2 47' N, 76 31' W) (Fig. 5-1)
was selected as a basis for the simulation study. Chapter 5 describes the farms physical
environment. Farm selection was based on several criteria. The size of the farm (5.2 ha,
of which 2.3 ha is cultivated with annual crops), its elevation (1650 m), topography and
soils, and the size of the farm household (6 members) are considered representative of
farms in the Cabuyal area. However, the farmer is regarded as an innovator. He has
cooperated with CIAT researchers by providing information, recording weather data, and
providing land for on-farm maize, bean and cassava trials.
The farm supports six family members: the farmer (Mr. Domingo), his mother,
wife, two daughters and a son. The eldest daughter is attending a state university, and the
second plans to soon. The Domingo family lives in a five-room stucco house of above-
average quality for the area. Access to the farm is by a well-maintained gravel road 2 km
from a small town and 4 km from the Panamericana highway. Coffee was the only
commercial enterprise until the farmer accepted the Colombian Coffee Federations
monetary incentive to remove his coffee plants in 1993. Since then, he has grown maize


162
incorporates a simple crop growth model that predicts canopy height and cover, biomass
production, and root growth based on thermal time (Alberts et al., 1989). However, it
possesses no mechanism for predicting impacts of soil erosion on crop production.
An early goal of this research was to link the DSSAT3 crop models with the
erosion component of WEPP. However, several points of incompatibility could not be
resolved within the time frame required. First, the crop models do not provide the
information about crop residues, canopy cover and tillage that the erosion component of
WEPP requires. The relative amount of crop residue standing, flat and buried can impact
predicted soil loss. Residue management is an important tool for managing erosion.
However, the crop models cannot simulate surface residue decomposition but rather
assume that all residue is buried at harvest time. Because the crop models running in
sequence apply all stover as buried residue at harvest time, mass balance errors occur if
either the harvest product is not completely removed or a portion of the stover is
removed. Although the input data for the crop models include tillage, tillage events are
ignored. Tillage affects erosion predictions by influencing surface roughness, infiltration
capacity and the proportion of residue remaining on the surface. The degree of canopy
cover, calculated as a function of leaf area index (LAI) and canopy dimensions, affects
predicted erosion. Although all of the models calculate LAI, only CROPGRO outputs
canopy height and width. The DSSAT3 crop models would require a surface residue
model, a tillage model, and a model of canopy dimensions before they would be
compatible with the erosion submodel of WEPP. Errors in the existing residue model
must also be addressed.


period, t
Figure 3-1. Relationship between time, distribution of state, and sustainability under (a) constant mean and high variability,
and (b) negative trend and low variability.
-U
o


177
in preliminary simulation analyses. In addition to monetary expenditures, the household
was assumed to consume 650 kg yr'1 of maize and 150 kg ha'1 of bean.
Although several cropping systems were included in the analyses (Fig. 6-2),
management of each crop was similar in all of the annual cropping systems. Detailed crop
management assumptions are given in the input files in Appendix D. I assumed (a) that
the farmer would hire day-laborers for field preparation (removing residues, fertilizer
applications and tillage), planting, weeding, tying tomato plants, pruning coffee, harvesting
and post-harvest processing, (b) that he would irrigate, spray pesticides, apply fertilizer as
side-dressing, prepare tomato seedlings and prune tomato himself if he had time, otherwise
hire day-laborers, and (c) that he would take care of marketing himself.
For each crop, a recovery factor accounts for harvest losses and yield-reducing
factors that the crop models do not account for (Table 6-6). The low recovery factor for
tomato is based on the difference between simulated yields and those reported by a farmer
in Siberia, about 2 km from the Domingo farm. It reflects the vulnerability of tomato to
loss from diseases and the inability to market deformed or damaged fruit. The recovery
factor for cassava (0.4) reflects the fact that the cassava model (CropSim CASSAVA v.
1.0, Hoogenboom et al, 1994) does not account for nitrogen stress. Ashby (1987)
showed that with given soil fertility inputs, bean yields were lower in farmer-managed
trials than in researcher-managed on-farm trials in the Cauca region (Fig. 6-3). The
difference reflected differences in intensity of weed and pest control. The magnitude of
the difference was not consistent among trials or fertility treatments. The recovery factor
of 0.8 is probably optimistic for bean and fresh maize.


146
In CROPGRO, nodule growth and N2 fixation are driven by C balance
(Hoogenboom et al., 1990). In a given day, C is allocated according to a priority scheme:
maintenance respiration > reproductive growth > vegetative growth > N2 fixation >
nodule growth. Carbon is allocated to N2 fixation based on N deficit only when
insufficient N prevents the plant from using all available C for growth. Nodule growth can
then occur only if inadequate nodule mass (a sink limitation) prevents using all of the C
allocated for N2 fixation. The sink limitation to nodule growth is based on current nodule
mass and a maximum daily relative growth rate. A result of the priority scheme is that
nodule growth, and therefore the capacity to fix N2 during seed formation, is extremely
sensitive to small changes in N supply early in the season (Fig. 5-19).
Tewari (1995) measured nodule growth and N2 fixation in beans grown with 0 and
275 kg ha'1 of N fertilizer in an experiment in Kuiaha, Hawaii. CROPGRO over predicted
the observed response to applied N, predicting nearly complete inhibition of nodule
growth and N2 fixation (Fig. 5-20a and c). I modified CROPGRO to force it to reserve a
minimum amount of C for nodule growth. The modification reserves an amount of C
(CNODMN) for nodule growth equivalent to a fixed fraction (FRCNOD) of the C
allocated to current root growth. Any surplus C remaining after N2 fixation is added to
CNODMN for nodule growth. Carbon allocated to nodule growth may be unused and
made available for vegetative growth if current nodule mass limits growth. Carbon is
allocated to CNODMN only up to first pod.
Three parameters were available for calibrating nodule growth in the modified
version of CROPGRO: FRCNOD, initial nodule mass (DWNODI), and maximum relative


234
Table B-15. Format of the HISTORICAL section of the price file.
Description
Header
Formal
Index of price
N
013
Name of price
Decsr
1 C 16
Transformation
Tr
1 C 1
N = no transformation
L = log10 transformation
Intercept of linear trend at the specified starting date
Intcp
0 R 8
Slope of linear trend relative to specified starting date
Slope
0 R 8
Year of starting date for trend calculation
StYr
1 14
Month of starting date for trend calculation
StMo
3 12
Name of text file containing historical series
File
1 C 12
f See footnote, Table B-l.


137
generated than with historical weather, distributions differed significantly (P = 0.05) only
for October-planted maize. The Kolmogorov-Smimov test did not identify differences
between generated and observed distributions of rainfall amounts grouped by calendar
month or by year (Table 5-12).
Only one of the four cases (October-planted bean) was consistent with the results
of Jones and Thornton (1993), who used CERES-Maize to test rainfall generation
techniques for conditions at the CIAT headquarters near Palmira, Colombia. In their
study, the variance of simulated yield distributions was significantly lower when WGEN
was used instead of historical weather data.
Table 5-11. Mean (x), standard deviation (s) skewness (a3), and Kolmogorov-Smimov
test statistic (D) for distributions of crop yield and maturity times simulated with observed
and simulated weather from La Florida, Popayan, Colombia.
Planting
date
Weather
source
X
s
X
Maize
S 3
D
3
D
Grain yield, kg ha'1:
October
observed
2223
209.0
0.126
0.343 *
4728
389.6 0.057
0.114 n.s.
generated
2336
162.6
0.284
4730
489.8 0.913
March
observed
1561
175.5
0.363
0.143 n.s.
4322
414.5 -0.059
0.286 n.s.
generated
1564
188.2
-0.139
4161
307.5 1.061
Days to physiological maturity:
October
observed
83
2.7
0.052
0.257 n.s.
149
6.8 -0.474
0.343 *
generated
83
1.6
0.509
148
3.2 -0.395
March
observed
81
2.7
0.092
0.286 n.s.
145
7.1 -0.161
0.229 n.s.
generated
81
1.2
-0.733
145
4.2 -0.191


89
Figure 4-4. Flow of information in FSS from the generation of a field operation to
its effect on resource accounting.


115
soils with crystalline mineralogy. In an incubation study, the decomposition rate of native
organic C was an average of eight times higher in non-allophanic than in allophanic soils
(Fig. 5-6, Martin et al, 1982). Additions of allophane to a sandy loam reduced the rate of
decomposition of wheat straw (Fig. 5-7, Zunino et al, 1982). Suppressed mineralization
in Andisols has been attributed to inadequacy of P (Munevar & Wollum, 1977) and
soluble C (Monreal et al, 1981) for microbial growth, and the formation of stable
complexes with Al and Fe (Zunino et al., 1982). The suppression of mineralization by
Figure 5-7. Decomposition of 14C-labeled wheat straw in Lo Aguire sandy loam
with added allophane. Data from Zunino et al, 1982.


Legend
nominal
participation
1
consultive
participation
i
decision-making
participation
Figure 6-3. Influence of type of farmer participation in on-farm trials on bean response to ground (a) and partially acidulated
rock phosphate (b), chicken manure (c), and 10-30-10, Pescador, Cauca, Colombia, 1982 and 1983. Data from Ashby (1987). ~
VO


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY
By
James William Hansen
May 1996
Chairperson: Dr. James W. Jones
Major Department: Agricultural and Biological Engineering
The potential usefulness of the concept of sustainability as a criterion for
evaluating and improving agricultural systems has been hindered by inadequacy of
approaches for its characterization. The approach presented here for characterizing farm
sustainability is based on a definition of sustainability as the ability of a system to continue
into the future, expressed by the probability that the system will continue without violating
failure thresholds during a particular future period. Characterization includes
quantification and diagnosis of constraints. Sustainability is quantified by using long-term,
stochastic simulation to sample realizations of the future behavior of a model of the
farming system. Sensitivity analysis provides a tool for testing hypotheses about
constraints to sustainability. An object-oriented farming system simulator developed for


5-7 Decomposition of 14C-labeled wheat straw in Lo Aguire sandy loam with
added allophane 115
5-8 Locations and elevations of weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation 123
5-9 Trend in the ratio of bean yields observed in on-farm trials and predicted
by CROPGRO, Domingo and Trujillo farms, Cauca, Colombia 131
5-10 Simulated and observed bean yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994, October 1994 and
March 1995 133
5-11 Simulated and observed maize grain yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994 and October 1994 .... 135
5-12 Distribution of simulated yields of October- and March-planted maize and
bean in response to historical and generated weather variability, La Florida,
Popayan, Colombia 138
5-13 Simulated grain and biomass yield response of maize and October- and
March-planted bean applied N using the default mineralization factor 140
5-14 Maize response to applied N on several volcanic soils in Nario, Colombia .. 141
5-15 Effect ofN mineralization factor, SLNF, on simulated maize response to
applied N 142
5-16 Simulated grain and biomass yield response of maize and October- and
March-planted bean to applied N using the adjusted mineralization factor .... 143
5-17 Grain and biomass yields of October- and March-planted bean simulated
by the original and modified versions of CROPGRO 144
5-18 Grain and biomass yields of irrigated and rainfed soybean simulated by the
original and modified versions of CROPGRO 145
5-19 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Domingo farm, Cauca, Colombia, 1994 147
xiv


176
Table 6-6, Adjustments to simulated yields and reported prices, Cauca, Colombia.
Product
Recovery*
Moisture
adjustment*
Quality
adjustment
Market
discount
coffee
1.00
1.00
0.90
1.00
cassava
0.40
2.86
1.00
0.20
fresh maize {choclo)
0.80
1.90
1.00
0.70
tomato (Manalucie)
0.40
16.67
1.00
0.70
bean (Radical)
0.80
1.00
1.00
0.65
* Ratio of expected harvest to potential yield, dry basis.
* Ratio of market to dry weight.
§ Discount due to difference in quality or cultivar.
11 Ratio of price paid to farmer to price paid by wholesaler.
Assumptions
Some of the information needed for simulating the farm was not readily available;
reasonable values had to be assumed. I assumed that the farmer initially had Col.$
4,000,000 (Col.$ 1,700,000 savings + 2,300,000 incentive payment for removing coffee)
in his operating fund in September 1994. He estimated his minimum annual household
expenses at Col.$ 3,000,000. This amount was adjusted down to Col.$ 2,250,000 on the
assumption that the family could endure more scarcity than he estimated. Discretionary
spending was assumed to be 20% of their liquid assets (i.e., monetary savings plus the
value of any stored material inputs and harvest products) per year. The Domingo familys
annual household expenses are probably high for the region because of college expenses
for the eldest daughter. Because household expenditure parameters could not be
measured, I adjusted them to obtain an intermediate level of fifteen-year farm sustainability


125
Soil data
Soils were sampled at eight locations across the Domingo farm (Fig. 5-2) on
January 24, 1994. At each location, soil was sampled to the depth of the Ap horizon, 30-
40 cm and 50-60 cm deep. Each Ap horizon sample was a composite from a center hole
plus four additional holes 2 m to each side of the center hole. Soil input data from a site 2
m from the edge of the bean trials served as the basis for all crop simulations presented in
this chapter. The soil analysis laboratory at CIAT measured the properties listed in Table
5-2. A moisture release curve was used to determine saturated water content (0 atm) and
lower limit (LL) of plant extractable water (-15 atm). Drained upper limit (DUL) of plant-
extractable water was measured in the field.
The soil data utility program, SDB3 (Hunt et al., 1994), was used to set up soil
profile input data. SDB3 uses soil texture and permeability and drainage codes to estimate
hydrological properties (SCS runoff curve numbers, saturated hydraulic conductivity,
evaporation parameters, and critical water contents) (Ritchie et al., 1990a). SDB3
determined root weighting factors from observed abundance of fine roots and from pH
and Al saturation. Ritchie et al. (1990a) warned that the relationships used by SDB3 to
estimate hydrological parameters should not be used on volcanic soils, therefore measured
critical water contents and conductivity values were used rather than the estimated values.


143
5000
nj4000

2
.3000
'5k
.=2000
(0
O
1000
0
Figure 5-16. Simulated grain and biomass yield response of maize (a, b) and October-
(c, d) and March-planted (e, f) bean to applied N using the adjusted mineralization
factor (SLNF = 0.35), Domingo farm, Cauca, Colombia. Mean SD of 10 replicates.


212
It is not necessary that objects in a collection derive from the same class. Collections have
iterator methods that can send messages to all or to a subset of collected objects. The
type and number of objects in a collection may not be determined until they are inserted at
runtime in response to, for example, interactive input or an input file. This is known as
late-binding, in contrast to early-binding in which procedure and function calls are
determined when the program is compiled. Using collection iterators to apply simulation
processes to a set of polymorphic model components results in a great deal of flexibility in
representing model structure.
An object-oriented approach offers several potential benefits including modularity,
comprehensibility and extensibility. Modularity results from decomposition of a program
into small, logical, loosely-coupled units. In OOP, modules take the form of objects, with
loose coupling between objects accomplished by messages. Modularity simplifies
collaborative development and code maintenance. Model components can be developed
and tested independently. Comprehensibility is enhanced by modularity. Furthermore, an
object-oriented model can be expressed clearly as a set of interacting objects with state
and behavior that are analogous to the components of the real system. The naming and
organization of classes and objects helps to communicate the structure of the system being
modeled. A third benefit of OOP is extensibility. Aggregation to higher hierarchical levels
is an important aspect of extensibility. The ability to create multiple instances of classes
mimics the occurrence of multiple subsystems in a system (eg., enterprises in a farm).
Program organization that provides for easy extension facilitates incremental development.


81
event handling, resource accounting and household consumption. The sections that follow
discuss these processes in more detail.
Overview
During a given simulation year, FSS performs the following steps. First, methods
reset strategies, resources and events in preparation for the new simulation year. Second,
prices advance one year. FSS generates monthly prices for the current year, and stores
them for use in linearly interpolating to daily values as needed. Third, crop activities are
simulated in order of landscape position. For each field, FSS calls external crop models to
simulate all crop and fallow activities for the current year. The event queue accumulates
field operations returned by the crop models for later processing. Fourth, any additional,
scheduled operations are retrieved from each enterprise and inserted into the event queue.
Fifth, events are executed. Household consumption and all resource transactions occur in
response to execution of events. Some information is written to output files by events as
they execute. The event queue checks conditions for farm failure with each event that it
executes. If event execution fails, then the current replicate is suspended and the failure is
recorded for later analysis. Finally, if the user requested resource output then the status of
each resource is written to a file.
At the beginning of each replicate of a scenario, prices, resources, enterprises and
the landscape are reinitialized so that starting dates and initial conditions are identical
among replicates.


259
Mullin, B.A., G.S. Abawi, M.A. Pastor-Corrales, and J.L. Kornegay. 1991. Root knot
nematodes associated with beans in Colombia and Peru and related yield loss.
Plant Disease 75:1208-1211.
Munevar, F. and A.G. Wollum. 1977. Effects of additions of phosphorus and inorganic
nitrogen on carbon and nitrogen mineralization in Andepts from Colombia. Soil
Sci. Soc. Am. J. 41:540-545.
Muoz, R.A. and A.P. Wieczoreck. 1978. Fertilizacin de maiz (Zea mays L.) en suelos
volcnicos de Nario, Colombia. Revista ICA (Bogot) 13:11-20.
Neher, D. 1992. Ecological sustainability in agricultural systems: definition and
measurement, p. 51-61. In R.K. Olson (ed.) Integrating Sustainable Agriculture,
Ecology, and Environmental Policy. Food Products Press, New York.
Neufeldt, E. (ed.). 1988. Websters New World Dictionary, Third College Edition. Simon
& Schuster, Inc., New York.
Norgaard, R.B. 1991. Sustainability: Three methodological suggestions for agricultural
economics. Can. J. Agrie. Econ. 39:637-645.
OConnell, P.F. 1990. Policy development for the low-input sustainable agriculture
program, p. 453-458. In C.A. Edwards, R. Lai, P. Madden, R.H. Miller, and G.
House (ed.) Sustainable Agricultural Systems. Soil and Water Conservation
Society, Ankeny, Iowa.
OConnell, P.F. 1992. Sustainable agriculture -- a valid alternative. Outlook on Agrie.
21(1):5-12.
Oram, P.A. 1988, Building the agroecological framework. Environment
30(9): 14-17,30-36.
Overman, A.J. 1991. Diseases caused by nematodes, p. 49-50. In J.B. Jones, J.P. Jones,
R.E. Stall and T.A. Zitter (ed.) Compendium of Tomato Diseases. APS Press, St.
Paul, Minnesota.
Pachico, D. (undated). Costs and returns in three bean systems, Nario. (unpublished
draft)
Pankratz, A. 1983. Forecasting With Univariate Box-Jenkins Models: Concepts and
Cases. John Wiley & Sons, New York.


99
Outputs
FSS creates five output files that relate to sustainability: (a) annual values of all
resources for the current scenario (RESOURCE. OUT), (b) sorted final values of a
particular resource specified in the scenario file for a set of scenarios (ENDWLTH.OUT),
Box Plot of UEALTH, Scenario GRAPH2
Year
Figure 4-15. Example of a resource box plot generated by FSS.


This dissertation was submitted to the Graduate Faculty of the College of
Agriculture and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
May, 1996
Dean, College of Agriculture
Dean, Graduate School


average component (Box and Jenkins, 1970) to simulate price cycles longer than a year.
The stochastic component of the model (w,) is calculated as,
p q
s t't-S >
P 6,
q < 6,
et ~ N(0, oE),
[4-2]
where 4>¡ and 0¡ are autoregressive and moving average coefficients for a lag of i months,
p and q are the maximum lag of the autoregressive and moving average components, S is
the lag of an optional multiplicative seasonal moving average component (0S), and et is a
random normal deviate. The simulated valuer, is the sum of the deterministic and
stochastic components, truncated if necessary to avoid unreasonably low values,
yx = max(wt + jtt, 0.05 xt),
[4-3]
or, if the data used to fit the model were log-transformed,
|Q(max(wt + xt, 0.05xt))
[4-4]
A set of ARMA prices constitutes a multivariate, contemporaneous ARMA process
that accounts for cross-price correlation. In a contemporaneous ARMA model, the
current value of a particular series is assumed to depend on prior values within the series
and on random shocks that influence other series, but not on prior values of other series


4-2 Description of resource classes 77
4-3 Format of the operations output file 87
5-1 Area in each slope class, Domingo farm, Cauca, Colombia 106
5-2 Properties of soil layers, site near on-farm trials, Domingo farm, Cauca,
Colombia 112
5-3 Spatially interpolated WGEN coefficients for the Domingo farm, Cauca,
Colombia 125
5-4 Planting information for crop simulation studies 126
5-5 Observed and predicted bean yields, Domingo and Trujillo farms, Cauca,
Colombia 130
5-6 Observed and predicted timing of phenological events for bean and maize,
Domingo farm, Cauca, Colombia, planted March 30, 1994 130
5-7 Treatment description and observed and predicted yields for the October 14,
1993 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132
5-8 Treatment description and observed and predicted yields for the March 30,
1994 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132
5-9 Observed and predicted maize yields, Domingo and Trujillo farms, Cauca,
Colombia 134
5-10 Observed and predicted cassava yields, Domingo and Trujillo farms, Cauca,
Colombia 135
5-11 Mean, standard deviation, skewness and Kolmogorov-Smirnov test statistic
for distributions of crop yield and maturity time simulated with observed and
simulated weather from La Florida, Cauca, Colombia 137
5-12 Mean, standard deviation, skewness, and Kolmogorov-Smirnov test statistic for
distributions of observed and simulated monthly and annual rainfall totals, La
Florida, Popayan, Colombia 139
6-1 Hypotheses related to determinants of farm sustainability 169
IX


VO
Figure 4-13. Forrester representation of variable cost linkages between a consumable resource (a) and three linked
consumable resources (b, c and d).


1S7
hypothesized constraints to sustainability of the model farming system in the context of its
model environment.
Results from the base scenario illustrate several perspectives from which farm
behavior can be viewed. A box-plot (Fig. 6-4) shows the magnitude, trends and changes
in dispersion of an aggregate measure of the status of a system, such as liquid assets. A
cumulative distribution plot (Fig. 6-5) gives a more complete but static picture of the
distribution of liquid assets at any point in time. Figures 6-4 and 6-5 show that the
dispersion of liquid assets and the cumulative probability of failure increase as the scenario
Figure 6-4. Box plot of liquid assets, base scenario, Domingo farm, Cauca, Colombia.


241
STRATEGIES
8
N
Ev
Name
i
1
MZ-
-BN-
-BN-
-CS-
t
1 ow
intensity
, phase a
e
N
Pe
Ac
Cl
CU
MP
MI
MF
MR
MC
MT
ME
MH
SM
LI
Recover
1
1
1
I
3
3
0
1
1
0
1
0
0
3
0
0.80
1
1
2
I
o
0
0
0
0
0
0
0
1
5
0
0.00
1
1
3
I
1
1
0
0
1
1
1
0
0
1
0
0.80
1
1
4
I
2
0
0
0
0
0
0
0
3
4
0
0.00
1
2
1
I
1
2
0
0
1
1
1
0
0
2
0
0.80
1
2
2
I
0
0
0
0
0
0
0
13
4
0
0.00
1
2
3
I
4
4
0
0
1
0
3
0
11
6
0
0.40
1
3
1
I
o
0
0
0
0
0
0
0
4
4
0
0.00
e
N
Ev
Name
2
1
MZ-
-BN-
-BN*
-CS-
low
intensity
, phase b
e
N
Pe
Ac
Cl
CU
MP
MI
MF
MR
MC
MT 1
ME
MH
SM
LI
Recover
2
1
1
I
1
12
0
0
1
1
1
0
0
2
0
0.80
2
1
o
L.
I
2
0
0
0
0
0
0
0
14
4
0
0.00
2
1
3
I
4
14
0
0
1
0
3
0
11
6
0
0.4 0
2
2
1
I
o
0
0
0
0
0
0
0
7
4
0
0.00
2
3
1
I
3
13
0
1
1
0
1
0
0
3
0
0.80
2
3
O
I
2
0
0
0
0
0
0
0
1
5
0
0.00
2
3
3
I
1
11
0
0
1
1
1
0
0
I
0
0.80
2
3
4
I
2
0
0
0
0
0
0
0
6
4
0
0.00
@
N
Ev
Name
3
2
MZ-
-BN-
-BN-
-CS-
low
intensity
, phase c
8
N
Pe
Ac
C1
CU
MP
MI
MF
MR
MC
MT 1
ME
MH
SM
LI
Recover
3
1
1
I
2
0
0
0
0
0
0
0
3
4
0
0.00
3
2
1
I
3
8
0
1
1
0
1
0
0
3
0
0.80
3
2
2
I
9
0
0
0
0
0
0
0
1
5
0
0.00
3
2
3
I
1
6
0
0
1
1
1
0
0
1
0
0.80
3
o
4
I
2
0
0
0
0
0
0
0
9
4
0
0.00
3
3
1
I
1
7
0
0
1
1
1
0
0
2
0
0.80
3
3
2
I
2
0
0
0
0
0
0
0
15
4
0
0.00
3
3
3
I
4
9
0
0
1
0
3
0
11
6
0
0.40
ENTERPRISES
@ N Name
1 Maize-Bean-Cassava A
@ Subfields
1 2
@ Strategies
1
8 N Name
2 Maize-Bean-Cassava B
@ Subfields
3 4
@ Strategies
2
@ N Name
3 Maize-Bean-Cassava C
@ Subfields
5 6
@ Strategies
3
CONSUMPTION DECISIONS
@
Resource
Subsist
Wealth
MPC
OPERATING FUND
2.30E
WEALTH
0.20
HA001
MZ IB0065
650
WEALTH
0.00
HA001
BN IB002
150
WEALTH
0.00
Name
Maize
Fallow!
Beanl
Fa11owl
Bean2
Fa11owl
Cassava
Fa11owl
Name
Bean2
Fallowl
Cassava
Fallowl
Maize
Fallow!
Beanl
Fallowl
Name
Fallowl
Maize
Fallow!
Beanl
Fallowl
Bean2
Fallowl
Cassava
Figure D-l, continued.


25
Resilience
Conway (1985) defined sustainability as resilience: . the ability of a system to
maintain productivity in spite of a major disturbance (p. 25). He suggested that
measurement of five system properties are necessary to characterize resilience: inertia,
elasticity, amplitude, hysteresis, and malleability (Conway, 1994). Cramb (1993) based
inferences about the sustainability of two shifting cultivation systems in eastern Malaysia
on both trends and resilience. Pepper production was determined to be sustainable
because production in 1989 recovered to its 1980 level in response to price recovery after
production diminished (in 1985) due to a period of low prices. Rubber production was
considered more sustainable at Batu Linang than at Nanga Tapih, as indicated by recovery
of depressed production in response to price recovery.
Like time trends, resilience can be viewed as an aggregate system response to
determinants of sustainability. However, inability to identify a single measure of resilience
(Conway, 1994) leads to the same problems of interpretation faced when using a diverse
set of indicators to characterize sustainability. Assumptions about the likelihood and
timing of disturbances have been avoided by interpreting sustainability as an intrinsic
property of an agricultural system in isolation from its environment (Conway and Barbier,
1990). However, York (1988) argued that (un)sustainability is not an intrinsic property
but rather a response to changing environmental and socioeconomic circumstances.
Predictions about future sustainability cannot be made in the absence of assumptions about
changes and variability in those higher-level systems that comprise a systems


Tables 5-7 and 5-8 show simulated and observed yields for the bean fertility
management trial on the Domingo farm. Observed sensitivity to fertilizer or manure
inputs was much greater than predicted by CROPGRO (Table 5-7). This is easily
explained by the confounding effect of P and other nutrients to which the crop models are
not sensitive. While CROPGRO simulates crop response to added N, a substantial
amount of P was added in the 10-30-10 fertilizer and manure treatments. The observed
treatment response was probably more a P than an N response, considering the high P
Jan-94 May-94 Sep-94 Jan-95 May-95
Harvest date
Figure 5-9. Trend in the ratio of bean yields observed in on-farm trials and predicted
by CROPGRO, Domingo and Trujillo farms, Cauca, Colombia.


CHAPTER 2
IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT?
Introduction
In literal English usage, sustainability is the ability to keep in existence; keep up;
maintain or prolong (Neufeldt, 1988, p. 1349). The variety of meanings acquired by
sustainability as applied to agriculture (Table 2-1) have been classified according to the
issues motivating concern (Douglass, 1984; Weil, 1990), their historical and ideological
roots (Kidd, 1992; Brklacich et al., 1991), and the hierarchical levels of systems
considered (Lowrance et al., 1986).
The distinction between sustainability as a system-describing and as a goal
prescribing concept (Thompson, 1992) identifies two current schools-of-thought that
differ in their underlying goals. The goal-prescribing concept interprets sustainability as an
ideological or management approach to agriculture. This concept developed in response
to concerns about negative impacts of agriculture, with the underlying goal of motivating
adoption of alternative approaches. The system-describing concept interprets
sustainability either as an ability to fulfill a diverse set of goals or as an ability to continue.
This concept can be related to concerns about impacts of global change on the viability of
4


168
exposed to erosive rains. Resulting soil erosion may irreversibly reduce crop growth and
yield. Furthermore, the Cauca River watershed is strategic as a source of water and
hydroelectric power for the city of Cali and for irrigated sugar production in the Cauca
Valley. There is growing concern that the shift from perennial crops will result in
increasing siltation and chemical pollution, and poorer regulation of flow rates in the
Cauca river.
On the positive side, the strategic Panamericana highway runs through the Cauca
river watershed, potentially providing access to the rapidly growing urban market of Cali.
In spite of the proximity of the Panamericana, most of the bean, maize and cassavathe
traditional crops for the regionare grown for subsistence consumption or for local
markets. Commercial production of specialty commodities, such as vegetables and cut
flowers is in an early stage of development.
Chapter 3 describes a framework for using long-term, stochastic simulation of a
farming system to characterize its sustainability. The purpose of the current chapter is to
apply that framework to a farm in the Cauca River watershed in order to gain insight into
the impact of cropping system, soil management, costs and prices, resources and sources
of risk on farm sustainability. This chapter has two specific objectives: (a) to demonstrate
and evaluate a systems approach for characterizing farm sustainability, and (b) to test a set
of hypotheses (Table 6-1) about determinants of sustainability of a particular hillside farm
in the Cauca watershed of Colombia.
A word of caution regarding interpretation of results is in order. Although this
study is based on an actual farm, it does not try to mimic all of the characteristics or


260
Perry, G.M., M.E. Rister, J.W. Richardson, W.R. Grant, and J.W. Sij, Jr. 1986. The
Impact of Tenure Arrangements and Crop Rotations on Upper Gulf Coast Rice
Farmers. B-1530. Texas Agricultural Experiment Station, College Station, Texas.
Plucknett, D.L. 1990. International goals and the role of the international agricultural
research centers, p. 33-49. In C.A. Edwards, R. Lai, P. Madden, R.H. Miller, and
G. House (ed.) Sustainable Agricultural Systems. Soil and Water Conservation
Society, Ankeny, Iowa.
Rai, R. 1992. Effect of nitrogen levels and Rhizobium strains on symbiotic N2 fixation and
grain yield of Phaseolns migar is L. genotypes in normal and saline-sodic soils.
Biol. Frtil. Soils 14:293-299.
Richardson, C.W. 1985. Weather simulation for crop management models. Trans. ASAE
18:1602-1606.
Richardson, J.W. and G.D. Condra. 1981. Farm size evaluation in the El Paso Valley: a
survival/success approach. Am. J. Agrie. Econ. 63:430-437.
Richardson, J.W. and C.L. Nixon. 1985. FLIPSIM v: a general firm level policy simulation
model. B-1528. Department of Agricultural Economics, Texas A&M Univ.
Rietveld, M.R. 1978. A new method for estimating the regression coefficients in the
formula relating solar radiation to sunshine. Agrie. Meteorology 19:243-252.
Ritchie, J.T., D.C. Godwin, and U. Singh. 1990a. Soil and water inputs for the IBSNAT
crop models, p.31-45. In International Benchmark Sites Network for
Agrotechnology Transfer (IBSNAT) Project. Proceedings of the IBSNAT
Symposium: Decision Support Systems for Agrotechnology Transfer, Las Vegas,
NV. 16-18 Oct., 1989. Part I: Symposium Proceedings. Dept, of Agronomy and
Soil Science, College of Tropical Agriculture and Human Resources, Univ.
Hawaii, Honolulu, Hawaii.
Ritchie, J.T., D.C. Godwin, and U. Singh. 1990b. The CERES models of crop growth and
yield. In International Benchmark Sites Network for Agrotechnology Transfer
(IBSNAT) Project. Proceedings of the IBSNAT Symposium: Decision Support
Systems for Agrotechnology Transfer, Las Vegas, NV. 16-18 Oct., 1989. Part II:
Posters. Dept, of Agronomy and Soil Science, College of Tropical Agriculture and
Human Resources, Univ. Hawaii, Honolulu, Hawaii.
Rodale, R. 1990. Finding the middle of the road on sustainability. J. Prod. Agrie.,
3:273-276.


D-l Farm scenario file used to simulate the base scenario 239
D-2 Crop management file used to simulate farm scenarios 242
D-3 Soil profile description file used to simulate farm scenarios 247
XVII


CHAPTER 3
A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM SUSTAINABILITY
Introduction
The potential benefits of applying the concept of agricultural sustainability-
providing feedback about future impacts of current decisions, and focusing research and
intervention by identifying constraintscan only be realized when sustainability of
particular systems is characterized. Characterization includes both quantification and
diagnosis of constraints.
In spite of the tremendous amount of concern about agricultural sustainability,
surprisingly few studies have attempted to characterize the sustainability of specific
agricultural systems. The methods which have been proposed or applied suffer from (a)
conceptual problems associated with interpreting sustainability as an approach rather than
a property of agriculture, and (b) practical difficulties that arise from the fact that
sustainability deals with the future (Chapter 2). Characterization based on management
practices either does not relate to a literal interpretation of sustainability or it leads to
circular logic. Attempts to characterize sustainability based on system response have
generally ignored or misinterpreted important system properties.
33


B-8 Format of the SCHEDULED OPERATIONS section of the scenario file .... 225
B-9 Format of the LANDSCAPE section of the scenario file .. 226
B-10 Format of the STRATEGIES section of the scenario file 228
B-ll Format of the ENTERPRISES section of the scenario file 230
B-12 Format of the CONSUMPTION DECISIONS section of the scenario file ... 231
B-13 Format of the CONSTANT section of the price file 232
B-12 Format of the ARMA section of the price file 233
B-12 Format of the HISTORICAL section of the price file 234
C-l Options available from FSS menu items 236
X!


11
Sustainability as a set of strategies
Francis and Youngberg (1990) described sustainable agriculture as a philosophy
that guides the creation of farming systems. Specific management strategies are often
suggested by ideological interpretations of sustainability. The strategies promoted as
sustainable (Table 2-3) are based on the types of problems emphasized and on views of
what would constitute an improvement.
Table 2-3. Strategies frequently associated with sustainability.
Strategy
References
Self-sufficiency through preferred use of on-farm or locally
available internal resources to purchased external resources.
a, b, g, d
Reduced use or elimination of soluble or synthetic fertilizers.
a, e, f, h, i, d, k
Reduced use or elimination of chemical pesticides, substituting
integrated pest management practices.
a, c, d, e, f, h, i, j, k
Increased or improved use of crop rotations for diversification,
soil fertility, and pest control.
a, c, d, e, f, h, j
Increased or improved use of manures and other organic
materials as soil amendments.
a, c, f, h, j, k
Increased diversity of crop (and animal) species.
a, d, g, i
Maintenance of crop or residue cover on the soil.
a, d, e
Reduced stocking rates for animals.
a, c, d
Lockeretz, 1988
b Harwood, 1990
c MacRae et al. 1990
d Neher, 1992
* Dobbs et al. 1991
f MacRae et al. 1989
8Gliessman, 1990
h Edwards, 1990
1 Hauptli et al. 1990
j OConnell, 1992
k Hill & MacRae, 1988


19
Approaches to Characterizing Sustainability
Characterization is a prerequisite to applying the concept of sustainability as a
criterion for identifying constraints, focusing research, and evaluating and improving
agricultural policy and practices. The conceptual problem of defining sustainability and
methodological problems imposed by its temporal nature have hindered development of
approaches to characterizing sustainability. Sustainability involves future outcomes that
cannot be observed in the time-frame required for intervention (Lynam and Herdt, 1989;
Harrington, 1992). For this reason, Conway (1994) argued that defining sustainability in
terms of preservation or duration has little practical value.
The variety of approaches reviewed here reflects the different interpretations of
sustainability and methodological difficulties that result from its temporal nature.
Characterization by adherence to prescribed approaches is based on an interpretation of
sustainability as an approach to agriculture. Characterization by multiple qualitative
indicators and attempts to integrate such indicators are consistent with interpreting
sustainability as an ability to satisfy diverse goals. Sustainability as an ability to continue is
usually characterized by time trends or resilience.
Adherence to prescribed approaches
In a study comparing conventional and sustainable farms in South Dakota, Dobbs
et al. (1991) identified farms as sustainable if they reduced chemical inputs relative to


O 10 20 30 40 50
Soil loss, cm
Figure 5-26. Simulated grain yields of October- (a) and March-planted (b)
bean in response to soil loss, Domingo Farm, Cauca, Colombia.


32
Conclusions
The importance and desirability of agricultural sustainability are generally
recognized. However, its potential as a criterion for guiding agricultures response to
change has not been realized. Characterization is a prerequisite to using sustainability as a
basis for guiding change. Logical inconsistencies limit the usefulness of characterization
of sustainability interpreted as an ideological or management approach to agriculture.
Interpreting sustainability as an ability to meet diverse goals suggests measuring sets of
system indicators consistent with those goals. However, these measurements have proven
difficult to integrate and interpret in a way that identifies constraints or focuses research.
Literal interpretations of sustainability as an ability to continue into the future suggest
measurable, integrated criteria for its characterization. However, applications of these
criteria-time trends and resilience--have ignored or misinterpreted important aspects of
system behavior. Criteria are needed that relate levels, trends and variability of long-term
system performance to the needs and goals of farmers and of society.
In order for sustainability to be a useful criterion for guiding change in agriculture,
its characterization should be literal, system-oriented, quantitative, predictive, stochastic
and diagnostic. These elements identify weaknesses in existing and proposed approaches,
suggest directions for future development of approaches, and together constitute a
systems approach for characterizing sustainability of agricultural systems. The tools of
system analysis and simulation must be part of approaches that incorporate these elements.


15
of chemical inputs, relatively deep, fertile soils, and relatively stable populations. In
contrast, many less developed tropical regions are characterized by lower levels of
resource consumption, frequent or chronic food shortages, lower levels of chemical
inputs, relatively fragile soils, and rapidly growing populations. Attempts to link strategies
to sustainability by definition fail to consider the need to match technologies to specific
environments.
Dicks (1992) argued that interpretations of sustainability in the U.S. have been
shaped by food surpluses. A shift in concern from global food security to environmental
quality in the 1980s (York, 1991) led to the perspective that . . the question is not can
we produce more food, but what are the ecological consequences of doing so?
(Douglass, 1984, p. 5). However, much of the concern about sustainability in less
developed countries is related to the need to increase productivity to meet future needs of
growing populations (Ruttan, 1988; York, 1988, 1991; Lynam and Herdt, 1989;
Plucknett, 1990). The potential for the desperation imposed by poverty to shorten
peoples planning horizons (Ashby, 1985) raises questions about the ecological
consequences of failing to produce more food (Mellor, 1988; Oram, 1988). The
alternative agriculture movement has not adequately addressed the need to feed rapidly
growing populations in order to prevent both human and ecological disaster.
The second problem is that a distorted caricature of conventional agriculture may
cause approaches that may enhance sustainability to be ignored or rejected because of
their association with conventional agricultural institutions. Although the philosophical
roots of the alternative agriculture movement formed outside of the academic community


37
Table 3-1. Description of symbols used.
Symbol Description
l, T Time variable and a particular time
D{t) Status of a system (1 continuing, 0=failed)
Tf Time to system failure
fir, F-j-p Density and cumulative distribution of Tv
S(T) Sustainability for the period (0, 7]
h(t) Sustainability hazard probability function
x, x State vector and a particular state variable
x0, x0 Failure threshold values for x and x
fx Fx>t Density and cumulative distribution of x at time 1
N Total number of realizations simulated
n(T) Number of realizations continuing at time T
§ S estimated from a finite number of realizations
SEs Standard error of S
z(t), oz A stationary stochastic process and its standard deviation
cl, P Intercept and slope of a deterministic trend
4! One period lag autocorrelation coefficient
e(/), oe A white-noise process and its standard deviation
n. n Base and alternate value of the /th hypothesized determinant of sustainability
4, S in response to the base and alternate value of the /th factor
Ri, r, Absolute and relative sensitivity to the /th factor
vab, N Frequency of a and b occurring, and sum of all frequencies
q Correction factor for continuity in G-tests
Gb Gladj Uncorrected and corrected statistic for tests of independent frequencies,
independent observations
GP, GP adj Uncorrected and corrected statistic for tests of independent frequencies,
paired observations


56
new domain of attraction. The ability to return to the original domain of attraction
depends on the relative stability of the two domains, the nature of the separatix between
them, and the existence of disturbances which could displace the system back across the
separatix. Failure of an agricultural system can be viewed as transition from a useful to a
less useful domain of attraction. The state threshold vector x0 forms the separatix between
the domains. Trenbath et al. (1990) used mathematical models to illustrate abrupt
transitions from useful to less useful domains in response to intensification of three
agricultural systems.
In many cases, farm failure criteria may be difficult to determine. However, it may
be possible to obtain meaningful insights into the relative impact of various stresses on a
farming system by assuming particular failure thresholds when those thresholds cannot be
measured.
Since increasingly restrictive failure thresholds increase the probability of system
failure, sustainability may be viewed as a non-increasing function of threshold levels. A
system is less able to continue at a high level than at a low level.
Determinants of farm sustainability
Sustainability is an aggregate response of a system to a range of external factors,
conditioned by internal characteristics of the system. Any factor that influences means,
trends, variability, autocorrelation or goal levels may influence sustainability. A host of
factors influences the balance between income and expenditure that determines the mean
level of farm wealth. Soil degradation, depletion of scarce resources, technological


261
Ruttan, V.W. 1988. Sustainability is not enough. Am. J. Alternative Agrie. 3:128-130.
Sands, G.R. and T.H. Podmore. 1993. Development of an environmental sustainability
index for irrigated agricultural systems, p. 71-80. In J.K. Mitchell (ed.) Integrated
Resource Management and Landscape Modification for Environmental Protection.
Proceedings of the International Symposium 13-14 December 1993, Chicago,
Illinois.
Scholberg, J.M.S., J.W. Jones, F.H. Beinroth, M.A. Vazquez, B.L. McNeal, and K.J.
Boote. 1995. Modeling Growth of Field-grown Tomatoes in Puerto Rico Using
CROPGRO: Model Validation Using a Data Input Set Collected at the Isabela
Substation, Puerto Rico. Unpublished research report, University of Florida,
Gainesville, Florida.
Scholberg, J.M.S., B.L. McNeal, J.W. Jones, and C.D. Stanley. 1993. Adaptation of a
generic crop-growth model (CROPGRO) for field-grown tomatoes and peppers.
Agronomy Abstracts.
Sentis, I.P. 1992. La erodabilidad de los Andisoles en Latino America. Suelos Ecuatoriales
22:33-43.
Singh, U. 1985. A crop growth model for predicting com performance in the tropics.
Ph.D. diss. University of Hawaii, Honolulu, Hawaii (Diss. Abstr. 85-20320).
Singh, U. and D.C. Godwin. 1990. Phosphorus dynamics in IBSNAT crop models. In
International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT)
Project. Proceedings of the IBSNAT Symposium: Decision Support Systems for
Agrotechnology Transfer, Las Vegas, NV. 16-18 Oct., 1989. Part II: Posters.
Dept, of Agronomy and Soil Science, College of Tropical Agriculture and Human
Resources, Univ. Hawaii, Honolulu, Hawaii.
Singh, U. and P.K. Thornton. 1992. Using crop models for sustainability and
environmental quality assessment. Outlook on Agriculture 21:209-218.
Singh, U., P.K. Thornton, A.R. Saka, and J.B. Dent. 1993. Maize modeling in Malawi: a
tool for soil fertility research and development, p.253-273. In F.W.T. Penning de
Vries, P.S. Teng, and K. Metselaar (ed.) Systems Approaches for Agricultural
Development, vol. 2. Kluwer Academic Publishers, Dordrecht, The Netherlands.
Snedecor, G.W. and W.G. Cochran. 1980. Statistical Methods, Seventh Edition. The Iowa
University Press, Aimes, Iowa.


224
G N
Type
CC
Methd
Pri
Win
Resourcel
BH
Hours
Resource2
BH
Hours
10
MRKT
MZ
0
7
14
JOSE
W
2.0
HIRED TRUCK
W
2.0
10
MRKT
MZ
0
7
14
HIRED LABOR
W
2.5
HIRED TRUCK
W
2.5
Figure B-6. Example item from the OPERATION REQUIREMENTS section.
SCHEDULED OPERATIONS section
The SCHEDULED OPERATIONS section specifies schedules of operations
which are not returned by the crop model (eg., marketing). It can be used to account for
production activities for which a process-level model is not available (eg., coffee).
Produce will not be marketed unless the SCHEDULED OPERATIONS section includes
marketing operations. The SCHEDULED OPERATIONS section contains multiple
items. Items with the same index are grouped together as an operation schedule.
The example in Fig. B-7 schedules a maize (MZ) marketing (MRKT)
operation to follow five days (indicated by BT = L) after each maize harvest operation
(item 7 in the *OPERATIONS section). The operation uses 0.8 times the amount of
maize harvest product (H MZ IB0063) that was harvested in the previous operation.
0 N
Cl
Type
CC
Methd
BT
LOp
Time
Resource
BA
Amount
1
O
MRKT
MZ
0
L
7
5
H MZ IB0063
O
0.8
Figure B-7. Example item from the SCHEDULED OPERATIONS section.


18
specification links criteria for determining sustainability to the goals and values of the
analyst or the author of a definition rather than to the agricultural system. At the farm
level and higher, goals belong to the actors within the system and are, therefore,
endogenous. Kidd (1992) argued that it is not helpful to use sustainability loosely as a
general purpose code word encompassing all of the aspects of agricultural policy that the
authors consider desirable (p. 24).
Sustainability as an ability to continue
The final concept interprets sustainability as a systems ability to continue through
time. Hildebrand (1990) suggested that sustainability may be interpreted as the length of
time that a system can be maintained. According to Hamblin (1992), sustainability implies
that agriculture remains the dominant land use. Lynam and Herdt (1989) and Jodha
(1990) expressed sustainability in terms of maintaining some level of output. Monteith
(1990) added consideration of the possible confounding interaction of changes in input
and output levels. The definitions of Fox (1991) and Hamblin (1992) emphasized the
continuing ability to meet human needs. Conway (1985), Conway and Barbier (1990) and
Altieri (1987) emphasized the ability to withstand disturbances.
Interpreting sustainability as an ability to continue is consistent with literal English
usage of sustain and its derivatives. Its potential usefulness comes from suggesting
criteria for characterizing sustainability, providing a basis for identifying constraints and
evaluating proposed approaches to its improvement. This potential usefulness has been
limited by inadequacy of current approaches for characterizing sustainability.


Outputs 99
Resource Status 99
Sustainability 101
Discussion 101
5 CROP SIMULATION FOR CHARACTERIZING SUSTAINABILITY
OF A COLOMBIAN HILLSIDE FARM 104
Introduction 104
A Colombian Hillside Environment 106
Crop Simulation 116
Approach 120
Weather Data 120
Soil Data 125
Simulation Conditions 126
Development and Yield 127
Response to Environmental Factors 128
Results and Discussion 129
Crop Development and Yield 129
Weather Variability 136
Response to Nitrogen Dynamics 139
Response to Soil Erosion 152
Linking Models 158
Issues in Linking Crop and Whole-farm Models 158
Issues in Linking Crop and Erosion Models 161
Conclusions 164
6 DETERMINANTS OF SUSTAINABILITY OF A COLOMBIAN
HILLSIDE FARM 167
Introduction 167
Approach 170
A Colombian Hillside Farm 170
Sources of Information 171
Assumptions 176
Scenarios 180
Simulation and Analysis 185
Results 186
Base Scenario 186
Cropping Systems 190
Soil Management 192
Costs and Prices 194
vi


14
accumulation while inadequate levels degrade resources through exhaustion. This concept
is in sharp contrast to the decreasing relationship between chemical input levels and
sustainability proposed by Stinner and House (1987) (Fig. 2-lb).
Studies in Mali, Benin, Zambia, and Tanzania provide examples of resource
degradation due to inadequate chemical inputs (Budelman and van der Pol, 1992). In each
case, supplies of soil nutrients were exhausted rapidly due to a combination of harvest,
erosion, leaching, denitrification and volatilization, with harvest being the greatest loss.
Nutrient budgets estimated for several crops in southern Mali were always negative for N
and K, and variable but generally better for P. The authors concluded that the only way to
make these cropping systems sustainable is with increased use of fertilizers.
Discussion
Interpreting sustainability as an approach to agriculture has been useful for
motivating change. Sustainability as an ideology has provided as a common banner for
various agricultural reform movements (Gips, 1988; Dahlberg, 1991). Research and
promotion of sustainability interpreted as a set of strategies has become part of policy in
the U.S. in the form of provisions in the 1990 Farm Bill (OConnell, 1992; Yetley, 1992).
Interpreting sustainability as an approach is not useful for guiding change in
agriculture for several reasons. First, approaches developed in response to problems in
North America and Europe may be inappropriate in regions where circumstances and
problems are different. The alternative agriculture movement has its roots primarily in
regions characterized by high levels of resource consumption, food surpluses, high levels


226
(Jones et al., 1994). They point to landscape information in the crop management file.
The current version can simulate constant annual soil loss. The LANDSCAPE section
contains multiple items.
The example in Fig. B-8 specifies subfield number 1 at the top (position 1) of
hillslope 1 with an area of 0.6 ha. Constant soil loss of 20.0 Mg ha'1 is applied. The item
points to field level 2 and initial conditions level 1 in the crop management file.
Table B-9. Format of the LANDSCAPE section of the scenario file.
Variable
Header
Format1
Plot index
N
013
Hillslope index
HS
013
Position along hillslope
PO
013
Field level from crop management file
FL
013
Initial conditions level
IC
013
Environmental modifications level
ME
013
Erosion model:
N = none
C = constant annual erosion
U = USLE {not implemented}
M = MUSLE {not implemented}
0 = Onstad-Foster modification {not implemented}
ER
2 C 1
Area of plot, ha
Area
0 R 8
Annual soil loss if ER = C, Mg ha'1
Loss
0 R 8
f See footnote, Table B-l.
@ N
HS
PO
FL
IC
ME
ER
Area
Loss
1
1
1
2
1
0
C
0.60
20.0
Figure B-8. Example item from the LANDSCAPE section.


166
greater for humified than for fresh organic material. Although the mineralization rate
factor needs to be calibrated for Andisols, standard data collection procedures do not
include a method for doing so (IBSNAT, 1989). This study used expected maize response
to applied N as a basis for calibrating SLNF. An approach is needed for adjusting
mineralization rates based on readily measured soil properties.
This study did not test the assumptions that NH4+ is retained by the exchange
complex and N03 moves freely with soil water. Since Andisols may have either net
negative or net positive charge depending on mineralogy, pH and soil amendments, both
of these assumptions are suspect. A more general model of ion movement might improve
predictions of N availability and leaching. Bowen et al. (1993) presented a modification
that accounts for N03' retention.
Adding P submodels, refining the soil N submodel and improving assumptions
about root physiology and response to soil properties are quite feasible. Reorganizing the
models along hierarchical boundaries is a more difficult task. Such restructuring would
allow the models to deal appropriately with farm-level resource allocation and would
facilitate linking the crop models with a two-dimensional hydrology and erosion model
such as WEPP. Previous experience demonstrated that reorganizing the models along
hierarchical boundaries is feasible and solves the problem of simulating sequences and
facilitates simulating interacting combinations of crops (Caldwell & Hansen, 1993).


165
conjunction with the WGEN weather generator appears to be useful for quantifying
weather-induced risk. However, comparison of multiple-year bean and maize trials with
simulation results suggests that there is substantial variability between years that the
models are not able to capture.
There appear to be several soil-related constraints to crop production in the study
area that the crop models do not address. These include low availability and high
buffering of soil P, nematode damage, impacts of soil erosion, and the threat of soil loss by
mass movements. Anticipated additions of soil and plant P submodels will enhance the
usefulness of the models on the Andisols of the Cauca region of Colombia.
The current source-driven root growth model is not adequate for simulating
response to soil loss. Its assumption that removing a favorable topsoil layer will force
deeper rooting is not consistent with what we know about root response to soil stresses.
A simple alternative model that reduces the absolute amount of root growth increased the
sensitivity of simulated bean yields to soil loss. More work is needed on the root growth
components of the crop models before they will realistically simulate crop response to
erosion.
Although the N component of the crop models has undergone extensive
development and testing, some of its assumptions do not hold well on Andisols.
Allophane retards mineralization of organic matter. This study confirms the need to
calibrate the N mineralization factor (SLNF). However, a single multiplier may not
adequately calibrate mineralization rate because the effect of allophane on mineralization is


250
Baily, D. von & J.W. Richardson. 1985. Analysis of selected marketing strategies: a
whole-farm simulation approach. Amer. J. Agr. Econ. 67:813-820.
Banco de la Repblica (various dates). Revista del Banco de la Repblica, various issues.
Bogot, Colombia.
Barlow, R.E., F. Proschan, and L.C. Hunter. 1965. Mathematical Theory of Reliability.
John Wiley & Sons, New York.
Batchelor, W.D., J.W. Jones, K.J. Boote and H.O. Pinnschmidt. 1993. Extending the use
of crop models to study pest damage. Trans. ASAE 36:551-558.
Batie, S.S. 1989. Sustainable development: challenges to the profession of agricultural
economics. Am. J. Agrie. Econ. 71:1083-1101.
Beus, C.E. and R.E. Dunlap. 1990. Conventional versus alternative agriculture: the
paradigmatic roots of the debate. Rural Sociology 55(4):590-616.
Beus, C.E. and R.E. Dunlap. 1991, Measuring adherence to alternative vs. conventional
agricultural paradigms: a proposed scale. Rural Sociology 56:432- 460.
Bidwell, O.W. 1986. Where do we stand on sustainable agriculture? J. Soil Water
Conserv. 41:317-320.
Borland International. 1992. Borland* Pascal with Objects. Borland International, Scotts
Valley, California.
Bowen, W.T., J.W. Jones, R.J. Carsky, and J.O. Quintana. 1993. Evaluation of the
nitrogen submodel of CERES-Maize following legume green manure
incorporation. Agron. J. 85:153-159.
Bowen, W.T., P.K. Thornton, and G. Hoogenboom. 1992. Crop sequencing using
IBSNAT crop models, p. 79. In Agronomy Abstracts. ASA, Madison, WI.
Box, G.E.P. and G.M. Jenkins. 1970. Time Series Analysis: Forecasting and Control.
Holden-Day, San Francisco, California.
Brklacich, M., C.R. Bryant, and B. Smit. 1991. Review and appraisal of the concept of
sustainable food production systems. Environ. Management 15( 1): 1 -14.
Budelman, A. and F. van der Pol. 1992. Farming system research and the quest for a
sustainable agriculture. Agroforestry Systems 19:187-206.


30
Finally, characterization of sustainability should be diagnostic. Sustainability is a
useful concept when its characterization focuses research and intervention by identifying
and prioritizing constraints. Diagnosis can be accomplished by testing hypotheses about
constraints based on a measure of sustainability that is both comprehensive and integrated.
Diagnosis is facilitated by use of a single measure of sustainability that combines the range
of possible determinants into a single, integrated measure of system response. An
integrated measure is necessary for comparing, for example, the relative impact of nitrate
leaching into aquifers and product price volatility on sustainability.
Weaknesses of the reviewed approaches for characterizing sustainability can be
related to their failure to incorporate the proposed elements (Table 2-6). Characterization
based on adherence to prescribed approaches fails because it is not founded on a literal
interpretation of sustainability. Lack of integration limits the usefulness of multiple
indicators of sustainability for diagnosing and prioritizing constraints. Integration of
indicators has been difficult because the underlying interpretation of sustainability as an
ability to meet diverse goals is not integrated. A time trend represents an integrated
system response that is potentially useful for diagnosis and can be predictive by
extrapolation, but it is not stochastic in the sense of accounting for variability. An
integrated measure of resilience has not yet been found. The assumptions about future
variability and disturbances necessary for resilience to be predictive and stochastic are
avoided in discussions of its use for characterizing sustainability. Simulated farm
survivability is the only approach reviewed that incorporates all of the elements listed.


53
the future behavior of system inputs, and hypothesizing constraints. The remainder of this
section examines these issues.
Selecting a time frame
Sustainability has meaning only in the context of a specific time frame. For
example, climate change, soil erosion, or extinction may be seen as irreversible threats to
the sustainability of an ecosystem in a time frame of decades or centuries, but as part of
natural cycles in a time frame of millennia or longer (Fresco and Kroonenberg, 1992).
From another perspective, sustainability can be viewed as a non-increasing function of
time. Considering the extreme cases, all existing agricultural systems can sustain
themselves for an arbitrarily short period. On the other hand, few agricultural systems can
be expected to continue in a recognizable form for tens of millennia. If F^t) is a true
probability distribution with a lower bound at t=0 then
lim
/- oo
Ftf(/) = 1
Although selecting a time period for analyzing farm sustainability is a subjective
decision, considerations of hierarchy, relevance, and realism suggest a range of about 10
to 15 years. Ecological hierarchy theory states that processes in higher-level systems
operate more slowly than in lower-level systems (Allen and Starr, 1988). The time frame
for analyzing sustainability of a farming system should therefore be longer than the several
months to a few years that are typical of crop and animal production cycles. Relevance


Probability Probability
1.0
0.8
0.6
0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
Figure 6-15. Distribution of liquid assets after 1 (a), 3 (b), 6 (c) and 9 years (d) for the source of risk scenarios.
3 6
Liquid assets, Col.$
(Millions)
Liquid assets, Col.$
(Millions)
200


152
Response to soil erosion
Simulated soil loss resulted in a small decrease in maize grain yield (Fig. 5-23a).
Simulated March-planted bean showed little response to soil loss, while October-planted
bean showed a slight increase in yield in response to soil loss (Fig. 5-23b).
Although experimental erosion response data from Andisols are not available,
there is reason to expect a substantial reduction of crop yields as the Ap horizon is eroded.
First, both quantities and availability of N and P are greater in the Ap than in the Bw
horizon (Table 5-2). Second, roots of existing vegetation are much more dense in the Ap
horizon. There is a sharp break in soil color, structure, and rooting between the Ap and
Bw horizons.
The lack of response to simulated erosion is a result of assumptions built into the
root growth model in the IBSNAT crop models. In these models, root growth is source-
driven, with adjustments for water and N stress. The soil data management program
(SDB3) used to build profile descriptions calculates root weighting factors for each soil
layer / (SRGF^ based on observed root density and modified by low pH or high A1
saturation. Aluminum toxicity as indicated by high A1 saturation inhibits root elongation.
Observed root abundance integrates the spatial partitioning of dry matter determined by
branching habit and meristem activity with a range of chemical and physical stresses that
reduce the quantity of roots grown. Hence, SRGF is determined by SDB3 partially as a
sink limitation. However, SRGF has no direct effect on total root growth in the crop
models. It determines only how growth is distributed within the soil profile. As a result,


216
Table B-2. Format of the OUTPUTS section of the scenario file.
Variable
Header
Format1
Farm failures output option
Y = create output file
N = do not create file
Fails
5 C 1
Monthly sustainability output option
Y = create output file
A = append to existing file
N = do not create file
Susta
5 C 1
Annual resource output option (Y or N)
Resou
5 C 1
Final wealth output option (Y, N, or A)
EndWl
5 C 1
Name of resource defining wealth (from RESOURCES
section)
WlthLink
1 C 15
Erosion output option (Y or N)
Eros
5 C 1
Event list output option (Y or N)
Evnt
5 C 1
Resource transaction list output option (Y or N)
Tms
5 C 1
T See footnote, Table B-l.
@Fail
Sust
Reso
EndW
WlthLink
Eros
Evnt
Trns
Y
A
Y
A
WEALTH
N
N
N
Figure B-2. Example OUTPUTS section.
ANALYSES section
The ANALYSES section controls graphical analysis at the end of a replicated
simulation run. It contains multiple items. Adjacent items with the same index will be
displayed in the same graph.
The example in Fig. B-3 specifies that sustainability time plots (S) for three
scenarios (CCJD01, CCJD02, and CCJD03) should be both (B) displayed as one


103
The versatility and realism of FSS is currently limited by its inability to simulate
production decisions, lack of feedback between within-season resource constraints and
crop production, an absence of a model of livestock production, and the over-simplicity of
the model of household consumption decisions. In spite of its limitations, FSS is a useful
tool for characterizing the sustainability of a farm operating under fixed management
scenarios. Its object-oriented design and data structures provide a flexible foundation for
enhancements that may correct some of its existing weaknesses.


135
Table 5-10. Observed and predicted cassava yields, Domingo and
Trujillo farms, Cauca, Colombia, 1995,
Plants
Root yield (kg ha1)
Farm
per ha
fresh
dry
obs.
pred.
Domingo
11,806
24,549
8,890
20,278
Trujillo
14,577
64,757
19,943
23,362
Figure 5-11. Simulated and observed maize grain yields, Domingo and Trujillo
farms, Cauca, Colombia, planted October 1993, March 1994 and October 1994.


75
Resources. The RESOURCES section specifies the particular resources that
constitute a farming systems state variables, and determines their initial characteristics and
supply. A hierarchy of nine resource classes is available in FSS to represent the various
types of farm resources (Table 4-2, Fig. 4-3). Two additional abstract ancestral classes
are used internally as templates for the other classes.
Resource
resource resource resource resource
Seasonal Capital Activity-linked
resource resource credit resource
Machine
resource
Figure 4-3. Tree of resource class hierarchy in FSS.


16
(Rodale, 1990; Bidwell, 1986), most of the practices it promotes as sustainable are largely
products of mainstream research and educational institutions (Francis and Sahs, 1988;
York, 1988).
Third, establishing the contribution of an approach to sustainability through
definition eliminates the perceived need to evaluate approaches that may be poor or
harmful in a particular context. If strategies are identified as sustainable based on their
effect on agricultural systems, and agricultural systems are then judged to be sustainable
based on their implementation of sustainable strategies, then a form of circular logic
results. It is logically impossible to evaluate the contribution of an approach to
sustainability when adherence to that approach has already been used as a criterion for
evaluating sustainability. This circular logic is a fourth reason why interpreting
sustainability as an approach is not useful for guiding change.
Because of the temporal nature of sustainability, errors of either ignoring
approaches that enhance sustainability or promoting approaches that threaten it may not
be obvious when the approaches are implemented. The evaluation needed to recognize
errors and improve approaches is not possible if sustainability is interpreted as a
philosophy or a set of strategies. Thompson (1992) warned that
Our society may collapse because of shortsighted stupidity on the part of
pro-growth, resource exploiting power elites, but the collapse will only be
tragic if it is shortsightedness or ignorance on the part of environmentally
and ethically concerned people that helps bring it about, (p. 19)


91
The Sell method (Fig. 4-6) disposes of an amount of a consumable resource by
exchanging with the first linked resource found in the list of variable costs (see next
section). For example, Sell may reduce the supply of a harvested product and increase
the supply of the operating fund. In resource classes other than consumable, Sell is a
dummy method that does nothing.
Algorithm
Set Amount = max(Amount, -Supply).
Set Sold - Amount.
If Sold > 0 then
request VariableCosts to execute Use(Sold).
Set Supply = Supply + Amount.
Purpose
Limit amount sold to amount available.
Positive Sold adds to linked resources.
Give Sold to the first linked resource.
Reduce Supply. (Amount should be negative.)
Figure 4-6. Pseudocode representation of the Sell method of the consumable resource
class.
The behavior of the Use method of timed resources and their descendants
(seasonal, capital and machine) is quite different from that of consumable resources.
Instead of consuming supply with use, each timed resource keeps a list of the amount of
time reserved each day. Use by a particular operation reduces the number of hours
available for use by subsequent operations on the same day (Fig. 4-7).
The USE method of credit and activity-linked credit is complicated by the need to
maintain a record of multiple loans from the single source (Fig. 4-8). If Amount is
negative, indicating borrowing, USE inserts a new loan for the appropriate amount into the
collection of loans or, if a previous loan was taken since the last scheduled payment date,
the new amount is added to the balance due on that loan. A positive Amount passed to
Use indicates an early, unscheduled loan repayment.


209
rotation and intensifying production. Agronomic trials combined with market research
might help identify promising high-valued vegetable crops. The severity and impacts of
soil erosion in the volcanic soils of the Cauca region represent an important knowledge
gap; we do not know how much erosion threatens farm sustainability. Some constraints
to sustainability are not under the farmer's control, and must be addressed through public
policy. Simulation results suggested that farm risk is controlled much more by prices than
by weather variability. Policies that reduce price volatility can be expected to enhance
farm sustainability.


254
Hamblin, A. 1992. How do we know when agricultural systems are sustainable? p. 90. In
A. Hamblin (ed.) Environmental Indicators for Sustainable Agriculture. Report on
a national workshop, November 28-29, 1991. Bureau of Rural Resources, Land
and Water Resource Research and Development Corporation, Grains Research
Corporation, Canberra, Australia.
Hansen, J.W., N.B. Pickering, J.W. Jones, C. Wells, H. Chan, and D.C. Godwin. 1994.
WeatherMan. p. 137-199. In G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT
Version 3, Vol. 3. International Benchmark Sites Network for Agrotechnology
Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Hansen, J.W., P.K. Thornton and J.W. Jones. 1995. Toward data standards for enterprise
and farm level analyses. Paper presented at the 2nd International Symposium on
Systems Approaches for Agricultural Development, 6-8 December, 1995,
International Rice Research Institute, Los Baos, The Philippines.
Harrington, L.W. 1992. Measuring sustainability: issues and alternatives. J. Farming
Systems Research Extension 3:1-20.
Harrington, L.W., P. Hobbs, T. Pokhrel, B. Sharma, S. Fujisaka, and S. Lightfoot. (1990)
The rice-wheat pattern in the Nepal Terai: issues in the identification and definition
of sustainability problems. J. Farming Systems Research Extension 1 (2): 1 -27.
Hart, R.D. 1986. Ecological framework for multiple cropping research, p. 40-56. In C.A.
Francis (ed.) Multiple Cropping Systems. Macmillian Publishing Co., New York.
Harwood, R.R. 1990. A history of sustainable agriculture, p. 3-19. In C.A. Edwards, R.
Lai, P. Madden, R.H. Miller, and G. House (ed.) Sustainable Agricultural Systems.
Soil and Water Conservation Society, Ankeny, Iowa.
Hauptli, H., D. Katz, B.R. Thomas, and R.M. Goodman. 1990. Biotechnology and crop
breeding for sustainable agriculture, p. 141-156. In C.A. Edwards, R. Lai, P.
Madden, R.H. Miller, and G. House (ed.) Sustainable Agricultural Systems. Soil
and Water Conservation Society, Ankeny, Iowa.
Held, L.J. and G.A. Helmers. 1981. Growth and survival in wheat farming: the impact of
land expansion and borrowing restraints. Western J. Agrie. Econ. 6:207-216.
Hernandez, M., L.J. Lane, and J.J. Stone. 1989. Surface runoff, p. 5.1-5.18. In L.J. Lane
and M.A. Nearing (ed.) USDA Water Erosion Prediction Project: Hillslope
Profile Model Documentation. NSERL Report No. 2. USDA-ARS National Soil
Erosion Laboratory, West Lafayette, Indiana.


205
models in this environment. Evidence (Chapter 5) suggests that the models underestimate
year-to-year variability, generally over-predict yields, and do not consider some important
yield-reducing stresses that are important in this environment. Further testing is certainly
needed. Third, FSS does not account for within-season resource constraints. Finally, FSS
does not simulate adaptive management of farming enterprises, but rather imposes fixed
management. An actual farmer would employ any of a number of strategies to control
downside risk under a threat of failure.


123
76.8C
76.6
76.4c
76.2
3.0
Santander de Quilichau(990m, P,T,S,R)
Japio (1015 m, P,T)
2.8
Mondomo (136Q m, P)
2.6'
DOMINGO FARM (1650 m)
Piendamo (184Q m, P)
2.4
Venta de Cajibio (1800
m, P.T.S)
La Florida, Popayan (1850 m, P,T,S)
Figure 5-8. Locations and elevations of weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation. Letters represent weather variables recorded (P = precipitation, T =
temperature, S = hours of bright sunshine and R = solar radiation).


Table 5-3. Spatially interpolated WGEN coefficients for the Domingo farm, Cauca, Colombia.
Month Solar radiation (MJ m 2 d1) Temperature (C) Rainfall
-- dry days
x s
-- wet days
x s
maxiii
dry days --
X s
iium
wet days --
x s
- minimum -
-- all days --
x s
at
Total
P{DW}}
No.
wet
days
Jan
18.5
4.0
15.5
4.1
26.1
1.7
24.7
1.9
14.5
1.2
0.875
178.5
0.293
13.2
Feb
19.1
4.0
15.5
4.2
26.2
1.8
24.7
1.9
14.7
1.2
0.888
181.5
0.294
13.1
Mar
18.7
4.6
15.2
4.7
26.1
1.6
25.0
2.1
14.8
1.2
0.780
220.3
0.337
15.1
Apr
17.7
4.5
14.3
4.0
25.8
1.5
24.8
1.8
14.9
1.0
0.898
221.5
0.416
16.1
May
17.0
4.4
14.0
3.8
25.8
1.5
24.5
1.8
14.9
1.1
0.740
183.1
0.363
16.0
Jun
17.8
3.6
14.3
3.5
26.0
1.5
24.4
1.7
14.4
1.1
0.803
93.3
0.238
11.2
Jul
18.1
3.7
15.1
3.1
26.2
1.6
24.9
1.6
13.8
1.4
0.857
68.8
0.161
7.8
Aug
18.3
4.0
15.1
3.8
27.0
1.7
24.9
1.9
13.9
1.3
0.743
79.8
0.157
7.6
Sep
18.1
4.2
16.2
4.1
26.5
1.9
25.1
1.9
14.2
1.1
0.935
119.2
0.259
11.1
Oct
16.9
4.4
14.9
4.1
25.4
1.7
24.7
1.8
14.5
1.2
0.865
245.9
0.454
18.0
Nov
17.1
4.1
14.4
3.9
25.4
1.5
24.5
1.8
14.8
1.3
0.872
274.5
0.449
17.9
Dec
17.6
3.9
14.8
3.8
25.6
1.4
24.7
1.7
14.7
1.2
0.951
189.7
0.342
14.2
r Alpha parameter of the gamma distribution.
* Probability that a day is wet and was preceeded by a dry day.


67
Overview of the Farming System Simulator
The Farming System Simulator (FSS) is an object-oriented, dynamic, stochastic,
discrete-event farm simulator. It is object-oriented in its design, and is implemented using
the object-oriented extensions of Borland Pascal (Borland International, 1992). The
description of FSS that follows cannot be fully understood without some familiarity with
object-oriented programming concepts (Appendix A). FSS is dynamic; it is capable of
simulating the operation of a farming system through many years. It is stochastic in the
sense that it is designed to simulate many replicates of a farm scenario with stochastic
inputs of weather and price data, and to present analyses based on the resulting
distributions. FSS runs external crop simulation models to simulate continuous
physiological and ecosystem processes. However, all farm-level processes (i.e.,
operations, production, management and consumption of resources, and failure) occur in
response to discrete events. A final characteristic of FSS is that it is primarily a resource
accounting model.
The object-oriented structure of FSS is based on a conceptual model of the
structure and function of a farming system. A farming system integrates ecological,
economic and social components (Fig 4-1). The ecological component of a farming
systema set of agricultural ecosystems, or agroecosystemsconsists of biotic
communities and the landscape that they inhabit, and can be delineated by field boundaries.
The economic component comprises the set of resources that are under the control of the


Canopy biomass, kg/ha Grain yield, kg/ha
Irrigated
Rainfed
Figure 5-18. Grain and biomass yields of irrigated (a and b) and rainfed (c and d) soybean (cv. Bragg) simulated
by the original and modified versions of CROPGRO, Gainesville, Florida, 1978.


154
removing the topsoil that is favorable to root growth forces more root growth lower in the
profile (Fig. 5-24). As soil is lost, the effect of a decreasing total supply of N is offset by
deeper rooting and better access to water and N lower in the soil. I modified the root
growth model in CROPGRO and CERES to increase sensitivity to soil loss. The modified
model attempts to divide the process of distributing new root growth into two
components: (a) partitioning among soil layers based on branching habit, and (b) reduction
of root elongation based on stresses encountered in each soil layer. The approach is to
compare the vertical distribution of new root growth specified by SRGF with a
distribution that would be expected in an ideal, homogeneous soil that imposes no
restrictions on root growth. A distribution factor, F, represents the vertical partitioning of
new root in this ideal soil. Jones et al. (1991a) used a function of the form,
F. = (1 r./300)*,
to represent the distribution of roots based on growth habit in the absence of soil stresses,
where i is soil layer number, z¡ is the depth of the middle of layer / (cm), and g is an
exponent that varies with plant species. The modified models use g = 2, which Jones et al.
recommended for maize and soybean.
The first step was to normalize SRGF for each layer / in the description of the
uneroded soil profile such that the relative weighting of the layers remains the same, but
SRGF, = 1:
SRGF. = SRGF. / SRGF,.


95
their ability to dispose of their supply. Priceab represents the number of units of resource b
that must be used for each unit of resource a that is replenished.
Algorithm
Purpose
Repeat
For each linked resource in the list...
Amount > 0 then
Positive Amount indicates obtaining.
set P = BuyMult Link.Price.Current.
Price P is todays price for the resource linkage,
adjusted with BuyMult.
otherwise
Negative Amount indicates disposing.
set P = SellMult Link.Price.CURRENT.
Adjust todays price with SellMult.
Set ,4 = Amount P.
Convert Amount to units of the linked resource.
Request Link to execute Use(j4).
Obtain from or dispose to the linked resource.
If A = 0 then
If linked resource exchanged all oL4 ...
set Amount = 0
Nothing remains to be exchanged.
otherwise
set Amount = Amount -A/P.
Determine Amount that still needs to be exchanged.
until Amount = 0.
Stop when full amount has been exchanged.
Figure 4-12. Pseudocode representation of the Use method of the cost list class.
Illustration: A fertilizer application operation
A fertilizer application operation illustrates the flow of information and sequence
of events that lead from the simulation of an operation by an external crop model to a
change in the operating fund (Fig. 4-14). In this fictitious example, a crop activity calls a
crop model, then uses information returned in the operations output file to insert a
fertilizer application operation into the event queue. After the remaining crop and fallow
activities have been simulated and scheduled operations have been inserted, the event
queue is executed. When the event queue executes the fertilizer application operation, the
operation searches the collection of operation requirements. The matching operation
requirement indicates that a fertilizer application using this particular method on this


127
Development and yield
In 1993, the Hillsides Program of CIAT began conducting trials on several farms in
the Cabuyal area, including the Domingo farm. These trials provide a basis for evaluating
development and yield predictions of the models for bean, maize and cassava. To date,
data are available from four bean harvests, three maize harvests, and one cassava harvest.
The maize and cassava trials each consisted of a single treatment with large inputs of soil
amendments (10-30-10, lime and chicken manure). Two bean trials were conducted. The
first trial consisted of a single fertility treatment. The second bean trial included several
levels of chemical fertilizer and manure. A split plot design was imposed during the
second season; half of each plot received additional amendments while the other half
benefitted only from any residual effect of previous amendments.
Results from the Trujillo farm are included here because they may better represent
attainable, water-limited yields than results of the trials on the Domingo farm. Yields of
all crops were higher on the Trujillo farm than on the other farms studied. Root knot
nematode damage was observed on the roots of beans at the Domingo farm during the
second (March 1994) growing season. Nematode damage was not observed on the
Trujillo farm.
Simulations were run for each treatment of each on-farm trial. The simulations
used observed plant densities, reported planting dates and fertilizer inputs, and measured
weather data. Except for cassava, all of the trials included more than one season and were
therefore simulated using the sequential driver in DSSAT3 (Thornton et al., 1994).


86
necessary because current IBSNAT crop model output files (Jones et al., 1994) do not
provide a complete list of operations or their timing. Furthermore, state events triggered
by field conditions (eg., automatic irrigation) cannot be inferred from information in the
crop management file.
Event handling
FSS is a discrete-event simulator. The state of the farming system, represented by
the supply of each of its resources, changes only in response to events. An event queue
accumulates, sorts, executes and monitors the status of events. FSS recognizes three
classes of events: field operations, schedided operations and monthly update events.
Field operations are returned by external production process models. Monthly update
events are inserted at the midpoint of each calendar month of the current simulation year,
and are responsible for fixed cost accounting, household consumption and for testing
criteria for farming system failure. Scheduled operations are described in the previous
section. Each time an external model simulates a production activity, resulting field
operations are inserted into the event queue. After the entire landscape is simulated for
the current year, any scheduled operations are added to the queue. Finally, twelve
monthly update events are inserted. Events are stored in order of date, then priority. This
gives high priority events first access to scarce resources. Monthly update events have
priority over all operations.


218
resources (i.e., variable costs) when use causes supply to fall below the minimum, and sells
to them if supply exceeds a maximum. If a linked resource is specified as a fixed cost,
expenses of ownership are charged to those costs each month. Most resource types
search the list of fixed or variable cost linkages to meet current requirements. Surpluses
are disposed in reverse order. Prices represent the amount of linked resource that can be
exchanged for a unit of the current resource. Price names must match time-series models
specified in the price file.
All material inputs and outputs returned by the crop models must be specified as
consumable resources. Pesticides, fertilizers, and organic amendments use the resource
names in Jones, et al. (1994, Appendix B). Planting material names consist of the planting
material code, crop code, and cultivar ID (eg., PM001 ML IB0063)- Similarly, harvest
product names use the harvest code, crop code, and cultivar ID (eg., H MZ IB0063).
Irrigation water should always use the name WATER.
The example in Fig. B-4 specifies a consumable resource with the name
OPERATING FUND and a value determined by the price, Unity. When the
OPERATING FUND falls below the minimum of 10,000, it will attempt to draw on any
variable cost linkages.
eci
Units
Name
Val
ValMult Supply Minimum Maximum IntRate . .
c
ColS
OPERATING FUND
Unity
1.00 S.0e6 10000 -90.0 0.026
Figure B-4. Example item from the RESOURCES section.


142
value indicates that N is released by mineralization at 35% of the rate simulated for a
comparable soil with crystalline mineralogy. Figure 5-16 shows the simulated response of
maize and bean to applied N with the reduced mineralization factor.
. Bean response to applied nitrogen. With the N mineralization calibration, bean
simulated with CROPGRO showed a dramatic and unexpected decrease in simulated
yields in response to small amounts of applied N (Figs. 5-17a and c). Simulated response
to applied N was more erratic for soybean grown on a sandy soil in Gainesville, Florida
(Fig. 5-18) than for bean at the study site.
Fig. 5-15. Effect ofN mineralization factor, SLNF, on simulated maize response to
applied N, Domingo farm, Cauca, Colombia.


223
Table B-6. Format of the OPERATION REQUIREMENTS section of the scenario file.
Variable
Header
Formal
Index of operation/crop/method combination
N
013
Operation type code
PLNT = planting
IRRT = irrigation
FERT = fertilizer application:
RESD = organic material application
CHEM = pesticide application
TILL = tillage
HARV = harvest
MRKT = marketing
Type
1 C 4
Crop code
CC
1 C 2
Method code
Methd
1 C5
Priority of operation
Pri
1 13
Time window for operation, days
Win
1 13
* See footnote, Table B-l.
The remaining columns are repeating blocks which specify the requirements for a
single timed resource.
Table B-7, A resource block within the OPERATION REQUIREMENTS section.
Variable
Header
Formal
Name of the /th resource (from RESOURCES section)
Resource/
1 C 15
Basis for specifying an operation time requirement:
E = per event
A = per unit area, Ha'1
W = per unit mass of material, Mg'1
BH
1 C 1
Hours of resource required
Hours
0 R 8
See footnote, Table B-l.


2
agriculture is to meet the needs of society as a wholeproducing food and other products
while protecting natural resourcesit must first meet the needs of the farmers who
implement and manage it. The farm level is also appropriate from a practical standpoint.
Since human goals are not intrinsic to fields or enterprises, it is difficult to identity what is
to be sustained. On the other hand, characterization of sustainability at system levels
higher than the farm is complicated by the complexity of the systems, the difficulty of
specifying system boundaries, and the often conflicting goals of multiple human actors.
The purpose of this study is to present and demonstrate a simulation-based systems
approach to characterizing sustainability of farming systems. Specific objectives of this
dissertation are as follows:
1. Review existing conceptual and methodological barriers to using the concept of
sustainability for guiding change in agriculture
2. Propose a set of elements necessary for an approach to characterizing
sustainability to provide a useful basis for guiding agriculture.
3. Present the logical and mathematical basis for a definition of sustainability that
applies generally to dynamic, hierarchical, stochastic, purposeful systems, and
relate this definition to farming systems.
4. Describe and illustrate a framework for using system simulation to quantify
sustainability and test hypotheses about its determinants.
5. Describe an object-oriented farming system simulator that was developed for the
purpose of characterizing farm sustainability, and present its data requirements.
6. Evaluate the compatibility of a set of crop models with requirements imposed by
a simulation study of sustainability in a hillside environment in Colombia.
7. Demonstrate and evaluate the approach for characterizing farm sustainability by
identifying determinants of the sustainability of a hillside farm in Colombia.


180
Scenarios
The framework for characterizing sustainability presented in Chapter 3 calls for the
use of sensitivity analysis to test hypothesized determinants of farm sustainability. Farm
scenarios (Table 6-7) were used to examine the role of cropping system, coffee yields,
nitrogen dynamics, soil erosion, prices, household consumption, resource endowment,
labor source and credit as determinants of sustainability. A base scenario served as a
basis for comparing each alternative scenario. Each scenario was simulated for the 15
year period beginning in September 1994, except as noted.
Base scenario. The base scenario incorporates the information and assumptions
presented in the previous sections. The cropping system in the base scenario is a three-
year rotation of maize, beans and cassava (Fig. 6-2). I attempted to design a rotation that
is diversified, avoids periods of excessive labor demand, avoids the problem of nematode
buildup in successive bean crops, and fits within the rainfall pattern. Tomato was not
included because it is somewhat speculative; few farmers in the region have had
experience with tomato. To maximize spatial diversification, the three phases of the
rotation were distributed equally among different fields. The farm scenario file in
Appendix D details the assumptions in the base scenario.
Cropping systems. Of the crops commonly grown in the Cauca region, simulation
models are available only for maize, bean, cassava and tomato. Scenarios incorporate
several sequences of these crops (Fig. 6-2) that were designed to be feasible based on
rainfall distribution (Fig. 5-3), distribution of labor requirements, and time required for


APPENDIX D
INPUT FILES USED FOR FARM SIMULATIONS
This appendix lists relevant portions of the base farm scenario file (Fig. D-l), the
crop management file (Fig. D-2) and soil profile file (Fig. D-3) used as input to the farm
simulations presented in Chapter 6. The scenario file format is documented in Appendix
B. Formats of the crop management and soil input files are documented in Jones et al.
(1994).
238


82
Random number sequences
When simulating a farm scenario, a user specifies the duration of the scenario and
the number of times the scenario is replicated. For consistency with the sequence analysis
programs in DSSAT3, FSS increments the initial random number seed by five at the
beginning of each replicate (Thornton et al., 1994). This permits the same set of
stochastic weather and prices to be used with different scenarios without pseudorandom
number sequences being affected by different timing of farm failures among different
scenarios.
Price generation
FSS implements three classes of prices: constant prices, historical prices and
stochastic prices simulated using an autoregressive-moving average (ARMA) model.
Constant prices always return a single value read from the price file.
The model used to generate an ARMA price sequence for a single commodity
includes a deterministic trend and seasonal component, and a stochastic component. The
deterministic component is,
*t = a + P' + Pm> [4-1]
where xt is the expected value at t months after a specified base month, a and P are the
slope and intercept of a linear trend, and pm is the mean deviation from the trend for
calendar month m. The ARMA price model can include a multiplicative seasonal moving


48
Sensitivity analysis
Sensitivity analysis is used to quantify the relative importance of hypothesized
constraints to sustainability. It involves changing the value of a factor a small amount in
the direction that would relax the hypothesized constraint relative to a base scenario which
represents existing or expected conditions, then simulating the modified scenario.
Hypothesized constraints can then be ranked based on either absolute,
[3-14]
or relative sensitivity,
r.
Y
i,0
S.-Sn
Y.-Y:
i.O I
[3-15]
where Yjfi is the value of the ;'th factor in the base scenario, Y¡ is its adjusted value, and S0
and £¡ are sustainability values estimated for the base and alternate scenarios. The
absolute value allows an increase in sustainability to result in r¡ > 0 regardless of the
direction of change in Yv Relative sensitivity is interpreted as the percent change in S in
response to a 1% change in Y. Comparisons may be made and ranks assigned among
discrete or among continuous factors. However, absolute sensitivity to discrete factors
cannot be compared with relative sensitivity to continuous factors.


161
IBSNAT crop models along hierarchical boundaries so that the ecosystem could run
continuously, and crop populations could be inserted or removed from the ecosystem at
any time. A farm-level implementation would need the ability to insert a number of
agroecosystems, and simulate all of them on a daily time step as in Fig. 5-27. This could
be accomplished most easily with an object-oriented design and implementation.
Issues in linking crop and erosion models
The IBSNAT models do not simulate soil loss. The most widely used soil erosion
prediction model is the universal soil loss equation (USLE, Wischmeier & Smith, 1978).
The USLE is a simple empirical model that aggregates the processes involved in soil
erosion. A single rainfall erosivity term based on energy and intensity aggregates the
effects of both rainfall and runoff. A single soil erodibility term aggregates susceptibility
to detachment by raindrop impact, detachment by rill streamflow, and transport by
streamflow. There is reason to expect the USLE to be a poor predictor of loss of
Andisols in steep, complex slopes, where resistance to detachment may be moderate to
poor, but runoff and therefore sediment transport is rare because of the very high
infiltration capacities of the soils.
The Water Erosion Prediction Project (WEPP) model (Lane & Nearing, 1989)
simulates the processes of interrill and rill detachment, transport and deposition, as well as
the hydrological processes that drive water erosion. Although it has not yet been tested
for steep, volcanic ash-derived soils, the abilities of WEPP to predict runoff and to
separate raindrop- and runoff-induced erosion are reason for optimism. WEPP


100
(c) the dates of all farm failures for the current scenario (FAILURES.OUT), (d) monthly
sustainability of a set of scenarios (SUSTAIN.OUT), and (e) the final status (0 = failed, 1
= continuing) of each replicate for a set of scenarios (STATUS.PRN). FSS uses
RESOURCE.OUT to create a table of percentiles for creating a resource box plot.
ENDWLTH.OUT is the basis for plotting the final distribution of a resource.
Cumulative Probability of Final Health
Ualue
Figure 4-16. Example of a final resource distribution plot generated by FSS.


140
Figure 5-13. Simulated grain and biomass yield response of maize (a, b) and October-
(c, d) and March-planted (e, f) bean to applied N using the default mineralization factor
(SLNF = 1.00), Domingo farm, Cauca, Colombia. Mean SD of 10 replicates.


102
to process-level biophysical models through an operations output file allows FSS to
respond to ecological determinants of sustainability. Its object-oriented design and input
data structures give FSS considerable flexibility for representing a range of farming system
types. The input data structures defined for FSS are sufficiently flexible to serve as a
starting point for developing data standards for farm-level systems analyses (Hansen,
1995).
Sustainability Tima Plot
95142 98046 316 3220 6124 9028 11298
Data
Figure 4-17. Example of a sustainability time plot generated by FSS.


74
scenario and the number of times it is replicated. The next two sectionsOUTPUTS and
ANALYSEScontrol file and graphical output. These are discussed later in this chapter.
Table 4-1. Sections in the farm scenario file.
Section
Purpose
SCENARIO
Specifies title, units, other files, and simulation control.
OUTPUTS
Controls farm-level file output.
ANALYSES
Controls graphic display and file output.
RESOURCES
Identifies farm resources.
LINKAGES
Defines interrelationships between resources.
OPERATION
REQUIREMENTS
Specifies types of operations, priorities, a time window, and
timed resources required.
SCHEDULED
OPERATIONS
Schedules operations which are not included in the experiment
file or returned by crop models.
LANDSCAPE
Identifies homogeneous fields and their positions and
characteristics.
LIVESTOCK
Not yet developed. When developed, will specify herd and
management information for grazing livestock.
STRATEGIES
Specifies sequences of management activities.
ENTERPRISES
Links management (i.e., strategies) with landscape (i.e.,
subfields).
PRODUCTION
DECISIONS
Not yet developed. When developed, will specify a model and
criteria for selecting strategies for each enterprise.
CONSUMPTION
DECISIONS
Specifies household subsistence and discretionary consumption
of resources.
The remaining sections of the scenario file define both the structure and the initial
conditions of a model of a particular farm.


156
The values of F were normalized in the same manner. The values of F and SRGF for the
top layer are thus set equal to each other (F, = SRGF¡ = 1.0) based on the assumption that
an uneroded topsoil does not inhibit root growth. We can then assume that if for a lower
layer i, SRGF¡ < F¡, the difference between SRGF¡ and F¡, represents some source of stress
relative to the ideal soil. This stress would reduce new root growth in layer i without
redistributing it to other soil layers.
The root models require two modifications. First, F replaces SRGF for
partitioning new growth among soil layers. Second, where SRGF < F, root growth is
reduced according to the ratio of the two distribution factors:
ARLD{ / A/ = min(S7?GF / F¡, 1) potential growth respiration,
where RLDt is root length density in layer i (cm root cm'3 soil), potential growth refers to
root growth apart from soil stress, and the units of potential growth and respiration are
cm cm'3 day'1.
Figure 5-25 illustrates how the modification increases sensitivity to simulated soil
loss on the Domingo farm. The shaded region between SRGF and F in Fig. 5-25a shows
the layers in which stress occurs in the uneroded soil. When the soil profile is truncated to
simulate erosion, the distribution of new root growth for an ideal soil (F) remains the
same, but the soil layers with reduced relative root growth indicated by low values of
SRGF become shallower. The area of stress between SRGF and F increases, particularly
in the upper layers where roots are concentrated (Fig. 5-25b and c). Reduced root growth
feeds back to shoot growth and grain yield by reduced ability to access water and N.


117
al., 1994) were selected for use in this study. The specific crop models used were
CROPGRO v.3.0 for bean and tomato, Generic-CERES v. 3.1 for maize, and CropSim
CASSAVA v. 1.0 for cassava. These models were selected because (a) they are sensitive
to weather and to soil nitrogen dynamics, (b) they conform to a common input and output
data standard (Jones et al., 1994), (c) they can run in sequence using a carryover file to
initialize ecosystem state variables based on their final values from a previous run (Bowen
et al., 1992), (d) source code, supporting software and technical information are readily
available for these models, and (e) they have been tested under a wider range of conditions
than most other crop models. The incorporation of an adaptation of the WGEN weather
generator (Richardson, 1985) into the IBSNAT models simplifies the process of sampling
stochastic weather sequences.
Capabilities and limitations. The DSSAT3 crop models simulate photosynthesis,
respiration, partitioning and development in response to daily weather inputs. They
simulate the soil water balance, evaporative demand and crop water stress response. They
also simulate soil N and organic C dynamics and, except for CropSim CASSAVA, plant N
status and stress response.
Some of the known limitations of the crop models are important for interpreting
results of this study. Nitrogen is the only soil nutrient that the crop models can simulate.
Versions that account for soil and plant P are undergoing development and testing (Singh
& Godwin, 1990; Bowen, W., Personal communication), but are not yet available for use.
Although coupling points and input file formats have been defined that allow the bean and


225
Table B-8. Format of the SCHEDULED OPERATIONS section of the scenario file.
Variable
Header
Format1
Event schedule index
N
013
Event class: 0 = field operation
Cl
2 C 1
Operation type code
Type
1 C 4
Crop code
CC
1 C 2
Method code
Methd
1 C 5
Basis for determining the event date:
A = absolute date (Yr, DOY)
R = relative date (Yr is incremented each simulation
year.)
L = days of lag following a specified operation
BT
2 C 1
Index of preceding operation from OPERATIONS section
if BT = L
LOp
113
Date of operation (interpreted according to value of BT)
Time
1 I 5
Name of consumable resource used
Resource
1 C 15
Basis for determining amount of material produced by event:
E = absolute amount per event
A = amount per ha land
0 = fraction of amount used in preceding operation
S = fraction of supply of consumable resource
R = amount to remain in supply of consumable resource
BA
2 C 1
Amount of consumable resource used (interpreted according
to the value of BA)
Amount
0 R 8
* See footnote, Table B-l.
LANDSCAPE section
The LANDSCAPE section identifies homogeneous units of land, the hillslope of
which they are a part, and their position within the hillslope. The use of FL, IC, and ME
are identical to their use in the TREATMENTS section of the crop management file


58
A similar issue arises when extending an analysis from a single farm to a group of
similar farms. A group of farmers could adversely affect common grazing land, water
resources, fuelwood, fishing areas, or forest by overuse, whereas a single farmers impact
might be negligible. Similarly, although an individual farmer is usually assumed to be a
price taker, a group of farmers may alter prices due to their aggregate effect on supply of
a product or demand for an input such as seasonal labor. Incorporating these effects
involves extending the analysis to include some processes (eg., market equilibria) above
the farming system level.
An Example: Sustainability of a Coastal Texas Rice Farm
Whole-farm simulation studies have examined the influence of factors such as
commodity price variability (Grant et al, 1984), farm size and beginning equity level
(Richardson and Condra, 1981), intergenerational estate transfer strategy (Walker et al.,
1979), and land tenure expansion strategy (Held and Helmers, 1981) on probability of
farm survival through various periods. Perry et al. (1986) conducted a more
comprehensive simulation study that examined the impacts of crop rotation, land tenure
arrangement, government programs, costs, labor availability, lenders policies, interest
rates, and the level and variability of crop yields and prices on rice farms in Texas. A
reinterpretation of the results of this study illustrates the use of long-term, stochastic
simulation to characterize farm sustainability.


Canopy biomass, kg/ha Grain yield, kg/ha
October planting
March planting
3000 '
4500
3500
2500
Figure 5-17. Grain and biomass yields of October- (a and b) and March-planted (c and d) bean simulated by the
original and modified versions of CROPGRO, Domingo farm, Cauca, Colombia.
4*.


236
FSS provides an industry-standard user interface. Screen items may be accessed
with a mouse, cursor keys (for menu items), the key (for dialog box items), or
-letter combinations for items with a highlighted letter. The main menu can be
activated from the keyboard by pressing . The key allows a user to back out
of any interface item. Dialog boxes contain a Cancel button that serves the same
function.
The main menu has four items: File, Run, Analyze, and Quit. Table C-l
summarizes the options available in each. The File menu provides features for accessing
the operating system, editing files, and exiting FSS.
Table C-l. Options available from FSS menu items.
Menu item
Explanation
File
Provides features for editing files, accessing the operating system,
displaying information about FSS, and exiting.
Edit
Accesses a text editor (TVED) so that data and output files can be
viewed and edited.
Change dir.
Opens a dialog box to change the current directory.
DOS shell
Leaves FSS temporary to access the operating system. Type EXIT to
return to FSS.
About
Displays information about FSS.
Exit
Exits the FSS program.
Run
Launches farm simulations.
Single scenario
Launches simulation of a single farm scenario.
Multiple scenarios
Allows a user to build a list of farm scenarios, then launches simulation
of the set of listed scenarios.
Analyze
Currently only offers display of graphical simulation results.
Graph
Lists available graphs and calls WMGRAF to display selected graphs.
Quit
Exits the FSS program.


262
Soil Survey Staff. 1990. Keys to Soil Taxonomy, Fourth Edition. SMSS Technical
Monograph No. 6. Soil Management Support Service, Blacksburg, Virginia.
Sokal, R.R. and F.J. Rohlf. 1981. Biometry, Second Edition. W.H. Freedman and
Company, New York.
Spencer, D.S.C. and M.J. Swift. 1992. Sustainable agriculture: definition and
measurement, p. 15-24. In K. Mulongoy, M. Gueye, and D.S.C. Spencer (ed.)
Biological Nitrogen Fixation and Sustainability of Tropical Agriculture.
Wiley-Sayce Co., West Sussex, UK.
Stinner, B.R. and G.J. House. 1987. Role of ecology in lower-input, sustainable
agriculture: an introduction. Am. J. Alternative Agrie. 2:146- 147.
Stinner, B.R. and G.J. House. 1989. The search for sustainable agroecosystems. J. Soil
Water Conserv. 44:111-116.
Stockle, C.O., R.I. Papendick, K.E. Saxton, G.S. Campbell, and F.K. van Evert. 1994. A
framework for evaluating the sustainability of agricultural production systems. Am.
J. Alternative Agrie. 9(l&2):45-50.
Taylor, D.C., Z.A. Mohammed, M.A. Shamsudin, M.G. Mohayidin, and E.F.C. Chiew.
1993. Creating a farmer sustainability index: a Malaysian case study. Am. J.
Alternative Agrie. 8:175-184.
Tewari, S. 1995. The relation between mode of legume nitrogen nutrition, yield
determinants and N assimilation efficiency. Ph.D. diss. Univ. of Hawaii, Honolulu
(Diss. Abstr. 95-32632).
Thompson, P.B. 1992. The varieties of sustainability. Agrie. Human Values 9(3); 11-19.
Thornton, P.K., J.W. Hansen, E.B. Knapp, and J.W. Jones. 1995. Designing optimal crop
management strategies. In J. Bouma, A. Kuyvenhoven, B.A.M. Bouman, J.C.
Luyten, and H.G. Zandstra (ed.) Ecoregional Approaches for Sustainable Land
Use and Food Production. Kluwer Academic Publishers, Dordrecht, The
Netherlands, (in press)
Thornton, P.K., G. Hoogenboom, P.W. Wilkens, and W.T. Bowen. 1995. A computer
program to analyze multiple-season crop model output. Agron. J. 87:131-136.


39
Finally, the last term in Eq. [3-3] is equivalent to sustainability from Eq. [3-1]. By
taking the limit of Eq. [3-3] as A t approaches 0 and substituting the simplified terms (Eq.
[3-4] and [3-5]), we obtain,
fTF(0 = Fx,,(*0) 5(0- [3-6]
We can now derive an expression of S as a function of only FX i. Differentiating Eq. [3-1]
gives,
dS(t)/dt = -fTF(0- [3-7]
Substituting Eq. [3-6] into [3-7] gives the differential equation,
dS(t)/dt = -Fxt(x0) S(t), [3-8]
which has the solution,
S(T) = exp | fFx,,(x0) d/j [3-9]
at time T. Thus, I have shown that sustainability is determined entirely by the probability
that a systems state falls below a threshold value during some time interval (0, 7].
Two examples illustrate how time, thresholds, and the distribution of system state
interact to determine sustainability. In the first example (Fig. 3-la), the variability ofx is
relatively high but its expected value, E[x(t)], remains constant. The system has a 0.15
probability of failing in any particular period. Sustainability declines exponentially with


59
Methods
Perry et al. studied a representative coastal Texas rice farm rather than an actual
farm. They examined four combinations of rotation and sharecropping arrangement.
Assumptions and initial conditions can be found in Perry et al. (1986).
Perry et al. used a modified version of the FLIPSIM (Firm Level Income Tax and
Farm Policy Simulator) simulation model (Richardson & Nixon, 1985) called RICESIM to
simulate the model farm. RICESIM is primarily a farm accounting model which randomly
samples from probability distributions as a proxy for the biological and ecological
processes involved in crop production.
The study examined two alternate criteria for system failure. First, probability of
survival was based on a threshold leverage ratio (total debt/total equity). Lenders were
assumed to force foreclosure by recalling loans if the leverage ratio exceeded 2.0. The
second criterion, a negative net present value (NPV) of future cash flow, requires a higher
level of minimum system performance to avoid failure. Failure indicated by a negative
NPV means that a secure, non-farm investment would be more profitable than farming.
Probabilities of survival and positive NPV were calculated by Eq. [3-12] from 50
replicates of each five-year scenario.
Perry et al. analyzed sensitivity of the probabilities of survival and positive NPV to
several factors. For our analysis, I selected only the soybean-soybean-rice rotation with
1/2 share of rice and 1/7 share of soybean going to the landowner. I tested those factors
included in the sensitivity analysis that I believed could constrain sustainability (Tables 3-5


55
Farm failure criteria
Failure criteria denote the minimum level of performance above which a system is
to be sustained. Since farmer livelihood is the primary purpose of most farming systems,
criteria for farm failure can be expressed in terms of minimum levels of livelihood goals.
Hamblin (1992) suggested that agriculture fails to sustain if production falls below the
levels necessary for profitability in a cash economy or survival in a subsistence economy.
In a subsistence economy, a level of poverty or malnutrition from which a farm family
cannot escape without outside intervention might indicate system failure. Lynam and
Herdt (1989) referred to famine as the ultimate indicator of unsustainability (p. 391). In
a cash economy, lenders may impose threshold leverage ratios above which they will force
foreclosure by recalling loans (Perry ei al., 1985). Failure could be expressed in several
forms such as farm abandonment, conversion of land to non-agricultural use, the need to
supplement income with off-farm employment, inability to meet critical goals such as
education of children, or major changes in farm enterprises, depending on the analysts
purpose.
Negative feedback loops between components of a system tend to counteract the
effects of disturbances and stabilize a system, resulting in a stable state or attractor.
Mathematical and empirical evidence suggests that ecosystems can possess multiple stable
states (May, 1977). The region about a stable state in state space is a domain of
attraction, and the boundary between adjacent domains of attraction is a separatix
(Trenbath ei al., 1990). If the state of a system is displaced across a separatix, it enters a


Rainfall, mm Temperature, C Solar radiation, MJ / m
no
Figure 5-3. Monthly climate statistics: mean daily solar radiation, mean daily
maximum and minimum temperature, and total rainfall, Domingo farm, Cauca,
Colombia.


71
farmer livelihood by accounting for all farm resources produced, used for production, or
consumed by the farm household. The fourth requirement was the ability to test
conditions for system failure. Failure can be based on insolvencythe inability to cover
fixed costs, obligations or minimum subsistence consumption requirements. Failure can
also occur when an individual or aggregate farm resource violates a user-specified
threshold value. Detailed resource accounting was a prerequisite to the ability to test for
conditions for farm failure. Finally, FSS offers several file and graphical outputs that are
relevant to the analysis of farm sustainability.
The current version of FSS possesses several important limitations. First, it does
not simulate adaptive management; it simulates a fixed, continuously repeating set of
management practices. However, an actual farm operating under stress would normally
employ a range of practices to avoid failure (Chapter 3). Second, FSS does not possess a
mechanism for adjusting crop management for within-season resource constraints. This
limitation is imposed by the need to simulate each crop for an entire season. Chapter 5
discusses the problem and a possible solution. Third, FSS can simulate only crop
production enterprises. No livestock model has yet been adapted for running under FSS.
Fourth, the household resource consumption model (Eq. [4-7]) is simplistic; it does not
consider the impact of risk or anticipated future lifestyle changes on consumption and
savings decisions. The remaining sections of this chapter focus on inputs, processes and
outputs of FSS.


Ill
Surface charge of these minerals depends on pH and soil solution activity.
Negative ApH (pH in KC1 pH in water) values (Table 5-2) indicate net negative charge,
and | ApH | >0.5 for each layer indicates that pH-dependent charge dominates (Uehara &
Gillman, 1981). However, cation retention is likely to be poor; the effective cation
exchange capacity (ECEC) is low in the Ap horizon and extremely low in the Bw horizon.
Raising the pH by liming would increase cation retention.
Figure 5-4. Mean monthly rainfall totals at weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation.


28
5). First, characterization should be based on a literal interpretation of sustainability (Fox,
1991). Regardless of the merits of goals and ideals frequently incorporated into
definitions of sustainability, if the idea of continuation through time is omitted then those
ideals and goals are something other than sustainability.
Table 2-5. Elements of a useful approach to characterizing sustainability of agricultural
systems.
Element
Explanation
Literal
Defines sustainability as an ability to continue through time, consistent
with literal English usage.
System-
oriented
Identifies sustainability as an objective property of a particular
agricultural system whose components, boundaries, and context in
hierarchy are clearly specified.
Quantitative
Treats sustainability as a continuous quantity, permitting comparisons of
alternative systems or approaches.
Predictive
Deals with the future rather than the past or present.
Stochastic
Treats variability as a determinant of sustainability and a component of
predictions.
Diagnostic
Uses an integrated measure of sustainability to identify and prioritize
constraints.
Second, characterization should be system-oriented. A literal interpretation
suggests that sustainability is an objective property of an agricultural system. It cannot be
a property of approaches to agriculture if it is to serve as a basis for evaluating and
improving approaches. Lynam and Herdt (1989) argued that sustainability is a relevant
criterion for evaluating technology only when the system is clearly specified, including its
boundaries, components, and context in hierarchy. Sustainability has meaning only in the
context of specific temporal and spatial scales. Fresco and Kroonenberg (1992) cited a


Probability Probability Probability
188
Liquid assets, Col.$ Liquid assets, Col.$
(Millions) (Millions)
Figure 6-5. Cumulative distribution of liquid assets after 1 (a), 3 (b), 6 (c), 9 (d), 12 (e)
and 15 (f) years, base scenario, Domingo farm, Cauca, Colombia.


126
Simulation conditions
The crop simulation analyses described in the remainder of this chapter are based
on the cultivar and planting information in Table 5-4 except as noted. The models were
run with soil water balance, soil N dynamics, canopy-level photosynthesis, and Priestly-
Taylor evaporation options enabled. Replicated simulations used weather data generated
by WGEN based on spatially interpolated parameters.
Table 5-4, Planting information for crop simulation studies.
Crop
Cultivar
method
Planting
date
density
(nr2)
depth
(cm)
Row
width
(cm)
October bean
ICA Caucaya*
seed
Oct 22, 1993
16.60
2
30
March bean
ICA Caucaya
seed
Mar 31, 1994
16.60
2
30
October maize
CIMCALP
seed
Oct 18, 1993
5.00
4
60
March maize
CIMCALI
seed
Mar 30, 1994
5.00
4
60
Cassava
MCol-1501*
cutting
Jan 10, 1993
0.70
10
100
Tomato
sunny, semi-
determinate5
transplant
Mar 30, 1994
0.75
10
150
i Genetic coefficients supplied with DSSAT3.
1 Genetic coefficients calibrated by E.B. Knapp from October 1993 planting.
§ Genetic coefficients supplied by J.M.S. Scholberg.
Initial KCl-extractable NH4+ measurements (Table 5-2) were suspiciously high:
about an order-of-magnitude higher than N03\ In similar soils on neighboring farms,
levels of extractable NH4+ were generally lower than N03'. Therefore, the models were
simulated for one year before each planting date to allow simulated exchangeable NH4+
and N03' to reach a balance with mineralization and other N transformation processes.


202
Constraints to sustainability
The results already presented demonstrate the potential for improving farm
sustainability through cropping system design, the role of weather and price risk as
sustainability constraints, the potential contribution of a supply of credit to farm
sustainability, and the potential loss of sustainability from failing to diversify crops or to
control erosion. Table 6-10 lists the factors included in this study that can be represented
as continuous quantities, and ranks their importance as constraints to sustainability. The
three factors that have a direct, proportional impact on farm income or expensesland
area cultivated, subsistence consumption, and mean commodity priceshad the greatest
impact on farm sustainability. Material input prices, wages, discretionary consumption
requirements, and initial funds also constrained sustainability significantly (P = 0.05).
Discussion
The simulation study of farm sustainability fulfilled the two objectives of this
chapter; it demonstrated the practical value of applying the systems framework and the
tools presented in previous chapters to an actual farming system, and it identified
determinants of sustainability of the Domingo farm. The approach was used to test
hypotheses about the role of factors as diverse as crop rotation, price volatility, soil
erosion, household consumption requirements and spatial diversity as determinants of


227
LIVESTOCK section
The LIVESTOCK section has not yet been developed. Its purpose is to provide a
place for information about herds of grazing livestock.
STRATEGIES section
The STRATEGIES section specifies all of the field (i.e., crop and livestock)
activities that are a part of a field strategy, and their place in the cropping sequence. Each
item defines a strategy. The use of CU, PL, MI, MF, MR, MC, MT, ME, MH, and SM
are identical to their use in the TREATMENTS section of the crop management file
(Jones et al., 1994). They point to management information in the crop management file.
The STRATEGIES section contains multiple items.
The example in Fig. B-9 specifies a bean-maize double cropping strategy, with two
crop and two fallow activities repeated every year. The activities all use IBSNAT crop
models (I), and initialize cultivar, planting, irrigation, residue, fertilizer and chemical
application, tillage, environmental modifications, and harvest information from the
specified levels of the crop management file identified in the SCENARIO section. None
includes livestock. The entire simulated bean yield and 80% of the maize yield are
recovered in harvest.


174
Table 6-3 Prices of production inputs, Cauca,
Colombia, December 1992.
Input
Unit
Price, Col.S
day labor wages1
hour
220.00
contracted plowing*
hour
1,430.00
maize seed
kg
900.00
bean seed
kg
1,600.00
tomato seed5
kg
2,600,000.00
cassava cuttings
kg
50.00
chicken manure5
kgN
1,422.00
10-30-10*
kgN
1,980.00
17-6-18-2*
kgN
1,006.50
Manzate*
kg
3,443.00
Roxin*
kg
5,324.00
Benomyl*
kg
16,487.00
Cu oxychloride*
kg
1,902.00
poles5
unit
13.00
irrigation water
mm-ha
100.00
*UMATA-DRI 1992. Programma agropecuario municipal.
Unpublished report.
* 1994. Coyuntura Colombiana 11(1).
5 E.B. Knapp, Cl AT (personal communication).
fitted parameters for the time series models used to generate crop prices. Multipliers used
to convert from wholesale to farmgate prices accounted for differences between dry
weight and reported moisture contents, for differences in quality or cultivar, and for
market discount accounted for differences between reported wholesale and farmgate
prices (Table 6-6). Market discounts were based on a farmer interview (maize) and on
discussions with CIAT economists familiar with the region (coffee, cassava and bean).


A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY
By
JAMES WILLIAM HANSEN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1996

Copyright 1996
by
James William Hansen

ACKNOWLEDGMENTS
I would like first to express my gratitude to my advisor and mentor, Dr. Jim Jones.
He took a risk by giving me the freedom to pursue an idea, then guided and encouraged
me through the process of making it reality. His competence, integrity and compassion
serve as an example to many. Dr. Robert Caldwell first suggested to me the idea that
became the basis for this study. Dr. Phillip Thornton helped me numerous times to clarify
and express my ideas. He and Drs. Peart, Boggess and Hildebrand provided
encouragement and wise guidance for this research. I owe a debt of gratitude to Dr. Ron
Knapp for his kindness and hospitality during my visits to CIAT, and for many long hours
of hunting down information for this research. His enthusiasm helped me to maintain
mine. Thanks goes also to his staff, especially Jorge and Jorge. José Domingo deserves
honor and thanks. His successful effort to extract a livelihood for his family and higher
education for his children from a few hectares of hilly land provided the basis for much of
this study.
I am grateful to my wife, Medie, who endured with me and remained my best
friend through the process. In her love I found strength as I worked and a refuge as I
rested. Thanks also to friends at North Central Baptist Church and others who supported
me in prayer. Ultimate thanks goes to the Lord Jesus who gave me strength to complete
this study and compassion that gives it purpose.
iii

TABLE OF CONTENTS
ACKNOWLEDGMENTS iii
LIST OF TABLES viii
LIST OF FIGURES xii
ABSTRACT xviii
CHAPTERS
1 INTRODUCTION 1
2 IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT? 4
Introduction 4
Sustainability as an Approach to Agriculture 6
Sustainability as an Alternative Ideology 7
Sustainability as a Set of Strategies 11
Discussion 14
Sustainability as a Property of Agriculture 17
Sustainability as an Ability to Satisfy Goals 17
Sustainability as an Ability to Continue 18
Approaches to Characterizing Sustainability 19
Adherence to Prescribed Approaches 19
Multiple Qualitative Indicators 20
Integrated, Quantitative Indicators 21
Time Trends 23
Resilience 25
System Simulation 26
Elements of a Useful Approach for Characterizing Sustainability 27
Conclusions 32
IV

3 A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM
SUSTAINABILITY 33
Introduction 33
Defining Sustainability 34
Quantifying Sustainability 36
Failure as Violation of State Thresholds 36
Sustainability Hazard 41
Simulating Sustainability 42
An Example: Sustainability of a Simple Time-series Model 44
Diagnosing Constraints to Sustainability 46
Sensitivity Analysis 48
Significance Tests 49
Sustainability Applied to Farming Systems 52
Selecting a Time Frame 53
Assumptions about Inputs 54
Farm Failure Criteria 55
Determinants of Farm Sustainability 56
An Example: Sustainability of a Coastal Texas Rice Farm 58
Methods 59
Results . 60
Discussion 62
4 AN OBJECT-ORIENTED REPRESENTATION OF A FARMING
SYSTEM 65
Introduction 65
Overview of the Farming System Simulator 67
Inputs 72
Scenario File 73
Price File 79
Crop Minimum Data Set . 80
Processes 80
Overview 81
Random Number Sequences 82
Price Generation 82
Crop and Ecosystem Processes 85
Event Handling 86
Resource Accounting 88
Illustration: A Fertilizer Application Operation 95
Household Consumption 98
v

Outputs 99
Resource Status 99
Sustainability 101
Discussion 101
5 CROP SIMULATION FOR CHARACTERIZING SUSTAINABILITY
OF A COLOMBIAN HILLSIDE FARM 104
Introduction 104
A Colombian Hillside Environment 106
Crop Simulation 116
Approach • ■ • 120
Weather Data 120
Soil Data 125
Simulation Conditions 126
Development and Yield 127
Response to Environmental Factors 128
Results and Discussion 129
Crop Development and Yield 129
Weather Variability 136
Response to Nitrogen Dynamics 139
Response to Soil Erosion 152
Linking Models 158
Issues in Linking Crop and Whole-farm Models 158
Issues in Linking Crop and Erosion Models 161
Conclusions 164
6 DETERMINANTS OF SUSTAINABILITY OF A COLOMBIAN
HILLSIDE FARM 167
Introduction 167
Approach 170
A Colombian Hillside Farm 170
Sources of Information 171
Assumptions 176
Scenarios 180
Simulation and Analysis 185
Results 186
Base Scenario 186
Cropping Systems 190
Soil Management 192
Costs and Prices 194
vi

Resources 196
Sources of Risk 198
Constraints to Sustainability 202
Discussion 202
Practical Implications 203
Limitations 204
7 SUMMARY AND CONCLUSIONS 206
APPENDICES
A OBJECT-ORIENTED PROGRAMMING CONCEPTS 210
B A MINIMUM DATA SET FOR SIMULATING FARM
SUSTAINABILITY 213
C FARMING SYSTEM SIMULATOR USER'S GUIDE 235
D INPUT FILES USED FOR FARM SIMULATIONS 238
BIBLIOGRAPHY 249
BIOGRAPHICAL SKETCH 265
vii

LIST OF TABLES
Table page
2-1 Interpretations of agricultural sustainability 5
2-2 Contrasting approaches of conventional and sustainable agriculture as
characterized by Hill and MacRae 9
2-3 Strategies frequently associated with sustainability 11
2-4 Contingency table for inferring sustainability based on trends of system inputs
and outputs 23
2-5 Elements of a useful approach to characterizing sustainability of agricultural
systems 28
2-6 Approaches to characterizing agricultural sustainability '31
3-1 Definitions of symbols used 37
3-2 Sensitivity of simulated sustainability to system properties 46
3-3 Frequency table for tests of difference between simulated sustainabilities,
independent observations 50
3-4 Frequency table for testing differences in sustainabilities: paired observations .. 51
3-5 Relative sensitivity of simulated five-year sustainability of a Texas rice farm
to continuous factors 60
3-6 Absolute sensitivity of simulated five-year sustainability of a Texas rice farm
to discrete factors 61
4-1 Sections in the farm scenario file 74
vm

4-2 Description of resource classes 77
4-3 Format of the operations output file 87
5-1 Area in each slope class, Domingo farm, Cauca, Colombia 106
5-2 Properties of soil layers, site near on-farm trials, Domingo farm, Cauca,
Colombia 112
5-3 Spatially interpolated WGEN coefficients for the Domingo farm, Cauca,
Colombia 125
5-4 Planting information for crop simulation studies 126
5-5 Observed and predicted bean yields, Domingo and Trujillo farms, Cauca,
Colombia 130
5-6 Observed and predicted timing of phenological events for bean and maize,
Domingo farm, Cauca, Colombia, planted March 30, 1994 130
5-7 Treatment description and observed and predicted yields for the October 14,
1993 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132
5-8 Treatment description and observed and predicted yields for the March 30,
1994 planting of the bean fertility trial, Domingo farm, Cauca, Colombia .... 132
5-9 Observed and predicted maize yields, Domingo and Trujillo farms, Cauca,
Colombia 134
5-10 Observed and predicted cassava yields, Domingo and Trujillo farms, Cauca,
Colombia 135
5-11 Mean, standard deviation, skewness and Kolmogorov-Smirnov test statistic
for distributions of crop yield and maturity time simulated with observed and
simulated weather from La Florida, Cauca, Colombia 137
5-12 Mean, standard deviation, skewness, and Kolmogorov-Smirnov test statistic for
distributions of observed and simulated monthly and annual rainfall totals, La
Florida, Popayan, Colombia 139
6-1 Hypotheses related to determinants of farm sustainability 169
IX

6-2 Estimates of labor requirements for annual crop production, Cauca,
Colombia 173
6-3 Prices of production inputs, Cauca, Colombia, December 1992 174
6-4 Fitted parameters for deterministic component of production commodity
price time series models 175
6-5 Fitted parameters for stochastic component of production commodity price
time series models 175
6-6 Adjustments to simulated crop yields and reported prices, Cauca, Colombia . . 176
6-7 Description of farm scenarios 181
6-8 Predicted 15 year sustainability of cropping system scenarios, and McNemar
test statistic for difference from the base scenario 192
6-9 Soil loss and predicted 15 year sustainability of erosion scenarios 195
6-10 Relative sensitivity of predicted 15 year sustainability to continuous factors,
and McNemar test statistic for difference from the base scenario 197
6-11 Predicted nine-year sustainability, McNemar test statistic for difference from
the base scenario, and standard deviation of liquid assets after three years for
sources of risk scenarios 199
A-l Object-oriented programming terms 211
B-l Format of the SCENARIO section of the scenario file 215
B-2 Format of the OUTPUTS section of the scenario file 216
B-3 Format of the ANALYSES section of the scenario file 217
B-4 Format of the RESOURCES section of the scenario file 219
B-5 Format of the LINKAGES section of the scenario file 221
B-6 Format of the OPERATION REQUIREMENTS section of the scenario file . . 223
B-7 A resource block within the OPERATION REQUIREMENTS section 223
x

B-8 Format of the SCHEDULED OPERATIONS section of the scenario file .... 225
B-9 Format of the LANDSCAPE section of the scenario file .. 226
B-10 Format of the STRATEGIES section of the scenario file 228
B-ll Format of the ENTERPRISES section of the scenario file 230
B-12 Format of the CONSUMPTION DECISIONS section of the scenario file ... 231
B-13 Format of the CONSTANT section of the price file 232
B-12 Format of the ARMA section of the price file 233
B-12 Format of the HISTORICAL section of the price file 234
C-l Options available from FSS menu items 236
X!

LIST OF FIGURES
Figure
page
2-1 Contrasting interpretations of the relationship between chemical input levels
and sustainability 13
3-1 Relationship between time, distribution of state and sustainability under
constant mean and high variability, and negative trend and low variability 40
3-2 Estimating sustainability by sampling a small number of simulated
realizations of future system behavior 43
3-3 Sustainability and hazard of an AR(1) process with constant parameters,
declining expected value, abrupt decrease and increase in mean, decrease
and increase in variance, and decrease and increase in autocorrelation 47
3-4 Simulated sustainability of a Texas rice farm under four scenarios 62
4-1 Parallel ecological, economic and social hierarchies of agricultural systems .... 68
4-2 Object representation of the main components of a farming system 70
4-3 Tree of resource class hierarchy inFSS 75
4-4 Flow of information in FSS from the generation of a field operation to its
effect on resource accounting 89
4-5 Pseudocode representation of the Use method of the consumable resource
class 90
4-6 Pseudocode representation of the SELL method of the consumable resource
class 91
4-7 Pseudocode representation of the USE method of the timed resource class .... 92
Xll

4-8
Pseudocode representation of the Use method of the credit resource class .... 92
4-9 Pseudocode representation of the Update method of the consumable
resource class 93
4-10 Pseudocode representation of the UPDATE method of the capital resource
class 93
4-11 Pseudocode representation of the UPDATE method of the credit resource
class 94
4-12 Pseudocode representation of the Use method of the cost list class 95
4-13 Forrester representation of variable cost linkages between a consumable
resource and three linked consumable resources 96
4-14 Flow of information from the generation of a fertilizer application operation
by a crop model to its effect on the operating fund 97
4-15 Example of a resource box plot generated by FSS .99
4-16 Example of a final resource distribution plot generated by FSS 100
4-17 Example of a sustainability time plot generated by FSS 102
5-1 Location map of the study area, Cauca, Colombia 107
5-2 Land use map of the Domingo farm, Cauca, Colombia ................. 108
5-3 Monthly climate statistics; mean daily solar radiation, mean daily maximum
and minimum temperature, and total rainfall, Domingo farm, Cauca,
Colombia 110
5-4 Mean monthly rainfall totals at weather station used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation 111
5-5 Phosphorus sorption isotherms for four soils with different mineralogy ...... 113
5-6 Decomposition of soil organic C in three allophanic and seven
non-allophanic soils 114
xiii

5-7 Decomposition of 14C-labeled wheat straw in Lo Aguire sandy loam with
added allophane 115
5-8 Locations and elevations of weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation 123
5-9 Trend in the ratio of bean yields observed in on-farm trials and predicted
by CROPGRO, Domingo and Trujillo farms, Cauca, Colombia 131
5-10 Simulated and observed bean yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994, October 1994 and
March 1995 133
5-11 Simulated and observed maize grain yields, Domingo and Trujillo farms,
Cauca, Colombia, planted October 1993, March 1994 and October 1994 .... 135
5-12 Distribution of simulated yields of October- and March-planted maize and
bean in response to historical and generated weather variability, La Florida,
Popayan, Colombia 138
5-13 Simulated grain and biomass yield response of maize and October- and
March-planted bean applied N using the default mineralization factor 140
5-14 Maize response to applied N on several volcanic soils in Nariño, Colombia .. 141
5-15 Effect ofN mineralization factor, SLNF, on simulated maize response to
applied N 142
5-16 Simulated grain and biomass yield response of maize and October- and
March-planted bean to applied N using the adjusted mineralization factor .... 143
5-17 Grain and biomass yields of October- and March-planted bean simulated
by the original and modified versions of CROPGRO 144
5-18 Grain and biomass yields of irrigated and rainfed soybean simulated by the
original and modified versions of CROPGRO 145
5-19 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Domingo farm, Cauca, Colombia, 1994 147
xiv

5-20 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Kuiaha, Hawaii, 1993 148
5-21 Plant and nodule mass at 35 days, and grain yield of container-grown
beans in response to applied N 150
5-22 Simulated 60 year sequences of maize grain yields at different levels of
applied N, Domingo farm, Cauca, Colombia 151
5-23 Simulated grain yields of maize and bean in response to soil loss, Domingo
farm, Cauca, Colombia 153
5-24 Simulated root distributions of October- and March-planted bean in response
to 0 cm, 20 cm, and 50 cm of soil loss, Domingo farm, Cauca, Colombia .... 155
5-25 Relative root distribution factors, F and SRGF after 0 cm, 20 cm and 50 cm
soil loss 157
5-26 Simulated grain yields of October- and March-planted bean in response to soil
loss, Domingo farm, Cauca, Colombia 159
5-27 Pseudocode representation of an algorithm for resolving resource conflicts
among crop enterprises 160
5-28 Pseudocode representation of an algorithm for simulating erosion and crop
growth on a complex hillslope 164
6-1 Historical crop wholesale prices, Cauca, Colombia 172
6-2 Cropping pattern included in farm scenario 178
6-3 Influence of type of farmer participation in on-farm trials on bean response
to ground and partially acidulated rock phosphate, chicken manure, and
10-30-10 179
6-4 Box plot of liquid assets, base scenario 187
6-5 Cumulative distribution of liquid assets after 1, 3, 6, 9, 12 and 15 years, base
scenario 188
6-6 Sustainability and hazard time plots of the base scenario 189
xv

6-7 Sustainability time plot of annual cropping systems 190
6-8 Sustainability time plot of base and coffee scenarios 191
6-9 Sustainability time plot of base and nitrogen management scenarios . 193
6-10 Sustainability time plot of base and soil erosion scenarios 194
6-11 Sustainability time plot of base and price scenarios 195
6-12 Sustainability time plot of base and household consumption scenarios 196
6-13 Sustainability time plot of base and resource scenarios 198
6-14 Sustainability time plot of base and credit scenarios 199
6-15 Distribution of liquid assets after 1, 3, 6 and 9 years for the source of risk
scenarios 200
6-16 Sustainability time plot of source of risk scenarios 201
B-l Example SCENARIO section 215
B-2 Example OUTPUTS section 216
B-3 Example ANALYSES section 217
B-4 Example item from the RESOURCES section 218
B-5 Example item from the LINKAGES section 221
B-6 Example item from the OPERATION REQUIREMENTS section 224
B-7 Example item from the SCHEDULED OPERATIONS section 224
B-8 Example item from the LANDSCAPE section 226
B-9 Example item from the STRATEGIES section 229
B-10 Example item from the ENTERPRISES section 229
B-l 1 Example item from the CONSUMPTION DECISIONS section 231
xvi

D-l Farm scenario file used to simulate the base scenario 239
D-2 Crop management file used to simulate farm scenarios 242
D-3 Soil profile description file used to simulate farm scenarios 247
XVII

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
A SYSTEMS APPROACH TO CHARACTERIZING FARM SUSTAINABILITY
By
James William Hansen
May 1996
Chairperson: Dr. James W. Jones
Major Department: Agricultural and Biological Engineering
The potential usefulness of the concept of sustainability as a criterion for
evaluating and improving agricultural systems has been hindered by inadequacy of
approaches for its characterization. The approach presented here for characterizing farm
sustainability is based on a definition of sustainability as the ability of a system to continue
into the future, expressed by the probability that the system will continue without violating
failure thresholds during a particular future period. Characterization includes
quantification and diagnosis of constraints. Sustainability is quantified by using long-term,
stochastic simulation to sample realizations of the future behavior of a model of the
farming system. Sensitivity analysis provides a tool for testing hypotheses about
constraints to sustainability. An object-oriented farming system simulator developed for

this purpose can simulate a replicated farm scenario with stochastic inputs of weather and
prices. The farm simulator simulates the balance of farm resources and accounts for
ecological determinants of crop production by calling external crop models. Several crop
simulation models were evaluated for compatibility with a simulation study of the
sustainability of a hillside farm in the Cauca region of Colombia. The crop models were
useful for capturing response to weather variability and N management, but did not
account for important yield determinants, including P deficiency and nematode damage.
Simple modifications improved simulated response to applied N and sensitivity to soil loss.
A study of a hillside farm in Colombia showed the practical value of using simulation to
characterize sustainability. Results identified cropping system, area under cultivation,
consumption requirements, crop prices, and soil erosion as important determinants of
sustainability. The study showed that price variability contributes more than weather
variability to farm risk in this location, and that spatial diversification reduces risk and
improves sustainability. Results suggest that the farmer can enhance farm sustainability by
diversifying and intensifying crop production. Researchers can contribute to sustainability
by identifying promising high-value crops, by testing the proposed cropping systems, and
by quantifying severity and impacts of erosion. Policy makers can address sustainability
constraints caused by price volatility and lack of affordable credit.
xix

CHAPTER 1
INTRODUCTION
The past decade has shown a shift of focus from productivity to sustainability of
agricultural systems. The growing emphasis on sustainability stems from concern about
both threats to, and negative impacts of agriculture, and from the realization that decisions
made now can have unforeseeable impacts on future generations. In spite of the growing
interest in agricultural sustainability, there is no generally accepted definition of what it is.
There is even less agreement about how to achieve it. Little progress has been made
toward developing methods for characterizing sustainability of particular agricultural
systems because of the conceptual problem of agreeing on a definition and the practical
problems that result from the fact that sustainability deals with the future and therefore
cannot be readily observed. Although some have argued that sustainability cannot and
should not be measured, there is growing realization that progress toward improving the
sustainability of agriculture is not possible without the ability to measure it and identify its
constraints.
Although concerns about agricultural sustainability focus on system levels ranging
from field to global, the farm is an appropriate level for dealing with sustainability. It is
appropriate from the standpoint of relevance; it is the farmer who makes the final
decisions about what to produce and what resources and methods will be employed. If
1

2
agriculture is to meet the needs of society as a whole—producing food and other products
while protecting natural resources—it must first meet the needs of the farmers who
implement and manage it. The farm level is also appropriate from a practical standpoint.
Since human goals are not intrinsic to fields or enterprises, it is difficult to identity what is
to be sustained. On the other hand, characterization of sustainability at system levels
higher than the farm is complicated by the complexity of the systems, the difficulty of
specifying system boundaries, and the often conflicting goals of multiple human actors.
The purpose of this study is to present and demonstrate a simulation-based systems
approach to characterizing sustainability of farming systems. Specific objectives of this
dissertation are as follows:
1. Review existing conceptual and methodological barriers to using the concept of
sustainability for guiding change in agriculture
2. Propose a set of elements necessary for an approach to characterizing
sustainability to provide a useful basis for guiding agriculture.
3. Present the logical and mathematical basis for a definition of sustainability that
applies generally to dynamic, hierarchical, stochastic, purposeful systems, and
relate this definition to farming systems.
4. Describe and illustrate a framework for using system simulation to quantify
sustainability and test hypotheses about its determinants.
5. Describe an object-oriented farming system simulator that was developed for the
purpose of characterizing farm sustainability, and present its data requirements.
6. Evaluate the compatibility of a set of crop models with requirements imposed by
a simulation study of sustainability in a hillside environment in Colombia.
7. Demonstrate and evaluate the approach for characterizing farm sustainability by
identifying determinants of the sustainability of a hillside farm in Colombia.

3
This dissertation is organized around these objectives. Chapter 2 critically reviews
existing interpretations of sustainability and approaches proposed for its characterization.
A conceptual and mathematical framework for characterizing farm sustainability is
presented in Chapter 3. Chapter 4 presents the farm simulator that was developed to
apply the proposed framework to a particular farm. The simulator uses process-level crop
simulation to characterize the contribution of ecosystem processes to farm production and
sustainability. Chapter 5 describes the physical environment in the region of Colombia
where the approach was applied, and evaluates the suitability of a set of crop models for
characterizing production and sustainability in that environment. Finally, Chapter 6
applies the approach to a particular farm in the lower Andes of southwestern Colombia.

CHAPTER 2
IS AGRICULTURAL SUSTAINABILITY A USEFUL CONCEPT?
Introduction
In literal English usage, sustainability is the ability to “keep in existence; keep up;
maintain or prolong” (Neufeldt, 1988, p. 1349). The variety of meanings acquired by
sustainability as applied to agriculture (Table 2-1) have been classified according to the
issues motivating concern (Douglass, 1984; Weil, 1990), their historical and ideological
roots (Kidd, 1992; Brklacich et al., 1991), and the hierarchical levels of systems
considered (Lowrance et al., 1986).
The distinction between sustainability as a system-describing and as a goal¬
prescribing concept (Thompson, 1992) identifies two current schools-of-thought that
differ in their underlying goals. The goal-prescribing concept interprets sustainability as an
ideological or management approach to agriculture. This concept developed in response
to concerns about negative impacts of agriculture, with the underlying goal of motivating
adoption of alternative approaches. The system-describing concept interprets
sustainability either as an ability to fulfill a diverse set of goals or as an ability to continue.
This concept can be related to concerns about impacts of global change on the viability of
4

5
Table 2-1. Interpretations of agricultural sustainability.
Sustainability as an ideology:
.. a philosophy and system of farming. It has its roots in a set of values that reflect a state of empowerment, of awareness
of ecological and social realities, and of one’s ability to take effective action.” (MacRae el at., 1990, p. 156)
.. an approach or a philosophy... that integrates land stewardship with agriculture. Land stewardship is the philosophy
that land is managed with respect for use by future generations.” (Neher, 1992, p. 54)
“... a philosophy based on human goals and on understanding the long-term impact of our activities on the environment
and on other species. Use of this philosophy guides our application of prior experience and the latest scientific advances to
create integrated, resource-conserving, equitable farming systems.” (Francis & Youngberg, 1990, p. 8)
“... farming in the image of Nature and predicated on the spiritual and practical notions and ethical dimensions of
responsible stewardship and sustainable production of wholesome food.” (Bidwell, 1986,p.317)
Sustainability as a set of strategies:
.. a management strategy which helps the producers to choose hybrids and varieties, a soil fertility package, a pest
management approach, a tillage system, and a crop rotation to reduce costs of purchased inputs, minimize the impact of the
system on the immediate and the off-farm environment, and provide a sustained level of production and profit from
farming.” (Francis, Sander & Martin, 1987, p. 12)
“... a loosely defined term for a range of strategies to cope with several agriculturally related problems causing increased
concern in the U.S. and around the world.” (Lockeretz, 1988, p. 174)
Farming systems are sustainable if “they minimize the use of external inputs and maximize the use of internal inputs already
existing on the farm.” (Carter, 1989, p. 16)
.. (a) the development of technology and practices that maintain and/or enhance the quality of land and water resources,
and (b) the improvements in plants and animals and the advances in production practices that will facilitate the substitution
of biological technology for chemical technology.” (Ruttan, 1988, p. 129)
Sustainability as the ability to fulfill a set ofgoals:
“A sustainable agriculture is one that, over the long term, enhances environmental quality and the resource base on which
agriculture depends, provides for basic human food and fiber needs, is economically viable, and enhances the quality of life
for fanners and society as a whole.” (American Society of Agronomy, 1989, p. 15):
“... agricultural systems that are environmentally sound, profitable, and productive and that maintain the social fabric of the
rural community.” (Keeney, 1989, p. 102)
“... an agrifood sector that over the long term can simultaneously (1) maintain or enhance environmental quality, (2)
provide adequate economic and social rewards to all individuals and firms in the production system, and (3) produce a
sufficient and accessible food supply.” (Brklacieh, Bryant & Smit, 1991, p. 10)
“... an agriculture that can evolve indefinitely toward greater human utility, greater efficiency of resource use, and a
balance with the environment that is favorable both to humans and to most other species.” (Harwood, 1990, p. 4)
Sustainability as the ability to continue:
“A system is sustainable over a defined period if outputs do not decrease when inputs are not increased.” (Monteith, 1990,
P- 91)
“... the ability of a system to maintain productivity in spite of a major disturbance, such as is caused by intensive stress or a
large perturbation.” (Conway, 1985, p. 35)
.. the maintenance of the net benefits agriculture provides to society for present and future generations.” (Gray, 1991, p.
628)
“Agriculture is sustainable when it remains the dominant land use over time and the resource base can continually support
production at levels needed for profitability (cash economy) or survival (subsistence economy).” (Hamblin, 1992, p. 90)

6
agriculture, and to the goal of using sustainability as a criterion for guiding agriculture as it
responds to rapid changes in its physical, social and economic environment.
Although the concept of sustainability has been useful for consolidating concerns
and motivating change, concrete examples of its use as an operational criterion for guiding
efforts to improve agricultural systems are difficult to identify. The objectives of this
chapter are (1) to examine conceptual and methodological barriers to using the concept of
sustainability for guiding change in agriculture, and (2) to propose a set of elements
necessary for an approach to characterizing sustainability to provide a useful criterion for
guiding agriculture.
Sustainability as an Approach to Agriculture
The sustainable agriculture movement evolved from several reform movements in
the U.S.A., Canada and Western Europe that developed in response to concerns about
impacts of agriculture such as depletion of nonrenewable resources, soil degradation,
health and environmental effects of agricultural chemicals, inequity, declining rural
communities, loss of traditional agrarian values, food quality, farm worker safety, decline
in self-sufficiency, and decreasing number and increasing size of farms. These problems
became associated with “conventional agriculture” that was perceived as unsustainable
(Dahlberg, 1991). “Alternative agriculture” is often equated with sustainable agriculture
(O’Connell, 1992; Madden, 1987; Harwood, 1990; Dahlberg, 1991; Bidwell, 1986) and
reflects the goal of promoting alternatives to conventional agriculture. Reviews by

7
Harwood (1990) and Kidd (1992) trace the historical development of the sustainable, or
alternative agriculture movement.
Differences in values and practices promoted as sustainable have been attributed to
differences in the problems emphasized (Carter, 1989) and to different visions of what
agriculture should be like (Thompson, 1992). “Originally, the advocates of alternative
approaches to agriculture-all united in their critique of industrial agriculture as being
unsustainable—debated among themselves the future direction and shape of agriculture”
(Dahlberg, 1991, p. 337). Some have focused on identifying sustainable alternatives to
existing management practices while others have advocated new philosophical orientations
toward agriculture.
Sustainability as an alternative ideology
MacRae et ah (1990), Neher (1992) and Francis and Youngberg (1990) defined
sustainable agriculture as a philosophy (Table 2-1). Ikerd (1991) described low-input,
sustainable agriculture (LISA) as more a philosophy than a practice. Examining the
concept of conventional agriculture is important since sustainable agriculture is often
described by its contrast with conventional agriculture (Lockeretz, 1988; MacRae et ah ,
1989; Hauptli etah, 1990; Dobbs etah, 1991; O’Connell, 1992; Hill and MacRae, 1988).
Conventional agriculture. The concept of conventional agriculture was apparently
developed in order to clarify, and justify alternative approaches to agriculture.
Conventional agriculture is characterized as “capital-intensive, large-scale, highly
mechanized agriculture with monocultures of crops and extensive use of artificial

8
fertilizers, herbicides and pesticides, with intensive animal husbandry” (Knorr and
Watkins, 1984, p. x) with a paradigm of “strength through exhaustion” (Bidwell, 1986, p.
317). Hill and MacRae (1988) contrasted approaches of conventional and sustainable
agriculture (Table 2-2). Based on a review of actual cropping practices in the U.S.,
Madden (1990) appropriately identified conventional agriculture as a caricature.
Beus and Dunlap (1990) identified centralization, dependence, competition,
domination of nature, specialization, and exploitation as key elements of conventional
agriculture from the writings of six conventional agriculture advocates. Although
“conventional agriculture” was applied to mainstream U.S. agriculture as one side of a
debate between competing paradigms, its description was admittedly a construct for the
purpose of “clarifying opposing positions, facilitating comparisons, and sharpening the
focus of the debate” (p. 597). A survey of agriculturalists in Washington state by the same
authors (Beus and Dunlap, 1991) suggested that the conventional agriculture that they
described does not represent mainstream U.S. agriculture. A random sample of farmers
and three of the four groups of conventional agriculturalists surveyed showed greater
agreement with the alternative than with the conventional agriculture paradigm.
Characterization of conventional agriculture extends to mainstream research and
education institutions where research has been described as too narrow, short-sighted,
biased by interests of agribusiness funding sources, and distorted by the values of scientists
to be able to deal with the issues necessary to achieve sustainability (Bidwell, 1986; Allen
and van Dusen, 1988; MacRae et al, 1989; Dahlberg, 1991; Kirschenmann, 1991; Hill and
MacRae, 1988). Beus and Dunlap (1990) and Dahlberg (1991) expressed concern that

9
Table 2-2. Contrasting approaches of conventional and sustainable agriculture as
characterized by Hill and MacRae (1988, p. 95).
Conventional agriculture
Sustainable agriculture
Symptoms
Causes, prevention
Reductionist
Holistic
Eliminate “Enemies”
Respond to indicators
Narrow focus (neglects side effects, health
& environmental costs ignored)
Broad focus (subcellular to all life to globe, all
costs internalized)
Instant
Long time frame (future generations)
Single, simple (magic bullet, single discipline)
Multifaceted, complex (multi- & trans-
disciplinary)
Temporary solutions
Permanent solutions
Unexpected disbenefits (to person & planet)
Unexpected benefits
High power (risk of overkill & errors/ accidents)
Low power (minimal risk)
Direct “attack”
Indirect, benign approaches (catalytic, multiplier,
synergistic effects)
Imported
Local solutions and materials
Products
Processes, services
Physico-chemical (often unnatural, synthetic)
Bio-ecological (natural)
Technology-intensive
Knowledge/skill intensive
Centralized
Decentralized (human scale)
Values secondary
Compatible with higher values
Expert, paternalistic (arrogant)
Individual/community responsibility (humble)
Dependent
Self-maintaining/regulating
Inflexible
Flexible
Ignores freedom of choice (unjust)
Respects freedom of choice (just)
Disempowering
Empowering
Competitive
Co-operative
Authored
Anonymous (seeking neither reward or fame)

10
such institutions threaten to dilute the concept of sustainable agriculture by co-opting it
while ignoring its more important and radical aspects.
Alternative values. Sustainable agriculture has been described as an umbrella term
encompassing several ideological approaches to agriculture (Gips, 1988) including organic
farming, biological agriculture, alternative agriculture, ecological agriculture, low-input
agriculture, biodynamic agriculture, regenerative agriculture, permaculture, and
agroecology (Carter, 1989; MacRae etal., 1989; Bidwell, 1986; O’Connell, 1992;
Kirschenmann, 1991; Dahlberg, 1991).
Beus and Dunlap (1990) listed decentralization, independence, community,
harmony with nature, diversity, and restraint as key values of alternative agriculture.
Social values such as equity, the value of traditional agricultural systems, self-sufficiency,
preservation of agrarian culture, and preference for small, owner-operated farms have
been incorporated into definitions of sustainability (Weil, 1990; Keeney, 1989; Bidwell,
1986; Francis and Youngberg, 1990). The concept of equity is extended to include future
generations (Batie, 1989; Norgaard, 1991). Environmental values associated with
sustainability include mimicry of nature and an “ecocentric” ethic. Hauptli et al. (1990)
described mimicry of nature: “. . . sustainable agriculture attempts to mimic the key
characteristics of a natural ecosystem ...” (p. 143). The ecocentric position—valuing
ecosystems or species without regard to their impact on human welfare—is illustrated by
Douglass (1984) who stated ecology-minded people “. . . define agricultural sustainability
in biophysical terms, and to allow its measurement to determine desirable population
levels” (p. 5).

11
Sustainability as a set of strategies
Francis and Youngberg (1990) described sustainable agriculture as a philosophy
that guides the creation of farming systems. Specific management strategies are often
suggested by ideological interpretations of sustainability. The strategies promoted as
sustainable (Table 2-3) are based on the types of problems emphasized and on views of
what would constitute an improvement.
Table 2-3. Strategies frequently associated with sustainability.
Strategy
References
Self-sufficiency through preferred use of on-farm or locally
available “internal” resources to purchased “external” resources.
a, b, g, d
Reduced use or elimination of soluble or synthetic fertilizers.
a, e, f, h, i, d, k
Reduced use or elimination of chemical pesticides, substituting
integrated pest management practices.
a, c, d, e, f, h, i, j, k
Increased or improved use of crop rotations for diversification,
soil fertility, and pest control.
a, c, d, e, f, h, j
Increased or improved use of manures and other organic
materials as soil amendments.
a, c, f, h, j, k
Increased diversity of crop (and animal) species.
a, d, g, i
Maintenance of crop or residue cover on the soil.
a, d, e
Reduced stocking rates for animals.
a, c, d
“Lockeretz, 1988
b Harwood, 1990
c MacRae et al. 1990
d Neher, 1992
* Dobbs et al. 1991
f MacRae et al. 1989
8Gliessman, 1990
h Edwards, 1990
1 Hauptli et al. 1990
j O’Connell, 1992
k Hill & MacRae, 1988

12
The strategy most frequently linked to sustainability is reduction or elimination of
the use of processed chemicals, particularly fertilizers and pesticides (Stinner and House,
1987; Lockeretz, 1988; Carter, 1989; Hauptli et al., 1990; Madden, 1990; Dobbs etal.,
1991). In 1988, the U.S. Department of Agriculture linked sustainability to levels of
inputs by establishing the LISA (low-input, sustainable agriculture) research program
(O’Connell, 1990; Dicks, 1992). Arguments for reducing chemical inputs include limited
supplies of fossil fuels, decreasing commodity prices necessitating reducing input costs, a
need for self-sufficiency, concerns about pollution, and health and safety concerns (Francis
and King, 1988; Carter, 1989; Stinner and House, 1989; Conway and Barbier, 1990;
MacRae e/ar/., 1990; Rodale, 1990).
York (1991) argued that fewer options exist for reducing fertilizer inputs than
pesticide inputs in agricultural systems while maintaining sustainable production. Unlike
pesticides, soil nutrient elements generally have no substitutes and are subjected to harvest
and other losses that must be replaced by weathering or imported from outside the system
if production is to be sustained. The high energy cost unique to N fertilizer production
and the potential for biological fixation suggest a need and potential for seeking
alternatives to synthetic N fertilizers that does not exist for mineral-based nutrients such as
P and K.
The important distinction between production systems that currently employ high
levels of chemical inputs and those that employ low levels (Weil, 1990) is often
overlooked. Zandstra (1994) described sustainability as a function of chemical input levels
(Fig. 2-la). Excessive input levels were said to degrade natural resources through

input level sustainability
B
sustainability
Figure 2.1. Contrasting interpretations of the relationship between chemical input
levels and sustainability by (a) Zandstra (1994) and by (b) Stinner and House (1987).

14
accumulation while inadequate levels degrade resources through exhaustion. This concept
is in sharp contrast to the decreasing relationship between chemical input levels and
sustainability proposed by Stinner and House (1987) (Fig. 2-lb).
Studies in Mali, Benin, Zambia, and Tanzania provide examples of resource
degradation due to inadequate chemical inputs (Budelman and van der Pol, 1992). In each
case, supplies of soil nutrients were exhausted rapidly due to a combination of harvest,
erosion, leaching, denitrification and volatilization, with harvest being the greatest loss.
Nutrient budgets estimated for several crops in southern Mali were always negative for N
and K, and variable but generally better for P. The authors concluded that the only way to
make these cropping systems sustainable is with increased use of fertilizers.
Discussion
Interpreting sustainability as an approach to agriculture has been useful for
motivating change. Sustainability as an ideology has provided as a common banner for
various agricultural reform movements (Gips, 1988; Dahlberg, 1991). Research and
promotion of sustainability interpreted as a set of strategies has become part of policy in
the U.S. in the form of provisions in the 1990 Farm Bill (O’Connell, 1992; Yetley, 1992).
Interpreting sustainability as an approach is not useful for guiding change in
agriculture for several reasons. First, approaches developed in response to problems in
North America and Europe may be inappropriate in regions where circumstances and
problems are different. The alternative agriculture movement has its roots primarily in
regions characterized by high levels of resource consumption, food surpluses, high levels

15
of chemical inputs, relatively deep, fertile soils, and relatively stable populations. In
contrast, many less developed tropical regions are characterized by lower levels of
resource consumption, frequent or chronic food shortages, lower levels of chemical
inputs, relatively fragile soils, and rapidly growing populations. Attempts to link strategies
to sustainability by definition fail to consider the need to match technologies to specific
environments.
Dicks (1992) argued that interpretations of sustainability in the U.S. have been
shaped by food surpluses. A shift in concern from global food security to environmental
quality in the 1980s (York, 1991) led to the perspective that “. . . the question is not can
we produce more food, but what are the ecological consequences of doing so?”
(Douglass, 1984, p. 5). However, much of the concern about sustainability in less
developed countries is related to the need to increase productivity to meet future needs of
growing populations (Ruttan, 1988; York, 1988, 1991; Lynam and Herdt, 1989;
Plucknett, 1990). The potential for the desperation imposed by poverty to shorten
people’s planning horizons (Ashby, 1985) raises questions about the ecological
consequences of failing to produce more food (Mellor, 1988; Oram, 1988). The
alternative agriculture movement has not adequately addressed the need to feed rapidly
growing populations in order to prevent both human and ecological disaster.
The second problem is that a distorted caricature of conventional agriculture may
cause approaches that may enhance sustainability to be ignored or rejected because of
their association with conventional agricultural institutions. Although the philosophical
roots of the alternative agriculture movement formed outside of the academic community

16
(Rodale, 1990; Bidwell, 1986), most of the practices it promotes as sustainable are largely
products of mainstream research and educational institutions (Francis and Sahs, 1988;
York, 1988).
Third, establishing the contribution of an approach to sustainability through
definition eliminates the perceived need to evaluate approaches that may be poor or
harmful in a particular context. If strategies are identified as sustainable based on their
effect on agricultural systems, and agricultural systems are then judged to be sustainable
based on their implementation of sustainable strategies, then a form of circular logic
results. It is logically impossible to evaluate the contribution of an approach to
sustainability when adherence to that approach has already been used as a criterion for
evaluating sustainability. This circular logic is a fourth reason why interpreting
sustainability as an approach is not useful for guiding change.
Because of the temporal nature of sustainability, errors of either ignoring
approaches that enhance sustainability or promoting approaches that threaten it may not
be obvious when the approaches are implemented. The evaluation needed to recognize
errors and improve approaches is not possible if sustainability is interpreted as a
philosophy or a set of strategies. Thompson (1992) warned that
Our society may collapse because of shortsighted stupidity on the part of
pro-growth, resource exploiting power elites, but the collapse will only be
tragic if it is shortsightedness or ignorance on the part of environmentally
and ethically concerned people that helps bring it about, (p. 19)

17
Sustainability as a Property of Agriculture
The concept of sustainability as an approach to agriculture evolved in parallel with
the concept of sustainability as a system property. While Dahlberg (1991) argued that
“sustainability” was first used by an emerging alternative agriculture movement to
prescribe a particular set of values, Kidd (1992) countered that the system describing
concept developed earlier but did not use the word “sustainability” until later. As a
property of agriculture, sustainability is interpreted as either the ability to satisfy a diverse
set of goals or an ability to continue through time.
Sustainability as an ability to satisfy goals
A sustainable agricultural system is often defined as one that fulfills a balance of
several goals through time. These goals generally include some expression of maintenance
or enhancement of the natural environment, provision of human food needs, economic
viability, and social welfare (Table 2-1).
Lynam and Herdt (1989) argued that an interpretation of sustainability based on
several qualitative goals fails to provide a criterion useful for guiding agricultural research.
If a system is defined as sustainable when it protects the natural environment, provides
adequate food, and maintains producer profitability, then there is no logical way to rank,
for example, the relative importance of commodity price variability and nitrate leaching
into aquifers as determinants of sustainability. Furthermore, the subjectivity of goal

18
specification links criteria for determining sustainability to the goals and values of the
• analyst or the author of a definition rather than to the agricultural system. At the farm
level and higher, goals belong to the actors within the system and are, therefore,
endogenous. Kidd (1992) argued that it is not helpful to “use sustainability loosely as a
general purpose code word encompassing all of the aspects of agricultural policy that the
authors consider desirable” (p. 24).
Sustainability as an ability to continue
The final concept interprets sustainability as a system’s ability to continue through
time. Hildebrand (1990) suggested that sustainability may be interpreted as the length of
time that a system can be maintained. According to Hamblin (1992), sustainability implies
that agriculture remains the dominant land use. Lynam and Herdt (1989) and Jodha
(1990) expressed sustainability in terms of maintaining some level of output. Monteith
(1990) added consideration of the possible confounding interaction of changes in input
and output levels. The definitions of Fox (1991) and Hamblin (1992) emphasized the
continuing ability to meet human needs. Conway (1985), Conway and Barbier (1990) and
Altieri (1987) emphasized the ability to withstand disturbances.
Interpreting sustainability as an ability to continue is consistent with literal English
usage of “sustain” and its derivatives. Its potential usefulness comes from suggesting
criteria for characterizing sustainability, providing a basis for identifying constraints and
evaluating proposed approaches to its improvement. This potential usefulness has been
limited by inadequacy of current approaches for characterizing sustainability.

19
Approaches to Characterizing Sustainability
Characterization is a prerequisite to applying the concept of sustainability as a
criterion for identifying constraints, focusing research, and evaluating and improving
agricultural policy and practices. The conceptual problem of defining sustainability and
methodological problems imposed by its temporal nature have hindered development of
approaches to characterizing sustainability. Sustainability involves future outcomes that
cannot be observed in the time-frame required for intervention (Lynam and Herdt, 1989;
Harrington, 1992). For this reason, Conway (1994) argued that defining sustainability in
terms of preservation or duration has little practical value.
The variety of approaches reviewed here reflects the different interpretations of
sustainability and methodological difficulties that result from its temporal nature.
Characterization by adherence to prescribed approaches is based on an interpretation of
sustainability as an approach to agriculture. Characterization by multiple qualitative
indicators and attempts to integrate such indicators are consistent with interpreting
sustainability as an ability to satisfy diverse goals. Sustainability as an ability to continue is
usually characterized by time trends or resilience.
Adherence to prescribed approaches
In a study comparing conventional and sustainable farms in South Dakota, Dobbs
et al. (1991) identified farms as sustainable if they reduced chemical inputs relative to

20
typical farms, and included rotations, legumes, tillage and cover crops for management of
fertility, erosion and weeds. Sustainable farms were sampled by sending questionnaires to
farmers they . . believed might be using greatly reduced or even zero levels of synthetic
chemicals in their farming operations” (p. 111). Cordray et al. (1993) characterized the
sustainability of farmers in Washington and Oregon based on changes in agricultural
chemical use and adoption of alternative production practices. Taylor et al. (1993)
developed a quantitative index of sustainability based on production practices of
Malaysian cabbage farmers. Practices were assigned values according to their “inherent
sustainability” determined by consensus of the research team, weighted by their expected
contribution to sustainability, then combined into a composite index evaluated for each
farm.
Taylor et al. (1993) were the only authors in the above studies to acknowledge the
necessity of assuming a relationship between farmer practices and future viability if
practices are to serve as a basis for characterizing sustainability. Even if evidence
supporting such a relationship were presented, circular logic prevents using sustainability
measured by adoption of particular practices as a criterion for evaluating and improving
agricultural practices.
Multiple qualitative indicators
Torquebiau (1992) used a set of indicators to characterize sustainability of tropical
agroforestry home gardens. Several system attributes that were believed to influence
sustainability were identified related to the resource base, system performance, and effects

21
on other systems. Measurable indicators were identified for each system attribute. A
negative change in an individual indicator indicated unsustainability. Jodha (1990)
developed a set of indicators of unsustainability of mountain agriculture in the Himalayan
region consisting of visible changes in natural resources and farming practices that indicate
system degradation, changes in farming practices that compensate for less visible changes
in the environment, and inappropriate development initiatives that may lead to negative
impacts. Neher (1992) described an approach that is being developed to monitor
agroecosystem health at regional and national scales in the U.S. A number of indicators of
environmental quality and agricultural performance are to be measured to provide a
baseline, then monitored to identify changes.
Monitoring sets of qualitative indicators is consistent with interpreting
sustainability as the ability to meet a diverse set of goals and the belief that no single
indicator can exist (Geng et al., 1990; Norgaard, 1991). However, diverse sets of
indicators are difficult to interpret and do not provide mechanisms for diagnosing causes
of unsustainability, or for evaluating effects of proposed interventions. Such indicators do
not facilitate establishing cause-effect relationships between diverse system properties.
Integrated, quantitative indicators
Increasing recognition of the need for quantification has motivated efforts to
combine diverse indicators ,of sustainability into integrated, quantitative measures. One
goal of the project described by Neher (1992) for monitoring agroecosystem health is to
combine indicators into an aggregate measure of agricultural sustainability in a manner

22
that balances productivity, environmental soundness and socioeconomic viability goals.
Lai (1991) proposed a sustainability coefficient as a function of output per unit of input at
optimal per capita productivity or profit, output per unit of decline in the most limiting or
least renewable resource, and the minimum assured output level. Sands and Podmore
(1993) proposed an environmental sustainability index as an aggregation of sub-indices of
soil productivity, ecosystem stability, and potential to degrade the environment. Selection
of components of the sub-indices and the form of aggregation functions were indicated as
important research topics. Stockle et al. (1994) proposed a framework for evaluating
sustainability based on nine system attributes: profitability, productivity, quality of soil,
water, and air, energy efficiency, fish and wildlife habitat, quality of life, and social
acceptance. Production system sustainability is determined by scoring attributes as
weighted functions of quantifiable, long-term constraints, then combining weighted
attribute scores into an integrated measure.
The consistent inability to specify aggregation functions in these studies points to
the weakness of interpreting sustainability as the ability to fulfill diverse sets of goals as a
conceptual foundation for characterization. Diagnosis is limited by need to decide a-priori
the relative importance of different types of constraints to sustainability. Stockle et al.
(1994) acknowledged and defended the subjectivity needed to aggregate diverse system
attributes into an integrated measure of sustainability.

23
Time trends
Time trend approaches express sustainability in terms of the direction and degree
of measurable changes in system properties through time. Lynam and Herdt (1989)
regarded a system as sustainable if there is a non-negative trend in its output. They
proposed total factor productivity, the total value of system outputs divided by the value
of system inputs, as the output criterion because it accounts for changes in the value of
inputs. Ehui and Spencer (1992) extended total factor productivity to account for changes
in the value of natural resource stocks, particularly soil nutrients. Hedgerow intercropping
in Nigeria was determined to be unsustainable without correction for soil nutrient flows,
but sustainable after accounting for nutrients. Monteith (1990) proposed determining
sustainability from a contingency table of trends of inputs and outputs (Table 2-4). Cereal
production was determined to be sustainable in the Karimnagar district in Andhra Pradesh,
India, based on increasing yields and decreasing land use during 27 years. Decreases in
both land use and yields in the Adilabad district prevented inference about sustainability.
Table 2-4. Contingency table for inferring sustainability based on trends of
system inputs and outputs (Monteith, 1990).
Outputs
decreasing
constant
increasing
decreasing
indeterminant
unsustainable
unsustainable
constant
sustainable
sustainable
unsustainable
increasing
sustainable
sustainable
indeterminant

24
Characterizing sustainability by time trends is appealing because of its simplicity.
The slope of the estimated trend line provides a quantitative index with an intuitive
interpretation as a rate of system deterioration or enhancement. Trends represent an
aggregate response to several determinants of sustainability, eliminating the need to devise
and defend aggregation schemes.
The assumption needed to infer sustainability from trends--that future rates of
system degradation can be approximated by past rates—is often difficult to defend.
Unsustainability can express itself either as a gradual change or as an abrupt collapse
(Conway, 1985; Trenbath et al., 1990). Furthermore, much of the concern about
sustainability comes from recognition that agriculture is being impacted by unprecedented
changes in population pressure, resource demands, market structures, and technology.
Another weakness is the manner in which time-trend approaches interpret temporal
variability. Variability tends to hinder sustainability by driving subsistence farmers to
desperation, leading to environmental degradation that may not recover during normal or
good periods (Mellor, 1988). Price and yield variability have also been shown to increase
the probability of farm failure in the U.S. (Grant et ah, 1984; Perry et al., 1986).
However, when characterization is based on time trends, either variability is ignored or it
implicitly enhances sustainability by reducing the probability of identifying a significant
negative trend.
A final criticism is that applications of time trends to sustainability have examined
levels of system performance without considering the levels of needs and goals of the
individuals or segments of society who decide on the fate of those systems.

25
Resilience
Conway (1985) defined sustainability as resilience: . . the ability of a system to
maintain productivity in spite of a major disturbance” (p. 25). He suggested that
measurement of five system properties are necessary to characterize resilience: inertia,
elasticity, amplitude, hysteresis, and malleability (Conway, 1994). Cramb (1993) based
inferences about the sustainability of two shifting cultivation systems in eastern Malaysia
on both trends and resilience. Pepper production was determined to be sustainable
because production in 1989 recovered to its 1980 level in response to price recovery after
production diminished (in 1985) due to a period of low prices. Rubber production was
considered more sustainable at Batu Linang than at Nanga Tapih, as indicated by recovery
of depressed production in response to price recovery.
Like time trends, resilience can be viewed as an aggregate system response to
determinants of sustainability. However, inability to identify a single measure of resilience
(Conway, 1994) leads to the same problems of interpretation faced when using a diverse
set of indicators to characterize sustainability. Assumptions about the likelihood and
timing of disturbances have been avoided by interpreting sustainability as an intrinsic
property of an agricultural system in isolation from its environment (Conway and Barbier,
1990). However, York (1988) argued that (un)sustainability is not an intrinsic property
but rather a response to changing environmental and socioeconomic circumstances.
Predictions about future sustainability cannot be made in the absence of assumptions about
changes and variability in those higher-level systems that comprise a system’s

26
environment. Resilience shares with time trend approaches the criticism that it ignores the
goals of the human actors within agricultural systems.
System simulation
Simulation has been used to characterize the sustainability of crop production in
response to soil dynamics. Singh and Thornton (1992) illustrated the use of long-term
simulation of crop sequences replicated with stochastic inputs of weather data to examine
trends and variability in yields. Lerohl (1991) used the Erosion-Productivity Impact
Calculator (EPIC) (Williams et al, 1984) to study the long-term impact of predicted soil
erosion on productivity of crop rotations on four soil types in Alberta, Canada.
Sustainability was inferred in all soil-rotation combinations because no negative trend in
crop yields could be detected during a simulated 100 year period.
Other studies have used crop simulation models to examine relationships between
production and environmental degradation. Singh and Thornton (1992) illustrated the use
of CERES-Maize (Jones and Kiniry, 1986) to simulate the effects of soil type and rate of
application of N fertilizer on distributions of maize grain yield and N03" leaching into
groundwater from upland fields in Chiang Mai, Thailand. Alocilja and Ritchie (1993) used
CERES-Maize and a multiple goal optimization technique to identify sets of N fertilization
schedules that were optimal in the sense that neither production nor water quality could be
improved without decreasing satisfaction of the other goal.
Several whole-farm simulation studies have looked at the effect of various factors
on farm survivability. For example, Perry et al. (1986) examined effects of production

27
costs, labor availability, rice grain quality, land tenure, trends and variability of rice and
soybean prices and yields, beginning equity, type of rotations, and participation and terms
of government farm programs on probability of farm survival during a five year period.
Production and environmental processes were not simulated. Although this and similar
studies have not generally used the word “sustainability,” the use of farm survival as a
criterion is consistent with an interpretation of sustainability as economic viability
(Madden, 1987; Lockeretz, 1988; Dicks, 1992;Neher, 1992). Survivability addresses
shortcomings of other approaches by integrating levels, trends and variability in system
performance with the needs and goals of farmers.
System simulation is a tool, and does not suggest a particular criterion for
evaluating sustainability. Simulation can be used to examine future impacts of alternative
interventions across the range of expected variability in a manner that is not possible with
empirical observation and experimentation. The value of simulation is limited by
capabilities of, and confidence in simulation models, by availability and reliability of input
data, and by lack of methods for designing and interpreting simulation studies for
characterizing sustainability. So far, there has been little integration of models of crop and
animal production, environmental degradation, economic processes, and farmer decisions.
Elements of a Useful Approach for Characterizing Sustainability
In order for sustainability to be a useful criterion for guiding change in agriculture,
several elements should be incorporated into approaches to its characterization (Table 2-

28
5). First, characterization should be based on a literal interpretation of sustainability (Fox,
1991). Regardless of the merits of goals and ideals frequently incorporated into
definitions of sustainability, if the idea of continuation through time is omitted then those
ideals and goals are something other than sustainability.
Table 2-5. Elements of a useful approach to characterizing sustainability of agricultural
systems.
Element
Explanation
Literal
Defines sustainability as an ability to continue through time, consistent
with literal English usage.
System-
oriented
Identifies sustainability as an objective property of a particular
agricultural system whose components, boundaries, and context in
hierarchy are clearly specified.
Quantitative
Treats sustainability as a continuous quantity, permitting comparisons of
alternative systems or approaches.
Predictive
Deals with the future rather than the past or present.
Stochastic
Treats variability as a determinant of sustainability and a component of
predictions.
Diagnostic
Uses an integrated measure of sustainability to identify and prioritize
constraints.
Second, characterization should be system-oriented. A literal interpretation
suggests that sustainability is an objective property of an agricultural system. It cannot be
a property of approaches to agriculture if it is to serve as a basis for evaluating and
improving approaches. Lynam and Herdt (1989) argued that sustainability is a relevant
criterion for evaluating technology only when the system is clearly specified, including its
boundaries, components, and context in hierarchy. Sustainability has meaning only in the
context of specific temporal and spatial scales. Fresco and Kroonenberg (1992) cited a

29
number of examples in which disturbances that threaten sustainability at one spatial and
temporal scale could be seen as natural cycles at broader scales. Both constraints to
sustainability and factors that can be managed for its enhancement depend on the level of
the system (Spencer and Swift, 1992). The objectivity that results from a system-oriented
approach is essential for guiding change, but may work against motivating change because
it may call prescribed approaches into question.
Third, an approach to characterizing sustainability should be quantitative.
Although MacRae et al. (1989) cited quantification as a barrier to sustainability, others see
it as a prerequisite to using sustainability as a criterion for evaluating and improving
agricultural systems (Monteith, 1990; Harrington, 1992). Sustainability is often treated as
a discrete property; “A farm is either sustainable or it’s not sustainable. Simply by
definition, you cannot create a system that is half sustainable” (Rodale, 1990, p. 273).
However, comparisons among agricultural systems or alternative approaches are possible
only when sustainability is treated as a continuous quantity.
Fourth, since sustainability deals with future changes, its characterization must be
predictive of the future rather than merely descriptive of the past or present (Harrington,
1992). Sustainability has little meaning after the fact. The deterministic view that . . a
farm will either last for a very long period, or it won’t” (Rodale, 1990, p. 273) does not
take into account the uncertainty of predictions resulting from the inherent variability of
the farming system’s environment. A stochastic approach, the fifth element, recognizes
variability as a determinant of sustainability and appropriately expresses predictions in
terms of probabilities.

30
Finally, characterization of sustainability should be diagnostic. Sustainability is a
useful concept when its characterization focuses research and intervention by identifying
and prioritizing constraints. Diagnosis can be accomplished by testing hypotheses about
constraints based on a measure of sustainability that is both comprehensive and integrated.
Diagnosis is facilitated by use of a single measure of sustainability that combines the range
of possible determinants into a single, integrated measure of system response. An
integrated measure is necessary for comparing, for example, the relative impact of nitrate
leaching into aquifers and product price volatility on sustainability.
Weaknesses of the reviewed approaches for characterizing sustainability can be
related to their failure to incorporate the proposed elements (Table 2-6). Characterization
based on adherence to prescribed approaches fails because it is not founded on a literal
interpretation of sustainability. Lack of integration limits the usefulness of multiple
indicators of sustainability for diagnosing and prioritizing constraints. Integration of
indicators has been difficult because the underlying interpretation of sustainability as an
ability to meet diverse goals is not integrated. A time trend represents an integrated
system response that is potentially useful for diagnosis and can be predictive by
extrapolation, but it is not stochastic in the sense of accounting for variability. An
integrated measure of resilience has not yet been found. The assumptions about future
variability and disturbances necessary for resilience to be predictive and stochastic are
avoided in discussions of its use for characterizing sustainability. Simulated farm
survivability is the only approach reviewed that incorporates all of the elements listed.

Table 2-6. Approaches to characterizing agricultural sustainability. A check (/) indicates the approach incorporated the specified
element. A question mark (?) indicates it addressed some aspect of the element.
. , „ - t u « System- Qaanti- Predic- Sto- Diag-
Approach Reference Literal *, , , , ,a . ^ j
r oriented tative tlve chastic nostic
Reduced use of chemicals relative to other farmers in South Dakota
Dobbs ero/., 1991
Quantitative index of cabbage farmer practices in Malaysia
Taylor eto/., 1993
?
Adoption of alternative practices and reduced chemical use on U.S. Pacific Northwest farms
Cordray ero/., 1993
/
Indicators of resource base, system performance &extemal effects in agroforestry gardens
Torquebiau, 1992
✓
/
Indicators of visible, masked, & potential degradation in Hymalaian fanning systems
Jodha, 1990
/
?
?
Indicators of regional "agroecosystem health"
Neher, 1992
?
Index of assured output, and output per unit input and per unit limiting resource decline
Lai, 1991
/
✓
Index of productivity, stability and degradivity
Sands & Podmore, 1993
✓
✓
Subjectively weighted index of constraints to nine system attributes
Stockle et al., 1994
✓
✓
Non-negative time trend in output
Lynam & Herdt, 1989
/
✓
✓
?
Total factor productivity accounting for natu-ral resources in cropping systems in Nigeria
Ehui & Spencer, 1992
/
✓
/
Regional trends in inputs and outputs in India
Monteith, 1990
✓
/
/
?
Properties representing resilience
Conway, 1994
/
✓
?
?
Production trends and resilience in shifting cultivation in Malaysia
Cramb, 1993
/
✓
?
Simulated trends and variability in crop production and NO, leaching
Singh & Thornton, 1992
/
/
/
/
✓
Simulated crop production trends in response to simulated soil erosion in Canada
Lerohl, 1991
/
✓
✓
/
Simulated survivability of Texas rice farms
Perry ero/., 1985
/
✓
/
/
✓
✓

32
Conclusions
The importance and desirability of agricultural sustainability are generally
recognized. However, its potential as a criterion for guiding agriculture’s response to
change has not been realized. Characterization is a prerequisite to using sustainability as a
basis for guiding change. Logical inconsistencies limit the usefulness of characterization
of sustainability interpreted as an ideological or management approach to agriculture.
Interpreting sustainability as an ability to meet diverse goals suggests measuring sets of
system indicators consistent with those goals. However, these measurements have proven
difficult to integrate and interpret in a way that identifies constraints or focuses research.
Literal interpretations of sustainability as an ability to continue into the future suggest
measurable, integrated criteria for its characterization. However, applications of these
criteria-time trends and resilience--have ignored or misinterpreted important aspects of
system behavior. Criteria are needed that relate levels, trends and variability of long-term
system performance to the needs and goals of farmers and of society.
In order for sustainability to be a useful criterion for guiding change in agriculture,
its characterization should be literal, system-oriented, quantitative, predictive, stochastic
and diagnostic. These elements identify weaknesses in existing and proposed approaches,
suggest directions for future development of approaches, and together constitute a
systems approach for characterizing sustainability of agricultural systems. The tools of
system analysis and simulation must be part of approaches that incorporate these elements.

CHAPTER 3
A SYSTEMS FRAMEWORK FOR CHARACTERIZING FARM SUSTAINABILITY
Introduction
The potential benefits of applying the concept of agricultural sustainability-
providing feedback about future impacts of current decisions, and focusing research and
intervention by identifying constraints—can only be realized when sustainability of
particular systems is characterized. Characterization includes both quantification and
diagnosis of constraints.
In spite of the tremendous amount of concern about agricultural sustainability,
surprisingly few studies have attempted to characterize the sustainability of specific
agricultural systems. The methods which have been proposed or applied suffer from (a)
conceptual problems associated with interpreting sustainability as an approach rather than
a property of agriculture, and (b) practical difficulties that arise from the fact that
sustainability deals with the future (Chapter 2). Characterization based on management
practices either does not relate to a literal interpretation of sustainability or it leads to
circular logic. Attempts to characterize sustainability based on system response have
generally ignored or misinterpreted important system properties.
33

34
Since sustainability deals with the future, it cannot be readily observed. Analysis
and simulation of a model system can compensate for the limitations of observation and
experimentation on a real agricultural system. In this chapter, I first present a definition of
sustainability that applies generally to dynamic, hierarchical, stochastic, purposeful
systems. I then describe a framework for using system simulation to quantify sustainability
and test hypotheses about its constraints. A Monte-Carlo study of a simple time-series
model demonstrates how sustainability relates important components of system behavior-
mean, trend, variability and autocorrelation—to threshold goal levels. I then discuss issues
that arise when applying the framework to farming systems. Finally, I illustrate an
application of the framework for characterizing farm sustainability using data from a
previously published farm simulation study.
Defining Sustainability
To sustain is literally “to keep in existence; keep up; maintain or prolong”
(Neufeldt, 1988, p. 1349). Sustainability can therefore be defined as the ability of a
system to continue into the future. Key words of this definition suggest a framework for
quantification.
First, the system addresses the question, “What is to be sustained?” Although the
system of interest in this paper is a farm, the concept of sustainability can be applied to any
system that is dynamic, stochastic, and purposeful. The ability to continue does not apply
to a static system. Furthermore, sustainability has no meaning unless some purpose or

35
threshold condition exists which distinguishes a system that is sustaining from one that has
failed. Finally, sustainability of a deterministic system is binary; in the absence of
uncertainty one can say that a system will either sustain itself or fail during some future
period.
Second, the word continue implies the possibility that a system can fail if some
criteria are met. Failure occurs when a system can no longer fulfill its purpose. Failure
implies irreversibility: a degree of stress from which the system cannot readily recover.
Criteria for failure address the question, “Above what minimum level is the system to be
sustained?”
Third, the future suggests a time period which extends from the present (t = 0) to
some future time (t = 7). For a stochastic system, the future also implies uncertainty;
uncertainty distinguishes the future from the past. The time period addresses the question,
“How long is the system to be sustained?”
Finally, the ability of the system to continue in the future is best expressed as a
probability. The suggestion that “A farm is either sustainable or it’s not sustainable ... A
farm will either last for a very long period or it won’t” (Rodale, 1990, p. 273) expresses a
common deterministic interpretation of sustainability. However, one cannot determine
with certainty whether a system will continue through some future period. Probability of
continuation provides a measure of sustainability with a zero-to-one range that addresses
the question, “With what degree of certainty will the system sustain itself?” The definition
of sustainability can be restated as the probability that a particular system will not meet
specified criteria for failure during a particular future period.

36
Quantifying Sustainability
Consider the status of a system, D(/), as a Bernoulli process with state space {0,
1} operating in the period from the present (t = 0) to some future time (/ = T) (See symbol
definitions in Table 3-1). Continuation is indicated by D(t) = 1 and failure by D(t) = 0.
The system is initially operating: D(0) = 1. Time of failure, TF, is then a random variable
with a probability density function, ^(0 = 'P{TF= /}, and a cumulative distribution, Ftf(0
= ¥{Tf <, t). The distribution Fjf applies to the population of possible time paths, or
realizations, of system behavior. Failure is irreversible such that if D(t) = 0 then D(t+At) =
0 for all At > 0. For the period (0, 7], sustainability, S, is defined as,
S(T) = 1 - Ftf(7). [3-1]
The definition given in Eq. [3-1] is equivalent to the survival function in mortality studies
(eg., Elandt-Johnson and Johnson, 1980) and to reliability in quality control literature (eg.,
Barlow, Proschan and Hunter, 1965).
Failure as violation of state thresholds
Time to failure, TF, is a random variable because the system’s state, x(/), behaves
as a stochastic process as it responds to a stochastic environment. Consider a single,
continuous state variable, x(t). At any given time t, x(t) has a probability density function,
fXjt(x), and a cumulative distribution, FX t(x) that apply to an initial population selected at

37
Table 3-1. Description of symbols used.
Symbol Description
l, T Time variable and a particular time
D{t) Status of a system (1 continuing, 0=failed)
Tf Time to system failure
fir, F-j-p Density and cumulative distribution of Tv
S(T) Sustainability for the period (0, 7]
h(t) Sustainability hazard probability function
x, x State vector and a particular state variable
x0, x0 Failure threshold values for x and x
fx„ Fx>t Density and cumulative distribution of x at time 1
N Total number of realizations simulated
n(T) Number of realizations continuing at time T
§ S estimated from a finite number of realizations
SEs Standard error of S
z(t), oz A stationary stochastic process and its standard deviation
cl, P Intercept and slope of a deterministic trend
4»! One period lag autocorrelation coefficient
e(/), oe A white-noise process and its standard deviation
n». n Base and alternate value of the /th hypothesized determinant of sustainability
4, S in response to the base and alternate value of the /th factor
Ri, r, Absolute and relative sensitivity to the /th factor
vab, N Frequency of a and b occurring, and sum of all frequencies
q Correction factor for continuity in G-tests
Gb Gladj Uncorrected and corrected statistic for tests of independent frequencies,
independent observations
GP, GP adj Uncorrected and corrected statistic for tests of independent frequencies,
paired observations

38
time 0. Assume that a threshold value, x0, exists such that the system will fail the first time
x ^ x0. If x0 represents a maximum threshold, a sign change is required to allow the
relationship to hold. We can express the relationship between x(t) and TF by considering
the probability that failure occurs during an interval (/, /+A/], At > 0:
P{/ t) P{TF > /}, [3.2]
and, since the condition TF < t+At is equivalent to x(/+A/) < x0, we can rewrite Eq. [3-2]
as,
P{/x0} (1-P{rF As At approaches 0, the left side of Eq. [3-3] becomes the probability density of time to
failure:
A'â„¢0P{rsrrs^A() = T{T,= I)
= fTF (0- [3-4]
Since Fx t is the distribution of x for the population of realizations that have continued to
time t, FX t(x0) expresses the probability that the system violates the threshold x0 at t, given
that it has not done so previously. Therefore, taking the limit of the first term on the right
side of Eq. [3-3] gives,
A/?0 PW/ + A/) - xo I x(0 > *o) = Fx,(*o)-
[3-5]

39
Finally, the last term in Eq. [3-3] is equivalent to sustainability from Eq. [3-1]. By
taking the limit of Eq. [3-3] as A t approaches 0 and substituting the simplified terms (Eq.
[3-4] and [3-5]), we obtain,
fTF(0 = Fx,,(*0) 5(0- [3-6]
We can now derive an expression of S as a function of only FX i. Differentiating Eq. [3-1]
gives,
dS(t)/dt = -fTF(0- [3-7]
Substituting Eq. [3-6] into [3-7] gives the differential equation,
dS(t)/dt = -Fxt(x0) S(t), [3-8]
which has the solution,
S(T) = exp | " f°Fx,,(x0) d/j [3-9]
at time T. Thus, I have shown that sustainability is determined entirely by the probability
that a system’s state falls below a threshold value during some time interval (0, 7].
Two examples illustrate how time, thresholds, and the distribution of system state
interact to determine sustainability. In the first example (Fig. 3-la), the variability ofx is
relatively high but its expected value, E[x(t)], remains constant. The system has a 0.15
probability of failing in any particular period. Sustainability declines exponentially with

period, t
Figure 3-1. Relationship between time, distribution of state, and sustainability under (a) constant mean and high variability,
and (b) negative trend and low variability.
-U
o

41
time as Eq. [3-9] predicts. In the second example (Fig. 3-lb), variability is lower, but
E[x(/)] decreases linearly with time. Here, the effect of the negative trend in x on
sustainability is not apparent until E[x(/)] approaches x0.
Several generalizations can be applied to the preceding discussion. First, a failure
threshold may exist for more than one system state variable. Second, the thresholds may
be dynamic. The state of a complex system and the corresponding set of minimum
threshold values could then be expressed as vector processes: x(/) and x0(t). The
threshold vector Xo(/) bounds the system’s state space in one or more dimension. Since
only a few of the state variables may be directly related to the ability of a system to
continue, many members of x0(/) may have values of -°° for all t. Third, the threshold
vector could be stochastic. The probability of violating system thresholds would then
depend not only on the behavior of x(t) but also on the dispersion of x0(/) and its
correlation with x(/). Finally, thresholds may apply to derived state variables such as a
sum (eg., aggregate wealth) or ratio (eg., a financial ratio) of basic state variables.
Sustainability hazard
The value, Fx t(x0), is the instantaneous probability of failure applied to the
population of realizations that have continued up to time t. It is referred to as a hazard
function,
Ki) =fTf(t)/(\ - FTF(/))
= fTf(t)/S(t)
[3-10]

42
(Barlow et al., 1965). Hazard is a probabilistic expression of the intensity of stress on a
system as a function of time; increasing h indicates an increasing stress or increasing threat
to sustainability. By rearranging and substituting Eq. [3-10] into [3-7], and integrating,
we arrive at,
[3-11]
We see that constant hazard results in exponentially declining sustainability. Although S(t)
increases monotonically from an arbitrary starting time (/ = 0), the threat to sustainability
expressed by h{t) is independent of starting time, and may increase or decrease.
Simulating Sustainability
Equations [3-1] and [3-9] cannot be used to calculate sustainability in practice
because the distributions of TF and x(?) are generally not known for a real system.
Harrington et al. (1990) expressed the need to artificially construct time paths for current
and alternative strategies to assess sustainability before long-term experiments or
monitoring could be completed. A set of replicated time paths, or realizations, can be
sampled by stochastic simulation of a model of the system.
Consider a set of N realizations simulated for the period (0, 7] (Fig 3-2). Although
the environment is sampled randomly, each replicate has the same initial conditions, x(0).

sustainability, S(t) state, x(t)
43
Figure 3-2. Estimating sustainability by sampling a small number (i.e. 5) of simulated
realizations of future system behavior.

44
Let n(í) be the number of realizations continuing at time t. Sustainability can then be
estimated by the relative frequency of surviving realizations, or
${T) = /?(7) /N. [3-12]
Because of the uncertainty associated with estimating the probability of success from a
small Bernoulli trial, sustainability estimated from Eq. [3-12] has a standard error
(Snedecor and Cochran, 1980) of
SE¿ = JS(T)(1 - S(T)) / N
= sJn{T)(N-n{T))IN\ [3-13]
An example: sustainability of a simple time-series model
The proposed definition (Eq. [3-1]) integrates several important aspects of system
behavior—means, trends, variability and autocorrelation—and relates them to levels of
goals expressed as system thresholds. I used Monte Carlo simulation of a simple, discrete¬
time, univariate time series model to demonstrate the role of these aspects of system
behavior in determining sustainability. The model consisted of a deterministic trend
component,
x(i) = a + p/ + s(t),

45
and a first-order autoregressive component,
= 1 + 8(0,
where a and p are intercepts and slope of a time trend, z(t) is the value of a stationary time
series in period t, $1 is a first-order autoregressive coefficient, |<{>i | < 1, and ~ N(0, oE) is
a random shock variable. The standard deviation of the random shock (oc) is related to
the asymptotic standard deviation of the generated time series (oz) by,
(Pankratz, 1983). I calculated *5by Eq. [3-12] from 10,000 replicated realizations, each
simulated for 120 periods.
Table 3-2 summarizes values of the parameters and the results of the sensitivity
analysis. Sustainability, ^120), was 0.567 ± 0.005 (± S.E.) in response to the base
parameters. Decreasing the mean, increasing the goal threshold, or increasing variability
reduced *5(120) by increasing the proportion of the population which fell below Xq at any
time before t=T. Decreasing autocorrelation reduced S( 120) because of its effect on Fxv
For discrete time periods, FXt in any period t is conditioned on the prior status of the
system and, therefore, on the previous value of the state variable. Finally, extending the
time period Tbeyond 120 decreased S(T).
Figure 3-3 shows how declining expected value, and abrupt changes of expected
value, variance and autocorrelation influence the sustainability and hazard time functions

46
of the time-series model described above. The general pattern of the sustainability time
function remains the same whether hazard is modified by a change in the expected value,
variance or autocorrelation (Fig. 3-3c-h).
Table 3-2. Sensitivity of simulated sustainability to system properties. 61(120)=0.567 for
the base scenario.
Para-
Base
Increased
Decreased
Property
meter
value
value
S{T)
value
%T)
mean
a
10.0
11.0
0.783
9.0
0.307
trend
a
P
10.0
0.0
9.0
+0.0167
0.566
11.0
-0.0167
0.517
variability
5.0
5.5
0.401
4.5
0.774
autocorrelation
4>i
0.8
0.88
0.924
0.72
0.235
goal threshold
*0
3.0
3.3
0.491
2.7
0.641
duration
T
120
132
0.535
108
0.604
Diagnosing Constraints to Sustainability
The potential value of the concept of sustainability lies in its ability to focus
research and intervention by identifying and ranking its constraints. Diagnosing
constraints entails a process of hypothesis formulation and testing using simulation of the
system model. Hypotheses should identify the current (or expected) value of a suspected
constraint, and a specific change that would relax the constraint. Sensitivity analysis then
provides the experimental tool for testing and ranking hypothesized constraints.

47
Figure 3-3. Sustainability and hazard of an AR(1) process with (a) constant parameters,
(b) declining expected value, (c) abrupt decrease and (d) increase in mean, (e) decrease
and (f) increase in variance, and (g) decrease and (h) increase in autocorrelation.
Parameter values were a = 10.0, p = 0.0, oz = 8.0, 4>i = 0.9, x0 = 4.0 and n = 50,000
except as indicated otherwise.
Hazard, h(t) (thousanths) Hazard, h(t) (thousanths) Hazard, h(t) (thousanths) Hazard, h(t) (thousanths)

48
Sensitivity analysis
Sensitivity analysis is used to quantify the relative importance of hypothesized
constraints to sustainability. It involves changing the value of a factor a small amount in
the direction that would relax the hypothesized constraint relative to a base scenario which
represents existing or expected conditions, then simulating the modified scenario.
Hypothesized constraints can then be ranked based on either absolute,
[3-14]
or relative sensitivity,
r.
Y
i,0
S.-Sn
Y.-Y:
i.O I
[3-15]
where Yjfi is the value of the ;'th factor in the base scenario, Y¡ is its adjusted value, and S0
and £¡ are sustainability values estimated for the base and alternate scenarios. The
absolute value allows an increase in sustainability to result in r¡ > 0 regardless of the
direction of change in Yv Relative sensitivity is interpreted as the percent change in S in
response to a 1% change in Y. Comparisons may be made and ranks assigned among
discrete or among continuous factors. However, absolute sensitivity to discrete factors
cannot be compared with relative sensitivity to continuous factors.

49
Significance tests
Independent observations. The frequencies (Vy) of failure and continuation in the
simulation of the base and alternate scenarios used for a hypothesis test can be represented
by a two-way contingency table arranged as in Table 3-3, where i is the row and j is the
column. Several tests are available for the null hypothesis that frequencies are
independent, equivalent to the null hypothesis that sustainability is the same in the two
scenarios. Sokal and Rohlf (1981) recommended a log likelihood ratio, or G-test, over
the more frequently used x2 test. One can calculate the expected frequencies of
continuation or failure from the observed frequencies based on the null hypothesis of
independence:
= (Vil+V¡2HVlj + V2j)/N-
The Gj statistic is then calculated from the observed and expected frequencies:
G, = 2 ¿ ¿(V¡¡ ln(vs/iL)), [3-16]
1 = 1 /= 1
and is corrected for continuity (Williams, 1976):
q = 1 + ((N/Vj, +N/v2t - l)(N/v,j + N/vt2
1))/6N,
[3-17]
Gi,adj = G/q.
[3-18]

50
The G statistic is distributed approximately %2 with one degree of freedom. If GUdj > x\i
then reject Ho: S0(T) = SIT).
Table 3-3. 2x2 frequency table for tests of difference between simulated
sustainabilities, independent observations.'
Scenario
continued
failed
Total
base
Vn
Vi* = Vn+V12
alternate
V21
V22
V2*= v21+v22
Total:
v*i=vn+v21
v*2=v12+v22
N - Evij
Paired observations. Different scenarios can be replicated under the same set of
environments by using the same pseudorandom number sequence for sampling stochastic
inputs. The additional information available from such a randomized block design permits
the use of a more powerful test. When data are arranged as in Table 3-4, the row or
column totals then represent the frequency of failure or continuation of the base (row
totals) or alternate (column totals) scenarios. Then vu is the number of replicated
environments in which both scenarios failed, v12 is the number in which the base scenario
failed but the alternate scenario continued, and so forth. The McNemar (1947) test as
adapted by Sokal and Rohlf (1981) uses a G statistic calculated as,

51
Gp = 2
V12 ln
2 v
12
V12 + V21
V21 ln
2 v.
21
V V12 + V21/
[3-19]
with the correction factor,
q = 1 + 1/2N,
[3-20]
applied as in Eq. [3-18] (Williams, 1976). Again, GP adj is compared to x2 with one degree
of freedom. The McNemar test statistic will be undefined if any of the frequencies (Table
3-3) have a value of zero. This can easily occur if, for example, none of the replicates that
continue in a base scenario fail in an alternate scenario (v12 = 0).
Table 3-4. 2x2 frequency table for tests of difference between simulated
sustainabilities, paired observations.
Base
--Alternate scenario—
scenario
continued
failed
Total
continued
Vn
v12
Vi*=V„+V12
failed
%
V22
V2* = V21+V22
Total:
V*1 = V„+V21
V*2 = V,2+V22
N = Zvii
It is important to keep in mind that statistical inferences based on simulated
sustainability apply to the model system and its environment. Extension to the actual
system depends on the validity of the model and assumptions about the future
environment.

52
Sustainability Applied to Farming Systems
Although sustainability is an important concern at several levels in the hierarchy of
agricultural systems (Lowrance etal., 1986; Lynam & Herdt, 1989), it is particularly
relevant at the farm level. If agriculture is to meet the needs of society—providing food
and other products while protecting natural resources—it must first meet the needs of the
farmers who implement and manage it.
The framework for characterizing sustainability applies to farming systems because
a farm is a dynamic, stochastic and purposeful system. Furthermore, sustainability is
characterized most easily at the farm level where system goals are more easily specified
and more consistent than at other system levels. For example, human goals are not
intrinsic to fields or enterprises. On the other hand, the emergence of many human actors
at levels higher than the farm leads to multiple and often conflicting goals. The continuing
and failed status would comprise fuzzy, nonexclusive sets at these higher system levels.
Sustainability expressed as a probability of continuation (Eq. [3-1]) must be
applied to some initial population. One could think of a population of farms and attempt
to predict the proportion that will survive through a future period. However, the
appropriate population to consider when characterizing farm sustainability is the
population of possible realizations of future behavior of an individual farm.
A farm’s context within hierarchy has implications for selecting an appropriate
time frame for sustainability analysis, identifying failure criteria, making assumptions about

53
the future behavior of system inputs, and hypothesizing constraints. The remainder of this
section examines these issues.
Selecting a time frame
Sustainability has meaning only in the context of a specific time frame. For
example, climate change, soil erosion, or extinction may be seen as irreversible threats to
the sustainability of an ecosystem in a time frame of decades or centuries, but as part of
natural cycles in a time frame of millennia or longer (Fresco and Kroonenberg, 1992).
From another perspective, sustainability can be viewed as a non-increasing function of
time. Considering the extreme cases, all existing agricultural systems can sustain
themselves for an arbitrarily short period. On the other hand, few agricultural systems can
be expected to continue in a recognizable form for tens of millennia. If F^t) is a true
probability distribution with a lower bound at t=0 then
lim
/- oo
Ftf(/) = 1
Although selecting a time period for analyzing farm sustainability is a subjective
decision, considerations of hierarchy, relevance, and realism suggest a range of about 10
to 15 years. Ecological hierarchy theory states that processes in higher-level systems
operate more slowly than in lower-level systems (Allen and Starr, 1988). The time frame
for analyzing sustainability of a farming system should therefore be longer than the several
months to a few years that are typical of crop and animal production cycles. Relevance

54
suggests that the time interval should be long enough to allow detection of important
threats to sustainability. For example, three or four years would reveal little about the
impact of soil erosion on sustainability. On the other hand, a study exceeding a century
would not be relevant to the livelihood goals of an individual farm. Lynam and Herdt
(1989) suggested that five to 20 years is a relevant time-frame for analysis of farming
system sustainability. Realism of assumptions about economic, policy, and technological
inputs to the farming system becomes increasingly difficult to defend past about 10 to 15
years.
Assumptions about inputs
The higher-level systems that comprise a farming system’s environment exert
control through inputs and through constraints to farmer decisions or farm outputs.
Analysis of farm sustainability requires assumptions about future behavior of inputs and
control mechanisms that are conceptually external to the farming system. The issues to
consider include (a) whether systematic trends or cycles are expected, (b) whether
variability is sufficient to warrant stochastic sampling, and (c) whether a feedback
mechanism exists which allows an input to respond to farm outputs. Although some farm
processes may be inherently stochastic, inputs of weather and prices are usually the major
sources of risk in a farming system. Catastrophic events such as disease epidemics, storms
or wars also represent important stochastic inputs to some farming systems.

55
Farm failure criteria
Failure criteria denote the minimum level of performance above which a system is
to be sustained. Since farmer livelihood is the primary purpose of most farming systems,
criteria for farm failure can be expressed in terms of minimum levels of livelihood goals.
Hamblin (1992) suggested that agriculture fails to sustain if production falls below the
levels necessary for profitability in a cash economy or survival in a subsistence economy.
In a subsistence economy, a level of poverty or malnutrition from which a farm family
cannot escape without outside intervention might indicate system failure. Lynam and
Herdt (1989) referred to famine as “the ultimate indicator of unsustainability” (p. 391). In
a cash economy, lenders may impose threshold leverage ratios above which they will force
foreclosure by recalling loans (Perry ei al., 1985). Failure could be expressed in several
forms such as farm abandonment, conversion of land to non-agricultural use, the need to
supplement income with off-farm employment, inability to meet critical goals such as
education of children, or major changes in farm enterprises, depending on the analyst’s
purpose.
Negative feedback loops between components of a system tend to counteract the
effects of disturbances and stabilize a system, resulting in a stable state or attractor.
Mathematical and empirical evidence suggests that ecosystems can possess multiple stable
states (May, 1977). The region about a stable state in state space is a domain of
attraction, and the boundary between adjacent domains of attraction is a separatix
(Trenbath ei al., 1990). If the state of a system is displaced across a separatix, it enters a

56
new domain of attraction. The ability to return to the original domain of attraction
depends on the relative stability of the two domains, the nature of the separatix between
them, and the existence of disturbances which could displace the system back across the
separatix. Failure of an agricultural system can be viewed as transition from a useful to a
less useful domain of attraction. The state threshold vector x0 forms the separatix between
the domains. Trenbath et al. (1990) used mathematical models to illustrate abrupt
transitions from useful to less useful domains in response to intensification of three
agricultural systems.
In many cases, farm failure criteria may be difficult to determine. However, it may
be possible to obtain meaningful insights into the relative impact of various stresses on a
farming system by assuming particular failure thresholds when those thresholds cannot be
measured.
Since increasingly restrictive failure thresholds increase the probability of system
failure, sustainability may be viewed as a non-increasing function of threshold levels. A
system is less able to continue at a high level than at a low level.
Determinants of farm sustainability
Sustainability is an aggregate response of a system to a range of external factors,
conditioned by internal characteristics of the system. Any factor that influences means,
trends, variability, autocorrelation or goal levels may influence sustainability. A host of
factors influences the balance between income and expenditure that determines the mean
level of farm wealth. Soil degradation, depletion of scarce resources, technological

57
innovation and trends in prices can affect trends in farm state variables. Variability is
influenced by weather patterns, price volatility and the occurrence of catastrophic events.
Credit availability, market access, and the ability to store agricultural products have
positive effects on autocorrelation of farm wealth. Finally, a household’s tolerance to
difficulty, alternative sources of livelihood and lenders’ policies can influence goal
thresholds.
Adaptive management generally serves to improve sustainability. Farmers employ
a range of management strategies, such as selling capital assets, reducing input use,
working off-farm, or shifting from cash to subsistence crops, to reduce the risk of failure
during difficult times. Although the proposed framework for characterizing sustainability
can account for adaptive strategies, the farmer decision process may be more difficult to
simulate than biological or economic processes. Simulating a fixed management strategy
may greatly overestimate the probability of farm failure.
Much of the concern about sustainability of agricultural production systems relates
to externalities: non-target outputs with costs (or benefits) which are not bom by an
individual farm but by society. A hierarchical perspective suggests that externalities
should have no direct impact on farm sustainability; the impact of an externality emerges
at a higher (eg., regional) system level. However, a feedback mechanism that either
charges the farming system for the externality or constrains the practice that produces it
can easily be incorporated into a simulation analysis. Such a deviation from a strict
hierarchical approach might be justified by assuming that society will attempt to correct
the perceived injustice of a negative externality.

58
A similar issue arises when extending an analysis from a single farm to a group of
similar farms. A group of farmers could adversely affect common grazing land, water
resources, fuelwood, fishing areas, or forest by overuse, whereas a single farmer’s impact
might be negligible. Similarly, although an individual farmer is usually assumed to be a
price taker, a group of farmers may alter prices due to their aggregate effect on supply of
a product or demand for an input such as seasonal labor. Incorporating these effects
involves extending the analysis to include some processes (eg., market equilibria) above
the farming system level.
An Example: Sustainability of a Coastal Texas Rice Farm
Whole-farm simulation studies have examined the influence of factors such as
commodity price variability (Grant et al, 1984), farm size and beginning equity level
(Richardson and Condra, 1981), intergenerational estate transfer strategy (Walker et al.,
1979), and land tenure expansion strategy (Held and Helmers, 1981) on probability of
farm survival through various periods. Perry et al. (1986) conducted a more
comprehensive simulation study that examined the impacts of crop rotation, land tenure
arrangement, government programs, costs, labor availability, lenders’ policies, interest
rates, and the level and variability of crop yields and prices on rice farms in Texas. A
reinterpretation of the results of this study illustrates the use of long-term, stochastic
simulation to characterize farm sustainability.

59
Methods
Perry et al. studied a “representative” coastal Texas rice farm rather than an actual
farm. They examined four combinations of rotation and sharecropping arrangement.
Assumptions and initial conditions can be found in Perry et al. (1986).
Perry et al. used a modified version of the FLIPSIM (Firm Level Income Tax and
Farm Policy Simulator) simulation model (Richardson & Nixon, 1985) called RICESIM to
simulate the model farm. RICESIM is primarily a farm accounting model which randomly
samples from probability distributions as a proxy for the biological and ecological
processes involved in crop production.
The study examined two alternate criteria for system failure. First, probability of
survival was based on a threshold leverage ratio (total debt/total equity). Lenders were
assumed to force foreclosure by recalling loans if the leverage ratio exceeded 2.0. The
second criterion, a negative net present value (NPV) of future cash flow, requires a higher
level of minimum system performance to avoid failure. Failure indicated by a negative
NPV means that a secure, non-farm investment would be more profitable than farming.
Probabilities of survival and positive NPV were calculated by Eq. [3-12] from 50
replicates of each five-year scenario.
Perry et al. analyzed sensitivity of the probabilities of survival and positive NPV to
several factors. For our analysis, I selected only the soybean-soybean-rice rotation with
1/2 share of rice and 1/7 share of soybean going to the landowner. I tested those factors
included in the sensitivity analysis that I believed could constrain sustainability (Tables 3-5

60
and 3-6). To test for differences in sustainabilities, I used a G-test that is based on
independent sampling (Eq. [3-16] to [3-18]) since the study did not provide paired data or
indicate the experimental design.
Results
The values of £at the end of the five-year scenario presented here were 0.50 ±
0.071 based on leverage ratio and 0.12 ± 0.046 (± S.E.) based on NPV. Figure 3-4 shows
Ü(t) of all four combinations of crop rotation and tenure arrangement based on the
threshold leverage ratio.
Table 3-5. Relative sensitivity, r (Eq. [3-15]), of simulated five-year sustainability, ^5), of
a Texas rice farm to continuous factors. For the base scenario, 5(5) was 0.50 based on
threshold leverage and 0.12 based on threshold NPV,
Factor
Base
value
Alternate
value
- Leverage < 2.0 -
S(5) r
—- NPV > 0.0 —-
S(5) r
variable costst
100%
90%
0.82
6.40 **
0.48
30.00 **
crop share
50%
45%
0.80
6.00 **
0.46
28.33 **
crop pricest
100%
110%
0.68
3.60 n.s.
0.38
21.67 **
mean rice yield1
100%
110%
0.60
2.00 n.s.
0.26
11.67 n.s.
mean soybean yield*
100%
110%
0.58
1.60 n.s.
0.26
11.67 n.s.
ratoon red rice
25%
15%
0.80
1.50 n.s.
0.32
4.17 n.s.
rice yield variance*
100%
75%
0.54
0.32 n.s.
0.12
0.00 n.s.
soybean yield variance*
100%
75%
0.52
0.16 n.s.
0.10
-0.67 n.s.
*Percent of base value.

61
Table 3-6. Absolute sensitivity, R (Eq. [3-14]), of simulated five-year sustainability, 51(5),
of a Texas rice farm to discrete factors. For the base scenario, £(5) was 0.50 based on
threshold leverage and 0.12 based on threshold NPV.
Base
Alternate
Leverage < 2.0
- NPV > 0.0 -
Factor
value
value
¿(5)
R
¿(5)
R
tenure
arrangement
1/2 share
100% owned
1.00
0.50 **
0.32
0.20 *
tenure
arrangement
1/2 share
50% owned
1.00
0.50 **
0.18
0.06 n.s.
soybean
irrigation
non-irrigated
irrigated
0.86
0.36 **
0.64
0.52 **
tenure
arrangement
1/2 share
1/7 share
0.82
0.32 **
0.52
0.40 **
tenure
arrangement
1/2 share
rent at
$74/ha
0.74
0.24 *
0.56
0.44 **
rotation
SSR
SR
0.72
0.22 *
0.20
0.08 n.s.
loan interest
rates
variable
2 points
lower
0.54
0.04 n.s.
0.22
0.01 n.s.
ratoon rice
quality
7% quality
discount
no quality
discount
0.50
0.00 n.s.
0.12
0.00 n.s.
Of the continuous factors tested in the sensitivity analysis, only variable costs and
the rice crop share showed a significant role in constraining sustainability (Table 3-5). The
ranking of continuous factors by relative sensitivity was the same for sustainability based
on the threshold leverage ratio or on positive NPV. Changes in most of the discrete
factors showed a significant improvement in sustainability based on leverage ratios (Table
3-6). However, the ranking of discrete factors was different when sustainability was based
on NPV. These results suggest that reducing variable costs, negotiating more favorable

62
tenure arrangements, and irrigating the soybean crop would be the most important
strategies for enhancing sustainability of the model farm.
Discussion
By defining sustainability as the ability of a dynamic, stochastic, purposeful system
to continue into the future, I arrived at a useful, quantitative expression of sustainability.
Figure 3-4. Simulated sustainability of a Texas rice farm under four scenarios: a
three year soybean-soybean-rice rotation with a 1/7 {SSR 1/7) and a 1/2 (SSR 1/2)
share arrangement, and a two year soybean-rice rotation with the same two share
arrangements {SR 1/7 and SR 1/2). Data from Perry el al., 1986.

63
Although sustainability of a real agricultural system cannot be observed because it deals
with the future, it can be estimated from simulation of a system model. Testing
hypotheses about constraints to system sustainability is then straightforward. Applying
this framework to farming systems results in an approach to characterizing sustainability
that is literal, system-oriented, quantitative, predictive, stochastic and diagnostic (Chapter
2). The use of such an approach could provide attempts to improve farm sustainability
with objective feedback.
The requirement for comprehensive and realistic farm simulation tools currently
limits application of the proposed approach. Most existing farm-level simulation models
are not sufficiently comprehensive; they do not integrate models of crop and animal
production, environmental degradation, economic processes, and farmer production and
consumption decisions. A study designed to examine a single constraint to farm
sustainability would be less demanding in its model and data requirements.
One could question the realism of assumptions about the future behavior of inputs
to a farming system that are required for characterizing its sustainability. However, all
approaches to characterizing sustainability involve inferences about the future. If high-
level systems change more slowly than lower-level systems, as ecological hierarchy theory
asserts (Allen & Starr, 1988), then future projections of inputs such as weather, prices,
infrastructure and technology are more defensible than extrapolation of past farming
system behavior.
The study by Perry et al. (1986) illustrates the utility of the framework presented
in this paper for characterizing sustainability. Although I interpreted their study beyond its

64
original purpose, their simulation analysis was comprehensive enough to illustrate how the
probability of continuation integrates the effects of a range of factors, and how sensitivity
analysis can be used to identify and rank those factors that constrain sustainability.

CHAPTER 4
AN OBJECT-ORIENTED REPRESENTATION OF A FARMING SYSTEM
Introduction
Chapter 3 presents a framework for applying simulation to characterize the
sustainability of a farming system. The framework calls for a farm simulation model that is
able to simulate the operation of a farm over an extended period and to replicate that
period with stochastic inputs of the important variables that contribute to farm risk,
particularly prices and weather.
A comprehensive review of existing farm simulation tools is beyond the scope of
this chapter. Fortunately, a few good reviews of farm modeling literature are available.
Klein and Narayanan (1992) provide an excellent summary of the history of farm modeling
efforts. Jones et al. (1995) derived several generalizations from a review of applications
of farm-scale models. Two are particularly relevant to this study. First, “.. . there is
usually a new model for each study, with little acknowledgment of the potential for using
the same model for different farms by emphasizing data requirements and collection for
implementation” (p. 5). One of the exceptions is the FLIPSIM family of models
(Richardson & Nixon, 1985), which has been used to study the impacts of farm size
(Richardson & Condra, 1981), price variability (Grant etal., 1984), marketing strategies
65

66
(von Bailey & Richardson, 1985), tenure arrangements and several other policy and
management factors (Perry et al., 1986) on farm viability. The first generalization
highlights the need for simulation tools and associated data standards that are generic,
flexible and extensible. The second generalization is that “.. . most farm-scale models
have been developed with a bias toward economics and limited consideration of the
biophysical components” (p. 5). Farm models often represent crop and animal response to
weather variability by sampling from probability distributions that were fitted to historical
or survey data. However, concerns about ecological threats to sustainability such as soil
degradation and climate change are better addressed by the integration of biophysical and
economic models. Recent farm models have incorporated crop simulation to capture the
impact of weather variability on farm risk (Dillon et al., 1989) and to relate global climate
change scenarios to farmer adaptation (Kaiser et al., 1993), for example.
A farm simulation model, the Farming System Simulator (FSS), was developed as
a tool for applying the framework presented in Chapter 3 for characterizing farm
sustainability. Its object-oriented design address the need for a generic simulator that can
represent very different farm types, and whose functionality can be extended relatively
easily. Linkage to process-level crop simulation models addresses the need for balanced
and integrated treatment of the ecological and economic components of a farming system.
The objectives of this chapter are (a) to describe the Farming System Simulator and (b) to
present data requirements for simulating the sustainability of a farming system. The
chapter first presents an overview of the design and functionality of FSS, then describes its
inputs, structure, processes and outputs.

67
Overview of the Farming System Simulator
The Farming System Simulator (FSS) is an object-oriented, dynamic, stochastic,
discrete-event farm simulator. It is object-oriented in its design, and is implemented using
the object-oriented extensions of Borland Pascal (Borland International, 1992). The
description of FSS that follows cannot be fully understood without some familiarity with
object-oriented programming concepts (Appendix A). FSS is dynamic; it is capable of
simulating the operation of a farming system through many years. It is stochastic in the
sense that it is designed to simulate many replicates of a farm scenario with stochastic
inputs of weather and price data, and to present analyses based on the resulting
distributions. FSS runs external crop simulation models to simulate continuous
physiological and ecosystem processes. However, all farm-level processes (i.e.,
operations, production, management and consumption of resources, and failure) occur in
response to discrete events. A final characteristic of FSS is that it is primarily a resource
accounting model.
The object-oriented structure of FSS is based on a conceptual model of the
structure and function of a farming system. A farming system integrates ecological,
economic and social components (Fig 4-1). The ecological component of a farming
system—a set of agricultural ecosystems, or agroecosystems—consists of biotic
communities and the landscape that they inhabit, and can be delineated by field boundaries.
The economic component comprises the set of resources that are under the control of the

LOCAL
INDUSTRY?
WATERSHED?
VILLAGE?
1
\
i
1
1
i
1
/
I
ecological economic social
systems systems systems
Figure 4-1. Parallel ecological, economic and social hierarchies of agricultural systems.
o\
00

69
farmer. The social component of a farming system is the farmer or farm household, and
includes goals and decision criteria. In FSS, objects represent each of these components:
fields in the farm landscape, a set of strategies that specify the sequence of crop activities
and their management, a set of enterprises that link management strategies to particular
fields within the landscape, and a set of farm resources (Fig. 4-2). The household is
represented by decision rules for consumption, production and farm failure. The object-
oriented design of FSS provides a flexible means of representing farm resources, possible
interactions among resources, and relationships between operations and the resources that
they use. The Farming System Simulator is a farm model only in a loose sense; the model
structure of a particular farm is specified at run-time by the resource, field, enterprise and
strategy objects that are initialized in response to input data.
The capabilities of FSS reflect its purpose as a tool for characterizing farm
sustainability based on the framework presented in Chapter 3. The first requirement was
the ability to replicate a long-term scenario with constant initial conditions but stochastic
inputs. FSS takes advantage of a stochastic weather generator, WGEN (Richardson,
1985), that has been incorporated into the crop models that are part of the Decision
Support System for Agrotechnology Transfer version 3 (DSSAT, Hoogenboom et al.,
1994). An analogous stochastic price generator is part of FSS. FSS addresses its second
requirement—an ability to simulate ecological processes of crop production—by calling
external crop simulation models. The crop models simulate weather variability, soil
dynamics and crop growth and development, then return the information that FSS requires
in the form of schedules of field operations. Third, FSS addresses the need to deal with

Farming
system
has a
has a
set of "as a
has a
set of
set of
l \ X I
Consumption
.
Production
Goal
Production
decision criteria
decision criteria
thresholds
strategies
Figure 4-2. Object representation of the main components of a farming system.

71
farmer livelihood by accounting for all farm resources produced, used for production, or
consumed by the farm household. The fourth requirement was the ability to test
conditions for system failure. Failure can be based on insolvency—the inability to cover
fixed costs, obligations or minimum subsistence consumption requirements. Failure can
also occur when an individual or aggregate farm resource violates a user-specified
threshold value. Detailed resource accounting was a prerequisite to the ability to test for
conditions for farm failure. Finally, FSS offers several file and graphical outputs that are
relevant to the analysis of farm sustainability.
The current version of FSS possesses several important limitations. First, it does
not simulate adaptive management; it simulates a fixed, continuously repeating set of
management practices. However, an actual farm operating under stress would normally
employ a range of practices to avoid failure (Chapter 3). Second, FSS does not possess a
mechanism for adjusting crop management for within-season resource constraints. This
limitation is imposed by the need to simulate each crop for an entire season. Chapter 5
discusses the problem and a possible solution. Third, FSS can simulate only crop
production enterprises. No livestock model has yet been adapted for running under FSS.
Fourth, the household resource consumption model (Eq. [4-7]) is simplistic; it does not
consider the impact of risk or anticipated future lifestyle changes on consumption and
savings decisions. The remaining sections of this chapter focus on inputs, processes and
outputs of FSS.

72
Inputs
Jones et al. (1995) cited a lack of emphasis on data standards as a barrier to
reusing farm models for different farms or applications. FSS input data structures and file
formats were designed to be flexible enough to be able to represent a range of farm types
and to accommodate possible future extensions of FSS. The general organization of data
and scheme for its use are being proposed as a starting point for developing data standards
for enterprise and farm-level systems analyses (Hansen etal., 1995). Detailed
presentation of input data requirements and formats in Appendix B supplements the
discussion in this section.
A farm scenario is the operation of a farm through a period of time with a given
set of initial conditions and rules for making decisions and scheduling activities. A
scenario may be replicated with stochastic sampling of input variables such as weather and
prices, but with the same initial conditions and decision rules for each replicate. At least
two files—a scenario file and a price file—are required to simulate a farm scenario. The
scenario file contains farm-level information and identifies the other input files. The price
file contains the parameters for models for generating sequences of prices. A scenario that
calls external IBSNAT crop simulation models also requires a minimum data set (MDS)
for each crop consisting of a crop management file, soil file, weather or climate file, and
genetic coefficient file (Tsuji et ah, 1994).

73
The set of input data required to run FSS serves three roles. First, inputs define
the structure of a farm model. The objects that define the model of a particular farm are
initialized from the input files. This is the topic of the next section. Second, inputs specify
initial conditions by specifying initial values of the state variables of objects. Third, inputs
provide the values of dynamic driving variables. The two types of driving variables—prices
and weather—can be represented either by inputs of actual historical series or as stochastic
time-series models.
The structure of the input files relates closely to the object-oriented structure of
FSS. In most cases, a line or section of an input file is an argument for the constructor
method (see Appendix A) that initializes an object. Multiple items within a section of a
file are used to initialize collections of similar objects. For example, each line in the
RESOURCES section of the farm scenario file is used to initialize and insert a resource
object into the collection of farm resources.
Scenario file
The farm scenario file contains the information necessary to initialize a farm model
and simulate a farm scenario. The scenario file also identifies other essential data files: a
crop management file, a soil file and a price file. Table 4-1 describes the sections of a farm
scenario file. Appendix B presents the format of each section.
Simulation control. The first three sections of the scenario file relate to simulation
control. The SCENARIO section identifies the farm scenario and specifies crop
management, price and soil data input files, units of wealth, and options for duration of the

74
scenario and the number of times it is replicated. The next two sections—OUTPUTS and
ANALYSES—control file and graphical output. These are discussed later in this chapter.
Table 4-1. Sections in the farm scenario file.
Section
Purpose
SCENARIO
Specifies title, units, other files, and simulation control.
OUTPUTS
Controls farm-level file output.
ANALYSES
Controls graphic display and file output.
RESOURCES
Identifies farm resources.
LINKAGES
Defines interrelationships between resources.
OPERATION
REQUIREMENTS
Specifies types of operations, priorities, a time window, and
timed resources required.
SCHEDULED
OPERATIONS
Schedules operations which are not included in the experiment
file or returned by crop models.
LANDSCAPE
Identifies homogeneous fields and their positions and
characteristics.
LIVESTOCK
Not yet developed. When developed, will specify herd and
management information for grazing livestock.
STRATEGIES
Specifies sequences of management activities.
ENTERPRISES
Links management (i.e., strategies) with landscape (i.e.,
subfields).
PRODUCTION
DECISIONS
Not yet developed. When developed, will specify a model and
criteria for selecting strategies for each enterprise.
CONSUMPTION
DECISIONS
Specifies household subsistence and discretionary consumption
of resources.
The remaining sections of the scenario file define both the structure and the initial
conditions of a model of a particular farm.

75
Resources. The RESOURCES section specifies the particular resources that
constitute a farming system’s state variables, and determines their initial characteristics and
supply. A hierarchy of nine resource classes is available in FSS to represent the various
types of farm resources (Table 4-2, Fig. 4-3). Two additional abstract ancestral classes
are used internally as templates for the other classes.
Resource
resource resource resource resource
Seasonal Capital Activity-linked
resource resource credit resource
Machine
resource
Figure 4-3. Tree of resource class hierarchy in FSS.

76
The LINKAGES section identifies the possible interactions among resources. Resource
linkages are initialized from the LINKAGES section and identify what resources can be
exchanged with a particular resource in response to a deficit, surplus, sale, or fixed costs
of ownership. Variable cost linkages identify resources that may be used to replenish a
deficit or dispose of a surplus. Fixed cost linkages identify resources that are charged
fixed costs of ownership for a particular resource. Resource linkages also identify the
components of a debt resource and components in the denominator or numerator of an
aggregate resource. Each resource linkage has a price from the price file that determines
the rate of exchange between the pair of resources. The price need not have monetary
units; it may represent, for example, a barter rate between two commodities.
Operations. The OPERATIONS REQUIREMENTS section relates field
operations to particular resources by identifying combinations of resources such as labor
and equipment that are required to complete each operation. Operations are either
returned by the crop models or specified in the SCHEDULED OPERATIONS section.
The resources required for an operation should be timed resources or their descendants
(seasonal, capital, or machine). The linkages that operation requirements establish
between operations and resources are analogous to the resource linkages that link
resources to other resources.
The SCHEDULED OPERATIONS section provides a flexible means of
incorporating management operations that the crop models do not consider, such as post¬
harvest processing, marketing harvested products or purchasing supplies. They can also
represent management and production of livestock or crops for which no model is

77
Table 4-2, Description of resource classes.
Class
Parent
Description
resource
none
An abstract resource used as a foundation for all other
resource classes.
simple
resource
An abstract resource built on the resource class used for
all resource classes except for aggregate and debt.
consumable
simple
A resource whose supply is reduced with use, that may
have minimum and maximum storage constraints, and
whose supply may change with decay or interest (eg.,
money, fertilizer).
timed
simple
A resource which is not consumed and whose use is
constrained by hours of availability per day (eg., labor).
seasonal
timed
A timed resource that is available only during a certain
period each year.
capital
timed
A timed resource with an initial value that may. depreciate
with time (eg., a building).
machine
capital
A capital resource that uses all variable costs each time it
is used. This permits proper accounting of simultaneous
use of fuel, maintenance, labor, and other costs of
machinery use.
credit
simple
A source of credit. A credit resource keeps track of
current debt, accrued interest, and payment schedules.
Using credit involves borrowing new loans or adding to
any loan that was borrowed during the current month.
activity-linked
credit
credit
A credit resource whose availability is based on the area
currently in a particular crop activity.
aggregate
resource
A resource that represents the sum (eg., liquid assets) or
ratio (eg., leverage ratio) of values of other resources for
accounting or output. Aggregate resources are used only
for consumption decisions, file output or analyses.
debt
aggregate
An aggregate resource that is used to keep track of
indebtedness associated with credit resources.

78
currently available. A scheduled operation includes information (type, crop and method
codes) used to identify the operation, a date, and information about any material resource
produced or consumed by the operation. The date of a schedided operation can be either
absolute, incremented each simulation year, or a fixed number of days relative to any other
operation. The material resource balance can be expressed per event, per unit of land, as a
fraction of the amount used on a preceding operation, or as a fraction of the current
supply of the resource.
The farm landscape. The LANDSCAPE describes farm fields. Fields are ordered
by hillslope and by position within each hillslope. The purpose of ordering fields by
hillslope position was to facilitate simulation of processes such as soil erosion for which
runoff from one field becomes an input for the adjacent, down slope field. Each field
description points to information about the physical environment that was initialized from
the crop management file.
Management. Each strategy initialized from the STRATEGIES section is a
repeating set of management practices consisting of a set of rotation periods and a
schedule of additional operations that are not returned by the crop models. A rotation
period is an ordered set of crop or fallow activities that occur within one year. An
activity represents a single crop season, fallow period, or livestock production period that
would be simulated by a single call to an external model. An activity points to
management information that was initialized from the crop management file, and
implements a driver that can run external crop simulation models. The current version of
FSS cannot simulate livestock activities.

79
The ENTERPRISES section links particular management strategies to particular
fields. FSS was originally designed to be able to select from a set of strategy alternatives
for each enterprise based on maximization of either profit or expected utility. However,
the current version does not simulate dynamic production decisions.
Consumption. The last section, CONSUMPTION DECISIONS, identifies a set of
resources consumed by the farm household, and subsistence and discretionary
consumption parameters for each. The consumption model is presented later (Eq. [4-7]).
Price file
The price file contains the information necessary to specify constant or
contemporaneous ARMA time series models (Eq. [4-1] to [4-4]) for a set of prices, or for
using a set of trend-adjusted historical prices as a proxy for future prices (Eq. [4-5] and
[4-6]). FSS recognizes four sections of a price file: CONSTANT, ARMA,
CORRELATION MATRIX, and HISTORICAL. Each item in the CONSTANT, ARMA
and HISTORICAL sections specifies a price series. The CORRELATION section
contains the cross-correlation matrix, in lower-triangular form, of the residuals of the price
models specified in the ARMA section. Variability between replicates can only be
obtained by ARMA prices. Historical prices are useful for removing price risk among
replicates while maintaining the historical pattern of prices to explore the relative
importance of weather and price variability. Constant prices can be used when trends,
seasonality and variability are not important. The format of each section of the price file is

80
presented in Appendix B. The models that they initialize for generating daily price values
are discussed in a later section.
Crop minimum data set
FSS obtains detailed information about crop management and the physical
environment from the crop management file, or FILEX (Jones et al., 1994). Thornton et
al. (1994) present relevant guidelines for preparing a crop management file for simulating
sequences. FSS ignores the TREATMENTS section of FILEX; the LANDSCAPE and
STRATEGIES sections of the scenario file serve the function of the TREATMENTS
section.
In addition to the crop management file, FSS requires a soil file, weather files or a
climate file, and genetic coefficient files. These data files are documented in Tsuji et al.
(1994). The climate file contains parameters for stochastic weather generators that are
incorporated into the crop models. Actual weather sequences can be used instead of
generated weather to eliminate weather risk in order to explore the relative importance of
weather and price variability.
Processes
FSS manages a number of processes, including managing random number
sequences, stochastic generation of prices, ecosystem dynamics and crop production,

81
event handling, resource accounting and household consumption. The sections that follow
discuss these processes in more detail.
Overview
During a given simulation year, FSS performs the following steps. First, methods
reset strategies, resources and events in preparation for the new simulation year. Second,
prices advance one year. FSS generates monthly prices for the current year, and stores
them for use in linearly interpolating to daily values as needed. Third, crop activities are
simulated in order of landscape position. For each field, FSS calls external crop models to
simulate all crop and fallow activities for the current year. The event queue accumulates
field operations returned by the crop models for later processing. Fourth, any additional,
scheduled operations are retrieved from each enterprise and inserted into the event queue.
Fifth, events are executed. Household consumption and all resource transactions occur in
response to execution of events. Some information is written to output files by events as
they execute. The event queue checks conditions for farm failure with each event that it
executes. If event execution fails, then the current replicate is suspended and the failure is
recorded for later analysis. Finally, if the user requested resource output then the status of
each resource is written to a file.
At the beginning of each replicate of a scenario, prices, resources, enterprises and
the landscape are reinitialized so that starting dates and initial conditions are identical
among replicates.

82
Random number sequences
When simulating a farm scenario, a user specifies the duration of the scenario and
the number of times the scenario is replicated. For consistency with the sequence analysis
programs in DSSAT3, FSS increments the initial random number seed by five at the
beginning of each replicate (Thornton et al., 1994). This permits the same set of
stochastic weather and prices to be used with different scenarios without pseudorandom
number sequences being affected by different timing of farm failures among different
scenarios.
Price generation
FSS implements three classes of prices: constant prices, historical prices and
stochastic prices simulated using an autoregressive-moving average (ARMA) model.
Constant prices always return a single value read from the price file.
The model used to generate an ARMA price sequence for a single commodity
includes a deterministic trend and seasonal component, and a stochastic component. The
deterministic component is,
*t = a + P' + Pm> [4-1]
where xt is the expected value at t months after a specified base month, a and P are the
slope and intercept of a linear trend, and pm is the mean deviation from the trend for
calendar month m. The ARMA price model can include a multiplicative seasonal moving

average component (Box and Jenkins, 1970) to simulate price cycles longer than a year.
The stochastic component of the model (w,) is calculated as,
p q
s t't-S >
P * 6,
q < 6,
et ~ N(0, oE),
[4-2]
where 4>¡ and 0¡ are autoregressive and moving average coefficients for a lag of i months,
p and q are the maximum lag of the autoregressive and moving average components, S is
the lag of an optional multiplicative seasonal moving average component (0S), and et is a
random normal deviate. The simulated valuer, is the sum of the deterministic and
stochastic components, truncated if necessary to avoid unreasonably low values,
yx = max(wt + jtt, 0.05 xt),
[4-3]
or, if the data used to fit the model were log-transformed,
|Q(max(wt + xt, 0.05xt))
[4-4]
A set of ARMA prices constitutes a multivariate, contemporaneous ARMA process
that accounts for cross-price correlation. In a contemporaneous ARMA model, the
current value of a particular series is assumed to depend on prior values within the series
and on random shocks that influence other series, but not on prior values of other series

84
(Hipel and McLeod, 1994). It is implemented by sampling ej t for each individual price j
from a multivariate normal distribution, e ~ N(0, S), where S is the variance-covariance
matrix of the residuals of the individual price series models. The Cholesky decomposition
is used to obtain the lower triangular matrix C such that CC = 2. Then e is obtained
from e = Cz, where z ~ N(0,1) is a vector of independent standard normal variates
calculated using the polar Box-Müller method. Dagpunar (1988) presents the algorithm in
detail.
At the beginning of each replicated scenario, the first g+S values of e and the first
p values of w are initialized from a random normal distribution to provide the past values
required before the first application of Eq. [4-2], Prices are then simulated for an arbitrary
number of months (240) to allow each price series to establish its own patterns consistent
with the model, and to allow the variance between replicates to approach its
asymptotically stable level. Monthly prices are simulated and stored at the beginning of
each simulated year, then linearly interpolated to daily values as needed.
A historical price object inserts a sequence of prices from a text file, beginning
with the first observation from the calendar month preceding the starting month of the
simulation. For example, if a scenario starts in September 1994, price observations are
read in sequence starting with the first August encountered. This avoids shifting any
seasonal patterns. Prices are then adjusted for any shift between the time of the
observation and the simulation time based on a trend:
+ PA/,
St = Tt-A,
[4-5]

85
or if the observed series was log-transformed,
Sx = io(1o6*-^pa°, [4-6]
where yt is the estimate of price at month t, A t is the number of months between the time
the price was observed and the time of the scenario for which the observed price is used as
a proxy, and p is the slope of the trend. Daily prices are linearly interpolated as required.
Crop and ecosystem processes
FSS is designed to run and communicate with external simulation models of the
ecosystem processes involved in agricultural production. It currently works with the
family of crop models in DSSAT3. Ecosystem processes are simulated in order of the
landscape position of fields, which are ordered by hillslope and by position within each
hillslope. The purpose of ordering fields by hillslope position was to facilitate simulation
of processes such as soil erosion for which runoff from one field becomes an input for the
adjacent, down-slope field.
An activity object implements a crop model driver that calls external programs to
simulate each production activity on each field. FSS communicates with IBSNAT crop
models by means of the temporary input file documented in Hoogenboom et al. (1994)
based on the data specified in the crop management file, or FILEX (Jones et al., 1994).
An operations output file (Table 4-3) supplies all of the information that FSS requires
from production process models. It consists of a schedule of field operations and the
material resource balance associated with each operation. The operations output file is

86
necessary because current IBSNAT crop model output files (Jones et al., 1994) do not
provide a complete list of operations or their timing. Furthermore, state events triggered
by field conditions (eg., automatic irrigation) cannot be inferred from information in the
crop management file.
Event handling
FSS is a discrete-event simulator. The state of the farming system, represented by
the supply of each of its resources, changes only in response to events. An event queue
accumulates, sorts, executes and monitors the status of events. FSS recognizes three
classes of events: field operations, schedided operations and monthly update events.
Field operations are returned by external production process models. Monthly update
events are inserted at the midpoint of each calendar month of the current simulation year,
and are responsible for fixed cost accounting, household consumption and for testing
criteria for farming system failure. Scheduled operations are described in the previous
section. Each time an external model simulates a production activity, resulting field
operations are inserted into the event queue. After the entire landscape is simulated for
the current year, any scheduled operations are added to the queue. Finally, twelve
monthly update events are inserted. Events are stored in order of date, then priority. This
gives high priority events first access to scarce resources. Monthly update events have
priority over all operations.

87
Table 4-3. Format of the operations output file.
Variable
Name
Formal
Date (year, day of year)
DATE
115
Days after planting
CDAY
1 15
Operation type:
“TILL” = tillage
“RESD” = residue application
“PLNT” = planting
“CHEM” = chemical application
“FERT” = fertilizer application
“IRRI” = irrigation
“HARV” = harvest
OPTY
2 C 4
Operation option (from simulation controls section)
OPOP
5 C 1
Operation method code (Jones et al., 1994, Appendix B)
OPME
1 C5
Material resource produced
RAMT
0 R 9 1
Unit of material resource
RUNT
1 C5
Name of material resource
RNAM
1 C5
Crop code
CC
4 C 5
Cultivar code
CLTV
1 C6
* Format descriptions consist of: number of leading spaces, variable type (C = character, I
= integer, R = real), field width, and (if real) number of decimal places.
The event queue executes all stored events that fall within the current simulation
year. The Execute method of a monthly update event sends messages to Update all
farm resources and CONSUME the resources specified in the household consumption
model. Failure to update resources or perform household consumption are conditions for
farm failure.
Figure 4-4 illustrates the flow of information as a field operation or scheduled
operation is created and later executed. When a production process model or an

88
operation schedule generates an operation, it is matched with a set of resource bundles
from the set of operation requirements. A resource bundle is a set of timed resources (or
their descendants) that must be used in a fixed proportion. On a mechanized farm, the
bundle will typically include a power unit, an implement and an operator. Operations
possess a mechanism for delaying their execution if resources are not available for their
completion on a particular day. When an operation executes, it first determines the
fraction that can be completed in the current day with the availability of the required set of
timed resources. It then adjusts any material resource requirement for the fraction that
can be completed. If the material resource is constraining, the fraction of the operation
that can be completed is again adjusted. The Execute method then uses the required
timed resources. If the operation could not be completed and the current date is within a
specified window, the Execute method returns a result of delayed. Otherwise, the result
is succeeded. If the operation is delayed, the event queue spawns another operation with
adjusted resource requirements that reflect the fraction of the operation remaining.
Resource accounting
Much of the flexibility of FSS comes from its polymorphic set of resource classes
(Table 4-2). Resource accounting is accomplished by three methods—USE, Sell and
Update—that are common to all resource classes. The behavior of these methods varies
depending on the resource class. These methods are presented in detail below. Familiarity
with these methods is essential to understanding how FSS accounts for resource use.

89
Figure 4-4. Flow of information in FSS from the generation of a field operation to
its effect on resource accounting.

90
Use and Sell methods. Most operations use or produce some material resource,
represented by a consumable resource object. Examples are water used for an irrigation
application and maize grain produced in a harvest. Most operations call the USE method
of any associated material resource, passing the amount of the resource produced as an
argument. A marketing operation is the exception; it calls the material resource’s SELL
method. The USE method adjusts supply according to the amount produced (Fig. 4-5). If
the amount produced is negative indicating consumption of the material, and the amount
used is greater than the available supply, then the USE method attempts to cover the
resulting deficit from resources identified in its list of variable costs. If use violates a
maximum storage constraint, then the USE method disposes of the resulting surplus by
selling to resources in its variable cost list.
Algorithm
Execute the method, Update.
Set Supply = Supply + Amount.
If Supply < max(Minimum, 0) then
begin
Set Deficit = mm(Supply - Minimum, 0.)
Request VariableCosts to execute UsE(Deficit).
If Deficit < Othen
set Amount = Amount - Deficit.
Set Supply = mm(Minimum, 0).
end
otherwise if Maximum is defined and
Supply > Maximum then
begin
Set Surplus = Maximum - Supply.
Request VariableCosts to
execute SEhh(Surplus).
Purpose
Make sure linked resources are up-to-date.
Change Supply according to Amount desired.
If the minimum storage constraint is violated...
Deficit represents a negative amount.
Try to meet deficit by using linked resources.
If VariableCosts could not eliminate the deficit...
Reduce Amount used by remaining Deficit.
Supply is exhausted.
The maximum storage constraint is violated.
Dispose of Surplus by selling to linked resources.
Figure 4-5. Pseudocode representation of the USE method of the consumable resource
class.

91
The Sell method (Fig. 4-6) disposes of an amount of a consumable resource by
exchanging with the first linked resource found in the list of variable costs (see next
section). For example, Sell may reduce the supply of a harvested product and increase
the supply of the operating fund. In resource classes other than consumable, Sell is a
dummy method that does nothing.
Algorithm
Set Amount = max(Amount, -Supply).
Set Sold - - Amount.
If Sold > 0 then
request VariableCosts to execute Use(Sold).
Set Supply = Supply + Amount.
Purpose
Limit amount sold to amount available.
Positive Sold adds to linked resources.
Give Sold to the first linked resource.
Reduce Supply. (Amount should be negative.)
Figure 4-6. Pseudocode representation of the Sell method of the consumable resource
class.
The behavior of the Use method of timed resources and their descendants
(seasonal, capital and machine) is quite different from that of consumable resources.
Instead of consuming supply with use, each timed resource keeps a list of the amount of
time reserved each day. Use by a particular operation reduces the number of hours
available for use by subsequent operations on the same day (Fig. 4-7).
The USE method of credit and activity-linked credit is complicated by the need to
maintain a record of multiple loans from the single source (Fig. 4-8). If Amount is
negative, indicating borrowing, USE inserts a new loan for the appropriate amount into the
collection of loans or, if a previous loan was taken since the last scheduled payment date,
the new amount is added to the balance due on that loan. A positive Amount passed to
Use indicates an early, unscheduled loan repayment.

92
Algorithm
Purpose
Execute the method, Update.
If Supply > 0 then
begin
Make sure linked resources are up-to-date.
Set Available = Supply.
Use a temporary variable. (Supply doesn’t change.)
Search UsageList for today’s usage.
Check if the resource is already being used today.
If today’s usage > 0 then
If it is already being used today ...
sel Available = Available - today’s usage.
Set Deficit = min(.Supply + Amount, 0).
Adjust availability for time already allocated.
If Deficit < 0 then
If more was requested than is available...
request VariableCosts to execute \3sE(Deficit).
Search for a source for hiring more resource.
Set UsedToday = max(-iSupply, Amount).
The portion of Amount that can be UsedToday.
Insert UsedToday into UsageList.
end.
Add the additional allocaton to the schedule.
Figure 4-7. Pseudocode representation of the Use method of the timed resource class.
Algorithm
If Amount < 0 then
begin
Set Amount = -Amount.
If Amount > Supply then
begin
Set Amount = Supply,
Set Supply = 0.
end
otherwise
set Supply - Supply - Amount.
Search Loans for a current Loan.
If found then
request Loan to add Amount to existing debt
otherwise
add a new Loan to the collection of Loans.
Set Debt = Debt + Amount.
Set Amount = -Amount.
end
otherwise
begin
Request Loans to execute REPMD(Amount).
If Reapid then
set Debt = 0
otherwise
set Debt = Debt - Amount.
end.
Purpose
Negative Amount indicates borrowing.
Work with positive Amount for convenience.
The amount requested is not available ...
Reduce the amount used to the amount availabe.
Supply is exhausted.
The amount requested is available...
If a Loan was already taken during the
current repayment interval then...
Add the new amount to the existing loan.
Take-out a new loan
Add the new Amount borrowed to existing Debt.
Change back to original sign.
Indicates an unscheduled repayment.
Attempt to repay existing Loans With Amount.
If Loans are all repaid then
cancel existing Debt.
Reduce Debt by the Amount paid.
Figure 4-8. Pseudocode representation of the USE method of the credit resource class.

93
Update method. The Update methods of consumable (Fig. 4-9) and capital
resources (Fig. 4-10) change supply or value in response to interest or depreciation, and
charge fixed costs of ownership. Credit resources use Update to make scheduled loan
payments. The Update method obtains the total payment required from its collection of
loans, then searches its collection of fixed costs to find a source of funds to complete the
required payment (Fig. 4-11).
Algorithm
Purpose
If the date is later than the last Update then
begin
Calculate A time.
The fraction of a year since the last Update.
Set Supply = Supply • exp {Interest • A time).
Adjust Supply for interest or depreciation.
Execute the method, Use(0).
Adjust for any new storage constraints.
Execute the method, UpdateFixedCosts.
Charge for any fixed costs of owning Supply.
end.
Figure 4-9. Pseudocode representation of the Update method of the consumable
resource class.
Algorithm
Purpose
If the date is later than the last Update then
begin
Calculate A time.
The fraction of a year since the last Update.
Set Value = Value • exp(Interest • A time).
Adjust Value for depreciation.
Execute the method, UpdateFixedCosts.
Charge any fixed costs of ownership.
end.
Figure 4-10. Pseudocode representation of the UPDATE method of the capital resource
class.

94
Aleorithm
If a loan payment is due then
begin
Purnose
Calculate A time
The fraction of a year since the last Update.
Request Loans to execute GetPayment
Payment = -GetPayment.
Repeat
Find the next linked resource in FixedCosts.
GetPayment returns the amount due on all loans.
Request linked resource to
execute USE(Payment).
Attempt to obtain Payment from the linked resource.
until Payment is reduced to 0.
... until Payment is met or all FixedCosts are tried.
Request Loans to execute Update
Update principal and number of payments remaining
on each loan, and dispose of repaid loans.
Set Supply = Initial supply - Debt.
end.
Calculate new Supply of available credit.
Figure 4-11. Pseudocode representation of the Update method of the credit resource
class.
Resource linkages. Resources interact and exchange with each other through
resource linkages. A set of linkages comprises a cost list. The Use method of a cost list
attempts to obtain or dispose of a specified amount by converting the amount to units of
the linked resource based on a price, executing the Use method of the linked resource,
reducing the amount by what the linked resource could use, then converting the amount
back to its original units (Fig. 4-12). The process repeats for each linkage or until the foil
amount has been exchanged.
A Forrester representation provides a clearer picture of how a resource linkage
operates. Each resource accesses an external source or sink. In the example in Fig. 4-13,
if the amount of resource a used is greater than the amount available, it will attempt to
draw on its source to replenish the resulting deficit. The amount that resource a can
access from its source is controlled by linkages to resources b, c and d, and is limited by

95
their ability to dispose of their supply. Priceab represents the number of units of resource b
that must be used for each unit of resource a that is replenished.
Algorithm
Purpose
Repeat
For each linked resource in the list...
É Amount > 0 then
Positive Amount indicates obtaining.
set P = BuyMult * Link.Price.Current.
Price P is today’s price for the resource linkage,
adjusted with BuyMult.
otherwise
Negative Amount indicates disposing.
set P = SellMult * Link.Price.CURRENT.
Adjust today’s price with SellMult.
Set ,4 = Amount * P.
Convert Amount to units of the linked resource.
Request Link to execute Use(j4).
Obtain from or dispose to the linked resource.
If A = 0 then
If linked resource exchanged all oL4 ...
set Amount = 0
Nothing remains to be exchanged.
otherwise
set Amount = Amount -A/P.
Determine Amount that still needs to be exchanged.
until Amount = 0.
Stop when full amount has been exchanged.
Figure 4-12. Pseudocode representation of the Use method of the cost list class.
Illustration: A fertilizer application operation
A fertilizer application operation illustrates the flow of information and sequence
of events that lead from the simulation of an operation by an external crop model to a
change in the operating fund (Fig. 4-14). In this fictitious example, a crop activity calls a
crop model, then uses information returned in the operations output file to insert a
fertilizer application operation into the event queue. After the remaining crop and fallow
activities have been simulated and scheduled operations have been inserted, the event
queue is executed. When the event queue executes the fertilizer application operation, the
operation searches the collection of operation requirements. The matching operation
requirement indicates that a fertilizer application using this particular method on this

VO
Figure 4-13. Forrester representation of variable cost linkages between a consumable resource (a) and three linked
consumable resources (b, c and d).

97
Fertilizer
application
operation
.QUEUE
EXECUTION OF
OPERATION
Operation
requirements
to
. u
JJ- o
o it
U] d
to O
a ÃœJ
o:
urea
requirement
operator
“FE005”
Linkage from
“FE005” to
“OPERATING
FUND”
>
r
“FRED”
>
f
Linkage from
“FRED” to
“OPERATING
FUND”
spreader
requirement
'tractor
requirement
“SPREADER”
Linkage from
“SPREADER” to
“OPERATING
FUND”
Linkage from
“TRACTOR1” to
‘OPERATING
FUND"
Linkage from
“TRACTOR1”
to
“FUEL"
“FUEL”
“OPERATING
FUND”
Linkage from
“FUEL” to
“OPERATING
FUND”
Figure 4-14. Flow of information from the generation of a fertilizer application
operation by a crop model to its effect on the operating fund.

98
particular crop requires an operator (“FRED”), an implement (“SPREADER”) and a
power unit (“TRACTOR 1”) in fixed proportions. Each of these timed resources has
associated variable costs that link usage to the operating fund. “FRED” requires hourly
wages from the operating fund. “TRACTOR 1” is linked to the operating fond both
directly for maintenance costs, and indirectly through “FUEL” that must be purchased.
The operation also consumes urea (“FE005”) that must be purchased through the
operating fond. If the required amounts of all five resources (urea, fuel, and the operator,
implement and power unit) are available or can be purchased or hired, then the resources
are used and operation can be completed in the current day. Otherwise, a portion of the
operation must be delayed.
Household consumption
Consumption occurs each month in response to a monthly update event. The
current version of FSS implements a very simple model of household consumption.
Consumption, C, of a given resource / in month m is given by
CUm = iM [4-7]
where is the annual subsistence consumption requirement, Dx is discretionary
consumption expressed as a fraction of the supply R} of possibly another resource j, and /m
is the length of month m as a fraction of a year. Consumption requirements can be
specified for any subset of farm resources.

99
Outputs
FSS creates five output files that relate to sustainability: (a) annual values of all
resources for the current scenario (RESOURCE. OUT), (b) sorted final values of a
particular resource specified in the scenario file for a set of scenarios (ENDWLTH.OUT),
Box Plot of UEALTH, Scenario GRAPH2
Year
Figure 4-15. Example of a resource box plot generated by FSS.

100
(c) the dates of all farm failures for the current scenario (FAILURES.OUT), (d) monthly
sustainability of a set of scenarios (SUSTAIN.OUT), and (e) the final status (0 = failed, 1
= continuing) of each replicate for a set of scenarios (STATUS.PRN). FSS uses
RESOURCE.OUT to create a table of percentiles for creating a resource box plot.
ENDWLTH.OUT is the basis for plotting the final distribution of a resource.
Cumulative Probability of Final Health
Ualue
Figure 4-16. Example of a final resource distribution plot generated by FSS.

101
Sustainability time plots are based on SUSTAIN.OUT. The McNemar test (Eq. [3-19]
and [3-20]) for differences in sustainabilities uses the occurrence of continuation matched
by replicate in STATUS.PRN. STATUS.PRN is formatted for importing into spreadsheet
software. Two additional files—a record of events (EVENT. OUT) and a record of
resource transactions (TRANSACT.OUT)—were designed for model testing. FSS uses an
external graphics program written for DSS AT3, WMGRAF, to display the three types of
graphs described below.
Resource status. A resource box plot (Fig. 4-15) shows the minimum, 25th, 50th,
75th percentiles and maximum values of a resource among realizations for each year of a
scenario. If an aggregate resource that represents the overall condition of the system (eg.,
liquid assets) is selected, then the box plot provides an overview of how the level and
dispersion of the system’s condition change through time. A final resource distribution
plot (Fig. 4-16) gives a more detailed picture of the distribution of the value of a resource
at a single time at the end of the scenario. The high probability of a zero value in Fig. 4-
16 reflects the realizations that failed before the end of the 15-year scenario.
Sustainability. FSS can create sustainability time plots (Fig. 4-17) that show the
probability of continuation from the start to each month of the scenario. Sustainability is
estimated from Eq. [3-12].
Discussion
The Farming System Simulator presented in this chapter possesses the capabilities
needed to simulate and analyze the sustainability of a range of farming systems. Linkage

102
to process-level biophysical models through an operations output file allows FSS to
respond to ecological determinants of sustainability. Its object-oriented design and input
data structures give FSS considerable flexibility for representing a range of farming system
types. The input data structures defined for FSS are sufficiently flexible to serve as a
starting point for developing data standards for farm-level systems analyses (Hansen,
1995).
Sustainability Tima Plot
95142 98046 316 3220 6124 9028 11298
Data
Figure 4-17. Example of a sustainability time plot generated by FSS.

103
The versatility and realism of FSS is currently limited by its inability to simulate
production decisions, lack of feedback between within-season resource constraints and
crop production, an absence of a model of livestock production, and the over-simplicity of
the model of household consumption decisions. In spite of its limitations, FSS is a useful
tool for characterizing the sustainability of a farm operating under fixed management
scenarios. Its object-oriented design and data structures provide a flexible foundation for
enhancements that may correct some of its existing weaknesses.

CHAPTER 5
CROP SIMULATION FOR CHARACTERIZING
SUSTAINABILITY OF A COLOMBIAN HILLSIDE FARM
Introduction
Many different factors can influence sustainability of an agricultural system,
including the ecological processes that affect crop production at a field scale. Process-
level crop simulation is a useful tool for quantifying and synthesizing effects of ecological
processes on levels, trends and stability of crop production. Since sustainability deals with
long-term, future behavior of systems, simulation may be an essential tool for studying
sustainability. Several authors have discussed possible roles for crop simulation in
addressing issues that threaten sustainability. Jones et al. (1993) proposed a framework
for applying crop simulation models to address sustainability of agricultural production
systems. The framework emphasizes providing decision makers with predictions about
various “sustainability indicators.” They illustrated the use of crop simulation to predict N
fertilizer required to restore maize yields after simulated soil loss, and to examine long¬
term stability and trends in yields of maize monoculture and a maize-soybean rotation.
Singh and Thornton (1992) discussed and demonstrated the use of crop simulation models
for studying soil nutrient losses and their impact on long-term yield trends and
104

105
environmental pollution. Jones et al. (1991b) reviewed studies that demonstrate various
capabilities of EPIC (Erosion-Productivity Impact Calculator) such as predicting crop
response to soil erosion, nitrogen inputs, weather variability and climate change scenarios.
They concluded that EPIC is an operational model for evaluating sustainability of cropping
systems because it addresses “constraints imposed by productivity, resource conservation,
protection of water quality, and socioeconomic considerations” (p. 347). While none of
these studies suggests specific criteria for interpreting simulation results in terms of
sustainability, all agree that crop simulation is a useful tool for studying crop response to
weather risk, and the agricultural and environmental impacts of long-term soil dynamics
that threaten sustainability of crop production systems.
Chapter 3 presents a framework for using long-term, stochastic simulation to
characterize farm sustainability. The Farming System Simulator that was developed as an
experimental tool for characterizing farm sustainability (Chapter 4) uses crop simulation to
capture the effects of agroecological processes and management on agricultural
production. In Chapter 6, the framework is applied to identifying determinants of
sustainability of a hillside farm in the Cauca region of Colombia. The purpose of this
chapter is to evaluate the compatibility of the IBSNAT family of crop models
(Hoogenboom et al., 1994) with requirements imposed by a simulation study of the
sustainability of a hillside farm in the Cauca region of Colombia. Specific objectives are
(a) to describe the physical environment of the study area, (b) to review the status of the
crop models, (c) to evaluate crop model predictions of development and yield in a
Colombian hillside environment in response to potential agroecological determinants of

106
sustainability (weather variability, soil nutrient dynamics and management, and soil
erosion), and (d) to discuss issues of compatibility between the crop models and the
broader modeling framework needed to address sustainability at a farm level.
A Colombian hillside environment
A farm (the Domingo farm) located in the Cabuyal River catchment within the
Cauca region of southwestern Colombia (2° 47' N, 76° 31' W) (Fig. 5-1) was selected as a
basis for a simulation study of determinants of sustainability (Chapter 6). The farm
occupies 5.16 ha ranging in elevation from 1660 m at the homesite to 1580 m at the
Cabuyal River (Fig. 5-2). The farm landscape is quite steep; about a third of the area has
slopes in excess of 24% (Table 5-1). Until July 1993, the major farm enterprise was
coffee production. The land that was in coffee production (2.38 ha) has since been used
for bean, both solecropped and intercropped with maize. CIAT personnel have managed
bean, maize and cassava trials on a small portion of the farm since October 1993. These
trials provide a basis for testing simulation models of these three crops.
Table 5-1. Area in each slope class, Domingo farm, Cauca, Colombia.
Slope -- Cultivated area -- -— Grass fallow -— Total farm
class area (ha) percent area (ha) percent area (ha) percent
0 - 12%
0.17
7%
0.16
12%
0.89
17%
12 - 24%
1.66
70%
0.48
36%
2.61
51%
24 - 36%
0.50
21%
0.45
34%
1.22
24%
> 36%
0.05
2%
0.24
18%
0.43
8%

Figure 5-1. Location map of the study area, Cauca, Colombia

108
to Pescador
Figure 5-2. Land use map of the Domingo farm, Cauca, Colombia.

109
Climate. Monthly climate statistics are given in Fig. 5-3. Temperatures at this
elevation are nearly ideal for many crops and vary little within the year. Rainfall follows a
bimodal distribution with peaks in October to November and in February to May and a
pronounced diy season between June and September. Rainfall is an important determinant
of cropping patterns in the absence of irrigation. The highest yields of maize and beans
are obtained with the lowest risk when they are planted at the beginning of the first rainy
season in October. March-planted maize is vulnerable to periods of severe water stress
during grain-fill. Rainfall quantities and seasonal patterns are surprisingly consistent
across locations in the Cauca watershed (Fig. 5-4), considering its hilly topography.
Soil characteristics. The soil on the Domingo farm belongs to the Pescador
association. Table 5-2 presents soil properties measured 2 m from the boundary of a bean
trial managed by CIAT personnel. Castro et al. (1992) described a soil in the Pescador
association with very similar properties, located at the same elevation and about 5 km
from the Domingo farm. They classified the soil as a Medial, Isothermic, Acrudoxic
Hydric Hapludand (Soil Survey Staff, 1990). The high organic C content and low bulk
density observed on the Domingo farm are consistent with classification as a Udand.
The unique properties of Andisols-their physical and hydrological characteristics,
pH dependent charge, tendency to accumulate organic matter, retarded N mineralization,
and high P fixation—challenge the assumptions and capabilities of existing crop simulation
models. These properties derive primarily from the noncrystalline (amorphous) materials—
allophane, imogolite, amorphous iron oxide and aluminum hydroxide gels, and aluminum
and iron humus complexes-that dominate the clay fraction of Andisols.

Rainfall, mm Temperature, °C Solar radiation, MJ / m
no
Figure 5-3. Monthly climate statistics: mean daily solar radiation, mean daily
maximum and minimum temperature, and total rainfall, Domingo farm, Cauca,
Colombia.

Ill
Surface charge of these minerals depends on pH and soil solution activity.
Negative ApH (pH in KC1 - pH in water) values (Table 5-2) indicate net negative charge,
and | ApH | >0.5 for each layer indicates that pH-dependent charge dominates (Uehara &
Gillman, 1981). However, cation retention is likely to be poor; the effective cation
exchange capacity (ECEC) is low in the Ap horizon and extremely low in the Bw horizon.
Raising the pH by liming would increase cation retention.
Figure 5-4. Mean monthly rainfall totals at weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation.

112
Table 5-2, Properties of soil layers, site near on-farm trials, Domingo farm.
Property
Ap
35 cm
55 cm
total N, percent
0.673
0.254
0.198
NH4+ - N, mg kg'1 soil
46.2
16.9
12.9
N03' - N, mg kg'1 soil
2.80
2.46
2.93
Bray II - extractable P, mg kg'1 soil
1.9
1.1
1.5
total P, mg kg"1 soil
334
148
128
organic P, mg kg'1 soil
200
108
96
organic C, percent
5.3
3.4
2.4
C/N ratio
8
14
12
KC1 - extractable K, cmolc kg'1 soil
0.18
0.05
0.05
KC1 - extractable Ca, cmolc kg'1 soil
1.88
0.30
0.11
KC1 - extractable Mg, cmolc kg'1 soil
0.43
0.07
0.05
KC1 - extractable acidity (H + Al), cmolc kg'1 soil
0.79
0.23
0.17
ECEC, cmolc kg'1 soil
3.28
0.65
0.38
base saturation, percent
76
65
55
pH in water
5.4
5.6
5.6
pH in KC1
4.7
5.0
5.1
sand, percent
55.8
silt, percent
27.7
clay, percent
16.7
bulk density, g cm'3
0.42
0.44
0.45
water retention at 0.0 atm, percent volume
57.5
55.4
62.2
water retention at -0.3 atm, percent volume
40.7
40.9
45.9
water retention at -1.0 atm, percent volume
38.0
38.4
42.6
water retention at -15 atm, percent volume
32.5
33.0
37.5
water drained upper limit, percent volume
46.7
48.9

113
The amorphous constituents of Andisols bind organic matter and inorganic anions
such as P. Because of P fixation, Andisols are usually deficient in P and require huge
additions of fertilizer P to bring soil solution concentrations to adequate levels. The high
capacity of Andisols for fixing P has been attributed to the high density of reactive A1 on
amorphous clays (van Wambeke, 1992), and entrapment in gels that deform and partially
liquefy when hydrated, occluding phosphate in voids that are removed from the soil
solution (Uehara & Gillman, 1981). The Andisol (Hydrandept) in Fig. 5-5 fixes several
times more P than the Oxisols or the Mollisol at any given soil solution concentration
Figure 5-5. Phosphorus sorption islotherms for four soils with different mineralogy.
Equilibrated for 6 days in 0.01 M CaCl2 at 25 °C. After Fox (1978).

114
(Fox, 1978). The slope of each sorption curve in Fig. 5-5 indicates both the amount of
fertilizer P needed to raise solution concentrations a given amount and the capacity of
adsorbed P to replenish solution P that is removed by roots. Fox (1978) showed that
crops respond to soil solution P concentrations consistently across different soil types.
Inhibited microbial decomposition causes many Andisols to accumulate large
reserves of organic matter. Although the supply of organic N may be high and climatic
conditions favorable to decomposition, N mineralization rates are generally lower than in
Figure 5-6. Decomposition of soil organic C in three allophanic and seven non-
allophanic soils. Data from Martin et al., 1982.

115
soils with crystalline mineralogy. In an incubation study, the decomposition rate of native
organic C was an average of eight times higher in non-allophanic than in allophanic soils
(Fig. 5-6, Martin et al, 1982). Additions of allophane to a sandy loam reduced the rate of
decomposition of wheat straw (Fig. 5-7, Zunino et al, 1982). Suppressed mineralization
in Andisols has been attributed to inadequacy of P (Munevar & Wollum, 1977) and
soluble C (Monreal et al, 1981) for microbial growth, and the formation of stable
complexes with Al and Fe (Zunino et al., 1982). The suppression of mineralization by
Figure 5-7. Decomposition of 14C-labeled wheat straw in Lo Aguire sandy loam
with added allophane. Data from Zunino et al, 1982.

116
allophane is apparently much greater for humified organic matter than for fresh residues
(Monreal etal., 1981).
The characteristic low bulk densities of Udands have been used to explain their
good workability, favorable infiltration and water-holding capacities, aeration and nutrient
availability, and their tendency to lose nutrients easily by leaching (van Wambeke, 1992).
Soil degradation. Soil erosion is a major concern in the Cauca watershed (Ashby,
1985). Annual crops are cultivated on very steep lands, often on slopes near 100%. Rains
are expected to be most erosive in November, before plant canopies have covered recently
tilled soil. However, a high infiltration capacity and the contribution of organic matter to
the formation of stable aggregates in the Ap moderate the erosion hazard somewhat.
The Andisols of Latin America are typically more vulnerable to mass movement
(i.e., landslides and slumping) in which the transport mechanism is gravity, than they are to
water erosion (Sentís, 1992). This vulnerability results from their high infiltration capacity
and rheological properties. A mass movement can occur during intense rain if antecedent
moisture is high, the subsoil is less permeable than the surface soil, and the subsoil has low
cohesiveness. Some amorphous clays behave as fluids when they are fully hydrated and
may act as a lubricant between surface and subsoil horizons. The landscape in the Cauca
region shows evidence of a great deal of degradation by mass movement.
Crop simulation
The crop simulation models (Hoogenboom et al., 1994) and support software in
the Decision Support System for Agrotechnology Transfer version 3 (DSSAT3, Tsuji et

117
al., 1994) were selected for use in this study. The specific crop models used were
CROPGRO v.3.0 for bean and tomato, Generic-CERES v. 3.1 for maize, and CropSim
CASSAVA v. 1.0 for cassava. These models were selected because (a) they are sensitive
to weather and to soil nitrogen dynamics, (b) they conform to a common input and output
data standard (Jones et al., 1994), (c) they can run in sequence using a carryover file to
initialize ecosystem state variables based on their final values from a previous run (Bowen
et al., 1992), (d) source code, supporting software and technical information are readily
available for these models, and (e) they have been tested under a wider range of conditions
than most other crop models. The incorporation of an adaptation of the WGEN weather
generator (Richardson, 1985) into the IBSNAT models simplifies the process of sampling
stochastic weather sequences.
Capabilities and limitations. The DSSAT3 crop models simulate photosynthesis,
respiration, partitioning and development in response to daily weather inputs. They
simulate the soil water balance, evaporative demand and crop water stress response. They
also simulate soil N and organic C dynamics and, except for CropSim CASSAVA, plant N
status and stress response.
Some of the known limitations of the crop models are important for interpreting
results of this study. Nitrogen is the only soil nutrient that the crop models can simulate.
Versions that account for soil and plant P are undergoing development and testing (Singh
& Godwin, 1990; Bowen, W., Personal communication), but are not yet available for use.
Although coupling points and input file formats have been defined that allow the bean and

118
tomato models to respond to observed pest populations or damage (Batchelor et ah,
1993), the crop models do not simulate pest populations.
The crop models are limited in their ability to handle crop residues and organic
amendments. They do not account for standing or surface residue, but assume that all
residue is incorporated at the time of application or harvest. The crop models cannot
simulate application of organic soil amendments (eg., manure) after the first season while
they are running in sequential mode. Finally, when the models are run in sequence, they
always incorporate all stover into the soil; harvesting part of the stover or failing to
remove some of the harvest product results in a mass balance error.
Prior testing. The CERES maize model has undergone extensive testing in
temperate North America and Europe (Kiniiy & Jones, 1986) and in various regions of
tropical Africa (Keating et al., 1991; Singh et ah, 1993), Asia and the Pacific (Singh,
1985) with generally acceptable results. Simulated maize yields match observed yields
closely in a Hawaiian Andisol (Hydric Dystrandept by the 1975 version of Soil Taxonomy)
(Ritchie et al., 1990b). Although CROPGRO is relatively new and has not yet been
thoroughly tested, a predecessor, BEANGRO v.1.01, has been tested in Colombia (White
et ah, 1995). Predictions of yield response to population density and water stress were
generally good. Phenology predictions were poorer. Scholberg (1993) adapted
CROPGRO for tomato. Validation work is in progress. Matthews & Hunt (1994)
described an earlier Pascal version of CropSim CASSAVA. Predictions of response to
temperature and photoperiod in Australia and to water stress in Colombia were generally

119
good. Unfortunately, CropSim CASSAVA does not include a plant N submodel; it
simulates the soil N balance but not plant response.
Although the crop models used in this study have been tested in multiple
environments, the demands inherent in characterizing farm sustainability and the
peculiarities of the Andean environment selected for this study necessitate additional
testing.
Model requirements. The application of crop simulation to characterizing farm
sustainability places several demands on the crop models. First, the models must provide
reasonable predictions of development and yields. Development determines the feasibility
of cropping patterns and the timing of resource use. Livelihood of a full-time farmer is
ultimately driven by the productivity of the farm’s crop and livestock populations.
Second, the models should respond realistically to weather variability.
Furthermore, the long-term, stochastic simulation that is needed to characterize farm
sustainability (Chapter 3) requires the ability to generate many sequences of weather data
with statistical properties comparable to those of the historical sequence. The stochastic
weather generator used to generate weather input data should be able to reproduce the
distribution of simulated crop yields obtained from a long sequence of historical data.
Third, the models should predict realistic response to long-term soil nutrient
dynamics and management. Phosphorus, N and erosion are probably the most critical soil-
based determinants of productivity and sustainability on these steep Andisols.
Water balance and N and C dynamics are the only soil processes that the DSSAT3
crop models can simulate. Although work is underway to incorporate P dynamics into the

120
DSSAT3 models, the modifications are unfortunately not yet ready for use. The simulated
water balance changes rapidly within specified, static limits. There is no direct feedback in
the models between soil C dynamics and crop production; soil C affects only the
mineralization or immobilization of N. Nitrogen dynamics is the only mechanism by which
the current models can produce a long-term trend in simulated crop yields. If the models
are to simulate yield trends realistically, they must realistically simulate both long-term N
dynamics and crop response to the N status of the soil.
Crop simulation models that are used to characterize sustainability should respond
realistically to soil loss. The DSSAT3 crop models do not predict soil loss. However,
their layered soil and root submodels provide a means for mimicking erosion by truncating
soil profile input data.
Approach
Weather data
Between November 10, 1993 and April 15, 1994 the farmer recorded temperature
from a portable minimum-maximum thermometer and rainfall from a plastic rain gauge.
After that, solar radiation and minimum and maximum temperatures were recorded by a
LI-COR® minimum data set weather station. The tipping bucket rain gauge was
connected to a separate pulse counter so rainfall intensities would be available for
applications such as erosion prediction. Because of difficulties in obtaining rainfall data

121
and converting it to daily totals, rainfall recorded manually from the plastic rain gauge was
used for simulations. Manually recorded rainfall agreed closely with data integrated from
the pulse counter for a period for which both were available.
Maximum temperatures recorded by the LI-COR® station increased about 3°C
relative to those measured by the portable thermometer. The discrepancy was apparently
due to the location of the sensor of the portable thermometer under the roof of the
farmer’s house, in an area shaded by trees. The LI-COR® sensors were in a standard
weather instrument enclosure located in an open field. Because of the discrepancy,
maximum temperature data recorded before the LI-COR® station was brought on-line
were increased by 3 °C.
Because long-term weather records were not available near the Domingo farm,
weather generator coefficients were estimated from records of surrounding stations using
inverse-squared-distance interpolation. The estimated value of a given weather parameter
at a particular target location (*) in month m was calculated as
Am = E "i-W [5-1]
z = l
where is the value ofy at station /' for month m, and n is the number of stations used
for interpolation. Whenj represented a mean temperature (°C), Eq. [5-1] was adjusted
for an adiabatic lapse rate of 6°C per 1000 m increase in elevation:
Am = ¿ w¡(Am + 0.006 (e.-e,)),
i - \
[5-2]

where em and el are elevation (m) at the target and the /th location, and 0.006 is the
adiabatic lapse rate (°C m'1). The weighting factor, w¡, for station i was calculated as
122
Wi
1 /<
£(l/<)
[5-3]
where is the distance from station j to the interpolated location. The distance, dti,
between any two locations, i and j, can be calculated from their longitude (long) and
latitude (lat) by
¿/¡j = 6366.2 arccos(cos(4) cos(2?)),
A = cos(0.017453 maxd/otfjl, \lat.^)) • 0.017453\long{ - longj|,
B = 0.017453 l/a/j-tajl
(P.G. Jones, 1994, personal communication). Figure 5-8 shows the weather stations used
to estimate monthly weather generator parameters (Table 5-3).
The WeatherMan software package (Hansen et al., 1994) was used to estimate the
missing solar radiation data, using the same adaptation of WGEN that is in the crop
models and the spatially interpolated parameters (Table 5-3). WGEN preserves the
dependence of solar radiation on observed rainfall by sampling from different distributions
on wet and dry days.

123
76.8C
76.6°
76.4c
76.2°
3.0
Santander de Quilichau(990m, P,T,S,R)
Japio (1015 m, P,T)
2.8°
Mondomo (136Q m, P)
2.6'
DOMINGO FARM (1650 m)
Piendamo (184Q m, P)
2.4°
Venta de Cajibio (1800
m, P.T.S)
La Florida, Popayan (1850 m, P,T,S)
Figure 5-8. Locations and elevations of weather stations used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation. Letters represent weather variables recorded (P = precipitation, T =
temperature, S = hours of bright sunshine and R = solar radiation).

Table 5-3. Spatially interpolated WGEN coefficients for the Domingo farm, Cauca, Colombia.
Month Solar radiation (MJ m 2 d1) Temperature (°C) Rainfall
-- dry days —
x s
-- wet days —
x s
maxiii
— dry days --
X s
iium
— wet days --
x s
- minimum -
-- all days --
x s
at
Total
P{DW}}
No.
wet
days
Jan
18.5
4.0
15.5
4.1
26.1
1.7
24.7
1.9
14.5
1.2
0.875
178.5
0.293
13.2
Feb
19.1
4.0
15.5
4.2
26.2
1.8
24.7
1.9
14.7
1.2
0.888
181.5
0.294
13.1
Mar
18.7
4.6
15.2
4.7
26.1
1.6
25.0
2.1
14.8
1.2
0.780
220.3
0.337
15.1
Apr
17.7
4.5
14.3
4.0
25.8
1.5
24.8
1.8
14.9
1.0
0.898
221.5
0.416
16.1
May
17.0
4.4
14.0
3.8
25.8
1.5
24.5
1.8
14.9
1.1
0.740
183.1
0.363
16.0
Jun
17.8
3.6
14.3
3.5
26.0
1.5
24.4
1.7
14.4
1.1
0.803
93.3
0.238
11.2
Jul
18.1
3.7
15.1
3.1
26.2
1.6
24.9
1.6
13.8
1.4
0.857
68.8
0.161
7.8
Aug
18.3
4.0
15.1
3.8
27.0
1.7
24.9
1.9
13.9
1.3
0.743
79.8
0.157
7.6
Sep
18.1
4.2
16.2
4.1
26.5
1.9
25.1
1.9
14.2
1.1
0.935
119.2
0.259
11.1
Oct
16.9
4.4
14.9
4.1
25.4
1.7
24.7
1.8
14.5
1.2
0.865
245.9
0.454
18.0
Nov
17.1
4.1
14.4
3.9
25.4
1.5
24.5
1.8
14.8
1.3
0.872
274.5
0.449
17.9
Dec
17.6
3.9
14.8
3.8
25.6
1.4
24.7
1.7
14.7
1.2
0.951
189.7
0.342
14.2
r Alpha parameter of the gamma distribution.
* Probability that a day is wet and was preceeded by a dry day.

125
Soil data
Soils were sampled at eight locations across the Domingo farm (Fig. 5-2) on
January 24, 1994. At each location, soil was sampled to the depth of the Ap horizon, 30-
40 cm and 50-60 cm deep. Each Ap horizon sample was a composite from a center hole
plus four additional holes 2 m to each side of the center hole. Soil input data from a site 2
m from the edge of the bean trials served as the basis for all crop simulations presented in
this chapter. The soil analysis laboratory at CIAT measured the properties listed in Table
5-2. A moisture release curve was used to determine saturated water content (0 atm) and
lower limit (LL) of plant extractable water (-15 atm). Drained upper limit (DUL) of plant-
extractable water was measured in the field.
The soil data utility program, SDB3 (Hunt et al., 1994), was used to set up soil
profile input data. SDB3 uses soil texture and permeability and drainage codes to estimate
hydrological properties (SCS runoff curve numbers, saturated hydraulic conductivity,
evaporation parameters, and critical water contents) (Ritchie et al., 1990a). SDB3
determined root weighting factors from observed abundance of fine roots and from pH
and Al saturation. Ritchie et al. (1990a) warned that the relationships used by SDB3 to
estimate hydrological parameters should not be used on volcanic soils, therefore measured
critical water contents and conductivity values were used rather than the estimated values.

126
Simulation conditions
The crop simulation analyses described in the remainder of this chapter are based
on the cultivar and planting information in Table 5-4 except as noted. The models were
run with soil water balance, soil N dynamics, canopy-level photosynthesis, and Priestly-
Taylor evaporation options enabled. Replicated simulations used weather data generated
by WGEN based on spatially interpolated parameters.
Table 5-4, Planting information for crop simulation studies.
Crop
Cultivar
method
Planting
date
density
(nr2)
depth
(cm)
Row
width
(cm)
October bean
ICA Caucaya*
seed
Oct 22, 1993
16.60
2
30
March bean
ICA Caucaya
seed
Mar 31, 1994
16.60
2
30
October maize
CIMCALP
seed
Oct 18, 1993
5.00
4
60
March maize
CIMCALI
seed
Mar 30, 1994
5.00
4
60
Cassava
MCol-1501*
cutting
Jan 10, 1993
0.70
10
100
Tomato
sunny, semi-
determinate5
transplant
Mar 30, 1994
0.75
10
150
i Genetic coefficients supplied with DSSAT3.
1 Genetic coefficients calibrated by E.B. Knapp from October 1993 planting.
§ Genetic coefficients supplied by J.M.S. Scholberg.
Initial KCl-extractable NH4+ measurements (Table 5-2) were suspiciously high:
about an order-of-magnitude higher than N03\ In similar soils on neighboring farms,
levels of extractable NH4+ were generally lower than N03'. Therefore, the models were
simulated for one year before each planting date to allow simulated exchangeable NH4+
and N03' to reach a balance with mineralization and other N transformation processes.

127
Development and yield
In 1993, the Hillsides Program of CIAT began conducting trials on several farms in
the Cabuyal area, including the Domingo farm. These trials provide a basis for evaluating
development and yield predictions of the models for bean, maize and cassava. To date,
data are available from four bean harvests, three maize harvests, and one cassava harvest.
The maize and cassava trials each consisted of a single treatment with large inputs of soil
amendments (10-30-10, lime and chicken manure). Two bean trials were conducted. The
first trial consisted of a single fertility treatment. The second bean trial included several
levels of chemical fertilizer and manure. A split plot design was imposed during the
second season; half of each plot received additional amendments while the other half
benefitted only from any residual effect of previous amendments.
Results from the Trujillo farm are included here because they may better represent
attainable, water-limited yields than results of the trials on the Domingo farm. Yields of
all crops were higher on the Trujillo farm than on the other farms studied. Root knot
nematode damage was observed on the roots of beans at the Domingo farm during the
second (March 1994) growing season. Nematode damage was not observed on the
Trujillo farm.
Simulations were run for each treatment of each on-farm trial. The simulations
used observed plant densities, reported planting dates and fertilizer inputs, and measured
weather data. Except for cassava, all of the trials included more than one season and were
therefore simulated using the sequential driver in DSSAT3 (Thornton et al., 1994).

128
Response to environmental factors
Tests of predicted crop responses to weather variability, N dynamics and
management, and soil loss consisted of simple simulation experiments in which the
environmental factor of interest was manipulated. These simulations were replicated with
stochastic weather sequences generated by WGEN. Tests of predicted response to
generated and observed weather variability, mean monthly weather parameters and soil
loss involved additional steps as described below.
Weather variability. I evaluated WGEN by compared simulated yield distributions
driven by historical weather data and driven by WGEN. Weather data were from La
Florida, Popayan (2° 26' N, 76° 35' W, 1850 m). Thirty-six years (1950-1986) of good
data for rainfall, minimum and maximum temperature, and hours of bright sunshine were
available for La Florida. The WeatherMan program estimated daily solar radiation from
bright sunshine duration by Rietveld’s (1978) method, filled-in gaps and errors in data, and
calculated monthly WGEN parameters. CERES-Maize and CROPGRO (Hoogenboom et
al., 1994) were used to simulate early (March) and late (October) plantings of maize and
beans. Each combination of crop and planting date was simulated with 35 years each of
historical and generated weather data. The Kolmogorov-Smirnov two-sample test served
as a basis for comparing yield distributions. Meinke, et al. (1995) used this approach to
compare simulated yield distributions as a means of evaluating weather generation
techniques for tropical and subtropical Australia.

129
Sensitivity to soil loss. Crop response to soil erosion was simulated by reducing
the depth of the top layer and the total profile by 10 cm increments in the soil profile
description and initial soil conditions. When the top layer was completely removed, the
number of layers was reduced and the next layer truncated.
Results and Discussion
Crop development and yield
Bean. Table 5-5 shows simulated and observed yields for the high fertility bean
trials on the Domingo and Trujillo farms. Simulated grain yields matched observed yields
well on the Trujillo farm, but were consistently higher than observed yields on the
Domingo farm. CROPGRO generally over predicted canopy mass and, therefore, under
predicted harvest index. Simulated time to flowering and maturity were earlier than the
observed timing (Table 5-6).
Yield predictions were much better for the Trujillo farm than for the Domingo
farm. Consistently higher yields of beans, maize and cassava in trials on the Trujillo farm
suggest a more favorable environment. Nematode damage observed on the Domingo farm
may account for the over prediction of bean yields.
A decline with time in the ratio of observed to simulated yields at both locations
(Figure 5-9) suggests a gradual buildup of some stress factor that CROPGRO does not
account for. This decline is consistent with increasing losses from the gradual buildup of

130
nematode populations with each successive bean crop. Mullin et al. (1991) identified
root-knot nematodes (Meloidogyne incognita and M. hapla) in soils from bean fields in
Cauca. They measured nematode-induced bean yield losses of 45-63% for cv. Calima and
26-32% for cv. PVA in an experiment at the CIAT headquarters near Palmira.
Table 5-5. Observed and predicted bean yields, Domingo and Trujillo farms, Cauca,
Colombia.
Farm
Planting
date
Plants
per ha
— Canopy mass --
(kg ha1)
obs. pred.
— Grain yield —
(kg ha1)
obs. pred.
Harvest index
obs. pred.
Domingo1
Oct 14,1993
176,389
5
4,964
2,600
2,913
0.587
Trujillo1
Oct 14,1993
117,824
5,057
5,725
3,315
2,679
0.656
0.468
Domingo
Mar 30,1994
89,286
4,494
990
2,100
0.467
Trujillo
Mar 28, 1994
137,269
4,596
5,613
2,644
2,468
0.575
0.440
Domingo
Oct 10, 1994
104,630
2,431
4,704
1,208
2,712
0.497
0.577
Trujillo
Oct 7, 1994
123,889
3,854
5,547
1,978
2,190
0.513
0.395
Domingo
Apr 1,1995
98,920
4,189
627
2,252
0.538
Trujillo
Mar 31,1995
138,819
3,608
5,247
1,727
1,956
0.479
0.373
1 Applied 250 kg 10-30-10 ha'1 each season.
* Applied 1500 kg 10-30-10 ha'1 each season.
§ Canopy dry mass was not recorded.
Table 5-6. Observed and predicted timing of phenological events (days
after planting) for bean and maize, Domingo farm, Cauca, Colombia,
planted March 30, 1994,
Crop
-- Emergence --
-- Flowering --
— Maturity —
obs.
pred.
obs.
pred.
obs.
pred.
bean
6
6
42
36
80
76
maize
t
6
69
66
133
124
f Maize emergence date was not recorded.

Tables 5-7 and 5-8 show simulated and observed yields for the bean fertility
management trial on the Domingo farm. Observed sensitivity to fertilizer or manure
inputs was much greater than predicted by CROPGRO (Table 5-7). This is easily
explained by the confounding effect of P and other nutrients to which the crop models are
not sensitive. While CROPGRO simulates crop response to added N, a substantial
amount of P was added in the 10-30-10 fertilizer and manure treatments. The observed
treatment response was probably more a P than an N response, considering the high P
Jan-94 May-94 Sep-94 Jan-95 May-95
Harvest date
Figure 5-9. Trend in the ratio of bean yields observed in on-farm trials and predicted
by CROPGRO, Domingo and Trujillo farms, Cauca, Colombia.

132
requirements of grain legumes and their ability to fix atmospheric N2. In the second
(March 1994) season of the bean fertility management trial, observed yields were a small
fraction of simulated yields (Table 5-8). Again, this may be due to nematode damage.
Table 5-7. Treatment description and observed and predicted yields for the October 14,
1993 planting of the bean fertility trial, Domingo farm, Cauca, Colombia.
Treatment
symbol
Soil amendments (kg ha'1)
10-30-10 manure
Plants
per ha
- Grain yield (kg ha'1) -
observed predicted
al
0
0
16.5
1,106
2,344
a2
500
0
16.2
1,970
2,550
a3
0
20,000
15.7
2,428
2,318
a4
250
10,000
17.6
2,600
2,914
Table 5-8. Treatment description and observed and predicted yields for the March 30,
1994 planting of the bean fertility trial, Domingo farm, Cauca, Colombia.
Soil amendments (kg ha'1)
- previous season - -- current season -- Plants Grain yield (kg ha'1)
Symbol 10-30-10 manure 10-30-10 manure per ha obs. pred.
bl
0
0
0
0
9.3
320
2,016
b2-
500
0
0
0
9.1
350
1,958
b2+
500
0
200
0
9.3
459
2,057
b3-
0
20,000
0
0
8.7
636
1,910
b3+
0
20,000
0
10,000
9.0
868
t
b4-
250
10,000
0
0
10.2
647
2,204
b4+
250
10,000
250
50,000
8.9
990
t
t CROPGRO cannot simulate manure application in sequential mode after the first season.

133
A plot of simulated vs. observed bean yields in all three bean trials (Fig. 5-10)
shows a poor fit. CROPGRO usually over predicted yields and could not account for the
wide range of observed yields. However, there is reason to believe that much of the
0 500 1000 1500 2000 2500 3000
Observed grain yield, kg/ha
Figure 5-10. Simulated and observed bean grain yields, Domingo (D) and Trujillo
(T) farms, Cauca, Colombia, planted October 1993 (a), March 1994 (b), October
1994 (c) and March 1995 (d). Numbered treatments are from the bean fertility trial
(Tables 5-7 and 5-8), Domingo farm.

134
observed yield variability can be attributed to two factors~P fertility and nematode
damage—to which CROPGRO is not sensitive.
Maize. Simulated and observed maize grain yields are shown in Table 5-9 and Fig.
5-11. Simulated and observed harvest indices were reasonably consistent. Predicted time
to silking and physiological maturity was earlier than observed (Table 5-6). Although
yield predictions were not obviously biased, CERES clearly did not account for the range
of observed variability of yields (Fig. 5-11). It is more difficult with maize than with bean
to identify the determinants of yield variability that the model does not account for.
Table 5-9. Observed and predicted maize yields, Domingo and Trujillo farms, Cauca,
Colombia.
Farm
Planting
date
Plants
per ha
— Canopy mass —
(kg ha1)
obs. pred.
— Grain yield —
(kg ha1)
obs. pred.
Harvest index
obs. pred.
Domingo
Oct 14, 1993
50,309
12,210
12,382
4,664
4,683
0.382
0.378
Trujillo
Oct 14, 1993
55,556
21,016
13,423
6,746
4,877
0.321
0.363
Domingo
Mar 30,1994
47,778
7,600
11,788
2,310
5,034
0.304
0.427
Trujillo
Mar 28, 1994
62,500
14,179
14,381
5,274
5,584
0.372
0.388
Domingo
Oct 10, 1994
48,750
12,557
11,444
5,186
4,585
0.413
0.401
Trujillo
Oct 7, 1994
50,000
16,129
11,617
6,048
4,625
0.375
0.398
Cassava. The simulated cassava yield was close to the observed yield on the
Trujillo farm, but over predicted yield on the Domingo farm (Table 5-10). Because there
was only one treatment grown for one season, few inferences can be made about the
performance of CropSim CASSAVA in this environment. However, farmers report

135
Table 5-10. Observed and predicted cassava yields, Domingo and
Trujillo farms, Cauca, Colombia, 1995,
Plants
Root yield (kg ha1)
Farm
per ha
fresh
dry —
obs.
pred.
Domingo
11,806
24,549
8,890
20,278
Trujillo
14,577
64,757
19,943
23,362
Figure 5-11. Simulated and observed maize grain yields, Domingo and Trujillo
farms, Cauca, Colombia, planted October 1993, March 1994 and October 1994.

136
typical yields of about 10 Mg ha'1 of dry root grown for the starch market (Knapp,
personal communication). CropSim CASSAVA does not simulate plant response to N
stress and therefore is expected to over predict yields under the low level of soil fertility
management typical for the region.
Tomato. Simulated fresh fruit yields were 24,833 ± 1729 kg ha'1 (±SD, n = 10)
based on 6% dry matter for tomato planted April 1, with 75 kg applied N ha'1 and
automatic irrigation. Tomato was not included in any of CIAT’s on-farm trials. However,
a farmer in Siberia, located about 2 km from the Domingo farm, reported obtaining yields
for irrigated tomato ranging from 7200 to 8400 kg fresh fruit ha'1. The farmer’s reported
yields were lower than expected; FEDCAFE (1986) reported that the average tomato
yield in the Cauca region was 27,000 kg ha'1 in 1983.
Tomato is susceptible to a range of diseases and pests. I observed severe wilting
and foliar necrosis in a field of tomatoes near the Domingo farm. These symptoms were
not consistent with physiological maturity. Furthermore, many tomato cultivars are
susceptible to the same root knot nematodes (Meloidogyne spp.) that apparently reduced
bean yields (Overman, 1991).
Weather variability
Table 5-11 and Figure 5-12 present distributions of simulated yields in response to
actual and generated weather variability. With the exception of October-planted bean,
yield distributions were not significantly affected by the source of weather data (P = 0.05).
Although standard deviations of simulated time to maturity was consistently lower with

137
generated than with historical weather, distributions differed significantly (P = 0.05) only
for October-planted maize. The Kolmogorov-Smimov test did not identify differences
between generated and observed distributions of rainfall amounts grouped by calendar
month or by year (Table 5-12).
Only one of the four cases (October-planted bean) was consistent with the results
of Jones and Thornton (1993), who used CERES-Maize to test rainfall generation
techniques for conditions at the CIAT headquarters near Palmira, Colombia. In their
study, the variance of simulated yield distributions was significantly lower when WGEN
was used instead of historical weather data.
Table 5-11. Mean (x), standard deviation (s) skewness (a3), and Kolmogorov-Smimov
test statistic (D) for distributions of crop yield and maturity times simulated with observed
and simulated weather from La Florida, Popayan, Colombia.
Planting
date
Weather
source
X
s
X
Maize —■
S «3
D
«3
D
Grain yield, kg ha'1:
October
observed
2223
209.0
0.126
0.343 *
4728
389.6 0.057
0.114 n.s.
generated
2336
162.6
0.284
4730
489.8 0.913
March
observed
1561
175.5
0.363
0.143 n.s.
4322
414.5 -0.059
0.286 n.s.
generated
1564
188.2
-0.139
4161
307.5 1.061
Days to physiological maturity:
October
observed
83
2.7
0.052
0.257 n.s.
149
6.8 -0.474
0.343 *
generated
83
1.6
0.509
148
3.2 -0.395
March
observed
81
2.7
0.092
0.286 n.s.
145
7.1 -0.161
0.229 n.s.
generated
81
1.2
-0.733
145
4.2 -0.191

Probability Probability
138
Figure 5-12. Distribution of simulated yields of October- and March-planted maize
(a) and bean (b) in response to historical and generated weather variability, La
Florida, Popayan, Colombia.

139
Table 5-12. Mean (x), standard deviation (5) skewness (a3), and Kolmogorov-Smimov
test statistic (.D) for distributions of observed and simulated monthly and annual rainfall
totals, La Florida, Popayan, Colombia.
Time
grouping
Observed-
x s
<*3
Generated
X s
«3
D
January
168.6
80.2
-0.062
162.0
65.6
0.312
0.135 n.s.
February
158.2
70.2
-0.190
153.6
60.9
0.903
0.216 n.s.
March
187.9
85.8
471
198.7
66.4
0.012
0.243 n.s.
April
191.5
69.8
552
190.1
61.7
0.451
0.108 n.s.
May
153.7
67.4
1.297
160.7
56.1
0.571
0.108 n.s.
June
94.3
48.1
0.199
82.8
30.6
-0.108
0.270 n.s.
July
42.6
44.3
2.593
42.4
18.7
0.148
0.189 n.s.
August
48.8
38.5
0.894
44.1
29.4
0.483
0.108 n.s.
September
104.2
54.6
0.628
94.0
36.7
0.858
0.243 n.s.
October
269.0
87.0
0.771
298.6
77.7
0.294
0.243 n.s.
November
293.4
88.4
0.007
304.1
93.1
0.496
0.131 n.s.
December
260.3
93.3
0.319
272.6
99.2
1.458
0.081 n.s.
Annual
1969
293.4
0.606
2005
253.7
0.075
0.162 n.s.
Response to nitrogen dynamics
Nitrogen mineralization. With the default N mineralization parameters, simulated
maize showed no yield response to N fertilizer applications whereas bean (using the
modified model described in the following section) showed a moderate response (Fig. 5-
13). No on-farm yield data were available for maize grown with adequate P and varying
levels of N fertilizer. Maize response to applied N varied among several volcanic soils in

140
Figure 5-13. Simulated grain and biomass yield response of maize (a, b) and October-
(c, d) and March-planted (e, f) bean to applied N using the default mineralization factor
(SLNF = 1.00), Domingo farm, Cauca, Colombia. Mean ± SD of 10 replicates.

141
Nariño (Muños & Wieczoreck, 1978); maize showed a strong response to N fertilizer in
only three of the six soils shown in Fig. 5-14.
I varied the mineralization rate factor (SLNF) in the soil profile description (Fig. 5-
15). In the absence of observed maize N response data for the study area, I selected a
value of 0.35 for SLNF because it produced a reasonable yield response curve based on
the experience of CIAT scientists (E.B. Knapp, 1994, personal communication). This
Figure 5-14. Maize response to applied N on several volcanic soils in Nariño,
Colombia. Data from Muñoz & Wieczoreck (1978).

142
value indicates that N is released by mineralization at 35% of the rate simulated for a
comparable soil with crystalline mineralogy. Figure 5-16 shows the simulated response of
maize and bean to applied N with the reduced mineralization factor.
. Bean response to applied nitrogen. With the N mineralization calibration, bean
simulated with CROPGRO showed a dramatic and unexpected decrease in simulated
yields in response to small amounts of applied N (Figs. 5-17a and c). Simulated response
to applied N was more erratic for soybean grown on a sandy soil in Gainesville, Florida
(Fig. 5-18) than for bean at the study site.
Fig. 5-15. Effect ofN mineralization factor, SLNF, on simulated maize response to
applied N, Domingo farm, Cauca, Colombia.

143
5000
nj4000
€
2
.¿3000
'5k
.=2000
(0
O
1000
0
Figure 5-16. Simulated grain and biomass yield response of maize (a, b) and October-
(c, d) and March-planted (e, f) bean to applied N using the adjusted mineralization
factor (SLNF = 0.35), Domingo farm, Cauca, Colombia. Mean ± SD of 10 replicates.

Canopy biomass, kg/ha Grain yield, kg/ha
October planting
March planting
3000 • ■'
4500
3500
2500
Figure 5-17. Grain and biomass yields of October- (a and b) and March-planted (c and d) bean simulated by the
original and modified versions of CROPGRO, Domingo farm, Cauca, Colombia.
4*.

Canopy biomass, kg/ha Grain yield, kg/ha
Irrigated
Rainfed
Figure 5-18. Grain and biomass yields of irrigated (a and b) and rainfed (c and d) soybean (cv. Bragg) simulated
by the original and modified versions of CROPGRO, Gainesville, Florida, 1978.

146
In CROPGRO, nodule growth and N2 fixation are driven by C balance
(Hoogenboom et al., 1990). In a given day, C is allocated according to a priority scheme:
maintenance respiration > reproductive growth > vegetative growth > N2 fixation >
nodule growth. Carbon is allocated to N2 fixation based on N deficit only when
insufficient N prevents the plant from using all available C for growth. Nodule growth can
then occur only if inadequate nodule mass (a sink limitation) prevents using all of the C
allocated for N2 fixation. The sink limitation to nodule growth is based on current nodule
mass and a maximum daily relative growth rate. A result of the priority scheme is that
nodule growth, and therefore the capacity to fix N2 during seed formation, is extremely
sensitive to small changes in N supply early in the season (Fig. 5-19).
Tewari (1995) measured nodule growth and N2 fixation in beans grown with 0 and
275 kg ha'1 of N fertilizer in an experiment in Kuiaha, Hawaii. CROPGRO over predicted
the observed response to applied N, predicting nearly complete inhibition of nodule
growth and N2 fixation (Fig. 5-20a and c). I modified CROPGRO to force it to reserve a
minimum amount of C for nodule growth. The modification reserves an amount of C
(CNODMN) for nodule growth equivalent to a fixed fraction (FRCNOD) of the C
allocated to current root growth. Any surplus C remaining after N2 fixation is added to
CNODMN for nodule growth. Carbon allocated to nodule growth may be unused and
made available for vegetative growth if current nodule mass limits growth. Carbon is
allocated to CNODMN only up to first pod.
Three parameters were available for calibrating nodule growth in the modified
version of CROPGRO: FRCNOD, initial nodule mass (DWNODI), and maximum relative

Figure 5-19. Effect of applied N on bean nodule growth (a, b) and cumulative N2 fixed (c, d) observed and
simulated with the original (a, c) and modified (b, d) versions of CROPGRO, Domingo farm, Cauca, Colombia,
1994.
147

O 20 40 60 80
Days after planting
Figure 5-20. Effect of applied N on bean nodule growth (a, b) and cumulative N2 fixed (c, d) observed and
simulated with the original (a, c) and modified (b, d) versions of CROPGRO, Kuiaha, Hawaii, 1993. Data from
Tewari (1995).
oo

149
growth rate (NODRGM). Criteria for calibrating these parameters were (a) to preserve
predicted biomass and grain yields in the absence of applied N, (b) to eliminate the
predicted drop in grain yield with the first increment of applied N, and (c) to reproduce the
suppression of nodule growth observed in the experimental study by Tewari (1995).
Based on these criteria, I obtained values of 0.003 for DWNODI, 0.17 for NODRGM, and
0.04 for FRCNOD. The modification reduced sensitivity of nodule growth to small
amounts of applied N and eliminated the delay in nodule growth (Fig. 5-19b). It also
improved predictions of nodule growth and N2 fixation observed in Tewari’s study,
although CROPGRO still tended to under predict both (Fig. 5-20b and d).
Figure 5-17 shows simulated bean response to applied N with the original model
and with the calibrated modifications. Simulated canopy biomass responded to increasing
levels of applied N as expected, increasing at a decreasing rate to a maximum, then
maintaining a plateau. Simulated soybean response to applied N was also more consistent
with expectations with the modified model (Fig 5-18).
The predicted suppression of grain yield of October planted bean with intermediate
levels of applied N simulated with the modified version of CROPGRO (Fig. 5-17) may be
reasonable. This yield reduction can be attributed to suppression of early nodule growth
and a resulting loss of capacity to fix N for reproductive growth. Larger applications of N
fertilizer provide sufficient N2 for reproductive growth. In a pot experiment with three
bean cultivars and three strains of Rhizobium inoculum, plant mass at 35 days increased
(Fig. 5-2la) and nodule mass decreased (Fig 5-2lc) in response to increasing amounts of

150
N applied, mg/kg soil
Figure 5-21. Plant (a) and nodule (b) mass at 35 days, and grain yield (c) of container-
grown beans in response to applied N. Mean of 3 cultivare x 3 Rhizobium strains. Data
from Rai (1992).

151
applied N (Rai, 1992). However, the final seed yield was consistently lower with 25 mg
of applied N per kg of soil than with either 12.5 or 40 mg kg'1 (Fig. 5-21b).
Long-term trends. Figure 5-22 shows mean simulated yields for 60-year
sequences of maize with different levels of applied fertilizer. Although ANOVA of results
indicated significant differences among years (P = 0.05), regression analysis did not
indicate a significant linear trend in yields for any fertilizer treatment.
Figure 5-22. Simulated 60 year sequences of maize grain yields at different levels of
applied N, Domingo farm, Cauca, Colombia. Mean of 20 replicates.

152
Response to soil erosion
Simulated soil loss resulted in a small decrease in maize grain yield (Fig. 5-23a).
Simulated March-planted bean showed little response to soil loss, while October-planted
bean showed a slight increase in yield in response to soil loss (Fig. 5-23b).
Although experimental erosion response data from Andisols are not available,
there is reason to expect a substantial reduction of crop yields as the Ap horizon is eroded.
First, both quantities and availability of N and P are greater in the Ap than in the Bw
horizon (Table 5-2). Second, roots of existing vegetation are much more dense in the Ap
horizon. There is a sharp break in soil color, structure, and rooting between the Ap and
Bw horizons.
The lack of response to simulated erosion is a result of assumptions built into the
root growth model in the IBSNAT crop models. In these models, root growth is source-
driven, with adjustments for water and N stress. The soil data management program
(SDB3) used to build profile descriptions calculates root weighting factors for each soil
layer / (SRGF^ based on observed root density and modified by low pH or high A1
saturation. Aluminum toxicity as indicated by high A1 saturation inhibits root elongation.
Observed root abundance integrates the spatial partitioning of dry matter determined by
branching habit and meristem activity with a range of chemical and physical stresses that
reduce the quantity of roots grown. Hence, SRGF is determined by SDB3 partially as a
sink limitation. However, SRGF has no direct effect on total root growth in the crop
models. It determines only how growth is distributed within the soil profile. As a result,

Soil loss, cm
Figure 5-23. Simulated grain yields of maize (a) and bean (b) in response to soil
loss, Domingo Farm, Cauca, Colombia. Mean of 10 replicates.

154
removing the topsoil that is favorable to root growth forces more root growth lower in the
profile (Fig. 5-24). As soil is lost, the effect of a decreasing total supply of N is offset by
deeper rooting and better access to water and N lower in the soil. I modified the root
growth model in CROPGRO and CERES to increase sensitivity to soil loss. The modified
model attempts to divide the process of distributing new root growth into two
components: (a) partitioning among soil layers based on branching habit, and (b) reduction
of root elongation based on stresses encountered in each soil layer. The approach is to
compare the vertical distribution of new root growth specified by SRGF with a
distribution that would be expected in an ideal, homogeneous soil that imposes no
restrictions on root growth. A distribution factor, F, represents the vertical partitioning of
new root in this ideal soil. Jones et al. (1991a) used a function of the form,
F. = (1 - r./300)*,
to represent the distribution of roots based on growth habit in the absence of soil stresses,
where i is soil layer number, z¡ is the depth of the middle of layer / (cm), and g is an
exponent that varies with plant species. The modified models use g = 2, which Jones et al.
recommended for maize and soybean.
The first step was to normalize SRGF for each layer / in the description of the
uneroded soil profile such that the relative weighting of the layers remains the same, but
SRGF, = 1:
SRGF. = SRGF. / SRGF,.

Depth, cm Depth, cm Depth, cm
155
October planting
Figure 5-24. Simulated root distributions of October- (a - c) and March-planted (d -f)
bean in response to 0 cm (a and d), 20 cm (b and e), and 50 cm (c and f) of soil loss,
Domingo farm, Cauca, Colombia. Mean of 10 replicates.
50 cm soil loss 20 cm soil loss 0 cm soil loss

156
The values of F were normalized in the same manner. The values of F and SRGF for the
top layer are thus set equal to each other (F, = SRGF¡ = 1.0) based on the assumption that
an uneroded topsoil does not inhibit root growth. We can then assume that if for a lower
layer i, SRGF¡ < F¡, the difference between SRGF¡ and F¡, represents some source of stress
relative to the ideal soil. This stress would reduce new root growth in layer i without
redistributing it to other soil layers.
The root models require two modifications. First, F replaces SRGF for
partitioning new growth among soil layers. Second, where SRGF < F, root growth is
reduced according to the ratio of the two distribution factors:
ARLD{ / A/ = min(S7?GF / F¡, 1) • potential growth - respiration,
where RLDt is root length density in layer i (cm root cm'3 soil), potential growth refers to
root growth apart from soil stress, and the units of potential growth and respiration are
cm cm'3 day'1.
Figure 5-25 illustrates how the modification increases sensitivity to simulated soil
loss on the Domingo farm. The shaded region between SRGF and F in Fig. 5-25a shows
the layers in which stress occurs in the uneroded soil. When the soil profile is truncated to
simulate erosion, the distribution of new root growth for an ideal soil (F) remains the
same, but the soil layers with reduced relative root growth indicated by low values of
SRGF become shallower. The area of stress between SRGF and F increases, particularly
in the upper layers where roots are concentrated (Fig. 5-25b and c). Reduced root growth
feeds back to shoot growth and grain yield by reduced ability to access water and N.

157
0.2 0.4 0.6 0.8
Relative distribution factor
Figure 5-25. Relative root distribution factors, F and SRGF, after 0 cm (a), 20 cm
(b) and 50 cm (c) of soil loss. SRGF was calculated by SDB3 for soil near the bean
trial, Domingo farm, Cauca, Colombia.

158
Bean yields in the uneroded soil were lower when simulated with the modified CROPGRO
than with the original version. The soil photosynthesis factor (SLPF) was increased from
its default value of 1.00 to 1.14 to raise yields simulated by the modified model to their
level before the modification. The CERES model does not respond to SLPF. However,
the effect of the modification on simulated maize yields was negligible.
The modified root growth model increased sensitivity of simulated bean to soil loss
(Fig. 5-26). The effect of the modification on simulated maize response to soil loss was
negligible. Apparently grain yields simulated by CERES are much less sensitive to root
length density than are bean yields simulated by CROPGRO. However, the response of
maize yields to soil loss simulated without the modification (Fig. 5-23a) was adequate to
demonstrate the impact of soil erosion on farm sustainability (Chapter 6).
Linking Models
Issues in linking crop and whole-farm models
The problem of enabling a farm model written in object-oriented Pascal (Chapter
4) to use crop models written in FORTRAN was easily solved by executing the crop
models from a call to the operating system, using text files to communicate between
models. The farm model assigns management strategies at the beginning of a simulation
year, passes the management information to a crop model when it is executed, obtains
results of the crop simulation, then processes those results. A problem arises when a

O 10 20 30 40 50
Soil loss, cm
Figure 5-26. Simulated grain yields of October- (a) and March-planted (b)
bean in response to soil loss, Domingo Farm, Cauca, Colombia.

160
strategy calls for a resource that is unavailable because of competing use by another
enterprise. The current model structure requires that each crop be simulated for an entire
season with a given set of resource requirements. There is no mechanism for feedback by
which resource constraints can alter crop performance. The appropriate way to deal with
the resource allocation problem is to resolve conflicts on a daily time step before
simulating crop and ecosystem processes (Fig. 5-27). This is not possible with the present
crop model structure.
For each simulation day
For each field
For each operation scheduled for today
Determine resources required.
For each resource
If there is a conflict (i.e., requirement > supply) then
Resolve the conflict through rescheduling or reallocation.
For each field
Simulate the adjusted operations.
Simulate crop and ecosystem processes,
end { for each field }.
end ( for each simulation day }.
Figure 5-27. Pseudocode representation of an algorithm for resolving resource conflicts
among crop enterprises.
The resource allocation problem could be solved by reorganizing the crop models
along hierarchical boundaries so that the ecosystem could function independently of crop
populations, then linking the models at code-level to a whole-farm model so that all
ecosystems on a farm could be simulated during any given simulation day. Caldwell and
Hansen (1993) demonstrated a hierarchical model structure. They reorganized a set of

161
IBSNAT crop models along hierarchical boundaries so that the ecosystem could run
continuously, and crop populations could be inserted or removed from the ecosystem at
any time. A farm-level implementation would need the ability to insert a number of
agroecosystems, and simulate all of them on a daily time step as in Fig. 5-27. This could
be accomplished most easily with an object-oriented design and implementation.
Issues in linking crop and erosion models
The IBSNAT models do not simulate soil loss. The most widely used soil erosion
prediction model is the universal soil loss equation (USLE, Wischmeier & Smith, 1978).
The USLE is a simple empirical model that aggregates the processes involved in soil
erosion. A single rainfall erosivity term based on energy and intensity aggregates the
effects of both rainfall and runoff. A single soil erodibility term aggregates susceptibility
to detachment by raindrop impact, detachment by rill streamflow, and transport by
streamflow. There is reason to expect the USLE to be a poor predictor of loss of
Andisols in steep, complex slopes, where resistance to detachment may be moderate to
poor, but runoff and therefore sediment transport is rare because of the very high
infiltration capacities of the soils.
The Water Erosion Prediction Project (WEPP) model (Lane & Nearing, 1989)
simulates the processes of interrill and rill detachment, transport and deposition, as well as
the hydrological processes that drive water erosion. Although it has not yet been tested
for steep, volcanic ash-derived soils, the abilities of WEPP to predict runoff and to
separate raindrop- and runoff-induced erosion are reason for optimism. WEPP

162
incorporates a simple crop growth model that predicts canopy height and cover, biomass
production, and root growth based on thermal time (Alberts et al., 1989). However, it
possesses no mechanism for predicting impacts of soil erosion on crop production.
An early goal of this research was to link the DSSAT3 crop models with the
erosion component of WEPP. However, several points of incompatibility could not be
resolved within the time frame required. First, the crop models do not provide the
information about crop residues, canopy cover and tillage that the erosion component of
WEPP requires. The relative amount of crop residue standing, flat and buried can impact
predicted soil loss. Residue management is an important tool for managing erosion.
However, the crop models cannot simulate surface residue decomposition but rather
assume that all residue is buried at harvest time. Because the crop models running in
sequence apply all stover as buried residue at harvest time, mass balance errors occur if
either the harvest product is not completely removed or a portion of the stover is
removed. Although the input data for the crop models include tillage, tillage events are
ignored. Tillage affects erosion predictions by influencing surface roughness, infiltration
capacity and the proportion of residue remaining on the surface. The degree of canopy
cover, calculated as a function of leaf area index (LAI) and canopy dimensions, affects
predicted erosion. Although all of the models calculate LAI, only CROPGRO outputs
canopy height and width. The DSSAT3 crop models would require a surface residue
model, a tillage model, and a model of canopy dimensions before they would be
compatible with the erosion submodel of WEPP. Errors in the existing residue model
must also be addressed.

163
Second, the hillslope profile version of WEPP is two-dimensional, while the
DSSAT3 crop models are one-dimensional (including the vertical dimension). WEPP
divides a hillslope into homogeneous overland flow elements (OFEs). Runoff calculations
are particularly problematic. The erosion submodel requires peak runoff rate and effective
runoff duration, which are derived from a hydrograph. The hydrograph of a given storm is
obtained by analytical or approximate solution of the kinematic wave equation, which is
continuous both in time and distance along the hillslope (Hernandez et al., 1989). A
complex slope with several OFEs is reduced to an equivalent plane for the purpose of
developing the hydrograph. Runoff is then estimated for particular points on the hillslope.
It is not possible to solve the kinematic wave equation for each OFE because the
kinematic wave equation assumes runoff depth of zero at the top of the slope. A result of
the method for obtaining the hydrograph is that runoff from a particular OFE is dependent
not only on rainfall, infiltration and run-on from the adjacent up-slope OFE, but on all of
the OFEs both uphill and downhill. Each simulation day, overland hydraulics simulated by
WEPP requires soil surface and antecedent water contents from the crop models, and the
soil water balance simulated by the crop models requires infiltration calculated by the
WEPP hydraulics model.
In their current form, the crop models must simulate an entire cropping season for
one OFE before simulating the next. However, because of the interdependence OFEs in a
hillslope, linking WEPP with process-level crop simulation models would require the
ability each simulation day to simulate overland hydraulics for the entire hillslope, then
simulate the current day's crop growth and development for each OFE (Fig. 5-28). This

164
could be accomplished by the type of restructuring of the crop models described in the
preceding section. However, instead of the farm landscape consisting of a collection of
field-scale agroecosystems, it would consist of a collection of hillslopes. Following the
algorithm in Fig. 5-28, each hillslope would simulate overland hydraulics, and then request
all of its OFE-scale agroecosystems to simulate soil and crop processes.
Algorithm
Model
For each simulation day
For each OFE
Obtain surface conditions and water contents.
crop model
Calculate hydrograph for the entire hillslope.
For each OFE
WEPP
Calculate peak runoff rate and duration.
WEPP
Determine infiltration.
WEPP
Calculate soil loss.
WEPP
Adjust soil depths.
crop model
Simulate soil water balance.
crop model
Simulate crop growth,
end {for each OFE}.
end {for each simulation day}.
crop model
Figure 5-28. Pseudocode representation of an algorithm for simulating erosion and crop
growth on a complex hillslope.
Conclusions
The task of characterizing sustainability of an Andean hillside farm challenge the
capabilities of crop simulation models. The DSSAT3 family of crop models provides a
useful starting point for exploring field-level constraints to sustainability. Their use in

165
conjunction with the WGEN weather generator appears to be useful for quantifying
weather-induced risk. However, comparison of multiple-year bean and maize trials with
simulation results suggests that there is substantial variability between years that the
models are not able to capture.
There appear to be several soil-related constraints to crop production in the study
area that the crop models do not address. These include low availability and high
buffering of soil P, nematode damage, impacts of soil erosion, and the threat of soil loss by
mass movements. Anticipated additions of soil and plant P submodels will enhance the
usefulness of the models on the Andisols of the Cauca region of Colombia.
The current source-driven root growth model is not adequate for simulating
response to soil loss. Its assumption that removing a favorable topsoil layer will force
deeper rooting is not consistent with what we know about root response to soil stresses.
A simple alternative model that reduces the absolute amount of root growth increased the
sensitivity of simulated bean yields to soil loss. More work is needed on the root growth
components of the crop models before they will realistically simulate crop response to
erosion.
Although the N component of the crop models has undergone extensive
development and testing, some of its assumptions do not hold well on Andisols.
Allophane retards mineralization of organic matter. This study confirms the need to
calibrate the N mineralization factor (SLNF). However, a single multiplier may not
adequately calibrate mineralization rate because the effect of allophane on mineralization is

166
greater for humified than for fresh organic material. Although the mineralization rate
factor needs to be calibrated for Andisols, standard data collection procedures do not
include a method for doing so (IBSNAT, 1989). This study used expected maize response
to applied N as a basis for calibrating SLNF. An approach is needed for adjusting
mineralization rates based on readily measured soil properties.
This study did not test the assumptions that NH4+ is retained by the exchange
complex and N03‘ moves freely with soil water. Since Andisols may have either net
negative or net positive charge depending on mineralogy, pH and soil amendments, both
of these assumptions are suspect. A more general model of ion movement might improve
predictions of N availability and leaching. Bowen et al. (1993) presented a modification
that accounts for N03' retention.
Adding P submodels, refining the soil N submodel and improving assumptions
about root physiology and response to soil properties are quite feasible. Reorganizing the
models along hierarchical boundaries is a more difficult task. Such restructuring would
allow the models to deal appropriately with farm-level resource allocation and would
facilitate linking the crop models with a two-dimensional hydrology and erosion model
such as WEPP. Previous experience demonstrated that reorganizing the models along
hierarchical boundaries is feasible and solves the problem of simulating sequences and
facilitates simulating interacting combinations of crops (Caldwell & Hansen, 1993).

CHAPTER 6
DETERMINANTS OF SUSTAINABILITY OF A COLOMBIAN HILLSIDE FARM
Introduction
The Cauca River watershed in the lower Andes of southwestern Colombia is
undergoing a transition that challenges its smallholder farmers. Coffee previously
provided a relatively high but variable level of income to farmers, while providing
continuous canopy and ground cover to protect the steep soils from erosion. Declining
world coffee prices and the spread of a new seed-boring insect pest, broca (Hypothenemus
hampei) have reduced the ability of coffee to provide for farmers’ livelihood. The
Colombian Coffee Federation in the early 1990s offered one-time monetary incentives for
farmers in marginal areas to abandon coffee production in order to slow the spread of
broca and to boost market prices. Farmers who participated found themselves with some
financial capital but few good alternative enterprises and a highly uncertain future.
Annual crops are now replacing coffee in many of the hillside farms in Colombia.
The shift from perennial to annual crops represents a shortening of farmers’ planning
horizons. Ashby (1985) argued that low profit margins and high price risk have forced
farmers in the region to give short-term returns priority over investing in the long-term
benefits of soil conservation. The annual crops that are replacing coffee leave steep lands
167

168
exposed to erosive rains. Resulting soil erosion may irreversibly reduce crop growth and
yield. Furthermore, the Cauca River watershed is strategic as a source of water and
hydroelectric power for the city of Cali and for irrigated sugar production in the Cauca
Valley. There is growing concern that the shift from perennial crops will result in
increasing siltation and chemical pollution, and poorer regulation of flow rates in the
Cauca river.
On the positive side, the strategic Panamericana highway runs through the Cauca
river watershed, potentially providing access to the rapidly growing urban market of Cali.
In spite of the proximity of the Panamericana, most of the bean, maize and cassava—the
traditional crops for the region—are grown for subsistence consumption or for local
markets. Commercial production of specialty commodities, such as vegetables and cut
flowers is in an early stage of development.
Chapter 3 describes a framework for using long-term, stochastic simulation of a
farming system to characterize its sustainability. The purpose of the current chapter is to
apply that framework to a farm in the Cauca River watershed in order to gain insight into
the impact of cropping system, soil management, costs and prices, resources and sources
of risk on farm sustainability. This chapter has two specific objectives: (a) to demonstrate
and evaluate a systems approach for characterizing farm sustainability, and (b) to test a set
of hypotheses (Table 6-1) about determinants of sustainability of a particular hillside farm
in the Cauca watershed of Colombia.
A word of caution regarding interpretation of results is in order. Although this
study is based on an actual farm, it does not try to mimic all of the characteristics or

169
Table 6-1. Hypotheses related to determinants of farm sustainability, Domingo farm,
Cauca, Colombia.
Determinant
Hypothesis
Cropping system
Farm sustainability is dependent on cropping system.
Diversified annual crop rotations contribute to a more sustainable
farming system than do monocultures.
Incorporating a high-valued, irrigated vegetable (i.e., tomato) into a
diversified rotation of traditional crops enhances farm sustainability.
Sustainability of a farm in coffee monoculture is positively related
to coffee yield.
Soil management
Either excessive or insufficient N fertilizer inputs reduce farm
sustainability.
Soil erosion reduces farm sustainability.
Costs and prices
Low prices for crop products constrain farm sustainability.
High material input prices constrain farm sustainability.
High wages for hired labor constrain farm sustainability.
A high subsistence spending requirement constrains farm
sustainability.
A high level of discretionary spending constrains farm
sustainability.
Resources
The decision to not cultivate the degraded land now in permanent
grass fallow constrains farm sustainability.
Limited initial savings constrains farm sustainability.
Access to credit enhances farm sustainability.
When credit is available, farm sustainability is negatively related to
loan interest rate.
Sources of risk
Conventional credit enhances farm sustainability more than credit
packages given for specific crops.
Weather and price variability contribute unequally to whole-farm
risk and to the probability of failure.
Spatial diversity reduces farm risk and enhances sustainability.

170
predict the sustainability of that farm. Inferences apply not to the real farming system, but
to a model of the system that is forced to follow the management practices defined in each
scenario.
Approach
A Colombian hillside farm
A farm located in the Cabuyal River catchment (2° 47' N, 76° 31' W) (Fig. 5-1)
was selected as a basis for the simulation study. Chapter 5 describes the farm’s physical
environment. Farm selection was based on several criteria. The size of the farm (5.2 ha,
of which 2.3 ha is cultivated with annual crops), its elevation (1650 m), topography and
soils, and the size of the farm household (6 members) are considered representative of
farms in the Cabuyal area. However, the farmer is regarded as an innovator. He has
cooperated with CIAT researchers by providing information, recording weather data, and
providing land for on-farm maize, bean and cassava trials.
The farm supports six family members: the farmer (Mr. Domingo), his mother,
wife, two daughters and a son. The eldest daughter is attending a state university, and the
second plans to soon. The Domingo family lives in a five-room stucco house of above-
average quality for the area. Access to the farm is by a well-maintained gravel road 2 km
from a small town and 4 km from the Panamericana highway. Coffee was the only
commercial enterprise until the farmer accepted the Colombian Coffee Federation’s
monetary incentive to remove his coffee plants in 1993. Since then, he has grown maize

171
for sale as green ear (choclo) for the fresh market and beans for market and household
consumption. Because of a shortage of family labor, he routinely hires day-laborers on a
contract basis for field operations. The farmer’s level of technical knowledge is high, but
he lacks capital—he has little savings, and credit is scarce and terms are unfavorable.
Production inputs such as seed, organic and inorganic fertilizer, and pesticides are readily
available.
Sources of information
Chapter 5 discusses how soil descriptions and weather data were obtained. CIAT
contracted a professional surveyor to map the farm’s topography and land use. The
description of the farm household, resources and management of crop enterprises came
from a formal survey conducted by CIAT in September 1993, and from subsequent
interviews with the farmer. Additional information about crop management came from
literature, unpublished enterprise budgets, an interview with a nearby tomato farmer, and
from discussions with CIAT personnel familiar with the region. Table 6-2 presents several
estimates of labor requirements for crop management operations.
Historical price data are from wholesaler records for cassava, maize and tomato
(CAVASA, 1994), from a study of retail prices of beans (Castillo, 1990, 1993), and from
published sources for chemical inputs (Anonymous, 1994), coffee and consumer price
indices (CPI) (Banco de la República, various dates). Prices were deflated to a constant
December 1992 basis using monthly CPI for the city of Cali. Prices are presented in Fig.
6-1 for crops and in Table 6-3 for purchased inputs and labor. Tables 6-4 and 6-5 show

172
1600
'! 1400
o
O) 1200 •
¡» 1000
3 800 -
600
500
o
p
IA
<1>
400
a> 300
ó
O
200 â– 
100 â– 
«o 200
v
Í 175
O) 150
o>
¿í 125 •
— 100
o
° 75 J
500
¿ 400
O)
300 -

Col
200 -
100 -
<
Q)
a>

1600 -
1400 -
O)
1200 -
«(»
1000 -
o
O
800 -
600 -
Oct-69
bean
cassava
maize
(choclo)
tomato
coffee
i t i
Oct-73
\J\
date
Figure 6-1. Historical crop wholsale prices, Cauca Valley, Colombia.

Table 6-2, Estimates of labor requirements (man-days ha'1 unless otherwise indicated) for annual crop production, Cauca, Colombia
Operation
Maize
a
g
Tomato
k
a
b
c
d
e
/
g
h
a
g
i
j
Land preparation:
32.5
43.7
land clearing
17.0
5.0
40.0
16.0
20.0
8.0
5.0
10.0
primary tillage
8.0
6.2
4.0
6.2
8.0
6.2
8.0
6.0
8.0
Planting
8.0
5.8
10.5
12.2
14.8
12.0
6.2
20.0
12.4
11.0
9.0
6.0
Fertilizer app.
9.0
7.5
9.4
6.2
6.2
10.0
10.0
Pest control:
20.8
26.5
18.5
weeding
13.0
30.8
26.9
16.9
15.0
40.0
15.0
40.0
46.8
20.0
20.0
pesticide app.
5.0
3.6
3.2
8.8
2.5
3.0
8.0
20.0
(app1)
Harvest
15.0
9.1
111
30.7
9.2
30.0
30.0
4.6
(Mg’1)
(Mg'1)
(Mg’1)
Pruning
10.0
Tying
12.0
Processing
15.4
(Mg'1)
a UMATA-DRI, 1992.
b Pachico, n.d.
c Ashby, 1987. (1982 season)
d Ashby, 1987. (1983 season)
e Hudgens, 1978. (traditional system)
f Hudgens, 1978. (improved system)
8E.B. Knapp, 1994. Personal communication, (assumes 1 man-day = 6.5 man-hour)
h J. Domingo, 1994. Personal communication.
1 Ashby & de Jong, 1982. (after fallow)
j Ashby & de Jong, 1982. (after cultivation)
k J. Alonso, 1994. Personal communication, (from a tomato farmer in Siberia)

174
Table 6-3 Prices of production inputs, Cauca,
Colombia, December 1992.
Input
Unit
Price, Col.S
day labor wages1
hour
220.00
contracted plowing*
hour
1,430.00
maize seed
kg
900.00
bean seed
kg
1,600.00
tomato seed5
kg
2,600,000.00
cassava cuttings
kg
50.00
chicken manure5
kgN
1,422.00
10-30-10*
kgN
1,980.00
17-6-18-2*
kgN
1,006.50
Manzate*
kg
3,443.00
Roxión*
kg
5,324.00
Benomyl*
kg
16,487.00
Cu oxychloride*
kg
1,902.00
poles5
unit
13.00
irrigation water
mm-ha
100.00
*UMATA-DRI 1992. Programma agropecuario municipal.
Unpublished report.
* 1994. Coyuntura Colombiana 11(1).
5 E.B. Knapp, Cl AT (personal communication).
fitted parameters for the time series models used to generate crop prices. Multipliers used
to convert from wholesale to farmgate prices accounted for differences between dry
weight and reported moisture contents, for differences in quality or cultivar, and for
market discount accounted for differences between reported wholesale and farmgate
prices (Table 6-6). Market discounts were based on a farmer interview (maize) and on
discussions with CIAT economists familiar with the region (coffee, cassava and bean).

Table 6-4. Fitted parameters for deterministic component (Eg. f4-l]) of production commodity price time series models.
Product Trend Seasonal
a
P
Pj.„
Pro.
Pm,,
PApr
Milay
Pjm
Pm
M'Aitg
Psep
Poet
f*Nov
Pd«
coffee
3.0035
-0.0001
0.0005
-0.0070
-0.0031
0.0008
-0.0019
0.0028
-0.0030
0.0006
0.0002
0.0035
0.0034
0.0035
cassava
2.2553
0.00073
0.0143
0.0056
0.0060
0.0134
0.0112
0.0373
0.0231
0.0091
0.0015
0.0055
0.0117
0.0247
choclo
1.9643
0.00055
0.0754
0.0044
0.0501
0.0634
0.0189
0.0665
0.0461
0.0136
0.0796
0.0695
0.0081
0.0659
tomato
2.5335
-0.0003
0.0075
0.0395
0.0637
0.0711
0.0225
0.0296
0.0273
0.0364
0.0343
0.0118
0.0343
0.0244
bean
3.0079
-9.1E-5
0.0064
0.0106
0.0104
0.0021
0.0075
0.0127
0.0078
0.0021
0.0090
0.0016
0.0052
-0.0003
Table 6-5, Fitted parameters for stochastic component (Eg, [4-2]) of production commodity price time series models,
Product Shock Autoregressive terms Moving average terms Seasonal — Cross-correlation coefficients
ot
♦.
0j
04
$5
«
e,
02
0j
04
05
s
0s
^"coffee
r
* cassava
^choclo
^tomato
coffee
0.01813
0.069
0.871
0.261
0.271
0.000
0.000
1.034
0.308
0.000
0.000
0.000
62
0.126
1.00
cassava
0.03458
1.014
0.024
0.630
0.664
0.000
0.000
0.043
0.123
0.602
0.173
0.000
0
0.000
0.10
1.00
choclo
0.05115
0.387
0.445
0.000
0.000
0.000
0.000
0.959
0.048
0.017
0.086
0.167
0
0.000
-0.06
-0.13
1.00
tomato
0.07315
0.000
0.000
0.000
0.000
0.000
0.000
0.955
0.555
0.246
0.000
0.000
0
0.000
0.13
0.04
0.00
1.00
bean
0.02098
0.916
0.110
0.109
0.016
0.005
0.229
0.000
0.000
0.000
0.000
0.000
0
0.000
0.32
-0.03
0.11
0.14 1.00

176
Table 6-6, Adjustments to simulated yields and reported prices, Cauca, Colombia.
Product
Recovery*
Moisture
adjustment*
Quality
adjustment
Market
discount
coffee
1.00
1.00
0.90
1.00
cassava
0.40
2.86
1.00
0.20
fresh maize {choclo)
0.80
1.90
1.00
0.70
tomato (‘Manalucie’)
0.40
16.67
1.00
0.70
bean (‘Radical’)
0.80
1.00
1.00
0.65
* Ratio of expected harvest to potential yield, dry basis.
* Ratio of market to dry weight.
§ Discount due to difference in quality or cultivar.
11 Ratio of price paid to farmer to price paid by wholesaler.
Assumptions
Some of the information needed for simulating the farm was not readily available;
reasonable values had to be assumed. I assumed that the farmer initially had Col.$
4,000,000 (Col.$ 1,700,000 savings + 2,300,000 incentive payment for removing coffee)
in his operating fund in September 1994. He estimated his minimum annual household
expenses at Col.$ 3,000,000. This amount was adjusted down to Col.$ 2,250,000 on the
assumption that the family could endure more scarcity than he estimated. Discretionary
spending was assumed to be 20% of their liquid assets (i.e., monetary savings plus the
value of any stored material inputs and harvest products) per year. The Domingo family’s
annual household expenses are probably high for the region because of college expenses
for the eldest daughter. Because household expenditure parameters could not be
measured, I adjusted them to obtain an intermediate level of fifteen-year farm sustainability

177
in preliminary simulation analyses. In addition to monetary expenditures, the household
was assumed to consume 650 kg yr'1 of maize and 150 kg ha'1 of bean.
Although several cropping systems were included in the analyses (Fig. 6-2),
management of each crop was similar in all of the annual cropping systems. Detailed crop
management assumptions are given in the input files in Appendix D. I assumed (a) that
the farmer would hire day-laborers for field preparation (removing residues, fertilizer
applications and tillage), planting, weeding, tying tomato plants, pruning coffee, harvesting
and post-harvest processing, (b) that he would irrigate, spray pesticides, apply fertilizer as
side-dressing, prepare tomato seedlings and prune tomato himself if he had time, otherwise
hire day-laborers, and (c) that he would take care of marketing himself.
For each crop, a recovery factor accounts for harvest losses and yield-reducing
factors that the crop models do not account for (Table 6-6). The low recovery factor for
tomato is based on the difference between simulated yields and those reported by a farmer
in Siberia, about 2 km from the Domingo farm. It reflects the vulnerability of tomato to
loss from diseases and the inability to market deformed or damaged fruit. The recovery
factor for cassava (0.4) reflects the fact that the cassava model (CropSim CASSAVA v.
1.0, Hoogenboom et al, 1994) does not account for nitrogen stress. Ashby (1987)
showed that with given soil fertility inputs, bean yields were lower in farmer-managed
trials than in researcher-managed on-farm trials in the Cauca region (Fig. 6-3). The
difference reflected differences in intensity of weed and pest control. The magnitude of
the difference was not consistent among trials or fertility treatments. The recovery factor
of 0.8 is probably optimistic for bean and fresh maize.

178
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1994
1995
1996
L
maize
r
bean
7 L
cassava
cassava
7
maize-bean-cassava
1994
1995
1996
1997
L
maize
maize,
L
bean
7
L
bean
7
Z
cassava
cassava /
maize-bean-bean-cassava
1994
1995
1996
1997
L
maize
maize
Ü7 [
tomato
/ bean /
Z
cassava
cassava /
maize-bean-tomato-cassava
1994
1995
maize
ize/ Í
maize
7
maize monoculture
L
maize
1994
1995
/
bean
7
maize-bean
L
maize
1994 / cassava
1995
cassava
7
1996 /
cassava
1997 cassava /
cassava monoculture
Figure 6-2. Cropping patterns included in farm scenarios.

Legend
nominal
participation
1
consultive
participation
i
decision-making
participation
Figure 6-3. Influence of type of farmer participation in on-farm trials on bean response to ground (a) and partially acidulated
rock phosphate (b), chicken manure (c), and 10-30-10, Pescador, Cauca, Colombia, 1982 and 1983. Data from Ashby (1987). ~
VO

180
Scenarios
The framework for characterizing sustainability presented in Chapter 3 calls for the
use of sensitivity analysis to test hypothesized determinants of farm sustainability. Farm
scenarios (Table 6-7) were used to examine the role of cropping system, coffee yields,
nitrogen dynamics, soil erosion, prices, household consumption, resource endowment,
labor source and credit as determinants of sustainability. A base scenario served as a
basis for comparing each alternative scenario. Each scenario was simulated for the 15
year period beginning in September 1994, except as noted.
Base scenario. The base scenario incorporates the information and assumptions
presented in the previous sections. The cropping system in the base scenario is a three-
year rotation of maize, beans and cassava (Fig. 6-2). I attempted to design a rotation that
is diversified, avoids periods of excessive labor demand, avoids the problem of nematode
buildup in successive bean crops, and fits within the rainfall pattern. Tomato was not
included because it is somewhat speculative; few farmers in the region have had
experience with tomato. To maximize spatial diversification, the three phases of the
rotation were distributed equally among different fields. The farm scenario file in
Appendix D details the assumptions in the base scenario.
Cropping systems. Of the crops commonly grown in the Cauca region, simulation
models are available only for maize, bean, cassava and tomato. Scenarios incorporate
several sequences of these crops (Fig. 6-2) that were designed to be feasible based on
rainfall distribution (Fig. 5-3), distribution of labor requirements, and time required for

181
Table 6-7. Description of farm scenarios
Scenario Description
base scenario
A three-year maize-bean-bean-cassava rotation (Fig 6-2) used as a basis for
comparing other scenarios in sensitivity analysis.
maize-becm
A one-year double-crop (Fig 6-2).
maize-tomato-bean-cassava
A three-year rotation, with irrigation for tomato and bean (Fig 6-2).
maize monoculture
A one-year maize-maize double-crop (Fig 6-2).
cassava monoculture
A three-year double crop (Fig 6-2).
maize-bean-cassava
A two-year rotation (Fig*6-2).
coffee @ 1.75 Mg/ha
Coffee monoculture with an annual yield of 1.75 Mg ha'1.
coffee @ 2.00 Mg/ha
Coffee monoculture with an annual yield of 2.00 Mg ha'1.
coffee @ 2.25 Mg/ha
Coffee monoculture with an annual yield of 2.25 Mg ha'1.
coffee @ 2.50Mg/ha
Coffee monoculture with an annual yield of 2.50 Mg ha'1.
less Nfertilizer
10-30-10 applied at 50% of base scenario (25 kg N ha'1 split application to
maize and 12.5 kg N ha'1 to bean).
more Nfertilizer
10-30-10 applied at 200% of base scenario (100 kg N ha'1 split application
to maize and 50 kg N ha'1 to bean).
erosion @ 0 Mg ha'1 yf1
erosion @ 25 Mg ha'1 yr'1
erosion @ 50 Mg ha'1 yr1
erosion @ 100 Mg ha'1ytrl
erosion @ 150 Mg ha'1 yr1
Constant annual soil loss of 0 Mg ha'1 yr'1
Constant annual soil loss of 25 Mg ha'1 yr'1
Constant annual soil loss of 50 Mg ha"1 yr'1
Constant annual soil loss of 100 Mg ha'1 yr'1
Constant annual soil loss of 150 Mg ha'1 yr"1
higher commodity prices
Ten percent higher production commodity prices.
lower input prices
Twenty percent lower material input prices.
lower labor prices
Twenty percent lower prices for labor and contracted plowing.
less subsistence spending
Ten percent lower subsistence requirement for money, maize and bean.
less discretionary spending
more initialfunds
Twenty percent lower discretionary spending.
Twenty percent higher initial operating fund.
more cultivated land
Additional land (0.23 ha, or 10% of currently cultivated area) brought into
cultivation.
credit @ 19%
Col.S 2,000,000 available at 19% interest, 24 month repayment schedule.
credit @ 9.5%
Col.$ 2,000,000 available at 9.5% interest, 24 month repayment schedule.
more credit
Col.$ 3,000,000 available at 19% interest, 24 month repayment schedule.

182
field preparation and harvest. Because beans and tomato are susceptible to nematodes and
soil-borne diseases (Chapter 5), no more than two bean and/or tomato crops were
simulated consecutively, and each béan or tomato crop required rotation with at-least one
season of maize, cassava or fallow. To maximize spatial diversity, the phases of each
multiple-year rotation were distributed among fields of equal size. Four additional
scenarios represented coffee monoculture with different assumed yields. Since no
process-level simulation model was available for coffee, its production was simulated by a
fixed schedule of field operations and fixed harvest amounts.
There is little a priori basis for hypothesizing that one particular crop contributes
more to farm sustainability than the others. However, the annual cropping scenarios were
used to test the hypothesis that diversification of crop enterprises leads to a more
sustainable farming system than any monoculture. They were also used to examine the
role of high value vegetables, represented by irrigated tomato.
Soil management. Soil processes can impact farm sustainability by modifying crop
yields. Soil management can also impose a cost on the farming system. The capabilities
of the crop models (Chapter 5) limit the soil-related determinants of sustainability that we
can test. This study includes scenarios designed to test the hypotheses that (a)
sustainability can be reduced by either excessive or insufficient N fertilizer inputs, and that
(b) soil erosion has an adverse impact on farm sustainability. Although the crop models
cannot predict soil erosion, Chapter 5 demonstrates a method for simulating the impact of
an assumed annual soil loss on crop production. The study includes erosion scenarios that
assume five constant levels (0, 25, 50, 100 and 150 Mg ha*1 yr*1) of annual soil loss.

183
Costs and prices. The Domingo farm is integrated into the market economy.
Although the family consumes much of the maize and bean grown on the farm, normally a
greater portion is sold for cash. Prices are important to sustainability because they convert
crop yields and the use of production inputs into cash flow. The higher commodity prices
scenario tests the hypothesis that low prices for crop products constrain sustainability.
The lower input prices and lower labor prices scenarios test the hypothesis that high input
prices limit sustainability by elevating production costs. Two scenarios—less subsistence
spending and less discretionary spending—test the role of household consumption in
determining farm sustainability.
Resources. Several scenarios were designed to examine the influence of access to
land and financial resources on farm sustainability. The more cultivated land scenario
assumes that a portion (0.23 ha) of land currently in permanent grass fallow is brought
into cultivation so that the total area cultivated is increased by 10%. The additional area is
distributed equally among the three phases of rotation. This scenario is used to test the
hypothesis that the land area under cultivation constrains sustainability.
This study considers several means of accessing additional operating funds. The
hypothesis that limited initial savings constrains sustainability is tested by the more initial
funds scenario in which the initial operating fund is increased by 20%. Scenarios are
included that assume access to credit. The two conventional credit scenarios assume that
Col.$ 2,000,000 (50% of initial operating fond) of credit is available on a two year
monthly repayment schedule at either 9.5% or 19% inflation-adjusted interest rate, and
that the farmer borrows only in response to a financial deficit. The 19% rate was selected

184
because it matches reported interest rates for crop loan packages after adjusting for
inflation. An additional credit scenario increases the amount of credit available by 50%.
Sources of risk. A set of four scenarios was designed to examine the impact of
weather and price risk, and of diversification in space on farm risk and sustainability. The
relevant hypotheses are (a) weather and price variability contribute unequally to whole-
farm risk and to the probability of failure, and (b) spatial diversification reduces risk and
the probability of failure.
Price risk was eliminated in the no price risk scenario by using historical prices
observed from August 1979 to September 1988 as a proxy for future prices. Because of
gaps in the historical price sequences, bean prices used were from September 1978 to June
1980 and July 1984 to September 1988 and maize prices were from August 1979 to
September 1989. FSS used Eq. [4-6] to adjusted prices for the trends given in Table 6-3.
Risk here refers to the probability distribution of future outcomes that results from
a range of possible realizations of the future. The procedure for removing variability
between replicates eliminates risk because it constrains the future to only one possible,
deterministic realization. However, the procedure does not remove variability between
years. The pattern of variability of prices through time does affect farm performance and
vulnerability to weather-induced production risk. Removing price risk in this manner is
analogous to obtaining a contract to sell produce according to predetermined, fixed
schedule of future prices.
Weather risk was eliminated by simulating a modification of the base scenario with
an ample credit supply, determining which replicate of the scenario had the median value

185
of liquid assets at the end of nine years, determining the random number seed used to
generate weather for that replicate, then using that random number seed for every
replicate of the no weather risk scenario. The third scenario, no spatial diversification,
was a modification of the base scenario in which all of the cultivated area is in the same
phase of crop rotation (i.e., all in maize then bean then bean then cassava). These
scenarios were compared with the base scenario simulated for nine years.
The scenarios used to examine sources of risk were simulated for only three
rotation cycles (9 years) for two reasons. First, after bean and maize prices were shifted
as described above, only 11 full years of prices were available for all five crops. Second,
risk is easier to interpret if distributions are not severely truncated by a large proportion of
failures. The distribution of liquid assets is the basis for interpreting farm risk through
time for these scenarios.
Simulation and Analysis
The farming system simulator (FSS) described in Chapter 4 was used to simulate
the farm scenarios. FSS simulated 100 replicates of each 15-year scenario. For a given
scenario, the model farming system had the same initial state but different realizations of
weather and prices in each replicate. The same sample of price and weather realizations
was used for all scenarios.
FSS recorded minimum, 25th percentile, median, 75th percentile and maximum
supply of liquid assets among replicates at the end of each simulation year, the value of
each resource each year for each continuing replicate, and the occurrence and timing of

186
failures. Box plots and cumulative distribution plots of liquid assets provided perspectives
of changes in farm risk through time. Failure was indicated by insolvency: the inability to
cover fixed costs, scheduled loan payments or household subsistence consumption. Liquid
assets are defined here as monetary reserves plus the value of all stored material inputs and
harvested products. Liquid assets are used as a proxy for overall wealth because the farm
scenarios do not provide for selling capital assets, such as land, to cover expenses.
Estimated sustainability (S(T)) was calculated from simulation results using Eq. [3-
12] with its standard error (SEs) calculated from Eq. [3-13]. Relative sensitivity of
predicted sustainability to continuous factors was calculated by Eq. [3-15], Estimated
sustainabilities of alternative scenarios were compared based on the McNemar test statistic
(Gpadj, Eq. [3-19] and [3-20]). Susainabilities of two scenariot were considered different if
Gpadj > X2o.o5,i- When GP adj was undefined, simulated sustainabilities were compared based
on the G-test statistic for independent observations (Gladj, Eq. [3-16] to [3-18]).
Results and Discussion
Base scenario
Simulated sustainability of the base scenario was 0.64 ± 0.048 (± SE$) at the end
of 15 years. Because of the assumptions that the base scenario required, this value should
not be interpreted as a predicted probability that the actual farm will not fail within 15
years. Rather, it is a standard for comparing alternatives and for testing the effects of

1S7
hypothesized constraints to sustainability of the model farming system in the context of its
model environment.
Results from the base scenario illustrate several perspectives from which farm
behavior can be viewed. A box-plot (Fig. 6-4) shows the magnitude, trends and changes
in dispersion of an aggregate measure of the status of a system, such as liquid assets. A
cumulative distribution plot (Fig. 6-5) gives a more complete but static picture of the
distribution of liquid assets at any point in time. Figures 6-4 and 6-5 show that the
dispersion of liquid assets and the cumulative probability of failure increase as the scenario
Figure 6-4. Box plot of liquid assets, base scenario, Domingo farm, Cauca, Colombia.

Probability Probability Probability
188
Liquid assets, Col.$ Liquid assets, Col.$
(Millions) (Millions)
Figure 6-5. Cumulative distribution of liquid assets after 1 (a), 3 (b), 6 (c), 9 (d), 12 (e)
and 15 (f) years, base scenario, Domingo farm, Cauca, Colombia.

189
progresses. Finally Fig. 6-6 presents both sustainability and hazard time plots for the base
scenario. Chapter 3 discusses interpretation of the sustainability and hazard time
functions.
We can make at least two observations from Fig. 6-6. First, h(t) = 0 and S(t) = 1
for the first three years of the scenario. The high initial monetary supply protected the
farm from failing regardless of yields or prices. Second, the hazard function peaked in
years 6, 9 and 12 of the scenario. This was apparently an effect of the three-year crop
rotation. Although the phases of the rotation were distributed among equal field areas,
productivity of the six fields differed. Failure was most likely to occur during the rainy
Figure 6-6. Sustainability and hazard time plots of the base scenario.
Hazard, h(t)

190
season following the years in which maize and bean were harvested from the least
productive fields, during the months (October to December) between planting and the
bean harvest.
Cropping systems
Simulated sustainabilities of the annual cropping system scenarios are shown in
Fig. 6-7 and Table 6-8. The most sustainable cropping system was the two-year maize-
bean-cassava rotation. It was also the most intensive rotation simulated, with very little
Figure 6-7. Sustainability time plot of annual cropping system scenarios.

191
fallow time. The results do not support the hypothesis that diversified rotations are
consistently more sustainable than monocultures. Although the three most sustainable
annual cropping systems were diversified rotations, sustainability of the cassava
monoculture (0.47) was higher than that of the maize-bean rotation (0.04). The maize
monoculture was the least sustainable annual cropping system scenario.
The simulated sustainability of coffee monoculture was positively related to yields
(Fig. 6-8). Coffee production was more sustainable than the base scenario when yields
were at least 2.25 Mg ha'1 yr'1. Coffee yields can be expected to yield between about 2.0
Figure 6-8. Sustainability time plot of base and coffee scenarios.

192
and 2.5 Mg ha'1 yr'1 in the area of the Domingo farm. However, uncertainties about coffee
yields and the variability of results obtained for different crop rotations limit
generalizations about the relative sustainability of coffee and annual crop production.
Table 6-8. Predicted 15 year sustainability (Ü(15)±SE$) of cropping system scenarios, and
ifivnvuiui.A^P.adiJ ^ V^Ladi/
Scenario
$15)±SE
f^P.adj
^I,adj
coffee @ 2.50 Mg ha'1
0.99±0.010
... .1
49.0
**
irrigated maize-tomato-bean-cassava (3 yr)
0.99±0.010
41.9 **
49.0
**
maize-bean-cassava (2 yr)
0.95 ±0.022
t
32.1
**
coffee @ 2.25 Mg ha'1
0.92 ±0.027
24.7 **
24.0
**
coffee @ 2.00 Mg ha'1
0.70 ±0.046
1.0 n.s.
0.8
n.s.
maize-bean-bean-cassava (3 yr)
(base scenario)
0.64 ±0.048
t
t
cassava monoculture (3 yr)
0.47 ±0.050
8.6 **
5.8
*
coffee @ 1.75 Mg ha'1
0.30 ±0.046
26.5 **
23.5
**
maize-bean (1 yr)
0.04 ±0.020
75.3 **
91.4
**
maize monoculture (1 yr)
0.00 ±0.000
t
t
T Undefined.
t Comparison does not apply to the base scenario.
Soil management
The impacts of rates of N fertilizer on simulated farm sustainability are shown in
Fig. 6-9. Either increasing or decreasing the amount of N applied to beans and maize
reduced sustainability compared with the base scenario. The reduction of sustainability
with less applied N supports the hypothesis that inadequate N diminishes sustainability by

193
reducing crop yields. The level of applied N affected immediate crop yields, but did not
result in discemable long-term trends due to buildup or depletion of soil N (Chapter 5).
The reduction of sustainability with more applied N supports the hypothesis that excessive
N decreases sustainability by increasing costs.
Simulation results support the hypothesis that increasing soil erosion decreases
farm sustainability (Fig. 6-10, Table 6-9). Although the fixed amounts of soil loss applied
to the farm do not represent the erosion process realistically, this procedure does
demonstrate the influence of soil loss on farm sustainability, and the potential value of
Figure 6-9. Sustainability time plot of base and nitrogen management scenarios.

194
further developing the root components of the crop models and linking them with a
process-oriented erosion model.
Costs and prices
Simulated farm sustainability was influenced by prices (Fig. 6-11). Increasing the
price that the farmer receives for crops or reducing the price of material or labor inputs
improved sustainability. However, sustainability was more sensitive to crop prices than to
input prices (Table 6-10). Sensitivity to material input prices and to the price of labor was
Figure 6-10. Sustainability of soil erosion scenarios.

195
Table 6-9. Soil loss and predicted 15 year sustainability ($15)±SEs) of erosion scenarios
and G-test statistic (G^j) for difference from the erosion @ 0 Mg ha'1 yr'1 scenario. The
McNemar test statistic was always undefined.
Scenario
annual
- Soil loss (cm)
total Í(15) ±
SE$
Gi,,dj
erosion @ 0 Mg ha'1 yr'1
0.00
0.00
0.677 ±
0.047
t
erosion @ 25 Mg ha'1 yr'1
0.56
8.33
0.60 ±
0.049
1.0 n.s.
erosion @ 50 Mg ha'1 yr'1
1.11
16.67
0.54 ±
0.050
3.5 n.s.
erosion @ 100 Mg ha'1 yr'1
2.22
33.33
0.29 ±
0.045
29.5 **
erosion @ 150 Mg ha'1 yr'1
3.33
50.00
0.16 ±
0.037
56.2 **
1 Comparison does not apply to the erosion @ 0 Mg ha'1 yr'1 scenario.
Figure 6-11. Sustainability time plot of base and price scenarios.

196
similar. Reducing household expenditures also improved sustainability (Fig. 6-12).
Simulated sustainability was much more sensitive to subsistence requirements than to
discretionary spending (Table 6-10).
Resources
Figure 6-13 shows the simulated sustainability of scenarios related to access to
resources. Increasing the total land area under cultivation by 10% increased simulated
sustainability to 1.0. This suggests that the land area that a farmer can cultivate is an
Figure 6-12. Sustainability time plot of base and household consumption scenarios.

197
important determinant of farm sustainability. The initial supply of cash was a significant (P
= 0.05) but much less important of sustainability. Access to Col.$ 2,000,000 of credit at
an interest rate of 19% improved sustainability a small but significant amount (Fig. 6-14).
Increasing the amount of credit available to Col.$ 3,000,000 did not further improve
sustainability, but reducing the interest rate to 9.5% did improve sustainability.
Table 6-10. Relative sensitivity (r) of predicted 15 year sustainability (Ü(15)±SE$) to
continuous factors, and McNemar (GPadj) and G-test (G^) statistics for difference from
the base scenario.
Factor
Base Adjusted
¿1(15) ±SEs
r
Gp,adj
Gl,adj
area cultivated (ha)
2.34
2.57
1.00 ±0.000
5.72
...t
...t
mean commodity prices1
100%
110%
0.98 ±0.014
5.31
t
43.6 **
subsistence consumption
(million Col.$ yr'1)
2.75
2.48
0.97 ±0.017
5.25
t
39.1 **
material input prices1
100%
80%
0.80 ±0.040
1.25
t
6.4 *
hired labor wages*
100%
80%
0.79 ±0.041
1.17
...t
5.5 *
discretionary consumption
(percent liquid assets yr'1)
20%
16%
0.78 ±0.041
1.09
14.6 **
4.7 *
initial funds (million Col.$)
4.00
4.40
0.69 ±0.046
0.78
3.9 *
0.6 n.s.
N fertilizer (kgN ha'1 yr'1)8
25.0
50.0
0.49 ±0.050
-0.23
10.6 **
4.6 *
N fertilizer (kg N ha'1 yr'1)8
25.0
12.5
0.29 ±0.045
-1.09
41 9 **
25.0 **
t Undefined.
* Percent of base scenario.
§ Average among the three phases of rotation.

198
Sources of risk
Eliminating either weather or price risk reduced overall farm risk after 6 or 9 years
(Fig. 6-15c and d). Removing price risk improved sustainability at nine years while the
effect of removing weather risk was not significant (P = 0.05) (Fig. 6-16, Table 6-11).
The effect of eliminating price risk was much greater than the effect of eliminating weather
risk, supporting the hypothesis that price variability contributes more to farm risk than
weather. This suggests that farmers in the Cauca region are more vulnerable to the
uncertainties of their economic environment than to their physical environment.
Figure 6-13. Sustainability time plot of base and resource scenarios.

Sustainability
199
Table 6-11. Predicted nine-year sustainability (3(9) ± SE$), McNemar test
statistic (GPadj) for difference from the base scenario, and standard
Scenario
$9) ± SE<¡
Gp,adj
SD
base
0.79 ± 0.040
t
699,504
no weather risk
0.85 ± 0.035
1.2 n.s.
659,063
no price risk
1.00 ± 0.000
t
181,131
not diversified
0.10 ±0.030
106.2 **
§
T Undefined.
* Comparison does not apply to the base scenario.
5 Not determined because the distribution was truncated by failures.
Figure 6-14. Sustainability time plot of base and credit scenarios.

Probability Probability
1.0
0.8
0.6
0.4
0.2
0.0
1.0
0.8
0.6
0.4
0.2
0.0
Figure 6-15. Distribution of liquid assets after 1 (a), 3 (b), 6 (c) and 9 years (d) for the source of risk scenarios.
3 6
Liquid assets, Col.$
(Millions)
Liquid assets, Col.$
(Millions)
200

201
Results of the no spatial diversification scenario confirm the role of diversification
in moderating farm risk. Although the maximum value of liquid assets was much higher
when the entire farm was planted in the same phase of rotation, a very high probability of
failure resulted compared with the base scenario at 3, 6 or 9 years (Fig. 6-15 b, c and d).
The highest rates of failure (i.e., greatest hazard) occurred in years 3, 6 and 9 while
waiting for the cassava harvest (Fig. 6-16). An important observation is that farm risk is
reduced by diversification in space, but not necessarily by diversification in time. The no
spatial diversification scenario followed the same diversified rotation as the base scenario.
1.0
0.8 --
>N
•g 0.6
TO
c
‘ro
w
W 0.4
0.2
0.0
\
no price risk
\ \
\ no weather risk
1
\
base scenario
\
V\
-\
\
i 1
\
no spatial diversification
1
1995
1998 2001
Year
2004
Figure 6-16. Sustainability time plot of base and source of risk scenarios.

202
Constraints to sustainability
The results already presented demonstrate the potential for improving farm
sustainability through cropping system design, the role of weather and price risk as
sustainability constraints, the potential contribution of a supply of credit to farm
sustainability, and the potential loss of sustainability from failing to diversify crops or to
control erosion. Table 6-10 lists the factors included in this study that can be represented
as continuous quantities, and ranks their importance as constraints to sustainability. The
three factors that have a direct, proportional impact on farm income or expenses—land
area cultivated, subsistence consumption, and mean commodity prices—had the greatest
impact on farm sustainability. Material input prices, wages, discretionary consumption
requirements, and initial funds also constrained sustainability significantly (P = 0.05).
Discussion
The simulation study of farm sustainability fulfilled the two objectives of this
chapter; it demonstrated the practical value of applying the systems framework and the
tools presented in previous chapters to an actual farming system, and it identified
determinants of sustainability of the Domingo farm. The approach was used to test
hypotheses about the role of factors as diverse as crop rotation, price volatility, soil
erosion, household consumption requirements and spatial diversity as determinants of

203
sustainability. The results have important practical implications to the farmer, researchers
and policy makers who are concerned with improving the sustainability of the Domingo
farm and similar farming systems in the Cauca region of Colombia.
Practical implications
Of the factors implicated as determinants of sustainability, cropping system and N
management are the ones most easily controlled by the farmer. By accepting the
Colombian Coffee Federation’s offer, the farmer eliminated the option of producing coffee
in the near future. The most promising alternatives appear to be intensifying production
with a two-year maize-bean-cassava rotation and incorporating tomato or another high¬
valued vegetable crop into the rotation. The study of sources of risk demonstrated the
value of staggering the phases of a multiple-year rotation among fields of similar size.
Although I did not attempt to find an optimum fertilizer management scheme in this study,
sustainability was higher with the N fertilizer rates used in the base scenario than when N
was applied at either 50% or 200% of those rates. Increasing the proportion of income
that the farm family saves would improve farm sustainability; reducing subsistence
expenditures ten percent increased simulated sustainability 53% (Table 6-10). However,
there may be little latitude for reducing household spending given the farmer’s goals for
his children’s education.
It would be risky for the farmer on the basis of simulation results alone to intensify
cropping, start producing tomato, or invest in an irrigation system. Intensifying the
cropping system might have unforeseeable effects on pest populations or soil fertility.

204
Vulnerability to disease and the farmer’s lack of experience makes tomato production
risky, On-farm agronomic research could reduce the uncertainty associated with these
innovations. Simulation results showed how erosion can reduce farm sustainability by
reducing yields. However, researchers have not yet quantified the severity of erosion and
resulting yield reductions on the volcanic soils in the Cauca region. The result of the
maize-bean-tomato-cassava scenario suggests that formers may benefit from producing
high-value vegetables. Enterprise budget analysis and investigations of productivity and
markets might identify other promising vegetable crops.
Some of the constraints to sustainability identified in this study can be addressed
only through public policy. Price volatility contributed much more than weather variability
to farm risk. Policies that reduce commodity price volatility should enhance farm
sustainability. There may also be some opportunity for improving farmgate prices
through, for example, marketing cooperatives. However, current farmgate prices do not
appear to be much lower than wholesale prices, Finally, policy makers might improve
farm sustainability by improving access to credit at reasonable interest rates.
Limitations
Several limitations should be considered when interpreting and applying the results
of this simulation study. First, I was unable to determine the family’s actual spending
habits and subsistence requirements. Simulated sustainability was quite sensitive to these
factors. Several other model parameters could not be measured and had to be assumed.
Second, on-farm trials raised questions about the performance of the crop simulation

205
models in this environment. Evidence (Chapter 5) suggests that the models underestimate
year-to-year variability, generally over-predict yields, and do not consider some important
yield-reducing stresses that are important in this environment. Further testing is certainly
needed. Third, FSS does not account for within-season resource constraints. Finally, FSS
does not simulate adaptive management of farming enterprises, but rather imposes fixed
management. An actual farmer would employ any of a number of strategies to control
downside risk under a threat of failure.

CHAPTER 7
SUMMARY AND CONCLUSIONS
This study presented a systems approach to characterizing farm sustainability. The
presentation included (a) a literature review that showed the inadequacy of existing
approaches, (b) the development of a conceptual framework, (c) implementation of farm-
level simulation tools, (d) evaluation of component crop simulation models, and (e)
application of the approach to a Colombian hillside farm. This chapter summarizes briefly
the main conclusions of each phase of this study as they relate to the objectives outlined in
Chapter 1.
In spite of a growing consensus about the importance of agricultural sustainability,
its potential usefulness as a criterion for guiding change in agriculture has not been
realized. The characterization that is necessary for using the concept of sustainability as a
criterion for evaluating and improving agricultural systems has been hindered by
conceptual and methodological problems. In order for sustainability to be a useful
criterion for evaluating and improving agriculture, it should be characterized in a way that
is literal, system-oriented, quantitative, predictive, stochastic and diagnostic. These
elements identify weaknesses of proposed approaches and suggest a direction for the
development of a systems approach for characterizing sustainability of agricultural
systems. The approach developed here incorporates these elements.
206

207
A quantitative expression was obtained by defining sustainability as the ability of a
dynamic, stochastic, purposeful system to continue into the future. Although the
sustainability of a farming system cannot be observed because it deals with the future, it
can be estimated from long-term, stochastic simulation of a system model. System
simulation can be used as an experimental tool for testing hypothesized constraints to
sustainability. Application of this framework places great demands on data, conceptual
models and simulation tools.
The Farming System Simulator was developed as a tool for simulating farm
sustainability. It can simulate a farm's operation and resource balance over an extended
period, and replicating a scenario with stochastic inputs of weather and prices. Links to
external crop models provide a mechanism for representing the ecological processes of
crop production. Its object-oriented design and flexible data structures permit it to
represent a variety of types of farms. In its current form it has several important
limitations, including inability to simulate production decisions, lack of feedback between
within-season resource constraints and crop production, absence of a model of livestock
production, and an overly-simplistic model of household consumption. However, it is
useful for characterizing the sustainability of a farm under fixed management.
The task of characterizing farm sustainability in an Andean hillside environment
challenges the capabilities of the crop models linked to the Farming System Simulator.
The models are useful for capturing the impact of weather variability, although some
evidence suggests that they under predict year-to-year variability. Several soil-related
constraints to crop production-low availability and high fixation of P, nematode damage,

208
im pacts of erosion, and the potential for soil loss by mass movement—are not addressed
by the crop models. Simple modifications improved response to applied N and sensitivity
to simulated soil loss in the bean model. The study highlighted the need to find a simple
method to calibrate the model of N mineralization for Andisols.
A simulation study of a hillside farm in the Cauca River watershed of southwestern
Colombia showed the practical value of using long-term, stochastic simulation to
characterize farm sustainability. Simulated sustainability was quite sensitive to cropping
patterns. The two most sustainable simulated cropping systems were an intensive, two-
year rotation of maize, bean and cassava, and a three-year rotation that incorporated
tomato as an example of a high-value vegetable crop. Of the factors that could be
expressed as continuous quantities, those that had a direct bearing on farm income or
expenses, such as the land area cultivated, subsistence spending requirements and average
commodity prices, were the most important determinants of simulated farm sustainability.
The study showed the potential impact of soil erosion on farm sustainability. A
component of the study that focused on sources of risk showed that prices contribute
much more than weather to farm risk and, therefore, sustainability in this location. It also
showed the value of spatial diversification for reducing risk and improving sustainability.
The simulation study has important practical implications for the farmer, for
researchers and for policy makers. It suggests that the farmer may contribute to the
sustainability of his farm by intensifying and diversifying production of annual crops and
by including tomato or other high-value vegetables in his crop rotations. On-farm
agronomic research could reduce the uncertainty associated with changing the crop

209
rotation and intensifying production. Agronomic trials combined with market research
might help identify promising high-valued vegetable crops. The severity and impacts of
soil erosion in the volcanic soils of the Cauca region represent an important knowledge
gap; we do not know how much erosion threatens farm sustainability. Some constraints
to sustainability are not under the farmer's control, and must be addressed through public
policy. Simulation results suggested that farm risk is controlled much more by prices than
by weather variability. Policies that reduce price volatility can be expected to enhance
farm sustainability.

APPENDIX A
OBJECT-ORIENTED PROGRAMMING CONCEPTS
A few object-oriented programming concepts (Table A-l) are essential to
understanding the design and operation of FSS. In the procedural approach to
programming, data structures are defined and made generally available to some portion of
the program, and procedures are written to manipulate the data structures. However, in
object-oriented programming (OOP), data and methods are encapsulated into a unit (an
object). Encapsulation allows the programmer to control which data and methods are
available only within an object, and which are accessible to other objects. A class serves
as a template for creating objects; an object is an instance of a class. The class provides
methods to the object, but defines only the potential instance variables (data fields).
Much of the power of OOP derives from inheritance and polymorphism. A class
specified as a descendant of another class inherits all of its ancestor’s data and methods.
The set of all classes that descend from a common ancestor is known as a class hierarchy.
A class can add data and methods to those that it inherits, and can redefine its inherited
methods. A message sent to objects instantiated from different classes within a class
hierarchy can result in different behavior. This is known as polymorphism.
210

Table A-l. Object-oriented programming terms.
Term
Explanation
object
The basic structural unit of an object-oriented program.
class
A template for forming an object.
data
An object’s set of variables. The set of data defines the state of an
object.
methods
Algorithms that act on an object’s data. A method is analogous to a
procedure in procedural programming.
encapsulation
The incorporation of data and methods into a single unit with a single
memory address.
message
The means by which public data and methods are accessed from
outside an object.
instantiation
The operation of creating an object at runtime from a class.
constructor
A method that creates an object in memory and initializes its data.
destructor
A method that disposes of an object from memory.
inheritance
The ability of classes and their derived objects to obtain data and
methods from their ancestral classes.
class hierarchy
A set of classes that descends from a common ancestral class.
polymorphism
The ability of different objects derived from a given class hierarchy to
exhibit different behavior in response to a single message.
collection
An object that implements a dynamically sizeable array of pointers to
objects.
early binding
In early binding, the code that will execute in response to a
procedural call is determined at compile time.
late binding
In late binding, the code that will execute in response to a procedural
call is determined at execution time.
The benefits of polymorphism are best seen in the context of collections. A
collection is an object that implements a dynamically sizeable array of pointers to objects

212
It is not necessary that objects in a collection derive from the same class. Collections have
iterator methods that can send messages to all or to a subset of collected objects. The
type and number of objects in a collection may not be determined until they are inserted at
runtime in response to, for example, interactive input or an input file. This is known as
late-binding, in contrast to early-binding in which procedure and function calls are
determined when the program is compiled. Using collection iterators to apply simulation
processes to a set of polymorphic model components results in a great deal of flexibility in
representing model structure.
An object-oriented approach offers several potential benefits including modularity,
comprehensibility and extensibility. Modularity results from decomposition of a program
into small, logical, loosely-coupled units. In OOP, modules take the form of objects, with
loose coupling between objects accomplished by messages. Modularity simplifies
collaborative development and code maintenance. Model components can be developed
and tested independently. Comprehensibility is enhanced by modularity. Furthermore, an
object-oriented model can be expressed clearly as a set of interacting objects with state
and behavior that are analogous to the components of the real system. The naming and
organization of classes and objects helps to communicate the structure of the system being
modeled. A third benefit of OOP is extensibility. Aggregation to higher hierarchical levels
is an important aspect of extensibility. The ability to create multiple instances of classes
mimics the occurrence of multiple subsystems in a system (eg., enterprises in a farm).
Program organization that provides for easy extension facilitates incremental development.

APPENDIX B
A MINIMUM DATA SET FOR SIMULATING FARM SUSTAINABILITY
The data set required for using the Farming System Simulator (FSS, Chapter 4) to
simulate sustainability of a farm consists of three components: a farm scenario file, a price
file and a crop minimum data set. The farm scenario file contains the farm-level
information needed to initialize a farm model and simulate a farm scenario. Dynamic price
%
inputs to the farming system are obtained from the price file as either parameters for
stochastic time-series models, historical sequences or constant values.
The crop MDS comprises field-level information about crop genotypes, crop
management, and soil. It also includes stochastic or historical weather inputs to the
farming system. A crop MDS consists of a crop management file, or FILEX, a soil file, a
set of weather files (Jones et ah, 1994) or a climate file (Hansen et al., 1994), and genetic
coefficient files (Hoogenboom et al., 1994).
Farm Scenario File
The farm scenario file is a text file and follows IBSNAT data file conventions
(Jones et ah, 1994). It is divided into sections, each beginning with a Header lines
begin with a Comment lines begin with a “!”. Sections in a scenario file should
occur in the order listed in this appendix. Every scenario file must contain at least the
213

214
SCENARIO, RESOURCES, LINKAGES, OPERATION REQUIREMENTS,
LANDSCAPE, STRATEGIES, ENTERPRISES, and CONSUMPTION DECISIONS
sections. Sections may contain either a single item or a list of items. Items in a list that
have an integer index, N, as their first field must be numbered consecutively.
To provide flexibility while avoiding redundancy, the sections of a scenario file
often relate to other sections or to items in the price or crop management files. Items in
the RESOURCES, LINKAGES, OPERATION REQUIREMENTS, SCHEDULED
OPERATIONS, LANDSCAPE, LIVESTOCK, STRATEGIES, ENTERPRISES, and
ANALYSES sections have an integer index, N, as their first field. Text names also
identify RESOURCES and price models from the price file. Items in one section often use
a name or index number to point to items in another section. For example, an enterprise
includes a set of indices to strategies from the STRATEGIES section and a set of field
indices from the LANDSCAPE section. Strategies point to the LIVESTOCK and
SCHEDULED OPERATIONS sections, and to the various crop management sections in
the experiment file.
SCENARIO section
The purpose of the SCENARIO section is to identify the farm scenario and to
specify crop management, prices and soil data input files, units of wealth, and simulation
control options. The SCENARIO section contains a single item.

215
Table B-l. Format of the SCENARIO section of the scenario file.
Variable
Header
Format1
Scenario title
Title
OC 8
Crop management file prefix (Has an “FMX” extension.)
ExpFile
1 C 8
Price file prefix (Has a “.PRI” extension.)
PricFile
1 C 8
Soil file prefix (Has a “.SOL” extension.)
SoilFile
1 C 8
Unit of value
ValUnit
1 C8
Starting random number seed
RandSeed
019
Scenario duration, years
Yrs
015
Number of replicates
Reps
015
Year of starting date
StYr
015
Day of year of starting date
StDy
015
T Format descriptions consist of: number of leading spaces, variable type (C = character, I
= integer, R = real), field width, and (if real) number of decimal places.
*SCENARIO ! MZ-BN-BN-CS, Domingo farm.
@ Title ExpFile PricFile ValUnit RandSeed
CCJD03 CCJD01 CABUYAL CO$ 335238
Yrs Reps StYr
10 20 1989
StDy
260
Figure B-l. Example SCENARIO section.
OUTPUTS section.
The OUTPUTS section controls creation of farm-level output files. The
OUTPUTS section contains a single item.
The example in Fig. B-2 specifies that new files (“Y”) should be created for
failures and yearly resource supplies, and that values of monthly sustainability and final
wealth as specified by the resource named WEALTH should be appended to existing files
(“A”).

216
Table B-2. Format of the OUTPUTS section of the scenario file.
Variable
Header
Format1
Farm failures output option
“Y” = create output file
“N” = do not create file
Fails
5 C 1
Monthly sustainability output option
“Y” = create output file
“A” = append to existing file
“N” = do not create file
Susta
5 C 1
Annual resource output option (“Y” or “N”)
Resou
5 C 1
Final wealth output option (“Y,” “N,” or “A”)
EndWl
5 C 1
Name of resource defining wealth (from RESOURCES
section)
WlthLink
1 C 15
Erosion output option (“Y” or “N”)
Eros
5 C 1
Event list output option (“Y” or “N”)
Evnt
5 C 1
Resource transaction list output option (“Y” or “N”)
Tms
5 C 1
T See footnote, Table B-l.
@Fail
Sust
Reso
EndW
WlthLink
Eros
Evnt
Trns
Y
A
Y
A
WEALTH
N
N
N
Figure B-2. Example OUTPUTS section.
ANALYSES section
The ANALYSES section controls graphical analysis at the end of a replicated
simulation run. It contains multiple items. Adjacent items with the same index will be
displayed in the same graph.
The example in Fig. B-3 specifies that sustainability time plots (“S”) for three
scenarios (“CCJD01,” “CCJD02,” and “CCJD03”) should be both (“B”) displayed as one

217
graph at the end of the current run, and saved for plotting after completion of model runs.
Table B-3. Format of the ANALYSES section of the scenario file.
Variable
Header
Format1
Index of graph
N
013
Class of analysis:
“S” = sustainability time plot
“B” = resource box plot
“C” = cumulative distribution of final wealth
Cl
2 C 1
Analysis option
“D” = display graph at end of run
“S” = save files for later display
“B” = display graph and save files
Op
2 C 1
Scenario title (Cl = “S” or “C”) or name of resource (Cl =
“B”) to be analyzed and plotted.
Item
1 C
f See footnote, Table B-l.
e n
Cl
Op
Item
i
s
B
CCJD01
i
s
B
CCJD02
i
s
B
CCJD03
Figure B-3. Example ANALYSES section.
RESOURCES section
The RESOURCES section defines the structure of the farm by identifying state
variables. The RESOURCES section contains multiple items. Each item identifies a
particular resource. The farm model recognizes several resource classes (Table B-4), each
with its own initialization variables. The next section (LINKAGES) specifies how
resources are linked to other resources. A resource buys from one or more linked

218
resources (i.e., variable costs) when use causes supply to fall below the minimum, and sells
to them if supply exceeds a maximum. If a linked resource is specified as a fixed cost,
expenses of ownership are charged to those costs each month. Most resource types
search the list of fixed or variable cost linkages to meet current requirements. Surpluses
are disposed in reverse order. Prices represent the amount of linked resource that can be
exchanged for a unit of the current resource. Price names must match time-series models
specified in the price file.
All material inputs and outputs returned by the crop models must be specified as
consumable resources. Pesticides, fertilizers, and organic amendments use the resource
names in Jones, et al. (1994, Appendix B). Planting material names consist of the planting
material code, crop code, and cultivar ID (eg., “PM001 ML IB0063”)- Similarly, harvest
product names use the harvest code, crop code, and cultivar ID (eg., “H MZ IB0063”).
Irrigation water should always use the name “WATER.”
The example in Fig. B-4 specifies a consumable resource with the name
OPERATING FUND and a value determined by the price, “Unity.” When the
OPERATING FUND falls below the minimum of 10,000, it will attempt to draw on any
variable cost linkages.
eci
Units
Name
Val
ValMult Supply Minimum Maximum IntRate . . .
c
ColS
OPERATING FUND
Unity
1.00 S.0e6 10000 -90.0 0.026
Figure B-4. Example item from the RESOURCES section.

219
Table B-4. Format of the RESOURCES section of the scenario file.
Variable
Header
Formal
Resource class:
Cl
2 C 1
“C” = consumable
“T” = timed
“K” = capital
“S” = seasonal
“D” = credit
“L” = activity-linked credit
“A” = aggregate
“B” = debt
Unit of resource
Units
1 C 8
Name of resource
Name
1 C 15
Name of price that defines value
Val
1 C 15
Multiplier for price which defines value
ValMult
0 R 8
Initial supply (consumable or credit) or availability, hr day'1
{timed, capital, machine or seasonal)
Supply
0 R 8
Minimum supply {consumable)
Minimum
0 R 8
Maximum supply {consumable)
Maximum
0 R 8
Interest, decay, or depreciation rate {consumable, credit,
activity-linked credit, capital or machine)
IntRate
0 R 8
Initial value {capital or machine)
Value
0 R 8
Beginning of available period, day of year {seasonal)
Stt
1 13
End of available period, day of year {seasonal)
Stp
1 I 3
Initial debt {credit, activity-linked credit)
Debt
0 R 8
Term of repayment, months {credit, activity-linked credit)
Trm
113
Payment interval, months {credit, activity-linked credit)
Frq
1 13
T See footnote, Table B-l.

220
LINKAGES section
The LINKAGES section specifies the interrelationships between resources as
described in the preceding (RESOURCES) section. Five types of linkages are currently
defined: fixed cost, variable cost, numerator, denominator and debt. Fixed costs specify
resource use associated with owning a particular resource. Scheduled debt repayment
associated with a credit resource is also charged to a resource linked by a fixed cost.
Variable costs identify resources to exchange with when a resource is sold or purchased,
or when use violates minimum or maximum storage constraints. Numerator and
denominator linkages identify sets of resources used by an aggregate resource to calculate
the sum of values or a ratio of summed values. Debt linkages are used by a debt resource
to sum the levels of indebtedness associated with a set of linked credit or activity-linked
credit resources. The price specifies a model in the price file identified in the SCENARIO
section. Price multipliers permit the prices to be modified without altering the price file,
and allow different purchase and sale prices.
The example in Fig. B-5 indicates that the bean harvest product (“HA001 BN
IB0026”) is linked to the OPERATING FUND as a variable cost. When consumption
causes the bean supply to go below the minimum storage constraint, it draws from the
operating fund to purchase the required amount. When a surplus is sold, its supply is
reduced and the operating fund is increased. The price identified by “Bean” multiplied by
the appropriate adjustment factor (BuyMult or SelMult) determines the ratio of the change
in the operating fund to the change in the supply of beans. In this example, beans are

221
purchased at the specified price and sold at 65% of that price.
Table B-5. Format of the LINKAGES section of the scenario file.
Variable
Header
Formal
Name of the resource
Resource
OC 15
Name of the linked resource
Source
1 C 15
Type of linkage:
“F” = fixed cost
“V” = variable cost
“N” = numerator (aggregate resource )
“D” = denominator (aggregate resource)
“B” = debt {debt resource)
Cl
1 C 1
Name of price which defines the exchange rate with the
linked resource
Price
1 C 15
Price multiplier for purchasing resource
BuyMult
0 R 8
Price multiplier for selling resource
SelMult
0 R 8
T See footnote, Table B-l.
@Resource
Source
Cl
Price
BuyMult
SelMult
HA001 BN IB0026
OPERATING FUND
V
Bean
1.00
0.65
Figure B-5. Example item from the LINKAGES section.
OPERATION REQUIREMENTS section
The OPERATION REQUIREMENTS section specifies how operations use
resources. The OPERATION REQUIREMENTS section contains multiple items. An
item specifies a unique combination of an operation type, crop type and method, and
assigns a priority to the combination. A time window indicates how long completion of
the operation can be delayed before it is abandoned. The remaining columns represent a

222
bundle of resources required to complete the operation. The resources should be timed
resources or descendants (i.e., capital, machine, or seasonal resources). A line contains
repeating blocks of resource requirements which represent a combination of mutually
dependent resources. A subsequent line with the same index and combination of
operation, crop and method represents an alternate resource bundle. An operation uses
resource bundles sequentially in the specified proportions until the operation is complete
or until the resource bundles are exhausted. While the operation date is within the time
window, any unfinished portion of the operation is deferred to the next day.
Eight operation type codes-“PLNT,” “IRRT , “FERT,” “RESD,” “CHEM,”
“TILL,” “HARV” and “MRKT”—are predefined. The farm model will recognize any
additional operation codes found in the OPERATION REQUIREMENTS section. Jones
et al. (1994, Appendix B) lists method codes. Planting (“PLNT”) uses the planting
material codes (eg., “PM001” indicates dry seeding). Method codes for application of
fertilizer (“FERT”), organic material (“RESD”) and pesticides (“CHEM”) use the
chemical applications codes. Tillage methods (“TILL”) are specified by implement codes.
Harvest (“HARV”) methods are “HA001" for harvest product, “HA002" for leaves, and
HA003 for canopy.
The example in Fig. B-6 specifies the requirements for marketing (“MRKT”) maize
(“MZ”). This operation has a priority of seven, can be spread across up to 14 days, and
' requires either two hours of both “JOSE” and “HIRED TRUCK,” or 2.5 hours of both
“HIRED LABOR” and the truck per Mg of grain.

223
Table B-6. Format of the OPERATION REQUIREMENTS section of the scenario file.
Variable
Header
Formal
Index of operation/crop/method combination
N
013
Operation type code
“PLNT” = planting
“IRRT = irrigation
“FERT” = fertilizer application:
“RESD” = organic material application
“CHEM” = pesticide application
“TILL” = tillage
“HARV” = harvest
“MRKT” = marketing
Type
1 C 4
Crop code
CC
1 C 2
Method code
Methd
1 C5
Priority of operation
Pri
1 13
Time window for operation, days
Win
1 13
* See footnote, Table B-l.
The remaining columns are repeating blocks which specify the requirements for a
single timed resource.
Table B-7, A resource block within the OPERATION REQUIREMENTS section.
Variable
Header
Formal
Name of the /th resource (from RESOURCES section)
Resource/
1 C 15
Basis for specifying an operation time requirement:
“E” = per event
“A” = per unit area, Ha'1
“W” = per unit mass of material, Mg'1
BH
1 C 1
Hours of resource required
Hours
0 R 8
See footnote, Table B-l.

224
G N
Type
CC
Methd
Pri
Win
Resourcel
BH
Hours
Resource2
BH
Hours
10
MRKT
MZ
0
7
14
JOSE
W
2.0
HIRED TRUCK
W
2.0
10
MRKT
MZ
0
7
14
HIRED LABOR
W
2.5
HIRED TRUCK
W
2.5
Figure B-6. Example item from the OPERATION REQUIREMENTS section.
SCHEDULED OPERATIONS section
The SCHEDULED OPERATIONS section specifies schedules of operations
which are not returned by the crop model (eg., marketing). It can be used to account for
production activities for which a process-level model is not available (eg., coffee).
Produce will not be marketed unless the SCHEDULED OPERATIONS section includes
marketing operations. The SCHEDULED OPERATIONS section contains multiple
items. Items with the same index are grouped together as an operation schedule.
The example in Fig. B-7 schedules a maize (“MZ”) marketing (“MRKT”)
operation to follow five days (indicated by BT = “L”) after each maize harvest operation
(item 7 in the *OPERATIONS section). The operation uses 0.8 times the amount of
maize harvest product (“H MZ IB0063”) that was harvested in the previous operation.
0 N
Cl
Type
CC
Methd
BT
LOp
Time
Resource
BA
Amount
1
O
MRKT
MZ
0
L
7
5
H MZ IB0063
O
0.8
Figure B-7. Example item from the SCHEDULED OPERATIONS section.

225
Table B-8. Format of the SCHEDULED OPERATIONS section of the scenario file.
Variable
Header
Format1
Event schedule index
N
013
Event class: “0” = field operation
Cl
2 C 1
Operation type code
Type
1 C 4
Crop code
CC
1 C 2
Method code
Methd
1 C 5
Basis for determining the event date:
“A” = absolute date (Yr, DOY)
“R” = relative date (Yr is incremented each simulation
year.)
“L” = days of lag following a specified operation
BT
2 C 1
Index of preceding operation from OPERATIONS section
if BT = “L”
LOp
113
Date of operation (interpreted according to value of BT)
Time
1 I 5
Name of consumable resource used
Resource
1 C 15
Basis for determining amount of material produced by event:
“E” = absolute amount per event
“A” = amount per ha land
“0” = fraction of amount used in preceding operation
“S” = fraction of supply of consumable resource
“R” = amount to remain in supply of consumable resource
BA
2 C 1
Amount of consumable resource used (interpreted according
to the value of BA)
Amount
0 R 8
* See footnote, Table B-l.
LANDSCAPE section
The LANDSCAPE section identifies homogeneous units of land, the hillslope of
which they are a part, and their position within the hillslope. The use of FL, IC, and ME
are identical to their use in the TREATMENTS section of the crop management file

226
(Jones et al., 1994). They point to landscape information in the crop management file.
The current version can simulate constant annual soil loss. The LANDSCAPE section
contains multiple items.
The example in Fig. B-8 specifies subfield number 1 at the top (position 1) of
hillslope 1 with an area of 0.6 ha. Constant soil loss of 20.0 Mg ha'1 is applied. The item
points to field level 2 and initial conditions level 1 in the crop management file.
Table B-9. Format of the LANDSCAPE section of the scenario file.
Variable
Header
Format1
Plot index
N
013
Hillslope index
HS
013
Position along hillslope
PO
013
Field level from crop management file
FL
013
Initial conditions level
IC
013
Environmental modifications level
ME
013
Erosion model:
“N” = none
“C” = constant annual erosion
“U” = USLE {not implemented}
“M” = MUSLE {not implemented}
“0” = Onstad-Foster modification {not implemented}
ER
2 C 1
Area of plot, ha
Area
0 R 8
Annual soil loss if ER = “C,” Mg ha'1
Loss
0 R 8
f See footnote, Table B-l.
@ N
HS
PO
FL
IC
ME
ER
Area
Loss
1
1
1
2
1
0
C
0.60
20.0
Figure B-8. Example item from the LANDSCAPE section.

227
LIVESTOCK section
The LIVESTOCK section has not yet been developed. Its purpose is to provide a
place for information about herds of grazing livestock.
STRATEGIES section
The STRATEGIES section specifies all of the field (i.e., crop and livestock)
activities that are a part of a field strategy, and their place in the cropping sequence. Each
item defines a strategy. The use of CU, PL, MI, MF, MR, MC, MT, ME, MH, and SM
are identical to their use in the TREATMENTS section of the crop management file
(Jones et al., 1994). They point to management information in the crop management file.
The STRATEGIES section contains multiple items.
The example in Fig. B-9 specifies a bean-maize double cropping strategy, with two
crop and two fallow activities repeated every year. The activities all use IBSNAT crop
models (“I”), and initialize cultivar, planting, irrigation, residue, fertilizer and chemical
application, tillage, environmental modifications, and harvest information from the
specified levels of the crop management file identified in the SCENARIO section. None
includes livestock. The entire simulated bean yield and 80% of the maize yield are
recovered in harvest.

228
Table B-10. Format of the STRATEGIES section of the scenario file.
Variable
Header
Formal
Line 1:
Strategy index
N
013
Index of an event schedule (from SCHEDULED
OPERATIONS section)
Ev
013
Strategy name
Name
1 C
All other lines:
Strategy index
N
013
Period (i.e., year) of rotation
Pe
013
Activity number within period
Ac
013
Crop activity class:
“I” = IBSNAT crop model
Cl
2 C 1
Cultivar level in the crop management file
CU
013
Planting level
MP
013
Irrigation management level
MI
013
Fertilizer management level
MF
013
Residue management level
MR
013
Pesticide management level
MC
013
Tillage management level
MT
013
Environmental modifications level
ME
013
Harvest level
MH
013
Simulation control level
SM
013
Livestock management level (from LIVESTOCK section)
LI
013
Harvest recovery multiplier
Recover
0 R 8
Activity name
Name
1 C
f See footnote, Table B-l.

229
0
N
Ev
Name
1
1
Maize/soybean
doublecrop
0
N
Pe
Ac
Cl
cu
MP
MI
MF
MR
MC
MT
ME
MH
SM
LI
Recover
Name
1
1
1
I
1
2
1
0
1
0
2
0
0
1
0
1.00
Bean
1
1
2
I
2
0
1
0
0
0
0
0
2
3
0
0.00
Fallow
1
1
1
3
I
3
1
1
1
0
0
1
0
0
2
0
0.80
Maize
1
1
4
I
2
0
1
0
0
0
0
0
1
3
0
0.00
Fallow
2
Figure B-9. Example item from the STRATEGIES section.
ENTERPRISES section
The ENTERPRISES section combines strategies with plots within the landscape.
Each section identifies an enterprise that will exist throughout the scenario. When
implemented, the production decision model will allow selection among strategies within
each enterprise. The ENTERPRISES section contains multiple items.
The example in Fig. B-10 links subfields 1 and 2 with strategy 1 to form a
bean-maize doublecrop enterprise.
@ N Name
1 Bean-Maize Doublecrop
@ Subfields
1 2
@ Strategies
1
Figure B-10. Example item from the ENTERPRISES section.

230
Table B-l 1. Format of the ENTERPRISES section of the scenario file.
Variable
Header
Formal
Line 1:
Enterprise index
N
013
Enterprise name
Name
1 C
Line 2 contains a
repeating set of subfield indices from the LANDSCAPE section:
Plot index
Subfields
1 I
Line 3 contains a repeating set of strategy indices front the STRATEGIES section:
Strategy index
Strategies
1 I
* See footnote, Table B-l.
PRODUCTION DECISIONS section
The PRODUCTION DECISIONS section has not yet been developed. However,
the section heading should still appear in the scenario file. Its purpose is to provide a
place for decision rules or criteria for selecting among alternative strategies for each
enterprise.
CONSUMPTION DECISIONS section.
The CONSUMPTION DECISIONS section specifies a set of resources, minimum
subsistence consumption levels, and the slope of a relationship between additional
(discretionary) consumption and a resource that represents wealth. The CONSUMPTION
DECISIONS section contains multiple items. Wealth is normally specified by an
aggregate resource.

The example in Fig. B-l 1 specifies consumption of the “OPERATING FUND”
and of maize (“HA001 MZ IB0065”) and bean (“HA001 BN IB0026”). A minimum of
2.75E6 (i.e., 2.75* 106) Col.$ must be consumed annually, plus annual discretionary
consumption of 0.2 times the value of the resource named “LIQUID ASSETS.”
Table B-12. Format of the CONSUMPTION DECISIONS section of the scenario file.
Variable
Header
Formal
Name of resource to be consumed (from RESOURCES
section)
Re
0C 15
Annual subsistence consumption
Subsist
0 R 8
Name of resource defining wealth (from RESOURCES
section)
Wth
1 C 15
Annual discretionary consumption as a fraction of wealth
MPC
0 R 8
1 See footnote, Table B-l.
@Re Subsist Wth
MPC
@Resource
Subsist
Wealth
MPC
OPERATING FUND
2.75E6
LIQUID ASSETS
0.20
Figure B-l 1. Example item from the CONSUMPTION DECISIONS section.
Price File
The price file contains the information necessary to specify constant or
contemporaneous ARMA time series models (Eq. [4-1] to [4-4]) for a set of prices, or for
using a set of trend-adjusted historical prices as a proxy for future prices (Eq. [4-5] and
[4-6]). Four sections are currently defined for the price file: CONSTANT, ARMA,

232
CORRELATION MATRIX, and HISTORICAL. Each item in the CONSTANT, ARMA
and HISTORICAL sections specifies a price series.
A price model initialized from the CONSTANT section of the price file always
returns the same specified value. Table B-13 presents the format of the CONSTANT
section.
Table B-13. Format of the CONSTANT section of the price file.
Description
Header
Formal
Index of price
N
013
Name of price
Descr
1 C 16
Constant value of price
Price
1 R 10
* See footnote, Table B-l.
Stochastic (ARMA) prices require both the ARMA (Table B-14) and
CORRELATION sections. The ARMA section provides the parameters used by Eq. [4-1]
and [4-2] to generate monthly prices. The CORRELATION section contains the cross¬
correlation matrix, in lower-triangular form, of the residuals of the price models specified
in the ARMA section. It is required to reproduce the cross-correlation structure of the
entire set of ARMA prices. Each cross-correlation coefficient has the format, 0 R 5 2 (see
footnote, Table B-l).

233
Table B-14. Format of the ARMA section of the price file.
Description
Variable1
Header
Format1
Index of price
N
013
Name of price
Decsr
1 C 16
Transformation
Tr
1 C 1
“N” = no transformation
“L” = log,0 transformation
Intercept of linear trend at the specified starting date
a
Intcp
0 R 8
Slope of linear trend relative to specified starting date
P
Slope
0 R 8
Year of starting date for trend calculation
StYr
114
Month of starting date for trend calculation
StMo
312
Mean deviation from trend in month m (m = Jan, ..., Dec)
S m
0 R 8
Standard deviation of random shock
StdDev
0 R 8
Autoregressive coefficient for a lag of / months (i = 1,..., 6)
4>,
AR;
0 R 7
Moving average coefficient for a lag of i months (/ = 1,..., 6)
0,
MA;
0 R 7
Lag (months) for multiplicative seasonal moving average temí
s
Lag
014
Multiplicative seasonal moving average coefficient
es
MALag
0 R 7
1 From Eq. [4-1] and [4-2].
* See footnote, Table B-l.
A price initialized from the HISTORICAL section of the price file returns a
historical series of prices read from the data file, adjusted for the specified trend (Eq. [4-5]
or [4-6]). Table B-l 5 gives the format of the HISTORICAL section. The test file
containing the historical series should contain two columns: month number (relative to the
specified starting date) and the value of the price.

234
Table B-15. Format of the HISTORICAL section of the price file.
Description
Header
Formal
Index of price
N
013
Name of price
Decsr
1 C 16
Transformation
Tr
1 C 1
“N” = no transformation
“L” = log10 transformation
Intercept of linear trend at the specified starting date
Intcp
0 R 8
Slope of linear trend relative to specified starting date
Slope
0 R 8
Year of starting date for trend calculation
StYr
1 14
Month of starting date for trend calculation
StMo
3 12
Name of text file containing historical series
File
1 C 12
f See footnote, Table B-l.

APPENDIX C
FARMING SYSTEMS SIMULATOR USER'S GUIDE
This appendix describes the operation of the Farming System Simulator (FSS)
presented in Chapter 4. The input files documented in Appendix B are essential for
running FSS as simulations are run in batch mode and controlled entirely by the scenario
file selected. Operating FSS consists of selecting one or more farm scenario files,
simulating the scenarios, then viewing the results.
FSS is available both as an overlaid DOS real mode program (FSS.EXE) and a
DOS protected mode program (FSSP.EXE). They are identical except that FSSP.EXE
can access up to 16 Mb of RAM while FSS.EXE is limited by available conventional
memory (up to the 640 kb limit imposed by DOS). The protected-mode version generally
leaves more conventional memory available for running external crop models. FSS.EXE
requires the overlay file, FSS.OVR. FSSP.EXE requires the DOS protected mode
interface (DPMI) files, RTM.EXE and DPMI16BI.OVL, or the DPMI provided by
Microsoft-Windows® or IBM’s OS2®. FSS was designed to work with DSSAT3 (Tsuji
et al., 1994) and requires it to be installed on the same hard drive under the directory
\DSSAT3. FSS requires the text editor, TVED.EXE to view and edit test files from the
FSS manu. Finally, to display graphs, FSS requires the graphics program,
WMGRAF.EXE, and its initialization file, GRAPH.INI.
235

236
FSS provides an industry-standard user interface. Screen items may be accessed
with a mouse, cursor keys (for menu items), the key (for dialog box items), or
-letter combinations for items with a highlighted letter. The main menu can be
activated from the keyboard by pressing . The key allows a user to back out
of any interface item. Dialog boxes contain a Cancel button that serves the same
function.
The main menu has four items: File, Run, Analyze, and Quit. Table C-l
summarizes the options available in each. The File menu provides features for accessing
the operating system, editing files, and exiting FSS.
Table C-l. Options available from FSS menu items.
Menu item
Explanation
File
Provides features for editing files, accessing the operating system,
displaying information about FSS, and exiting.
Edit
Accesses a text editor (TVED) so that data and output files can be
viewed and edited.
Change dir.
Opens a dialog box to change the current directory.
DOS shell
Leaves FSS temporary to access the operating system. Type “EXIT” to
return to FSS.
About
Displays information about FSS.
Exit
Exits the FSS program.
Run
Launches farm simulations.
Single scenario
Launches simulation of a single farm scenario.
Multiple scenarios
Allows a user to build a list of farm scenarios, then launches simulation
of the set of listed scenarios.
Analyze
Currently only offers display of graphical simulation results.
Graph
Lists available graphs and calls WMGRAF to display selected graphs.
Quit
Exits the FSS program.

237
Farm simulations are launched from the Run menu. The Run menu contains two
items: Single scenario and Multiple scenarios. The Single scenario item provides a file
dialog box that lists existing farm scenario files based on their .FMI extension.
Highlighting a scenario file name and selecting the OK button launches a simulation. The
Multiple scenarios item allows a user to launch a set of simulation runs. It opens a dialog
box that presents an empty list. Buttons allow a user to Add or Delete scenarios from the
list. The OK button causes FSS to simulate all of the scenarios in the list.
The Analyze menu currently contains only one item: Graph. The Graph item
opens a dialog box that lists the graphs that were generated by simulations during the
current FSS session. The list shows the graph attribute file used by WMGRAF, the type
of graph, and the scenarios included in the graph. Highlighting a graph description and
selecting the Display button causes FSS to call WMGRAF to display the graph. A file
distributed with DSSAT3 (WMGRAF.DOC) documents features that are available while
in WMGRAF. Although the list of graphs is lost when FSS is exited, the data and
attribute files are retained in the FSS directory.
FSS can be exited by selecting the Quit main menu item, by selecting File|Exit, or
by pressing -X.

APPENDIX D
INPUT FILES USED FOR FARM SIMULATIONS
This appendix lists relevant portions of the base farm scenario file (Fig. D-l), the
crop management file (Fig. D-2) and soil profile file (Fig. D-3) used as input to the farm
simulations presented in Chapter 6. The scenario file format is documented in Appendix
B. Formats of the crop management and soil input files are documented in Jones et al.
(1994).
238

239
♦SCENARIO ! J. Domingo farm, Base scenario: MZ-BN-BI1-CS
@ Title ExpFile PricFile SoilFile ValUnit RandSeed Yrs Reps StYr StDy
CCJD01 CCJD02 CABUYAL CABUYAL CO$ 335238 15 100 1994 255
*OUTPUTS
SFail Sust Reso EndW WLnk Eros Evnt Trns
Y Y Y Y WEALTH N N N
♦ANALYSES
í N Cl Op Item
IBS WEALTH
2 S S CCJD01
3 C S CCJD01
♦RESOURCES
@C1
Units
Name
Val
ValMult
Supply
Minimum
Maximum
Value
Stt Stp Debt
Trm Frq
A
Col$
WEALTH
C
Col$
OPERATING FUND
Unity
1.00
4.0e6
0.0
-99
T
hr/day
JOSE
1.00
8.0
T
hr/day
HIRED
LABOR
1.00
0.0
T
hr/day
HIRED
OX
1.00
0.0
C
kg
RE001
1.00
100000
0.0
-99
C
kg
PM001
BN
IB0026
BN seed
1.00
0.0
0.0
-99
C
kg
HA001
BN
IB0026
bean
0.65
0.0
0.0
-99
C
ka
PM001
MZ
IB0065
MZ seed
1.00
0.0
0.0
-99
C
kg
HA001
MZ
IB0065
choclo
1.33
0.0
0.0
-99
C
kg
PM003
cs
UC0022
CS cuttings
1.00
0.0
0.0
-99
C
kg
HA001
cs
UC0022
cassava
0.57
0.0
0.0
-99
c
kg
PM002
TM
TM0002
TM seed
1.00
0.0
0.0
-99
c
kg
HA001
TM
TM0002
tomato
11.67
0.0
0.0
-99
c
kg
HA001
CF
IB0001
coffee
0.90
0.0
0.0
-99
c
kg
FE007
10-30-10
1.00
0.0
0.0
-99
c
kg
FE005
Chicken manure
1.00
0.0
0.0
-99
c
mm-ha
WATER
Water
1.00
0.0
0.0
-99
c
kg
CH054
Manzate
1.00
0.0
0.0
-99
c
kg
CH021
Roxión
1.00
0.0
0.0
-99
c
kg
CH052
Benomyl
1.00
0.0
0.0
-99
c
kg
CH055
Cu Oxychloride
1.00
0.0
0.0
-99
c
kg
POLES
Poles
1.00
0.0
0.0
-99
♦LINKAGES
e
Resource
Source
Cl
Price
BuyMult
SelMult
WEALTH
OPERATING
FUND
N
-99
-SS
WEALTH
HA001 BN :
IB0026
N
-99
-99
WEALTH
HA001 MZ :
IB0065
N
-99
-99
WEALTH
HA001 TM TM0002
N
-95
-99
WEALTH
HA001 CF :
IB0001
N
-99
-99
HIRED
LABOR
OPERATING
FUND
V
HL Wages
1.00
1.00
HIRED
OX
OPERATING
FUND
V
Plowing
1.00
1.00
PM001
BN
IB0026
OPERATING
FUND
V
BN seed
1.00
1.00
HA001
BN
IB0026
OPERATING
FUND
V
bean
1.00
0.65
PM001
MZ
IB0065
OPERATING
FUND
V
MZ seed
1.00
1.00
h'AOOl
MZ
IB0065
OPERATING
FUND
V
choclo
1.69
1.33
PM003
CS
UC0022
OPERATING
FUND
V
CS cuttings
1.00
1.00
HAD 01
CS
UC0022
OPERATING
FUND
V
cassava
0.57
0.57
PM002
TM
TM0002
OPERATING
FUND
V
TM seed
1.00
1.00
HA001
TM
TM0002
OPERATING
FUND
V
tomato
11 .O'7
11 -ÓT
HA001
CF
IB0001
OPERATING
FUND
V
coffee
0.90
0.90
FE007
OPERATING
FUND
V
10-30-10
1.00
1.00
FE005
OPERATING
FUND
V
Chicken manure
1.00
1.00
WATER
OPERATING
FUND
V
Water
1.00
1.00
CH054
OPERATING
FUND
V
Manzate
1.00
1.00
CH021
OPERATING
FUND
V
Roxión
1.00
1.00
CH052
OPERATING
FUND
V
Benomyl
1.00
1.00
CH055
OPERATING
FUND
V
Cu Oxychloride
1.00
1.00
POLES
OPERATING
FUND
V
Pc’l es
1.00
1.00
IntRate
0.000
-0.50
0.00
-0.20
0.00
-0.20
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Figure D-l. Farm scenario file used to simulate the base scenario.

240
‘OPERATION REQUIREMENTS
@ N
Type
cc
Methd
Pri
Win
Resourcel
BH
Hours
1
PLNT
MZ
PM001
5
7
HIRED
LABOR
A
40.0
2
PLNT
BN
PM001
5
14
HIRED
LABOR
A
96.0
3
PLNT
cs
PM003
8
30
HIRED
LABOR
A
8 0. C'
4
PLNT
TM
PM002
6
21
HIRED
LABOR
A
39.0
5
IRRI
IR005
1
3
JOSE
A
2.0
5
IRRI
IR005
1
3
HIRED
LABOR
A
2.0
6
TILL
TI021
4
21
HIRED
OX
A
24.0
7
TILL
CS
TI023
7
21
HIRED
LABOR
A
60.0
8
FERT
AP002
7
7
HIRED
LABOR
A
2 4.0
9
HARV
MZ
HA001
2
10
HIRED
LABOR
W
20.0
10
HARV
BN
HA001
3
14
HIRED
LABOR
W
60.0
11
HARV
CS
HA001
8
30
HIRED
LABOR
w
30.0
12
HARV
TM
HA001
2
14
HIRED
LABOR
w
40.0
13
HARV
CF
HA001
3
14
HIRED
LABOR
K
40.0
14
MRKT
MZ
0
1
5
JOSE
W
6.0
15
MRKT
BN
0
3
14
JOSE
W
3.0
16
MRKT
CS
0
6
21
JOSE
W
9.0
17
MRKT
TM
0
1
3
JOSE
W
9.0
18
MRKT
CF
0
6
28
JOSE
W
3.0
19
RESD
AP002
8
5
HIRED
LABOR
A
15.0
20
SPRP
TM
1
4
10
JOSE
A
12.0
20
SPRP
TM
1
4
10
HIRED
LABOR
A
12.0
21
CLEA
0
4
21
HIRED
LABOR
A
60.0
22
CHEM
AP006
3
14
JOSE
A
8.0
22
CHEM
AP006
3
14
HIRED
LABOR
A
8.0
23
PRUN
CF
0
8
28
HIRED
LABOR
A
150.0
24
FERT
AP003
8
4
JOSE
A
8.0
24
FERT
AP003
8
4
HIRED
LABOR
A
8.0
25
STAK
TM
0
3
4
HIRED
LABOR
A
72.0
26
PRUN
TM
0
6
4
JOSE
A
15.0
26
PRUN
TM
0
6
4
HIRED
LABOR
A
15.0
27
TILL
TM
TI023
7
21
HIRED
LABOR
A
120.0
28
TILL
MZ
TI023
7
21
HIRED
LABOR
A
120.0
29
TILL
BN
TI023
7
21
HIRED
LABOR
A
120.0
30
PROC
BN
0
5
28
HIRED
LABOR
W
100.0
31
PROC
CF
0
5
45
HIRED
LABOR
W
4 00.0
‘SCHEDULED OPERATIONS
@ N
Cl Type
CC Methd BT LOp Time Resource
BA
Amount
1
O CLEA
MZ
0
L
9
1
A
1.0
1
O PROC
BN
0
L
10
1
A
1.0
1
O MRKT
MZ
0
L
9
5 HA001
MZ
IB0C
65 R
4 87.5
1
O MRKT
BN
0
L
10 14 HAG 01
BN
IBOO
26 R
112.5
1
O MRKT
CS
Cl
L
11
5 HAÚ01
CS
UCOO
22 S
-1.0
2
O CLEA
MZ
0
L
9
i
A
1.0
2
O PROC
BN
0
L
10
i
A
1.0
O MRKT
MZ
0
T
9
5 HAG01
MZ
IB00Ó5 R
487.5
2
O MRKT
BN
0
L
10 14 HA001
BN
IBOO
26 R
112.5
2
O MRKT
CS
0
L
11
5 HA001
CS
UCOO
22 S
-1.0
O
O HARV
CS HA001
A
0 95110 HA001
CS
UC0022 A
10000
‘LANDSCAPE
@ N
HS PO FL IC ME
ER
Area
Los s
1
J. _
1 0
N
v. 6 0
0.0
2
1 5
3 1 0
N
0.18
0.0
3
2 2
11
i 0
N
3.60
0.0
4
2 3
i:
1 0
N
0.18
o. c*
5
3 2
c
1 0
N
0.60
0.0
6
3 3
6 10
N
0.18
0.0
Figure D-l, continued.

241
♦STRATEGIES
8
N
Ev
Name
i
1
MZ-
-BN-
-BN-
-CS-
’ t
1 ow
intensity
, phase a
e
N
Pe
Ac
Cl
CU
MP
MI
MF
MR
MC
MT
ME
MH
SM
LI
Recover
1
1
1
I
3
3
0
1
1
0
â–  1
0
0
3
0
0.80
1
1
2
I
o
0
0
0
0
0
0
0
1
5
0
0.00
1
1
3
I
1
1
0
0
1
1
1
0
0
1
0
0.80
1
1
4
I
2
0
0
0
0
0
0
0
3
4
0
0.00
1
2
1
I
1
2
0
0
1
1
1
0
0
2
0
0.80
1
2
2
I
0
0
0
0
0
0
0
13
4
0
0.00
1
2
3
I
4
4
0
0
1
0
3
0
11
6
0
0.40
1
3
1
I
o
0
0
0
0
0
0
0
4
4
0
0.00
e
N
Ev
Name
2
1
MZ-
-BN-
-BN*
-CS-
low
intensity
, phase b
e
N
Pe
Ac
Cl
CU
MP
MI
MF
MR
MC
MT 1
ME
MH
SM
LI
Recover
2
1
1
I
1
12
0
0
1
1
1
0
0
2
0
0.80
2
1
o
L.
I
2
0
0
0
0
0
0
0
14
4
0
0.00
2
1
3
I
4
14
0
0
1
0
3
0
11
6
0
0.4 0
2
2
1
I
o
0
0
0
0
0
0
0
7
4
0
0.00
2
3
1
I
3
13
0
1
1
0
1
0
0
3
0
0.80
2
3
O
I
2
0
0
0
0
0
0
0
1
5
0
0.00
2
3
3
I
1
11
0
0
1
1
1
0
0
I
0
0.80
2
3
4
I
2
0
0
0
0
0
0
0
6
4
0
0.00
@
N
Ev
Name
3
2
MZ-
-BN-
-BN-
-CS-
low
intensity
, phase c
8
N
Pe
Ac
C1
CU
MP
MI
MF
MR
MC
MT 1
ME
MH
SM
LI
Recover
3
1
1
I
2
0
0
0
0
0
0
0
3
4
0
0.00
3
2
1
I
3
8
0
1
1
0
1
0
0
3
0
0.80
3
2
2
I
9
0
0
0
0
0
0
0
1
5
0
0.00
3
2
3
I
1
6
0
0
1
1
1
0
0
1
0
0.80
3
o
4
I
2
0
0
0
0
0
0
0
9
4
0
0.00
3
3
1
I
1
7
0
0
1
1
1
0
0
2
0
0.80
3
3
2
I
2
0
0
0
0
0
0
0
15
4
0
0.00
3
3
3
I
4
9
0
0
1
0
3
0
11
6
0
0.40
♦ENTERPRISES
@ N Name
1 Maize-Bean-Cassava A
@ Subfields
1 2
@ Strategies
1
8 N Name
2 Maize-Bean-Cassava B
@ Subfields
3 4
@ Strategies
2
@ N Name
3 Maize-Bean-Cassava C
@ Subfields
5 6
@ Strategies
3
♦CONSUMPTION DECISIONS
@
Resource
Subsist
Wealth
MPC
OPERATING FUND
2.30EÓ
WEALTH
0.20
HA001
MZ IB0065
650
WEALTH
0.00
HA001
BN IB002Ó
150
WEALTH
0.00
Name
Maize
Fallow!
Beanl
Fa11owl
Bean2
Fa11owl
Cassava
Fa11owl
Name
Bean2
Fallowl
Cassava
Fallowl
Maize
Fallow!
Beanl
Fallowl
Name
Fallowl
Maize
Fallow!
Beanl
Fallowl
Bean2
Fallowl
Cassava
Figure D-l, continued.

242
♦EXP.DETAILS: CCJD940FM J.Domingo farm sustainability study
*GENERAL
6PEOPLE
J.W. Hansen, E.B. Knapp
8ADDRESS
ClAT, COLOMBIA
SSITE
José Domingo farm
*CULTIVARS
@C CR INGENO CNAME
1 BN IB0026 ICA Caucaya
2 FA IB0001
3 MZ IB0065 CIMCALI
4 CS UC0022 MCol-1501
5 TM TM0002 SUNNY TOM DET
*FIELDS
@L
ID FIELD WSTA....
FLSA
FLOB
FLDT
FLDD
FLDS
FLST
SLTX
SLDP
ID SOIL
1
CCJD06A CCJD
3 SE
3
0
0
0
0
0
-99
CCBN94 0001
o
CCJD18A CCJD
10 SE
10
0
0
0
0
0
-99
CCBN940002
3
CCJD30A CCJD
17 SE
17
0
0
0
0
0
-99
CCBN 94 0003
4
CCJD06B CCJD
3 SE
3
0
0
0
0
0
-99
CCBN940004
5
CCJD18B CCJD
10 SE
10
0
0
0
0
0
-99
CCBN 940008
6
CCJD30B CCJD
17 SE
17
0
0
0
0
0
-99
CCBN 94 0005
7
CCJD06C CCJD
3WSW
3
0
0
0
0
0
-99
CCBN940006
8
CCJD18C CCJD
10WSW
10
0
0
0
0
0
-99
CCBN94 0007
9
CCJD30C CCJD
17WSW
17
0
0
0
o
0
-99
CCBN94 0007
10
CCJD06D CCJD
3 SE
3
0
0
0
0
0
-99
CCBN940001
11
CCJD18D CCJD
10 SE
10
0
0
0
0
0
-99
CCBN94 0002
12
CCJD30D CCJD
17 SE
17
0
0
0
0
0
-99
CCBN940003
13
CCJD06E CCJD
3 SE
3
0
0
0
0
o
-99
CCBN 94 0004
14
CCJD18E CCJD
10 SE
10
0
0
0
0
0
-99
CCBN940008
15
CCJD30E CCJD
17 SE
17
0
0
0
0
0
-99
CCBN94 0005
♦INITIAL
CONDITIONS
@C
PCR
ICDAT ICRT
ICND
ICRN
ICRE
1
BN
94255 500.
50.
1.00
1.00
@C
ICBL
SH20 SNH4
SN03
1
20
.396 1.5
4.5
1
50
.458 0.8
2.1
1
200
.458 0.6
1.5
@c
PCR
ICDAT ICRT
ICND
ICRN
ICRE
2
BN
94255 500.
50.
1.00
1.00
@C
ICBL
SH20 SNH4
SN03
2
20
.396 1.0
2.5
o
50
.458 0.5
1.2
2
200
.456 0.4
0.8
♦PLANTING
; DETAILS
@P
PDATE
EDATE PPOP
PPOE
PLME
PLDS
PLRS
F'LRD
PLDP
PLWT
PAGE
PENV PLPH
1
95074
-99 15.0
15.0
S
R
60
0
5.9
-99
-99
-99 -99
o
9527 3
-99 15.0
15.0
S
R
60
0
5.9
-99
-99
-99 -99
3
94273
-99 5.0
5.0
5
R
30
0
5.9
- 99
-99
-99 -99
4
96010
-0 9 0.7
0.7
r
r.
10 0
n
10.0
1 c r,
-95
-99 -99
5
95074
-99 0.75
0.75
T
R
150
0
1.0
20
28
25.0 1.0
6
96074
-99 15.0
15.0
S
R
60
0
5.9
-99
-99
-99 -99
7
96273
-99 15.0
15.0
£
R
60
0
5.9
-99
-99
-99 -99
8
95273
-99 5.0
5.0
S
R
80
0
5.9
-95
-99
-99 -99
9
97010
-99 0.7
0.7
c
R
100
0
10.0
150
-99
-99 -99
10
96074
-99 0.75
0.75
T
R
150
0
1.0
20
28
25.0 1.0
11
97074
-99 15.0
15.0
c
R
60
0
5.9
-99
-99
-99 -99
Figure D-2. Crop management file used to simulate farm scenarios.

243
12
94273
-99
15.0
15.0
c
R
60
0
5.9
-99
-99
-99
-99
13
96273
-99
5.0
5.0
s
R
80
o
5.9
-99
-99
-99
-99
14
95010
-99
0.7
0.7
c
R
100
0
10.0
150
-99
-99
-99
15
97074
-99
0.75
0.75
T
R
150
0
1.0
20
28
25.0
1.0
16
95074
-99
30.0
30.0
S
R
25
0
5.9
-99
-99
-99
-99
17
95273
-99
30.0
30.0
s
R
25
0
5.9
-99
-99
-99
-99
18
94273
-99
8.0
8.0
s
R
50
0
5.9
-99
-99
-99
-99
19
96074
-99
30.0
30.0
s
R
25
0
5.9
-99
-99
-99
-99
20
96273
-99
30.0
30.0
s
R
25
0
5.9
-99
-99
-99
-99
21
95273
-99
8.0
8.0
s
R
50
0
5.9
-99
-99
-99
-99
22
97074
-99
30.0
30.0
s
R
25
0
5.9
-99
-99
-99
-99
23
94273
-99
30.0
30.0
s
R
25
0
5.9
-99
-99
-99
-99
24
96273
-99
8.0
8.0
s
R
50
0
5.9
-99
-99
-99
-99
25
95074
-99
5.0
5.0
s
R
100
0
5.9
-99
-99
-99
-99
26
94273
-99
0.7
0.7
c
R
100
0
20.0
150
-99
-99
-99
27
95273
-99
0.7
0.7
c
R
100
0
20.0
150
-99
-99
-99
28
96273
-99
0.7
0.7
c
R
100
0
20.0
150
-99
-99
-99
29
95180
-99
0.7
0.7
c
R
100
0
20.0
150
-99
-99
-99
30
96180
-99
0.7
0.7
c
R
100
0
20.0
150
-99
-99
-99
*FERTILIZERS (INORGANIC)
@F
FDATE
FMCD
FACD
FDEP
FAMN
FAMP
FAMK
FAMC
FAMO
FOCD
1
1
FE007
AP002
10
25
150
50
-99
-99
-99
1
28
FE007
AP003
10
25
150
50
-99
-99
-99
2
1
FE007
AP002
10
75
225
75
-99
-99
-99
2
28
FE007
AP003
10
75
225
75
-99
-99
-99
3
1
FE007
AP002
10
25
75
25
-99
-99
-99
4
1
FE007
AP002
10
25
"7 C
OC
-99
-99
-99
5
1
FE007
AP002
10
75
o o c,
75
-99
-99
-99
6
1
FE007
AP002
10
100
225
75
-95
-99
-95
■••RESIDUES AND OTHER ORGANIC
MATERIALS
@R
RDATE
RCOD
RAMT
RESN
RESP
RESK
RINP
RDEP
1
94260
RE001
2000
4.00
2.00
1.50
20
10
*CHEMICAL APPLICATIONS
ec
CDATE
CHCOD
CHAMT
CHME
CHDEP
CHT
1
20
CH052
1.0
APC06
0.0
-99
1
20
CH054
3.0
AP006
0.0
-99
1
34
CH054
3.0
AP006
0.0
-95
1
48
CH054
3.0
AP006
0.0
-99
20
CH052
1.0
AP00 6
0.0
-99
2
20
CH054
3.0
AP006
0.0
-99
2
27
CH054
3.0
AP006
0.0
-99
o
¿.
34
CH054
3.0
AP006
0.0
-99
41
CH054
3.0
AP006
0.0
-99
2
48
CH054
3.0
AP006
0.0
-99
o
55
CH054
3.0
AP006
0.0
-99
2
69
CH054
3.0
AP006
0.0
-99
2
83
CH054
3.0
AP006
0.0
-99
2
97
CH054
3.0
AF006
0.0
-99
*TILLAGE
@ T
TDATE
TIMPL
TDEP
1
-7
TI021
15.0
1
20
TI023
2.0
2
-7
TI021
15.0
O
20
T10 2 3
2.0
60
TI 02 3
2.0
3
-7
TI021
15.0
3
20
TI023
2.0
3
60
TI023
2.0
3
120
TI023
2.0
3
180
TI023
2.0
Figure D-2, continued.

244
*HARVEST DETAILS
@R
HDATE HSTG
HCOM
HSIZ
HPC
1
14 -99
H
-99
100.0
2
95100 -99
H
-99
100.0
3
95255 -99
H
-99
100.0
4
97255 -99
H
-99
100.0
5
96100 -99
H
-99
100.0
6
96255 -99
H
-99
100.0
7
98255 -99
H
-99
100.0
8
97100 -99
H
-99
100.0
9
97255 -99
H
-99
100.0
10
99255 -99
H
-99
100.0
11
425 -99
H
-99
100.0
12
30 -99
H
-99
100.0
13
96009 -99
H
-99
100.0
14
97009 -99
H
-99
100.0
15
95009 -99
H
-99
100.0
♦SIMULATION CONTROLS
@N
GENERAL
NYERS
NREPS
START
SDATE
RSEED
SHAME
1
GE
10
1
P
93255
2150
Bean '
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
1
OP
Y
Y
Y
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
1
ME
W
S
E
R
S
C
6N
MANAGEMENT
PLANT
IRRIG
FERTI
RES ID
HARVS
1
MA
A
N
D
A
M
@N
OUTPUTS
XCODE
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
1
OU
N
N
A
10
N
N
N
N
N
N
Y
e
AUTOMATIC MANAGEMENT
0N
PLANTING PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
1
PL
65
120
4 0
100
15
4 0
o
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
1
IR
30
50
100
GS001
IR005
0
1.00
SN
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
1
NI
30
50
25
FE001
GS001
SN
RESIDUES
RIPCN
RTIME
RIDEP
1
RE
100
1
20
@N
HARVEST HFIRST
HLAST
HPCNP
HPCNR
1
HA
0
365
100
o
@N
GENERAL
NYERS
NREPS
START
SDATE
RSEED
SHAME
2
GE
10
1
P
93255
2150
Bean 2
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
2
OP
Y
Y
Y
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
2
ME
W
S
E
R
O
C
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RES ID
HARVS
o
MA
A
N
D
A
M
@N
OUTPUTS
XCODE
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
2
OU
N
N
A
10
N
N
N
N
N
N
Y
@
AUTOMATIC MANAGEMENT
@N
PLANTING PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
2
PL
260
305
40
100
15
4 0
0
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
2
IR
30
50
100
GS001
IR005
0
1.00
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
2
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIFCN
RTIME
RIDEP
o
RE
100
1
20
@N
HARVEST HFIRST
HLAST
HPCNP
HPCNR
2
HA
0
365
100
0
Figure D-2, continued.

245
@N
GENERAL
NYERS
NREPS
START
SDATE
RSEED
SHAME
3
GE
10
1
P
93255
2150
Mai se
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
3
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
3
ME
W
S
E
R
S
C
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RES ID
HARVS
3
MA
A
N
D
A
M
@N
OUTPUTS
XCODE
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
3
OU
N
N
A
10
N
N
N
N
N
N
Y
@
AUTOMATIC MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
3
PL
260
305
35
100
15
40
0
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
3
IR
30
50
100
GS001
IR005
0
1.00
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
3
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
3
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
3
HA
0
365
100
0
@N
GENERAL
NYERS
NREPS
START
YRDAY
RSEED
SNAME
4
GE
1
1
S
93255
2150
Fallow 1
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
4
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
4
ME
w
S
E
R
S
c
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RESID
HARVS
4
MA
A
N
N
A
R
@N
OUTPUTS
FNAME
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
4
OU
N
N
N
10
N
N
N
N
N
N
Y
e
AUTOMATIC MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
4
PL
1
366
0
100
15
50
0
@N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
4
IR
30
50
100
GS001
IR001
10
0.75
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
4
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
4
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
4
HA
0
365
100
0
@N
GENERAL
NYERS
NREPS
START
YRDAY
RSEED
SNAME
5
GE
1
J.
1
£
93255
2150
Fallow 2
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
5
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
5
ME
W
c
E
R
S
r
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RESID
HARVS
5
MA
E.
N
N
A
D
@N
OUTPUTS
FNAME
OVVEW
SUMRY
FROPT
GROTH
CARBN
WATER
NITRO
MINER
DISES
LONG
5
OU
N
K
N
10
N
N
N
N
N
N
Y
Figure D-2, continued.

246
8
AUTOMATIC
MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
5
PL
1
3 66
0
100
15
50
0
0N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
5
IR
30
50
ICO
GS001
IR005
10
0.75
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
5
NI
30
50
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
5
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
5
HA
0
365
100
0
@N
GENERAL
NYERS
NREPS
START
YRDAY
RSEED
SNAME
6
GE
1
1
S
93255
2150
Cassava
@N
OPTIONS
WATER
NITRO
SYMBI
PHOSP
POTAS
DISES
6
OP
Y
Y
N
N
N
N
@N
METHODS
WTHER
INCON
LIGHT
EVAPO
INFIL
PHOTO
6
ME
W
S
E
R
S
C
@N
MANAGEMENT
PLANT
IRRIG
FERTI
RES ID
HARVS
6
MA
A
N
D
A
D
@N
OUTPUTS
FNAME
OWEW
SUMRY
FROPT
GROTH
CARBN
WATER NITRO
MINER
DISES
LONG
6
OU
N
N
A
10
N
N
N N
N
N
Y
@
AUTOMATIC MANAGEMENT
@N
PLANTING
PFIRST
PLAST
PH20L
PH20U
PH20D
PSTMX
PSTMN
6
PL
10
70
30
100
15
40
10
8N
IRRIGATION
IMDEP
ITHRL
I THRU
IROFF
IMETH
IRAMT
IREFF
6
IR
30
50
100
GS001
IR005
10
0.75
@N
NITROGEN
NMDEP
NMTHR
NAMNT
NCODE
NAOFF
6
NI
30
50
25
FE001
GS001
@N
RESIDUES
RIPCN
RTIME
RIDEP
6
RE
100
1
20
@N
HARVEST
HFIRST
HLAST
HPCNP
HPCNR
6
HA
0
365
100
0
Figure D-2, continued.

*CCBN940001
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 1A
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
@ SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
76
0.35
1.00
IB001
IB001
IB001
@ SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
20 AP
0.325
0.467
0.575
0.75
25.0
0.45
11.70
16.2
25.8
0.0
0.78
5.3
4.5
3.3
50 B
0.375
0.541
0.622
0.15
5.0
0.48
3.10
75.0
15.0
0.0
0.27
5.4
5.0
0.8
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.10
75.0
15.0
0.0
0.20
5.6
5.1
0.8
*CCBN940002
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO IB
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
22 AP
0.325
0.467
0.575
0.75
25.0
0.45
10.00
16.1
28.3
0.0
0.73
4.8
4.5
2.6
50 B
0.375
0.541
0.622
0.15
5.0
0.48
7.20
75.0
15.0
0.0
0.30
5.0
4.9
0.7
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.50
75.0
15.0
0.0
0.24
5.3
5.0
0.6
♦CCBN940003
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 1C
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.5
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
23 AP
0.325
0.467
0.575
0.75
25.0
0.45
8.40
18.8
29.7
0.0
0.61
4.9
4.4
2.2
50 B
0.375
0.541
0.622
0.15
5.0
0.48
3.10
75.0
15.0
0.0
0.25
5.0
4.6
1.0
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.10
75.0
15.0
0.0
0.19
5.2
4.7
0.7
♦CCBN940004
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 2A
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
19 AP
0.325
0.467
0.575
0.75
25.0
0.45
3.90
16.1
29.7
0.0
0.55
5.2
4.7
2.8
50 B
0.375
0.541
0.622
0.15
5.0
0.48
2.80
75.0
15.0
0.0
0.24
5.2
5.1
1.2
200 B
0.375
0.541
0.622
0.10
5.0
0.48
1.70
75.0
15.0
0.0
0.15
5.6
5.2
0.7
Figure D-3. Soil profile data file used for simulating farm scenarios.
247

*CCBN940005
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 2B
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
@ SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.2
0.80
83
0.35
1.00
IB001
IB001
IB001
@ SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
20 AP
0.325
0.467
0.575
0.75
25.0
0.45
4.90
14.7
25.6
0.0
0.71
5.2
4.6
2.7
50 B
0.375
0.541
0.622
0.15
5.0
0.48
4.90
75.0
15.0
0.0
0.38
5.3
5.0
0.8
200 B
0.375
0.541
0.622
0.10
5.0
0.48
3.20
75.0
15.0
0.0
0.27
5.5
5.0
0.6
*CCBN940006
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 3A
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
Y 0.16
9.6
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
25 B
0.330
0.489
0.554
0.50
10.0
0.48
5.80
20.6
32.7
0.0
0.39
5.4
4.8
2.1
50 B
0.375
0.541
0.622
0.15
5.0
0.48
2.10
21.8
16.2
0.0
0.16
5.5
5.0
0.7
200 B
0.375
0.541
0.622
0.10
5.0
0.48
1.10
79.7
18.5
0.0
0.07
5.2
4.7
1.3
*CCBN940007
SCS
lo
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 3B
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
13 AP
0.325
0.467
0.575
1.00
25.0
0.45
5.10
16.7
24.8
0.0
0.65
5.4
4.7
5.0
50 B
0.375
0.541
0.622
0.15
5.0
0.48
2.40
75.0
15.0
0.0
0.21
5.8
5.1
0.5
200 B
0.375
0.541
0.622
0.10
5.0
0.48
1.60
75.0
15.0
0.0
0.13
5.9
5.1
0.4
*CCBN940008
SCS
lio
200
UNKNOWN
0SITE
COUNTRY
LAT
LONG
SCS FAMILY
JDOMINGO 3C
COLOMBIA
2.
783 -
76.520
OXIC DYSTROPEPT
0 SCOM SALB
SLU1
SLDR
SLRO
SLNF
SLPF
SMHB
SMPX
SMKE
BK 0.11
9.3
0.80
83
0.35
1.00
IB001
IB001
IB001
0 SLB SLMH
SLLL
SDUL
SSAT
SRGF
SSKS
SBDM
SLOC
SLCL
SLSI
SLCF
SLNI
SLHW
SLHB
SCEC
15 AP
0.325
0.467
0.575
0.75
25.0
0.45
5.30
16.7
27.7
0.0
0.67
5.4
4.7
3.3
50 B
0.375
0.541
0.622
0.15
5.0
0.48
3.40
75.0
15.0
0.0
0.25
5.6
5.0
0.6
200 B
0.375
0.541
0.622
0.10
5.0
0.48
2.40
75.0
15.0
0.0
0.20
5.6
5.1
0.4
Figure D-3, continued.
248

BIBLIOGRAPHY
Alberts, E.E., M.A. Weltz and F. Ghidey. 1989. Plant growth component, p. 8.1-8.38. In
L.J. Lane and M.A. Nearing (ed.) USDA - Water Erosion Prediction Project:
Hillslope Profile Model Documentation. NSERL Report No. 2. USDA-ARS
National Soil Erosion Laboratory, West Lafayette, Indiana.
Allen, P. and D. van Dusen. 1988. Sustainable agriculture: choosing the future, p. 1-13. In
P. Allen and D. van Dusen (ed.) Global Perspectives on Agroecology and
Sustainable Agricultural Systems. Proceedings of the Sixth International
Conference of the International Federation of Organic Agriculture Movements.
University of California, Santa Cruz, California.
Allen, T.F.H. and T.B. Starr. 1988. Hierarchy: Perspectives for Ecological Complexity.
The University of Chicago Press, Chicago.
Alocilja, E.C. and J.T. Ritchie. 1993. Multicriteria optimization for a sustainable
agriculture, p. 381-396. In F.W.T. Penning de Vries, P.S. Teng, and K. Metselaar
(ed.) Systems Approaches for Agricultural Development, v. 2. Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Altieri, M.A. 1987. Agroecology: The Scientific Basis of Alternative Agriculture.
Westview Press, Boulder, Colorado.
American Society of Agronomy (ASA). 1989. Decision reached on sustainable
agriculture. Agron. News, January, p.15.
Anonymous. 1994. Anexo estadístico. Coyuntura Colombiana 11(1):218-219.
Ashby, J.A. 1985. The social ecology of soil erosion in a Colombian farming system. Rural
Sociology 50:377-396.
Ashby, J.A. 1987. The effects of different types of farmer participation on the management
of on-farm trials. Agrie. Admin. & Extension 25:235-252.
Ashby, J.A. and G. de Jong, G. 1982. Farmer filed preparation and tillage practices:
implications for fertilizer technology research. Soil & Tillage Research 2:331-346.
249

250
Baily, D. von & J.W. Richardson. 1985. Analysis of selected marketing strategies: a
whole-farm simulation approach. Amer. J. Agr. Econ. 67:813-820.
Banco de la República (various dates). Revista del Banco de la República, various issues.
Bogotá, Colombia.
Barlow, R.E., F. Proschan, and L.C. Hunter. 1965. Mathematical Theory of Reliability.
John Wiley & Sons, New York.
Batchelor, W.D., J.W. Jones, K.J. Boote and H.O. Pinnschmidt. 1993. Extending the use
of crop models to study pest damage. Trans. ASAE 36:551-558.
Batie, S.S. 1989. Sustainable development: challenges to the profession of agricultural
economics. Am. J. Agrie. Econ. 71:1083-1101.
Beus, C.E. and R.E. Dunlap. 1990. Conventional versus alternative agriculture: the
paradigmatic roots of the debate. Rural Sociology 55(4):590-616.
Beus, C.E. and R.E. Dunlap. 1991, Measuring adherence to alternative vs. conventional
agricultural paradigms: a proposed scale. Rural Sociology 56:432- 460.
Bidwell, O.W. 1986. Where do we stand on sustainable agriculture? J. Soil Water
Conserv. 41:317-320.
Borland International. 1992. Borland*’ Pascal with Objects. Borland International, Scotts
Valley, California.
Bowen, W.T., J.W. Jones, R.J. Carsky, and J.O. Quintana. 1993. Evaluation of the
nitrogen submodel of CERES-Maize following legume green manure
incorporation. Agron. J. 85:153-159.
Bowen, W.T., P.K. Thornton, and G. Hoogenboom. 1992. Crop sequencing using
IBSNAT crop models, p. 79. In Agronomy Abstracts. ASA, Madison, WI.
Box, G.E.P. and G.M. Jenkins. 1970. Time Series Analysis: Forecasting and Control.
Holden-Day, San Francisco, California.
Brklacich, M., C.R. Bryant, and B. Smit. 1991. Review and appraisal of the concept of
sustainable food production systems. Environ. Management 15( 1): 1 -14.
Budelman, A. and F. van der Pol. 1992. Farming system research and the quest for a
sustainable agriculture. Agroforestry Systems 19:187-206.

251
Caldwell, R.M. and J.W. Hansen. 1993. Simulation of multiple cropping systems with
CropSys. p. 397-412. In F.W.T. Penning de Vries, P. Teng & K. Metselaar (ed.)
Systems Approaches for Agricultural Development, Vol. 2. Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Carter, H.O. 1989. Agricultural sustainability: an overview and research assessment. Calif.
Agrie. 43(3): 16-18,37.
Casagrande, J.T. and M.C. Pike. 1978. An improved approximate formula for calculating
sample sizes for comparing two binomial distributions. Biometrics 34:483-486.
Castillo, W.C. 1990, 1993. Series estadísticas de precios promedios corrientes de frijol a
nivel de consumidor en Cali y Palmira Valle del Cauca, Centro Internacional de
Agricultura Tropical (CIAT). (unpublished report)
Castro, D.M., C.P. Roa and R.L. Rivera. 1992. Genesis y taxonomía de los Andisoles
Colombianos. Suelos Ecuatoriales 22:50-68.
Corporación de Abastecimentos del Valle del Cauca S.A. (CAVASA). 1994. Series
históricas de precios por kilogramo registradas en el mercado mayorista de
CAVASA. CAVASA, Cali, Colombia, (unpublished report)
Conway, G.R. 1985. Agroecosystems analysis. Agrie. Admin. 20:31-55.
Conway, G.R. 1994. Sustainability in agricultural development: trade-offs with
productivity, stability and equitibility. J. Farming Systems Research - Extension
4(2): 1-14.
Conway, G.R. and E.B. Barbier. 1990. After the green revolution: sustainable agriculture
for development. Earthscan Publications, London.
Cordray, S.M., L.S. Lev, R.P. Dick, and H. Murray. 1993. Sustainability of Pacific
Northwest horticultural producers. J. Prod. Agrie. 6:121-135.
Cramb, R.A. 1993. Shifting cultivation and sustainable agriculture in East Malaysia: a
longitudinal case study. Agrie. Systems 42:209-266.
Dagpunar, J. 1988. Principles of Random Variate Generation. Oxford Univ. Press, New
York.
Dahlberg, K.A. 1991. Sustainable agriculture -- fad or harbinger? BioScience
41(5):337-340.

252
Dent, J.B. and P.K. Thornton. 1988. The role of biological simulation models in farming
systems research. Agr. Admin. Extension 29:111-122.
Dicks, M.R. 1992. What will be required to guarantee the sustainability of U.S. agriculture
in the 21st century? Am. J. Alternative Agrie. 7:190-195.
Dillon, C.R., J.W. Mjelde and B.A. McCarl. 1989. Biophysical simulation in support of
crop production decisions: a case study in the Blacklands region of Texas.
Southern J. Agr. Econ. 21(l):73-85.
Dobbs, T.L., D.L. Becker, and D.C. Taylor. 1991. Sustainable agriculture policy analyses:
South Dakota On-Farm Case Studies. J. Farming Systems Research - Extension
2(2): 109-124.
Douglass, G.K. 1984. The meanings of agricultural sustainability, p. 1-29. In G.K.
Douglass (ed.) Agricultural Sustainability in a Changing World Order. Westview
Press, Boulder, Colorado.
Edwards, C.A. 1990. The importance of integration in sustainable agricultural systems, p.
249-264. In C.A. Edwards, R. Lai, P. Madden, R.H. Miller, and G. House (ed.)
Sustainable Agricultural Systems. Soil and Water Conservation Society, Ankeny,
Iowa.
Ehui, S.K. and D.S.C. Spencer. 1992. A general approach for evaluating the economic
viability and sustainability of tropical cropping systems, p. 110-119. In M. Bellamy
and B. Greenshields (ed.) Issues in Agricultural Development: Sustainability and
Cooperation. IAAE Occasional Paper No. 6. Dartmouth Publishing Co.,
Aldershot, UK.
Elandt-Johnson, R.C. and N.L. Johnson. 1980. Survival Models and Data Analysis. John
Wiley & Sons, New York.
Federación Nacional de Cafeteros de Colombia (FEDCAFE). 1986. El Cultivo del
Tomate. 2nd ed. Federación Nacional de Cafeteros de Colombia, Cali, Colombia.
Fox, G. 1991. Agriculture and the environment: economic dimensions of sustainable
agriculture. Can. J. Agrie. Econ. 39:647-653.
Fox, R.L. 1978. Studies on phosphorus nutrition in the tropics. In C.S. Andrew and E.J.
Kamprath (ed.) Mineral Nutrition of Legumes in Tropical and Subtropical Soils.
CSIRO, Brisbane, Australia.

253
Francis, C.A. and J.W. King. 1988. Cropping systems based on farm-derived, renewable
resources. Agrie. Systems 27:67-75.
Francis, C.A. and W.W. Sahs. 1988. Research for sustainable agriculture by U.S.
universities, p. 113-119. In P. Allen and D. van Dusen (ed.) Global Perspectives on
Agroecology and Sustainable Agricultural Systems. Proceedings of the Sixth
International Scientific Conference of the International Federation of Organic
Agriculture Movements. Agroecology Program, University of California, Santa
Cruz, California.
Francis, C.A., D. Sander, and A. Martin. 1987. Search for a sustainable agriculture:
reduced inputs and increased profits. Crops & Soils 39(9): 12-14.
Francis, C.A. and G. Youngberg. 1990. Sustainable agriculture — an overview, p. 1-23. In
C.A. Francis, C.B. Flora, and L.D. King (ed.) Sustainable Agriculture in
Temperate Zones. John Wiley & Sons, Inc., New York.
Fresco, L.O. and S.B. Kroonenberg. 1992. Time and spatial scales in ecological
sustainability. Land Use Policy, July 1992:155-168.
Geng, S., C.E. Hess, and J. Auburn. 1990. Sustainable agricultural systems: concepts and
definitions. J. Agronomy & Crop Science 165:73-85.
Gips, T. 1988. What is a sustainable agriculture? p. 63-74. In P. Allen and D. van Dusen
(ed.) Global Perspectives on Agroecology and Sustainable Agricultural Systems.
Proceedings of the Sixth International Conference of the International Federation
of Organic Agriculture Movements. Agroecology Program, University of
California, Santa Cruz, California.
Gliessman, S.R. 1990. Understanding the basis of sustainability for agriculture in the
tropics: experiences in Latin America, p. 378-390. In C.A. Edwards, R. Lai, P.
Madden, R.H. Miller and G. House (ed.) Sustainable Agricultural Systems. Soil
and Water Conservation Society, Ankeny, Iowa.
Grant, W.R., J.W. Richardson, B.W. Brorsen, and M.E. Rister. 1984. Economic impacts
of increased price variability: a case study with rice. Agrie. Econ. Res.,
36(4): 17-27.
Gray, R. 1991. Economic measures of sustainability. Can. J. Agrie. Econ. 39:627-635.

254
Hamblin, A. 1992. How do we know when agricultural systems are sustainable? p. 90. In
A. Hamblin (ed.) Environmental Indicators for Sustainable Agriculture. Report on
a national workshop, November 28-29, 1991. Bureau of Rural Resources, Land
and Water Resource Research and Development Corporation, Grains Research
Corporation, Canberra, Australia.
Hansen, J.W., N.B. Pickering, J.W. Jones, C. Wells, H. Chan, and D.C. Godwin. 1994.
WeatherMan. p. 137-199. In G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT
Version 3, Vol. 3. International Benchmark Sites Network for Agrotechnology
Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Hansen, J.W., P.K. Thornton and J.W. Jones. 1995. Toward data standards for enterprise
and farm level analyses. Paper presented at the 2nd International Symposium on
Systems Approaches for Agricultural Development, 6-8 December, 1995,
International Rice Research Institute, Los Baños, The Philippines.
Harrington, L.W. 1992. Measuring sustainability: issues and alternatives. J. Farming
Systems Research - Extension 3:1-20.
Harrington, L.W., P. Hobbs, T. Pokhrel, B. Sharma, S. Fujisaka, and S. Lightfoot. (1990)
The rice-wheat pattern in the Nepal Terai: issues in the identification and definition
of sustainability problems. J. Farming Systems Research - Extension 1 (2): 1 -27.
Hart, R.D. 1986. Ecological framework for multiple cropping research, p. 40-56. In C.A.
Francis (ed.) Multiple Cropping Systems. Macmillian Publishing Co., New York.
Harwood, R.R. 1990. A history of sustainable agriculture, p. 3-19. In C.A. Edwards, R.
Lai, P. Madden, R.H. Miller, and G. House (ed.) Sustainable Agricultural Systems.
Soil and Water Conservation Society, Ankeny, Iowa.
Hauptli, H., D. Katz, B.R. Thomas, and R.M. Goodman. 1990. Biotechnology and crop
breeding for sustainable agriculture, p. 141-156. In C.A. Edwards, R. Lai, P.
Madden, R.H. Miller, and G. House (ed.) Sustainable Agricultural Systems. Soil
and Water Conservation Society, Ankeny, Iowa.
Held, L.J. and G.A. Helmers. 1981. Growth and survival in wheat farming: the impact of
land expansion and borrowing restraints. Western J. Agrie. Econ. 6:207-216.
Hernandez, M., L.J. Lane, and J.J. Stone. 1989. Surface runoff, p. 5.1-5.18. In L.J. Lane
and M.A. Nearing (ed.) USDA - Water Erosion Prediction Project: Hillslope
Profile Model Documentation. NSERL Report No. 2. USDA-ARS National Soil
Erosion Laboratory, West Lafayette, Indiana.

255
Hildebrand, P.E. 1990. Agronomy’s role in sustainable agriculture: integrated farming
systems. J. Prod. Agrie. 3:285-288.
Hill, S.B. and R.J. MacRae. 1988. Developing sustainable agriculture education in
Canada. Agrie. Hum. Values 5(4):92-95.
Hipel, K.W. and A.I. McLeod. 1994. Time Series Modelling of Water Resources and
Environmental Systems. Elsevier, Amsterdam.
Hoogenboom, G., J.W. Jones and K.J. Boote. 1990. Nitrogen fixation, uptake, and
remobilization in legumes: a modeling approach, In International Benchmark Sites
Network for Agrotechnology Transfer (IBSNAT) Project. Proceedings of the
IBSNAT Symposium: Decision Support Systems for Agrotechnology Transfer,
Las Vegas, NV. 16-18 Oct., 1989. Part II: Posters. Dept, of Agronomy and Soil
Science, College of Tropical Agriculture and Human Resources, Univ. Hawaii,
Honolulu, Hawaii.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D. Batchelor, W.T. Bowen, L.A. Hunt,
N.B. Pickering, U. Singh, D.C. Godwin, B. Baer, K.J. Boote, J.T. Ritchie, and
J.W. White. 1994. Crop models, p. 95-244 In G.Y. Tsuji, G. Uehara, and S. Balas
(ed.) DSSAT Version 3, Vol. 2. International Benchmark Sites Network for
Agrotechnology Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Hudgens, R.E. 1978. Adapting agronomic technology for small farm bean production in
highland Colombia. Ph.D. diss., University of Florida, Gainesville, Florida.
Hunt, L.A., J.W., Jones, P.K. Thornton, G. Hoogenboom, D.T. Imamura, G.Y. Tsuji and
U. Singh. 1994. Accessing data, models & application programs, p. 21-110. In
G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT Version 1, Vol. 2. International
Benchmark Sites Network for Agrotechnology Transfer, Univ. of Hawaii,
Honolulu, Hawaii.
IBSNAT—International Benchmark Sites Network for Agrotechnology Transfer. 1990.
Technical Report 2: Field and Laboratory Methods for the Collection of the
IBSNAT Minimum Data Set, 1st ed. Department of Agronomy and Soil Science,
College of Tropical Agriculture and Human Resources, University of Hawaii,
Honolulu, Hawaii, USA.
Ikerd, J.E. 1991. Applying LISA concepts on southern farms. Southern J. Agrie. Econ.
23(l):43-52.
Janssen, W. and N. Ruiz do Londoño. 1994. Modernization of a peasant crop in
Colombia: evidence and implications. Agricultural Economics 10:13-25.

Jodha, N.S. 1990. Mountain agriculture: the search for sustainability. J. Farming Systems
Research - Extension 11:55-76.
Jones, C.A., W.L. Bland, J.T. Ritchie and J.R. Williams. 1991a. Simulation of root
growth, p. 91-123. In J. Hanks and J.T. Ritchie (ed.) Modeling Plant and Soil
Systems. Agronomy Monograph no. 31. ASA-CSSA-SSSA, Madison, Wisconsin.
Jones, C.A., P.T. Dyke, J.R. Williams, J.R. Kiniry, V.W. Benson, and R.H. Griggs. 1991b.
EPIC: An operational model for evaluation of agricultural sustainability. Agrie.
Sys. 37:341-350.
Jones, C.A. and J.R. Kiniry (ed.). 1986. CERES-Maize: A Simulation Model of Maize
Growth and Development. Texas A&M Univ. Press, College Station, Texas.
Jones, J.W., W.T. Bowen, W.G. Boggess, and J.T. Ritchie. 1993. Decision support
systems for sustainable agriculture, p. 123-138. In J. Ragland and R. Lai (ed.)
Technologies for Sustainable Agriculture in the Tropics. ASA Special Publication
Number 56. ASA-CSSA-SSSA, Madison, Wisconsin.
Jones, J.W., L.A. Hunt, G. Hoogenboom, D.C. Godwin., U. Singh, G.Y. Tsuji, N.B.
Pickering, P.K. Thornton, W.T. Bowen, K.J. Boote, and J.T. Ritchie. 1994. Input
and output files, p. 1-94. In G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT
Version 3, Vol. 2. International Benchmark Sites Network for Agrotechnology
Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Jones, J.W., P.K. Thornton and J.W. Hansen. 1995. Opportunities for systems approaches
at the farm scale. Paper presented at the 2nd International Symposium on Systems
Approaches for Agricultural Development, 6-8 December, 1995, International
Rice Research Institute, Los Baños, The Philippines, (submitted)
Jones, P.G. and P.K. Thornton. 1993. A rainfall generator for agricultural applications in
the tropics. Agricultural and Forest Meteorology 63:1-19.
Kaiser, H.M., S.J. Rhia, D.S. Wilks, D.G. Rossiter and R. Sampath. 1993. A farm-level
analysis of economic and agronomic impacts of gradual climate warming. Amer. J.
Agr. Econ. 75:387-398.
Kay, J.J. 1991. A nonequilibrium thermodynamic framework for discussing ecosystem
integrity. Environ. Manage. 15:483-495.

257
Keating, B.A., D.C. Godwin, and J.M. Watiki. 1991. Optimising nitrogen inputs in
response to climatic risk. p. 329-358. In R.C. Muchow and J.A. Bellamy (ed.)
Climatic Risk in Crop Production: Models and Management for the Semi-arid
Tropics and Subtropics. CAB International, Wallingford, Oxfordshire, UK.
Keeney, D.R. 1989. Toward a sustainable agriculture: need for clarification of concepts
and terminology. Am. J. Alternative Agrie. 4(3/4): 101-105.
Kidd, C.V. 1992. The evolution of sustainability. J. Agrie. Environ. Ethics 5:1-26.
Kiniry, J.R. and C.A. Jones. 1986. Model evaluation, p.l 13-144. In C.A. Jones and J.R.
Kiniry (ed.) CERES-Maize. A Simulation Model of Maize Growth and
Development. Texas University Press, College Station, Texas.
Kirschenmann, F. 1991. Fundamental fallacies of building agricultural sustainability. J. Soil
Water Conserv. 46:165-168.
Klein, K.K. and S. Narayanan. 1992. Farm level models; a review of developments,
concepts and applications in Canada. Can. J. Agr. Econ. 40:351-368.
Knorr, D. and T.R. Watkins (ed.). 1984. Alterations in Food Production. Van Nostrand
Reinhold Company, New York.
Lai, R. 1991. Soil structure and sustainability. J. Sustainable Agrie. 1(4):67-91.
Lane, L.J. and M.A. Nearing (ed.). 1989. USDA - Water Erosion Prediction Project:
Hillslope Profile Model Documentation. NSERL Report No. 2. USDA-ARS
National Soil Erosion Laboratoiy, West Lafayette, Indiana.
Lerohl, M.L. 1991. The sustainability of selected prairie crop rotations. Can. J. Agrie.
Econ. 39:667-676.
Lockeretz, W. 1988. Open questions in sustainable agriculture. Am. J. Alternative Agrie.
3:174-181.
Lowrance, R., P.F. Hendrix, and E.P. Odum. 1986. A hierarchical approach to sustainable
agriculture. Am. J. Alternative Agrie. 1:169-173.
Lynam, J.K. and R.W. Herdt. 1989. Sense and sustainability, sustainability as an objective
in international agricultural research. Agrie. Econ. 3:381- 398.

258
MacRae, R.J., S.B. Hill, J. Henning, and G.R. Mehuys. 1989. Agricultural science and
sustainable agriculture, a review of the existing scientific barriers to sustainable
food production and potential solutions. Biol. Agrie. Hort. 6:173-219.
MacRae, R.J., S.B. Hill, G.R. Mehuys, and J. Henning. 1990. Farm-scale agronomic and
economic conversion from conventional to sustainable agriculture. Adv. Agron.
43:155-198.
Madden, P. 1987. Can sustainable agriculture be profitable? Environment 29(4):19-34.
Madden, P. 1990. The economics of sustainable low-input farming systems, p. 315-341. In
C.A. Francis, C.B. Flora, and L.D. King (ed.) Sustainable Agriculture in
Temperate Zones. John Wiley & Sons, Inc., New York.
Martin, J.P., H. Zunino, P. Peirano, M. Caiozzi, and K. Haider. 1982. Decomposition of
14C-labeled lignins, model humic acid polymers, and fungal melanins in allophanic
soils. Soil Biol. Biochem. 14:289-293.
Matthews, R.B. and L.A. Hunt. 1994. GUMCAS: a model describing the growth of
cassava (Manihot escalenta L. Crantz). Field Crops Research 36:69-84.
May, R.M. 1977. Thresholds and breakpoints in ecosystems with a multiplicity of stable
states. Nature 269:471-477.
McNemar, Q. 1947. Note on the sampling error of the difference between correlated
proportions or percentages. Psychometrika 12:153-157.
Meinke, H., P.S., Carberry, M.R. McCaskill, M.A. Hills, and I. McLeod. 1995. Evaluation
of radiation and temperature data generators in the Australian tropics and sub¬
tropics using crop simulation models. Agrie. Forest Meteorol. 72:295-316.
Mellor, J.W. 1988. The intertwining of environmental problems and poverty. Environment
30(9):8-13,28-30.
Monreal, C.M., W.B. McGill, and J.D. Etchevers. 1981. Internal nitrogen cycling
compared in surface samples of an andept and a mollisol. Soil Biol. Biochem.
13:451-454.
Monteith, J.L. 1990. Can sustainability be quantified? ICRISAT Conference Paper No.
538, ICRISAT, Patancheru, Andhra Pradesh, India.

259
Mullin, B.A., G.S. Abawi, M.A. Pastor-Corrales, and J.L. Kornegay. 1991. Root knot
nematodes associated with beans in Colombia and Peru and related yield loss.
Plant Disease 75:1208-1211.
Munevar, F. and A.G. Wollum. 1977. Effects of additions of phosphorus and inorganic
nitrogen on carbon and nitrogen mineralization in Andepts from Colombia. Soil
Sci. Soc. Am. J. 41:540-545.
Muñoz, R.A. and A.P. Wieczoreck. 1978. Fertilización de maiz (Zea mays L.) en suelos
volcánicos de Nariño, Colombia. Revista ICA (Bogotá) 13:11-20.
Neher, D. 1992. Ecological sustainability in agricultural systems: definition and
measurement, p. 51-61. In R.K. Olson (ed.) Integrating Sustainable Agriculture,
Ecology, and Environmental Policy. Food Products Press, New York.
Neufeldt, E. (ed.). 1988. Webster’s New World Dictionary, Third College Edition. Simon
& Schuster, Inc., New York.
Norgaard, R.B. 1991. Sustainability: Three methodological suggestions for agricultural
economics. Can. J. Agrie. Econ. 39:637-645.
O’Connell, P.F. 1990. Policy development for the low-input sustainable agriculture
program, p. 453-458. In C.A. Edwards, R. Lai, P. Madden, R.H. Miller, and G.
House (ed.) Sustainable Agricultural Systems. Soil and Water Conservation
Society, Ankeny, Iowa.
O’Connell, P.F. 1992. Sustainable agriculture -- a valid alternative. Outlook on Agrie.
21(1):5-12.
Oram, P.A. 1988, Building the agroecological framework. Environment
30(9): 14-17,30-36.
Overman, A.J. 1991. Diseases caused by nematodes, p. 49-50. In J.B. Jones, J.P. Jones,
R.E. Stall and T.A. Zitter (ed.) Compendium of Tomato Diseases. APS Press, St.
Paul, Minnesota.
Pachico, D. (undated). Costs and returns in three bean systems, Nariño. (unpublished
draft)
Pankratz, A. 1983. Forecasting With Univariate Box-Jenkins Models: Concepts and
Cases. John Wiley & Sons, New York.

260
Perry, G.M., M.E. Rister, J.W. Richardson, W.R. Grant, and J.W. Sij, Jr. 1986. The
Impact of Tenure Arrangements and Crop Rotations on Upper Gulf Coast Rice
Farmers. B-1530. Texas Agricultural Experiment Station, College Station, Texas.
Plucknett, D.L. 1990. International goals and the role of the international agricultural
research centers, p. 33-49. In C.A. Edwards, R. Lai, P. Madden, R.H. Miller, and
G. House (ed.) Sustainable Agricultural Systems. Soil and Water Conservation
Society, Ankeny, Iowa.
Rai, R. 1992. Effect of nitrogen levels and Rhizobium strains on symbiotic N2 fixation and
grain yield of Phaseolns migar is L. genotypes in normal and saline-sodic soils.
Biol. Fértil. Soils 14:293-299.
Richardson, C.W. 1985. Weather simulation for crop management models. Trans. ASAE
18:1602-1606.
Richardson, J.W. and G.D. Condra. 1981. Farm size evaluation in the El Paso Valley: a
survival/success approach. Am. J. Agrie. Econ. 63:430-437.
Richardson, J.W. and C.L. Nixon. 1985. FLIPSIM v: a general firm level policy simulation
model. B-1528. Department of Agricultural Economics, Texas A&M Univ.
Rietveld, M.R. 1978. A new method for estimating the regression coefficients in the
formula relating solar radiation to sunshine. Agrie. Meteorology 19:243-252.
Ritchie, J.T., D.C. Godwin, and U. Singh. 1990a. Soil and water inputs for the IBSNAT
crop models, p.31-45. In International Benchmark Sites Network for
Agrotechnology Transfer (IBSNAT) Project. Proceedings of the IBSNAT
Symposium: Decision Support Systems for Agrotechnology Transfer, Las Vegas,
NV. 16-18 Oct., 1989. Part I: Symposium Proceedings. Dept, of Agronomy and
Soil Science, College of Tropical Agriculture and Human Resources, Univ.
Hawaii, Honolulu, Hawaii.
Ritchie, J.T., D.C. Godwin, and U. Singh. 1990b. The CERES models of crop growth and
yield. In International Benchmark Sites Network for Agrotechnology Transfer
(IBSNAT) Project. Proceedings of the IBSNAT Symposium: Decision Support
Systems for Agrotechnology Transfer, Las Vegas, NV. 16-18 Oct., 1989. Part II:
Posters. Dept, of Agronomy and Soil Science, College of Tropical Agriculture and
Human Resources, Univ. Hawaii, Honolulu, Hawaii.
Rodale, R. 1990. Finding the middle of the road on sustainability. J. Prod. Agrie.,
3:273-276.

261
Ruttan, V.W. 1988. Sustainability is not enough. Am. J. Alternative Agrie. 3:128-130.
Sands, G.R. and T.H. Podmore. 1993. Development of an environmental sustainability
index for irrigated agricultural systems, p. 71-80. In J.K. Mitchell (ed.) Integrated
Resource Management and Landscape Modification for Environmental Protection.
Proceedings of the International Symposium 13-14 December 1993, Chicago,
Illinois.
Scholberg, J.M.S., J.W. Jones, F.H. Beinroth, M.A. Vazquez, B.L. McNeal, and K.J.
Boote. 1995. Modeling Growth of Field-grown Tomatoes in Puerto Rico Using
CROPGRO: Model Validation Using a Data Input Set Collected at the Isabela
Substation, Puerto Rico. Unpublished research report, University of Florida,
Gainesville, Florida.
Scholberg, J.M.S., B.L. McNeal, J.W. Jones, and C.D. Stanley. 1993. Adaptation of a
generic crop-growth model (CROPGRO) for field-grown tomatoes and peppers.
Agronomy Abstracts.
Sentis, I.P. 1992. La erodabilidad de los Andisoles en Latino America. Suelos Ecuatoriales
22:33-43.
Singh, U. 1985. A crop growth model for predicting com performance in the tropics.
Ph.D. diss. University of Hawaii, Honolulu, Hawaii (Diss. Abstr. 85-20320).
Singh, U. and D.C. Godwin. 1990. Phosphorus dynamics in IBSNAT crop models. In
International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT)
Project. Proceedings of the IBSNAT Symposium: Decision Support Systems for
Agrotechnology Transfer, Las Vegas, NV. 16-18 Oct., 1989. Part II: Posters.
Dept, of Agronomy and Soil Science, College of Tropical Agriculture and Human
Resources, Univ. Hawaii, Honolulu, Hawaii.
Singh, U. and P.K. Thornton. 1992. Using crop models for sustainability and
environmental quality assessment. Outlook on Agriculture 21:209-218.
Singh, U., P.K. Thornton, A.R. Saka, and J.B. Dent. 1993. Maize modeling in Malawi: a
tool for soil fertility research and development, p.253-273. In F.W.T. Penning de
Vries, P.S. Teng, and K. Metselaar (ed.) Systems Approaches for Agricultural
Development, vol. 2. Kluwer Academic Publishers, Dordrecht, The Netherlands.
Snedecor, G.W. and W.G. Cochran. 1980. Statistical Methods, Seventh Edition. The Iowa
University Press, Aimes, Iowa.

262
Soil Survey Staff. 1990. Keys to Soil Taxonomy, Fourth Edition. SMSS Technical
Monograph No. 6. Soil Management Support Service, Blacksburg, Virginia.
Sokal, R.R. and F.J. Rohlf. 1981. Biometry, Second Edition. W.H. Freedman and
Company, New York.
Spencer, D.S.C. and M.J. Swift. 1992. Sustainable agriculture: definition and
measurement, p. 15-24. In K. Mulongoy, M. Gueye, and D.S.C. Spencer (ed.)
Biological Nitrogen Fixation and Sustainability of Tropical Agriculture.
Wiley-Sayce Co., West Sussex, UK.
Stinner, B.R. and G.J. House. 1987. Role of ecology in lower-input, sustainable
agriculture: an introduction. Am. J. Alternative Agrie. 2:146- 147.
Stinner, B.R. and G.J. House. 1989. The search for sustainable agroecosystems. J. Soil
Water Conserv. 44:111-116.
Stockle, C.O., R.I. Papendick, K.E. Saxton, G.S. Campbell, and F.K. van Evert. 1994. A
framework for evaluating the sustainability of agricultural production systems. Am.
J. Alternative Agrie. 9(l&2):45-50.
Taylor, D.C., Z.A. Mohammed, M.A. Shamsudin, M.G. Mohayidin, and E.F.C. Chiew.
1993. Creating a farmer sustainability index: a Malaysian case study. Am. J.
Alternative Agrie. 8:175-184.
Tewari, S. 1995. The relation between mode of legume nitrogen nutrition, yield
determinants and N assimilation efficiency. Ph.D. diss. Univ. of Hawaii, Honolulu
(Diss. Abstr. 95-32632).
Thompson, P.B. 1992. The varieties of sustainability. Agrie. Human Values 9(3); 11-19.
Thornton, P.K., J.W. Hansen, E.B. Knapp, and J.W. Jones. 1995. Designing optimal crop
management strategies. In J. Bouma, A. Kuyvenhoven, B.A.M. Bouman, J.C.
Luyten, and H.G. Zandstra (ed.) Ecoregional Approaches for Sustainable Land
Use and Food Production. Kluwer Academic Publishers, Dordrecht, The
Netherlands, (in press)
Thornton, P.K., G. Hoogenboom, P.W. Wilkens, and W.T. Bowen. 1995. A computer
program to analyze multiple-season crop model output. Agron. J. 87:131-136.

263
Thornton, P.K., P.W. Wilkens, G. Hoogenboom, and J.W. Jones. 1994. Sequence
analysis, p. 67-136. In G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT Version
3, Vol. 2. International Benchmark Sites Network for Agrotechnology Transfer,
Univ. of Hawaii, Honolulu, Hawaii.
Torquebiau, E. 1992. Are tropical agroforestry gardens sustainable? Agrie. Ecosys.
Environ. 41:189-207.
Trenbath, B.R., G.R. Conway, and L.A. Craig. 1990. Threats to sustainability in
intensified agricultural systems: analysis and implications for management, p.
337-365. In S.R. Gliessman (ed.) Agroecology: Researching the Ecological Basis
for Sustainable Agriculture. Springer-Verlag, New York.
Tsuji, G.Y., G. Uehara, and S. Balas (ed.) DSSAT Version 3. International Benchmark
Sites Network for Agrotechnology Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Uehara, G. and G.P. Gillman. 1981. The Mineralogy, Chemistry, and Physics of Tropical
Soils with Variable Charge Clays. Westview Press, Boulder, Colorado.
Walker, O.L., M.L. Hardin, H.P. Mapp Jr., and C.E. Roush. 1979. Farm growth and
estate transfer in an uncertain environment. Southern J. Agrie. Econ. 11:33-44.
Wambeke, A. van. 1992. Soils of the Tropics: Properties and Appraisal. McGraw-Hill,
New York.
Weil, R.R. 1990. Defining and using the concept of sustainable agriculture. J. Agron.
Educ. 19(2):126-130.
White, J.W., G. Hoogenboom, J.W. Jones, and K.J. Boote. 1995. Evaluation of the dry
bean model BEANGRO VI.01 for crop production research in a tropical
environment. Expl. Agrie. 31:241-254.
Williams, D.A. 1976. Improved likelihood ratio tests for complete contingency tables.
Biometrika 63:33-37.
Williams, J.R., C.A. Jones, and P.T. Dyke. 1984. A modeling approach to determining the
relationship between erosion and soil productivity. Trans. ASAE 27:129-144.
Wischmeier, W.H. and D.D. Smith, 1978. Predicting Rainfall Erosion Losses. Agricultural
Handbook No. 537. USDA, Washington D.C.
Yetley, 1992. Sustainable agriculture in the 1990 Farm Bill. Proceedings of the
Philadelphia Society for Promoting Agriculture 1991-1992. pp. 21-38.

264
York, E.T., Jr. 1988. Improving sustainability with agricultural research. Environment
30(9): 18-20,36-40.
York, E.T., Jr. 1991. Agricultural sustainability and its implications to the horticulture
profession and the ability to meet global food needs. HortScience 26:1252-1256.
Zandstra, H. 1994. Sustainability and productivity growth: Issues, objectives, and
knowledge needs -- guidelines for working groups. In Reconciling sustainability
with productivity growth. Report of a workshop, Gainesville, Florida. May 1993.
Univ. of Florida and Cornell Univ.
Zunino, H., F. Borie, S. Aguilera, J.P. Martin, and K. Haider. 1982. Decomposition of
14C-labeled glucose, plant and microbial products and phenols in volcanic ash-
derived soil of Chile. Soil Biol. Biochem. 14:37-43.

BIOGRAPHICAL SKETCH
James Hansen was born in Encino, California, on March 1960. He grew-up in
California, Hawaii and Guam. He received his high school diploma in May 1978 from
John F. Kennedy High School in Turnon, Guam, then worked as an aircraft mechanic
before starting studies at the University of Hawaii. From the University of Hawaii, he
received a Bachelor of Science degree in general tropical agriculture in May 1985, and a
Master of Science degree in agronomy and soil science with an emphasis in soil chemistry
in May 1989. For two years he worked at the University of Hawaii as a research associate
developing a multiple cropping systems model. He spent the year following September
1989 working with a Southern Baptist agricultural development project in the Philippines.
James moved to Gainesville, Florida in August 1991 to pursue a Ph.D. in agricultural
operations management. James received his degree in May 1996.
265

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
- i, :
//V»—
íes W. Jones^Chair
irofessor of Agricultural and Biological
Engineering
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
Robert M. Peart
Graduate Research Professor of
Agricultural and Biological Engineering
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.
William G. Boggess
Professor of Food and Resource
Economics
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is Mly-adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosopny.
Peter E. Hildebrand
Professor of Food and Resource
Economics
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy.'
Phillip K. Thornton
Senior Scientist, International Fertilizer
Development Center

This dissertation was submitted to the Graduate Faculty of the College of
Agriculture and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
May, 1996
Dean, College of Agriculture
Dean, Graduate School



184
because it matches reported interest rates for crop loan packages after adjusting for
inflation. An additional credit scenario increases the amount of credit available by 50%.
Sources of risk. A set of four scenarios was designed to examine the impact of
weather and price risk, and of diversification in space on farm risk and sustainability. The
relevant hypotheses are (a) weather and price variability contribute unequally to whole-
farm risk and to the probability of failure, and (b) spatial diversification reduces risk and
the probability of failure.
Price risk was eliminated in the no price risk scenario by using historical prices
observed from August 1979 to September 1988 as a proxy for future prices. Because of
gaps in the historical price sequences, bean prices used were from September 1978 to June
1980 and July 1984 to September 1988 and maize prices were from August 1979 to
September 1989. FSS used Eq. [4-6] to adjusted prices for the trends given in Table 6-3.
Risk here refers to the probability distribution of future outcomes that results from
a range of possible realizations of the future. The procedure for removing variability
between replicates eliminates risk because it constrains the future to only one possible,
deterministic realization. However, the procedure does not remove variability between
years. The pattern of variability of prices through time does affect farm performance and
vulnerability to weather-induced production risk. Removing price risk in this manner is
analogous to obtaining a contract to sell produce according to predetermined, fixed
schedule of future prices.
Weather risk was eliminated by simulating a modification of the base scenario with
an ample credit supply, determining which replicate of the scenario had the median value


114
(Fox, 1978). The slope of each sorption curve in Fig. 5-5 indicates both the amount of
fertilizer P needed to raise solution concentrations a given amount and the capacity of
adsorbed P to replenish solution P that is removed by roots. Fox (1978) showed that
crops respond to soil solution P concentrations consistently across different soil types.
Inhibited microbial decomposition causes many Andisols to accumulate large
reserves of organic matter. Although the supply of organic N may be high and climatic
conditions favorable to decomposition, N mineralization rates are generally lower than in
Figure 5-6. Decomposition of soil organic C in three allophanic and seven non-
allophanic soils. Data from Martin et al., 1982.


26
environment. Resilience shares with time trend approaches the criticism that it ignores the
goals of the human actors within agricultural systems.
System simulation
Simulation has been used to characterize the sustainability of crop production in
response to soil dynamics. Singh and Thornton (1992) illustrated the use of long-term
simulation of crop sequences replicated with stochastic inputs of weather data to examine
trends and variability in yields. Lerohl (1991) used the Erosion-Productivity Impact
Calculator (EPIC) (Williams et al, 1984) to study the long-term impact of predicted soil
erosion on productivity of crop rotations on four soil types in Alberta, Canada.
Sustainability was inferred in all soil-rotation combinations because no negative trend in
crop yields could be detected during a simulated 100 year period.
Other studies have used crop simulation models to examine relationships between
production and environmental degradation. Singh and Thornton (1992) illustrated the use
of CERES-Maize (Jones and Kiniry, 1986) to simulate the effects of soil type and rate of
application of N fertilizer on distributions of maize grain yield and N03" leaching into
groundwater from upland fields in Chiang Mai, Thailand. Alocilja and Ritchie (1993) used
CERES-Maize and a multiple goal optimization technique to identify sets of N fertilization
schedules that were optimal in the sense that neither production nor water quality could be
improved without decreasing satisfaction of the other goal.
Several whole-farm simulation studies have looked at the effect of various factors
on farm survivability. For example, Perry et al. (1986) examined effects of production


197
important determinant of farm sustainability. The initial supply of cash was a significant (P
= 0.05) but much less important of sustainability. Access to Col.$ 2,000,000 of credit at
an interest rate of 19% improved sustainability a small but significant amount (Fig. 6-14).
Increasing the amount of credit available to Col.$ 3,000,000 did not further improve
sustainability, but reducing the interest rate to 9.5% did improve sustainability.
Table 6-10. Relative sensitivity (r) of predicted 15 year sustainability ((15)SE$) to
continuous factors, and McNemar (GPadj) and G-test (G^) statistics for difference from
the base scenario.
Factor
Base Adjusted
1(15) SEs
r
Gp,adj
Gl,adj
area cultivated (ha)
2.34
2.57
1.00 0.000
5.72
...t
...t
mean commodity prices1
100%
110%
0.98 0.014
5.31
t
43.6 **
subsistence consumption
(million Col.$ yr'1)
2.75
2.48
0.97 0.017
5.25
t
39.1 **
material input prices1
100%
80%
0.80 0.040
1.25
t
6.4 *
hired labor wages*
100%
80%
0.79 0.041
1.17
...t
5.5 *
discretionary consumption
(percent liquid assets yr'1)
20%
16%
0.78 0.041
1.09
14.6 **
4.7 *
initial funds (million Col.$)
4.00
4.40
0.69 0.046
0.78
3.9 *
0.6 n.s.
N fertilizer (kgN ha'1 yr'1)8
25.0
50.0
0.49 0.050
-0.23
10.6 **
4.6 *
N fertilizer (kg N ha'1 yr'1)8
25.0
12.5
0.29 0.045
-1.09
41 9 **
25.0 **
t Undefined.
* Percent of base scenario.
§ Average among the three phases of rotation.


69
farmer. The social component of a farming system is the farmer or farm household, and
includes goals and decision criteria. In FSS, objects represent each of these components:
fields in the farm landscape, a set of strategies that specify the sequence of crop activities
and their management, a set of enterprises that link management strategies to particular
fields within the landscape, and a set of farm resources (Fig. 4-2). The household is
represented by decision rules for consumption, production and farm failure. The object-
oriented design of FSS provides a flexible means of representing farm resources, possible
interactions among resources, and relationships between operations and the resources that
they use. The Farming System Simulator is a farm model only in a loose sense; the model
structure of a particular farm is specified at run-time by the resource, field, enterprise and
strategy objects that are initialized in response to input data.
The capabilities of FSS reflect its purpose as a tool for characterizing farm
sustainability based on the framework presented in Chapter 3. The first requirement was
the ability to replicate a long-term scenario with constant initial conditions but stochastic
inputs. FSS takes advantage of a stochastic weather generator, WGEN (Richardson,
1985), that has been incorporated into the crop models that are part of the Decision
Support System for Agrotechnology Transfer version 3 (DSSAT, Hoogenboom et al.,
1994). An analogous stochastic price generator is part of FSS. FSS addresses its second
requirementan ability to simulate ecological processes of crop productionby calling
external crop simulation models. The crop models simulate weather variability, soil
dynamics and crop growth and development, then return the information that FSS requires
in the form of schedules of field operations. Third, FSS addresses the need to deal with


169
Table 6-1. Hypotheses related to determinants of farm sustainability, Domingo farm,
Cauca, Colombia.
Determinant
Hypothesis
Cropping system
Farm sustainability is dependent on cropping system.
Diversified annual crop rotations contribute to a more sustainable
farming system than do monocultures.
Incorporating a high-valued, irrigated vegetable (i.e., tomato) into a
diversified rotation of traditional crops enhances farm sustainability.
Sustainability of a farm in coffee monoculture is positively related
to coffee yield.
Soil management
Either excessive or insufficient N fertilizer inputs reduce farm
sustainability.
Soil erosion reduces farm sustainability.
Costs and prices
Low prices for crop products constrain farm sustainability.
High material input prices constrain farm sustainability.
High wages for hired labor constrain farm sustainability.
A high subsistence spending requirement constrains farm
sustainability.
A high level of discretionary spending constrains farm
sustainability.
Resources
The decision to not cultivate the degraded land now in permanent
grass fallow constrains farm sustainability.
Limited initial savings constrains farm sustainability.
Access to credit enhances farm sustainability.
When credit is available, farm sustainability is negatively related to
loan interest rate.
Sources of risk
Conventional credit enhances farm sustainability more than credit
packages given for specific crops.
Weather and price variability contribute unequally to whole-farm
risk and to the probability of failure.
Spatial diversity reduces farm risk and enhances sustainability.


191
fallow time. The results do not support the hypothesis that diversified rotations are
consistently more sustainable than monocultures. Although the three most sustainable
annual cropping systems were diversified rotations, sustainability of the cassava
monoculture (0.47) was higher than that of the maize-bean rotation (0.04). The maize
monoculture was the least sustainable annual cropping system scenario.
The simulated sustainability of coffee monoculture was positively related to yields
(Fig. 6-8). Coffee production was more sustainable than the base scenario when yields
were at least 2.25 Mg ha'1 yr'1. Coffee yields can be expected to yield between about 2.0
Figure 6-8. Sustainability time plot of base and coffee scenarios.


63
Although sustainability of a real agricultural system cannot be observed because it deals
with the future, it can be estimated from simulation of a system model. Testing
hypotheses about constraints to system sustainability is then straightforward. Applying
this framework to farming systems results in an approach to characterizing sustainability
that is literal, system-oriented, quantitative, predictive, stochastic and diagnostic (Chapter
2). The use of such an approach could provide attempts to improve farm sustainability
with objective feedback.
The requirement for comprehensive and realistic farm simulation tools currently
limits application of the proposed approach. Most existing farm-level simulation models
are not sufficiently comprehensive; they do not integrate models of crop and animal
production, environmental degradation, economic processes, and farmer production and
consumption decisions. A study designed to examine a single constraint to farm
sustainability would be less demanding in its model and data requirements.
One could question the realism of assumptions about the future behavior of inputs
to a farming system that are required for characterizing its sustainability. However, all
approaches to characterizing sustainability involve inferences about the future. If high-
level systems change more slowly than lower-level systems, as ecological hierarchy theory
asserts (Allen & Starr, 1988), then future projections of inputs such as weather, prices,
infrastructure and technology are more defensible than extrapolation of past farming
system behavior.
The study by Perry et al. (1986) illustrates the utility of the framework presented
in this paper for characterizing sustainability. Although I interpreted their study beyond its


233
Table B-14. Format of the ARMA section of the price file.
Description
Variable1
Header
Format1
Index of price
N
013
Name of price
Decsr
1 C 16
Transformation
Tr
1 C 1
N = no transformation
L = log,0 transformation
Intercept of linear trend at the specified starting date
a
Intcp
0 R 8
Slope of linear trend relative to specified starting date
P
Slope
0 R 8
Year of starting date for trend calculation
StYr
114
Month of starting date for trend calculation
StMo
312
Mean deviation from trend in month m (m = Jan, ..., Dec)
S m
0 R 8
Standard deviation of random shock
StdDev
0 R 8
Autoregressive coefficient for a lag of / months (i = 1,..., 6)
4>,
AR;
0 R 7
Moving average coefficient for a lag of i months (/ = 1,..., 6)
0,
MA;
0 R 7
Lag (months) for multiplicative seasonal moving average tem
s
Lag
014
Multiplicative seasonal moving average coefficient
es
MALag
0 R 7
1 From Eq. [4-1] and [4-2].
* See footnote, Table B-l.
A price initialized from the HISTORICAL section of the price file returns a
historical series of prices read from the data file, adjusted for the specified trend (Eq. [4-5]
or [4-6]). Table B-l 5 gives the format of the HISTORICAL section. The test file
containing the historical series should contain two columns: month number (relative to the
specified starting date) and the value of the price.


7
Harwood (1990) and Kidd (1992) trace the historical development of the sustainable, or
alternative agriculture movement.
Differences in values and practices promoted as sustainable have been attributed to
differences in the problems emphasized (Carter, 1989) and to different visions of what
agriculture should be like (Thompson, 1992). Originally, the advocates of alternative
approaches to agriculture-all united in their critique of industrial agriculture as being
unsustainabledebated among themselves the future direction and shape of agriculture
(Dahlberg, 1991, p. 337). Some have focused on identifying sustainable alternatives to
existing management practices while others have advocated new philosophical orientations
toward agriculture.
Sustainability as an alternative ideology
MacRae et ah (1990), Neher (1992) and Francis and Youngberg (1990) defined
sustainable agriculture as a philosophy (Table 2-1). Ikerd (1991) described low-input,
sustainable agriculture (LISA) as more a philosophy than a practice. Examining the
concept of conventional agriculture is important since sustainable agriculture is often
described by its contrast with conventional agriculture (Lockeretz, 1988; MacRae et ah ,
1989; Hauptli etah, 1990; Dobbs etah, 1991; OConnell, 1992; Hill and MacRae, 1988).
Conventional agriculture. The concept of conventional agriculture was apparently
developed in order to clarify, and justify alternative approaches to agriculture.
Conventional agriculture is characterized as capital-intensive, large-scale, highly
mechanized agriculture with monocultures of crops and extensive use of artificial


263
Thornton, P.K., P.W. Wilkens, G. Hoogenboom, and J.W. Jones. 1994. Sequence
analysis, p. 67-136. In G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT Version
3, Vol. 2. International Benchmark Sites Network for Agrotechnology Transfer,
Univ. of Hawaii, Honolulu, Hawaii.
Torquebiau, E. 1992. Are tropical agroforestry gardens sustainable? Agrie. Ecosys.
Environ. 41:189-207.
Trenbath, B.R., G.R. Conway, and L.A. Craig. 1990. Threats to sustainability in
intensified agricultural systems: analysis and implications for management, p.
337-365. In S.R. Gliessman (ed.) Agroecology: Researching the Ecological Basis
for Sustainable Agriculture. Springer-Verlag, New York.
Tsuji, G.Y., G. Uehara, and S. Balas (ed.) DSSAT Version 3. International Benchmark
Sites Network for Agrotechnology Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Uehara, G. and G.P. Gillman. 1981. The Mineralogy, Chemistry, and Physics of Tropical
Soils with Variable Charge Clays. Westview Press, Boulder, Colorado.
Walker, O.L., M.L. Hardin, H.P. Mapp Jr., and C.E. Roush. 1979. Farm growth and
estate transfer in an uncertain environment. Southern J. Agrie. Econ. 11:33-44.
Wambeke, A. van. 1992. Soils of the Tropics: Properties and Appraisal. McGraw-Hill,
New York.
Weil, R.R. 1990. Defining and using the concept of sustainable agriculture. J. Agron.
Educ. 19(2):126-130.
White, J.W., G. Hoogenboom, J.W. Jones, and K.J. Boote. 1995. Evaluation of the dry
bean model BEANGRO VI.01 for crop production research in a tropical
environment. Expl. Agrie. 31:241-254.
Williams, D.A. 1976. Improved likelihood ratio tests for complete contingency tables.
Biometrika 63:33-37.
Williams, J.R., C.A. Jones, and P.T. Dyke. 1984. A modeling approach to determining the
relationship between erosion and soil productivity. Trans. ASAE 27:129-144.
Wischmeier, W.H. and D.D. Smith, 1978. Predicting Rainfall Erosion Losses. Agricultural
Handbook No. 537. USDA, Washington D.C.
Yetley, 1992. Sustainable agriculture in the 1990 Farm Bill. Proceedings of the
Philadelphia Society for Promoting Agriculture 1991-1992. pp. 21-38.


9
Table 2-2. Contrasting approaches of conventional and sustainable agriculture as
characterized by Hill and MacRae (1988, p. 95).
Conventional agriculture
Sustainable agriculture
Symptoms
Causes, prevention
Reductionist
Holistic
Eliminate Enemies
Respond to indicators
Narrow focus (neglects side effects, health
& environmental costs ignored)
Broad focus (subcellular to all life to globe, all
costs internalized)
Instant
Long time frame (future generations)
Single, simple (magic bullet, single discipline)
Multifaceted, complex (multi- & trans-
disciplinary)
Temporary solutions
Permanent solutions
Unexpected disbenefits (to person & planet)
Unexpected benefits
High power (risk of overkill & errors/ accidents)
Low power (minimal risk)
Direct attack
Indirect, benign approaches (catalytic, multiplier,
synergistic effects)
Imported
Local solutions and materials
Products
Processes, services
Physico-chemical (often unnatural, synthetic)
Bio-ecological (natural)
Technology-intensive
Knowledge/skill intensive
Centralized
Decentralized (human scale)
Values secondary
Compatible with higher values
Expert, paternalistic (arrogant)
Individual/community responsibility (humble)
Dependent
Self-maintaining/regulating
Inflexible
Flexible
Ignores freedom of choice (unjust)
Respects freedom of choice (just)
Disempowering
Empowering
Competitive
Co-operative
Authored
Anonymous (seeking neither reward or fame)


189
progresses. Finally Fig. 6-6 presents both sustainability and hazard time plots for the base
scenario. Chapter 3 discusses interpretation of the sustainability and hazard time
functions.
We can make at least two observations from Fig. 6-6. First, h(t) = 0 and S(t) = 1
for the first three years of the scenario. The high initial monetary supply protected the
farm from failing regardless of yields or prices. Second, the hazard function peaked in
years 6, 9 and 12 of the scenario. This was apparently an effect of the three-year crop
rotation. Although the phases of the rotation were distributed among equal field areas,
productivity of the six fields differed. Failure was most likely to occur during the rainy
Figure 6-6. Sustainability and hazard time plots of the base scenario.
Hazard, h(t)


214
SCENARIO, RESOURCES, LINKAGES, OPERATION REQUIREMENTS,
LANDSCAPE, STRATEGIES, ENTERPRISES, and CONSUMPTION DECISIONS
sections. Sections may contain either a single item or a list of items. Items in a list that
have an integer index, N, as their first field must be numbered consecutively.
To provide flexibility while avoiding redundancy, the sections of a scenario file
often relate to other sections or to items in the price or crop management files. Items in
the RESOURCES, LINKAGES, OPERATION REQUIREMENTS, SCHEDULED
OPERATIONS, LANDSCAPE, LIVESTOCK, STRATEGIES, ENTERPRISES, and
ANALYSES sections have an integer index, N, as their first field. Text names also
identify RESOURCES and price models from the price file. Items in one section often use
a name or index number to point to items in another section. For example, an enterprise
includes a set of indices to strategies from the STRATEGIES section and a set of field
indices from the LANDSCAPE section. Strategies point to the LIVESTOCK and
SCHEDULED OPERATIONS sections, and to the various crop management sections in
the experiment file.
SCENARIO section
The purpose of the SCENARIO section is to identify the farm scenario and to
specify crop management, prices and soil data input files, units of wealth, and simulation
control options. The SCENARIO section contains a single item.


101
Sustainability time plots are based on SUSTAIN.OUT. The McNemar test (Eq. [3-19]
and [3-20]) for differences in sustainabilities uses the occurrence of continuation matched
by replicate in STATUS.PRN. STATUS.PRN is formatted for importing into spreadsheet
software. Two additional filesa record of events (EVENT. OUT) and a record of
resource transactions (TRANSACT.OUT)were designed for model testing. FSS uses an
external graphics program written for DSS AT3, WMGRAF, to display the three types of
graphs described below.
Resource status. A resource box plot (Fig. 4-15) shows the minimum, 25th, 50th,
75th percentiles and maximum values of a resource among realizations for each year of a
scenario. If an aggregate resource that represents the overall condition of the system (eg.,
liquid assets) is selected, then the box plot provides an overview of how the level and
dispersion of the systems condition change through time. A final resource distribution
plot (Fig. 4-16) gives a more detailed picture of the distribution of the value of a resource
at a single time at the end of the scenario. The high probability of a zero value in Fig. 4-
16 reflects the realizations that failed before the end of the 15-year scenario.
Sustainability. FSS can create sustainability time plots (Fig. 4-17) that show the
probability of continuation from the start to each month of the scenario. Sustainability is
estimated from Eq. [3-12].
Discussion
The Farming System Simulator presented in this chapter possesses the capabilities
needed to simulate and analyze the sustainability of a range of farming systems. Linkage


121
and converting it to daily totals, rainfall recorded manually from the plastic rain gauge was
used for simulations. Manually recorded rainfall agreed closely with data integrated from
the pulse counter for a period for which both were available.
Maximum temperatures recorded by the LI-COR station increased about 3C
relative to those measured by the portable thermometer. The discrepancy was apparently
due to the location of the sensor of the portable thermometer under the roof of the
farmers house, in an area shaded by trees. The LI-COR sensors were in a standard
weather instrument enclosure located in an open field. Because of the discrepancy,
maximum temperature data recorded before the LI-COR station was brought on-line
were increased by 3 C.
Because long-term weather records were not available near the Domingo farm,
weather generator coefficients were estimated from records of surrounding stations using
inverse-squared-distance interpolation. The estimated value of a given weather parameter
at a particular target location (*) in month m was calculated as
Am = E "i-W [5-1]
z = l
where is the value ofy at station /' for month m, and n is the number of stations used
for interpolation. Whenj represented a mean temperature (C), Eq. [5-1] was adjusted
for an adiabatic lapse rate of 6C per 1000 m increase in elevation:
Am = w¡(Am + 0.006 (e.-e,)),
i \
[5-2]


5-20 Effect of applied N on bean nodule growth and cumulative N2 fixed
observed and simulated with the original and modified versions of
CROPGRO, Kuiaha, Hawaii, 1993 148
5-21 Plant and nodule mass at 35 days, and grain yield of container-grown
beans in response to applied N 150
5-22 Simulated 60 year sequences of maize grain yields at different levels of
applied N, Domingo farm, Cauca, Colombia 151
5-23 Simulated grain yields of maize and bean in response to soil loss, Domingo
farm, Cauca, Colombia 153
5-24 Simulated root distributions of October- and March-planted bean in response
to 0 cm, 20 cm, and 50 cm of soil loss, Domingo farm, Cauca, Colombia .... 155
5-25 Relative root distribution factors, F and SRGF after 0 cm, 20 cm and 50 cm
soil loss 157
5-26 Simulated grain yields of October- and March-planted bean in response to soil
loss, Domingo farm, Cauca, Colombia 159
5-27 Pseudocode representation of an algorithm for resolving resource conflicts
among crop enterprises 160
5-28 Pseudocode representation of an algorithm for simulating erosion and crop
growth on a complex hillslope 164
6-1 Historical crop wholesale prices, Cauca, Colombia 172
6-2 Cropping pattern included in farm scenario 178
6-3 Influence of type of farmer participation in on-farm trials on bean response
to ground and partially acidulated rock phosphate, chicken manure, and
10-30-10 179
6-4 Box plot of liquid assets, base scenario 187
6-5 Cumulative distribution of liquid assets after 1, 3, 6, 9, 12 and 15 years, base
scenario 188
6-6 Sustainability and hazard time plots of the base scenario 189
xv


203
sustainability. The results have important practical implications to the farmer, researchers
and policy makers who are concerned with improving the sustainability of the Domingo
farm and similar farming systems in the Cauca region of Colombia.
Practical implications
Of the factors implicated as determinants of sustainability, cropping system and N
management are the ones most easily controlled by the farmer. By accepting the
Colombian Coffee Federations offer, the farmer eliminated the option of producing coffee
in the near future. The most promising alternatives appear to be intensifying production
with a two-year maize-bean-cassava rotation and incorporating tomato or another high
valued vegetable crop into the rotation. The study of sources of risk demonstrated the
value of staggering the phases of a multiple-year rotation among fields of similar size.
Although I did not attempt to find an optimum fertilizer management scheme in this study,
sustainability was higher with the N fertilizer rates used in the base scenario than when N
was applied at either 50% or 200% of those rates. Increasing the proportion of income
that the farm family saves would improve farm sustainability; reducing subsistence
expenditures ten percent increased simulated sustainability 53% (Table 6-10). However,
there may be little latitude for reducing household spending given the farmers goals for
his childrens education.
It would be risky for the farmer on the basis of simulation results alone to intensify
cropping, start producing tomato, or invest in an irrigation system. Intensifying the
cropping system might have unforeseeable effects on pest populations or soil fertility.


24
Characterizing sustainability by time trends is appealing because of its simplicity.
The slope of the estimated trend line provides a quantitative index with an intuitive
interpretation as a rate of system deterioration or enhancement. Trends represent an
aggregate response to several determinants of sustainability, eliminating the need to devise
and defend aggregation schemes.
The assumption needed to infer sustainability from trends--that future rates of
system degradation can be approximated by past ratesis often difficult to defend.
Unsustainability can express itself either as a gradual change or as an abrupt collapse
(Conway, 1985; Trenbath et al., 1990). Furthermore, much of the concern about
sustainability comes from recognition that agriculture is being impacted by unprecedented
changes in population pressure, resource demands, market structures, and technology.
Another weakness is the manner in which time-trend approaches interpret temporal
variability. Variability tends to hinder sustainability by driving subsistence farmers to
desperation, leading to environmental degradation that may not recover during normal or
good periods (Mellor, 1988). Price and yield variability have also been shown to increase
the probability of farm failure in the U.S. (Grant et ah, 1984; Perry et al., 1986).
However, when characterization is based on time trends, either variability is ignored or it
implicitly enhances sustainability by reducing the probability of identifying a significant
negative trend.
A final criticism is that applications of time trends to sustainability have examined
levels of system performance without considering the levels of needs and goals of the
individuals or segments of society who decide on the fate of those systems.


Copyright 1996
by
James William Hansen


88
operation schedule generates an operation, it is matched with a set of resource bundles
from the set of operation requirements. A resource bundle is a set of timed resources (or
their descendants) that must be used in a fixed proportion. On a mechanized farm, the
bundle will typically include a power unit, an implement and an operator. Operations
possess a mechanism for delaying their execution if resources are not available for their
completion on a particular day. When an operation executes, it first determines the
fraction that can be completed in the current day with the availability of the required set of
timed resources. It then adjusts any material resource requirement for the fraction that
can be completed. If the material resource is constraining, the fraction of the operation
that can be completed is again adjusted. The Execute method then uses the required
timed resources. If the operation could not be completed and the current date is within a
specified window, the Execute method returns a result of delayed. Otherwise, the result
is succeeded. If the operation is delayed, the event queue spawns another operation with
adjusted resource requirements that reflect the fraction of the operation remaining.
Resource accounting
Much of the flexibility of FSS comes from its polymorphic set of resource classes
(Table 4-2). Resource accounting is accomplished by three methodsUSE, Sell and
Updatethat are common to all resource classes. The behavior of these methods varies
depending on the resource class. These methods are presented in detail below. Familiarity
with these methods is essential to understanding how FSS accounts for resource use.


54
suggests that the time interval should be long enough to allow detection of important
threats to sustainability. For example, three or four years would reveal little about the
impact of soil erosion on sustainability. On the other hand, a study exceeding a century
would not be relevant to the livelihood goals of an individual farm. Lynam and Herdt
(1989) suggested that five to 20 years is a relevant time-frame for analysis of farming
system sustainability. Realism of assumptions about economic, policy, and technological
inputs to the farming system becomes increasingly difficult to defend past about 10 to 15
years.
Assumptions about inputs
The higher-level systems that comprise a farming systems environment exert
control through inputs and through constraints to farmer decisions or farm outputs.
Analysis of farm sustainability requires assumptions about future behavior of inputs and
control mechanisms that are conceptually external to the farming system. The issues to
consider include (a) whether systematic trends or cycles are expected, (b) whether
variability is sufficient to warrant stochastic sampling, and (c) whether a feedback
mechanism exists which allows an input to respond to farm outputs. Although some farm
processes may be inherently stochastic, inputs of weather and prices are usually the major
sources of risk in a farming system. Catastrophic events such as disease epidemics, storms
or wars also represent important stochastic inputs to some farming systems.


94
Aleorithm
If a loan payment is due then
begin
Purnose
Calculate A time
The fraction of a year since the last Update.
Request Loans to execute GetPayment
Payment = -GetPayment.
Repeat
Find the next linked resource in FixedCosts.
GetPayment returns the amount due on all loans.
Request linked resource to
execute USE(Payment).
Attempt to obtain Payment from the linked resource.
until Payment is reduced to 0.
... until Payment is met or all FixedCosts are tried.
Request Loans to execute Update
Update principal and number of payments remaining
on each loan, and dispose of repaid loans.
Set Supply = Initial supply Debt.
end.
Calculate new Supply of available credit.
Figure 4-11. Pseudocode representation of the Update method of the credit resource
class.
Resource linkages. Resources interact and exchange with each other through
resource linkages. A set of linkages comprises a cost list. The Use method of a cost list
attempts to obtain or dispose of a specified amount by converting the amount to units of
the linked resource based on a price, executing the Use method of the linked resource,
reducing the amount by what the linked resource could use, then converting the amount
back to its original units (Fig. 4-12). The process repeats for each linkage or until the foil
amount has been exchanged.
A Forrester representation provides a clearer picture of how a resource linkage
operates. Each resource accesses an external source or sink. In the example in Fig. 4-13,
if the amount of resource a used is greater than the amount available, it will attempt to
draw on its source to replenish the resulting deficit. The amount that resource a can
access from its source is controlled by linkages to resources b, c and d, and is limited by


4-8
Pseudocode representation of the Use method of the credit resource class .... 92
4-9 Pseudocode representation of the Update method of the consumable
resource class 93
4-10 Pseudocode representation of the UPDATE method of the capital resource
class 93
4-11 Pseudocode representation of the UPDATE method of the credit resource
class 94
4-12 Pseudocode representation of the Use method of the cost list class 95
4-13 Forrester representation of variable cost linkages between a consumable
resource and three linked consumable resources 96
4-14 Flow of information from the generation of a fertilizer application operation
by a crop model to its effect on the operating fund 97
4-15 Example of a resource box plot generated by FSS .99
4-16 Example of a final resource distribution plot generated by FSS 100
4-17 Example of a sustainability time plot generated by FSS 102
5-1 Location map of the study area, Cauca, Colombia 107
5-2 Land use map of the Domingo farm, Cauca, Colombia ................. 108
5-3 Monthly climate statistics; mean daily solar radiation, mean daily maximum
and minimum temperature, and total rainfall, Domingo farm, Cauca,
Colombia 110
5-4 Mean monthly rainfall totals at weather station used to estimate monthly
weather statistics and WGEN parameters for the Domingo farm by spatial
interpolation 111
5-5 Phosphorus sorption isotherms for four soils with different mineralogy ...... 113
5-6 Decomposition of soil organic C in three allophanic and seven
non-allophanic soils 114
xiii


119
good. Unfortunately, CropSim CASSAVA does not include a plant N submodel; it
simulates the soil N balance but not plant response.
Although the crop models used in this study have been tested in multiple
environments, the demands inherent in characterizing farm sustainability and the
peculiarities of the Andean environment selected for this study necessitate additional
testing.
Model requirements. The application of crop simulation to characterizing farm
sustainability places several demands on the crop models. First, the models must provide
reasonable predictions of development and yields. Development determines the feasibility
of cropping patterns and the timing of resource use. Livelihood of a full-time farmer is
ultimately driven by the productivity of the farms crop and livestock populations.
Second, the models should respond realistically to weather variability.
Furthermore, the long-term, stochastic simulation that is needed to characterize farm
sustainability (Chapter 3) requires the ability to generate many sequences of weather data
with statistical properties comparable to those of the historical sequence. The stochastic
weather generator used to generate weather input data should be able to reproduce the
distribution of simulated crop yields obtained from a long sequence of historical data.
Third, the models should predict realistic response to long-term soil nutrient
dynamics and management. Phosphorus, N and erosion are probably the most critical soil-
based determinants of productivity and sustainability on these steep Andisols.
Water balance and N and C dynamics are the only soil processes that the DSSAT3
crop models can simulate. Although work is underway to incorporate P dynamics into the


this purpose can simulate a replicated farm scenario with stochastic inputs of weather and
prices. The farm simulator simulates the balance of farm resources and accounts for
ecological determinants of crop production by calling external crop models. Several crop
simulation models were evaluated for compatibility with a simulation study of the
sustainability of a hillside farm in the Cauca region of Colombia. The crop models were
useful for capturing response to weather variability and N management, but did not
account for important yield determinants, including P deficiency and nematode damage.
Simple modifications improved simulated response to applied N and sensitivity to soil loss.
A study of a hillside farm in Colombia showed the practical value of using simulation to
characterize sustainability. Results identified cropping system, area under cultivation,
consumption requirements, crop prices, and soil erosion as important determinants of
sustainability. The study showed that price variability contributes more than weather
variability to farm risk in this location, and that spatial diversification reduces risk and
improves sustainability. Results suggest that the farmer can enhance farm sustainability by
diversifying and intensifying crop production. Researchers can contribute to sustainability
by identifying promising high-value crops, by testing the proposed cropping systems, and
by quantifying severity and impacts of erosion. Policy makers can address sustainability
constraints caused by price volatility and lack of affordable credit.
xix


255
Hildebrand, P.E. 1990. Agronomys role in sustainable agriculture: integrated farming
systems. J. Prod. Agrie. 3:285-288.
Hill, S.B. and R.J. MacRae. 1988. Developing sustainable agriculture education in
Canada. Agrie. Hum. Values 5(4):92-95.
Hipel, K.W. and A.I. McLeod. 1994. Time Series Modelling of Water Resources and
Environmental Systems. Elsevier, Amsterdam.
Hoogenboom, G., J.W. Jones and K.J. Boote. 1990. Nitrogen fixation, uptake, and
remobilization in legumes: a modeling approach, In International Benchmark Sites
Network for Agrotechnology Transfer (IBSNAT) Project. Proceedings of the
IBSNAT Symposium: Decision Support Systems for Agrotechnology Transfer,
Las Vegas, NV. 16-18 Oct., 1989. Part II: Posters. Dept, of Agronomy and Soil
Science, College of Tropical Agriculture and Human Resources, Univ. Hawaii,
Honolulu, Hawaii.
Hoogenboom, G., J.W. Jones, P.W. Wilkens, W.D. Batchelor, W.T. Bowen, L.A. Hunt,
N.B. Pickering, U. Singh, D.C. Godwin, B. Baer, K.J. Boote, J.T. Ritchie, and
J.W. White. 1994. Crop models, p. 95-244 In G.Y. Tsuji, G. Uehara, and S. Balas
(ed.) DSSAT Version 3, Vol. 2. International Benchmark Sites Network for
Agrotechnology Transfer, Univ. of Hawaii, Honolulu, Hawaii.
Hudgens, R.E. 1978. Adapting agronomic technology for small farm bean production in
highland Colombia. Ph.D. diss., University of Florida, Gainesville, Florida.
Hunt, L.A., J.W., Jones, P.K. Thornton, G. Hoogenboom, D.T. Imamura, G.Y. Tsuji and
U. Singh. 1994. Accessing data, models & application programs, p. 21-110. In
G.Y. Tsuji, G. Uehara, and S. Balas (ed.) DSSAT Version 1, Vol. 2. International
Benchmark Sites Network for Agrotechnology Transfer, Univ. of Hawaii,
Honolulu, Hawaii.
IBSNATInternational Benchmark Sites Network for Agrotechnology Transfer. 1990.
Technical Report 2: Field and Laboratory Methods for the Collection of the
IBSNAT Minimum Data Set, 1st ed. Department of Agronomy and Soil Science,
College of Tropical Agriculture and Human Resources, University of Hawaii,
Honolulu, Hawaii, USA.
Ikerd, J.E. 1991. Applying LISA concepts on southern farms. Southern J. Agrie. Econ.
23(l):43-52.
Janssen, W. and N. Ruiz do Londoo. 1994. Modernization of a peasant crop in
Colombia: evidence and implications. Agricultural Economics 10:13-25.


where em and el are elevation (m) at the target and the /th location, and 0.006 is the
adiabatic lapse rate (C m'1). The weighting factor, w¡, for station i was calculated as
122
Wi
1 /<
£(l/<)
[5-3]
where is the distance from station j to the interpolated location. The distance, dti,
between any two locations, i and j, can be calculated from their longitude (long) and
latitude (lat) by
/¡j = 6366.2 arccos(cos(4) cos(2?)),
A = cos(0.017453 maxd/otfjl, \lat.^)) 0.017453\long{ longj|,
B = 0.017453 l/a/j-tajl
(P.G. Jones, 1994, personal communication). Figure 5-8 shows the weather stations used
to estimate monthly weather generator parameters (Table 5-3).
The WeatherMan software package (Hansen et al., 1994) was used to estimate the
missing solar radiation data, using the same adaptation of WGEN that is in the crop
models and the spatially interpolated parameters (Table 5-3). WGEN preserves the
dependence of solar radiation on observed rainfall by sampling from different distributions
on wet and dry days.


149
growth rate (NODRGM). Criteria for calibrating these parameters were (a) to preserve
predicted biomass and grain yields in the absence of applied N, (b) to eliminate the
predicted drop in grain yield with the first increment of applied N, and (c) to reproduce the
suppression of nodule growth observed in the experimental study by Tewari (1995).
Based on these criteria, I obtained values of 0.003 for DWNODI, 0.17 for NODRGM, and
0.04 for FRCNOD. The modification reduced sensitivity of nodule growth to small
amounts of applied N and eliminated the delay in nodule growth (Fig. 5-19b). It also
improved predictions of nodule growth and N2 fixation observed in Tewaris study,
although CROPGRO still tended to under predict both (Fig. 5-20b and d).
Figure 5-17 shows simulated bean response to applied N with the original model
and with the calibrated modifications. Simulated canopy biomass responded to increasing
levels of applied N as expected, increasing at a decreasing rate to a maximum, then
maintaining a plateau. Simulated soybean response to applied N was also more consistent
with expectations with the modified model (Fig 5-18).
The predicted suppression of grain yield of October planted bean with intermediate
levels of applied N simulated with the modified version of CROPGRO (Fig. 5-17) may be
reasonable. This yield reduction can be attributed to suppression of early nodule growth
and a resulting loss of capacity to fix N for reproductive growth. Larger applications of N
fertilizer provide sufficient N2 for reproductive growth. In a pot experiment with three
bean cultivars and three strains of Rhizobium inoculum, plant mass at 35 days increased
(Fig. 5-2la) and nodule mass decreased (Fig 5-2lc) in response to increasing amounts of


219
Table B-4. Format of the RESOURCES section of the scenario file.
Variable
Header
Formal
Resource class:
Cl
2 C 1
C = consumable
T = timed
K = capital
S = seasonal
D = credit
L = activity-linked credit
A = aggregate
B = debt
Unit of resource
Units
1 C 8
Name of resource
Name
1 C 15
Name of price that defines value
Val
1 C 15
Multiplier for price which defines value
ValMult
0 R 8
Initial supply (consumable or credit) or availability, hr day'1
{timed, capital, machine or seasonal)
Supply
0 R 8
Minimum supply {consumable)
Minimum
0 R 8
Maximum supply {consumable)
Maximum
0 R 8
Interest, decay, or depreciation rate {consumable, credit,
activity-linked credit, capital or machine)
IntRate
0 R 8
Initial value {capital or machine)
Value
0 R 8
Beginning of available period, day of year {seasonal)
Stt
1 13
End of available period, day of year {seasonal)
Stp
1 I 3
Initial debt {credit, activity-linked credit)
Debt
0 R 8
Term of repayment, months {credit, activity-linked credit)
Trm
113
Payment interval, months {credit, activity-linked credit)
Frq
1 13
T See footnote, Table B-l.


106
sustainability (weather variability, soil nutrient dynamics and management, and soil
erosion), and (d) to discuss issues of compatibility between the crop models and the
broader modeling framework needed to address sustainability at a farm level.
A Colombian hillside environment
A farm (the Domingo farm) located in the Cabuyal River catchment within the
Cauca region of southwestern Colombia (2 47' N, 76 31' W) (Fig. 5-1) was selected as a
basis for a simulation study of determinants of sustainability (Chapter 6). The farm
occupies 5.16 ha ranging in elevation from 1660 m at the homesite to 1580 m at the
Cabuyal River (Fig. 5-2). The farm landscape is quite steep; about a third of the area has
slopes in excess of 24% (Table 5-1). Until July 1993, the major farm enterprise was
coffee production. The land that was in coffee production (2.38 ha) has since been used
for bean, both solecropped and intercropped with maize. CIAT personnel have managed
bean, maize and cassava trials on a small portion of the farm since October 1993. These
trials provide a basis for testing simulation models of these three crops.
Table 5-1. Area in each slope class, Domingo farm, Cauca, Colombia.
Slope -- Cultivated area -- - Grass fallow - Total farm
class area (ha) percent area (ha) percent area (ha) percent
0 12%
0.17
7%
0.16
12%
0.89
17%
12 24%
1.66
70%
0.48
36%
2.61
51%
24 36%
0.50
21%
0.45
34%
1.22
24%
> 36%
0.05
2%
0.24
18%
0.43
8%


77
Table 4-2, Description of resource classes.
Class
Parent
Description
resource
none
An abstract resource used as a foundation for all other
resource classes.
simple
resource
An abstract resource built on the resource class used for
all resource classes except for aggregate and debt.
consumable
simple
A resource whose supply is reduced with use, that may
have minimum and maximum storage constraints, and
whose supply may change with decay or interest (eg.,
money, fertilizer).
timed
simple
A resource which is not consumed and whose use is
constrained by hours of availability per day (eg., labor).
seasonal
timed
A timed resource that is available only during a certain
period each year.
capital
timed
A timed resource with an initial value that may. depreciate
with time (eg., a building).
machine
capital
A capital resource that uses all variable costs each time it
is used. This permits proper accounting of simultaneous
use of fuel, maintenance, labor, and other costs of
machinery use.
credit
simple
A source of credit. A credit resource keeps track of
current debt, accrued interest, and payment schedules.
Using credit involves borrowing new loans or adding to
any loan that was borrowed during the current month.
activity-linked
credit
credit
A credit resource whose availability is based on the area
currently in a particular crop activity.
aggregate
resource
A resource that represents the sum (eg., liquid assets) or
ratio (eg., leverage ratio) of values of other resources for
accounting or output. Aggregate resources are used only
for consumption decisions, file output or analyses.
debt
aggregate
An aggregate resource that is used to keep track of
indebtedness associated with credit resources.


57
innovation and trends in prices can affect trends in farm state variables. Variability is
influenced by weather patterns, price volatility and the occurrence of catastrophic events.
Credit availability, market access, and the ability to store agricultural products have
positive effects on autocorrelation of farm wealth. Finally, a households tolerance to
difficulty, alternative sources of livelihood and lenders policies can influence goal
thresholds.
Adaptive management generally serves to improve sustainability. Farmers employ
a range of management strategies, such as selling capital assets, reducing input use,
working off-farm, or shifting from cash to subsistence crops, to reduce the risk of failure
during difficult times. Although the proposed framework for characterizing sustainability
can account for adaptive strategies, the farmer decision process may be more difficult to
simulate than biological or economic processes. Simulating a fixed management strategy
may greatly overestimate the probability of farm failure.
Much of the concern about sustainability of agricultural production systems relates
to externalities: non-target outputs with costs (or benefits) which are not bom by an
individual farm but by society. A hierarchical perspective suggests that externalities
should have no direct impact on farm sustainability; the impact of an externality emerges
at a higher (eg., regional) system level. However, a feedback mechanism that either
charges the farming system for the externality or constrains the practice that produces it
can easily be incorporated into a simulation analysis. Such a deviation from a strict
hierarchical approach might be justified by assuming that society will attempt to correct
the perceived injustice of a negative externality.


130
nematode populations with each successive bean crop. Mullin et al. (1991) identified
root-knot nematodes (Meloidogyne incognita and M. hapla) in soils from bean fields in
Cauca. They measured nematode-induced bean yield losses of 45-63% for cv. Calima and
26-32% for cv. PVA in an experiment at the CIAT headquarters near Palmira.
Table 5-5. Observed and predicted bean yields, Domingo and Trujillo farms, Cauca,
Colombia.
Farm
Planting
date
Plants
per ha
Canopy mass --
(kg ha1)
obs. pred.
Grain yield
(kg ha1)
obs. pred.
Harvest index
obs. pred.
Domingo1
Oct 14,1993
176,389
5
4,964
2,600
2,913
0.587
Trujillo1
Oct 14,1993
117,824
5,057
5,725
3,315
2,679
0.656
0.468
Domingo
Mar 30,1994
89,286
4,494
990
2,100
0.467
Trujillo
Mar 28, 1994
137,269
4,596
5,613
2,644
2,468
0.575
0.440
Domingo
Oct 10, 1994
104,630
2,431
4,704
1,208
2,712
0.497
0.577
Trujillo
Oct 7, 1994
123,889
3,854
5,547
1,978
2,190
0.513
0.395
Domingo
Apr 1,1995
98,920
4,189
627
2,252
0.538
Trujillo
Mar 31,1995
138,819
3,608
5,247
1,727
1,956
0.479
0.373
1 Applied 250 kg 10-30-10 ha'1 each season.
* Applied 1500 kg 10-30-10 ha'1 each season.
§ Canopy dry mass was not recorded.
Table 5-6. Observed and predicted timing of phenological events (days
after planting) for bean and maize, Domingo farm, Cauca, Colombia,
planted March 30, 1994,
Crop
-- Emergence --
-- Flowering --
Maturity
obs.
pred.
obs.
pred.
obs.
pred.
bean
6
6
42
36
80
76
maize
t
6
69
66
133
124
f Maize emergence date was not recorded.


5
Table 2-1. Interpretations of agricultural sustainability.
Sustainability as an ideology:
.. a philosophy and system of farming. It has its roots in a set of values that reflect a state of empowerment, of awareness
of ecological and social realities, and of ones ability to take effective action. (MacRae el at., 1990, p. 156)
.. an approach or a philosophy... that integrates land stewardship with agriculture. Land stewardship is the philosophy
that land is managed with respect for use by future generations. (Neher, 1992, p. 54)
... a philosophy based on human goals and on understanding the long-term impact of our activities on the environment
and on other species. Use of this philosophy guides our application of prior experience and the latest scientific advances to
create integrated, resource-conserving, equitable farming systems. (Francis & Youngberg, 1990, p. 8)
... farming in the image of Nature and predicated on the spiritual and practical notions and ethical dimensions of
responsible stewardship and sustainable production of wholesome food. (Bidwell, 1986,p.317)
Sustainability as a set of strategies:
.. a management strategy which helps the producers to choose hybrids and varieties, a soil fertility package, a pest
management approach, a tillage system, and a crop rotation to reduce costs of purchased inputs, minimize the impact of the
system on the immediate and the off-farm environment, and provide a sustained level of production and profit from
farming. (Francis, Sander & Martin, 1987, p. 12)
... a loosely defined term for a range of strategies to cope with several agriculturally related problems causing increased
concern in the U.S. and around the world. (Lockeretz, 1988, p. 174)
Farming systems are sustainable if they minimize the use of external inputs and maximize the use of internal inputs already
existing on the farm. (Carter, 1989, p. 16)
.. (a) the development of technology and practices that maintain and/or enhance the quality of land and water resources,
and (b) the improvements in plants and animals and the advances in production practices that will facilitate the substitution
of biological technology for chemical technology. (Ruttan, 1988, p. 129)
Sustainability as the ability to fulfill a set ofgoals:
A sustainable agriculture is one that, over the long term, enhances environmental quality and the resource base on which
agriculture depends, provides for basic human food and fiber needs, is economically viable, and enhances the quality of life
for fanners and society as a whole. (American Society of Agronomy, 1989, p. 15):
... agricultural systems that are environmentally sound, profitable, and productive and that maintain the social fabric of the
rural community. (Keeney, 1989, p. 102)
... an agrifood sector that over the long term can simultaneously (1) maintain or enhance environmental quality, (2)
provide adequate economic and social rewards to all individuals and firms in the production system, and (3) produce a
sufficient and accessible food supply. (Brklacieh, Bryant & Smit, 1991, p. 10)
... an agriculture that can evolve indefinitely toward greater human utility, greater efficiency of resource use, and a
balance with the environment that is favorable both to humans and to most other species. (Harwood, 1990, p. 4)
Sustainability as the ability to continue:
A system is sustainable over a defined period if outputs do not decrease when inputs are not increased. (Monteith, 1990,
P- 91)
... the ability of a system to maintain productivity in spite of a major disturbance, such as is caused by intensive stress or a
large perturbation. (Conway, 1985, p. 35)
.. the maintenance of the net benefits agriculture provides to society for present and future generations. (Gray, 1991, p.
628)
Agriculture is sustainable when it remains the dominant land use over time and the resource base can continually support
production at levels needed for profitability (cash economy) or survival (subsistence economy). (Hamblin, 1992, p. 90)


181
Table 6-7. Description of farm scenarios
Scenario Description
base scenario
A three-year maize-bean-bean-cassava rotation (Fig 6-2) used as a basis for
comparing other scenarios in sensitivity analysis.
maize-becm
A one-year double-crop (Fig 6-2).
maize-tomato-bean-cassava
A three-year rotation, with irrigation for tomato and bean (Fig 6-2).
maize monoculture
A one-year maize-maize double-crop (Fig 6-2).
cassava monoculture
A three-year double crop (Fig 6-2).
maize-bean-cassava
A two-year rotation (Fig*6-2).
coffee @ 1.75 Mg/ha
Coffee monoculture with an annual yield of 1.75 Mg ha'1.
coffee @ 2.00 Mg/ha
Coffee monoculture with an annual yield of 2.00 Mg ha'1.
coffee @ 2.25 Mg/ha
Coffee monoculture with an annual yield of 2.25 Mg ha'1.
coffee @ 2.50Mg/ha
Coffee monoculture with an annual yield of 2.50 Mg ha'1.
less Nfertilizer
10-30-10 applied at 50% of base scenario (25 kg N ha'1 split application to
maize and 12.5 kg N ha'1 to bean).
more Nfertilizer
10-30-10 applied at 200% of base scenario (100 kg N ha'1 split application
to maize and 50 kg N ha'1 to bean).
erosion @ 0 Mg ha'1 yf1
erosion @ 25 Mg ha'1 yr'1
erosion @ 50 Mg ha'1 yr1
erosion @ 100 Mg ha'1ytrl
erosion @ 150 Mg ha'1 yr1
Constant annual soil loss of 0 Mg ha'1 yr'1
Constant annual soil loss of 25 Mg ha'1 yr'1
Constant annual soil loss of 50 Mg ha"1 yr'1
Constant annual soil loss of 100 Mg ha'1 yr'1
Constant annual soil loss of 150 Mg ha'1 yr"1
higher commodity prices
Ten percent higher production commodity prices.
lower input prices
Twenty percent lower material input prices.
lower labor prices
Twenty percent lower prices for labor and contracted plowing.
less subsistence spending
Ten percent lower subsistence requirement for money, maize and bean.
less discretionary spending
more initialfunds
Twenty percent lower discretionary spending.
Twenty percent higher initial operating fund.
more cultivated land
Additional land (0.23 ha, or 10% of currently cultivated area) brought into
cultivation.
credit @ 19%
Col.S 2,000,000 available at 19% interest, 24 month repayment schedule.
credit @ 9.5%
Col.$ 2,000,000 available at 9.5% interest, 24 month repayment schedule.
more credit
Col.$ 3,000,000 available at 19% interest, 24 month repayment schedule.


62
tenure arrangements, and irrigating the soybean crop would be the most important
strategies for enhancing sustainability of the model farm.
Discussion
By defining sustainability as the ability of a dynamic, stochastic, purposeful system
to continue into the future, I arrived at a useful, quantitative expression of sustainability.
Figure 3-4. Simulated sustainability of a Texas rice farm under four scenarios: a
three year soybean-soybean-rice rotation with a 1/7 {SSR 1/7) and a 1/2 (SSR 1/2)
share arrangement, and a two year soybean-rice rotation with the same two share
arrangements {SR 1/7 and SR 1/2). Data from Perry el al., 1986.


Sustainability
199
Table 6-11. Predicted nine-year sustainability (3(9) SE$), McNemar test
statistic (GPadj) for difference from the base scenario, and standard
Scenario
$9) SE<¡
Gp,adj
SD
base
0.79 0.040
t
699,504
no weather risk
0.85 0.035
1.2 n.s.
659,063
no price risk
1.00 0.000
t
181,131
not diversified
0.10 0.030
106.2 **
§
T Undefined.
* Comparison does not apply to the base scenario.
5 Not determined because the distribution was truncated by failures.
Figure 6-14. Sustainability time plot of base and credit scenarios.


29
number of examples in which disturbances that threaten sustainability at one spatial and
temporal scale could be seen as natural cycles at broader scales. Both constraints to
sustainability and factors that can be managed for its enhancement depend on the level of
the system (Spencer and Swift, 1992). The objectivity that results from a system-oriented
approach is essential for guiding change, but may work against motivating change because
it may call prescribed approaches into question.
Third, an approach to characterizing sustainability should be quantitative.
Although MacRae et al. (1989) cited quantification as a barrier to sustainability, others see
it as a prerequisite to using sustainability as a criterion for evaluating and improving
agricultural systems (Monteith, 1990; Harrington, 1992). Sustainability is often treated as
a discrete property; A farm is either sustainable or its not sustainable. Simply by
definition, you cannot create a system that is half sustainable (Rodale, 1990, p. 273).
However, comparisons among agricultural systems or alternative approaches are possible
only when sustainability is treated as a continuous quantity.
Fourth, since sustainability deals with future changes, its characterization must be
predictive of the future rather than merely descriptive of the past or present (Harrington,
1992). Sustainability has little meaning after the fact. The deterministic view that . a
farm will either last for a very long period, or it wont (Rodale, 1990, p. 273) does not
take into account the uncertainty of predictions resulting from the inherent variability of
the farming systems environment. A stochastic approach, the fifth element, recognizes
variability as a determinant of sustainability and appropriately expresses predictions in
terms of probabilities.


46
of the time-series model described above. The general pattern of the sustainability time
function remains the same whether hazard is modified by a change in the expected value,
variance or autocorrelation (Fig. 3-3c-h).
Table 3-2. Sensitivity of simulated sustainability to system properties. 61(120)=0.567 for
the base scenario.
Para-
Base
Increased
Decreased
Property
meter
value
value
S{T)
value
%T)
mean
a
10.0
11.0
0.783
9.0
0.307
trend
a
P
10.0
0.0
9.0
+0.0167
0.566
11.0
-0.0167
0.517
variability
5.0
5.5
0.401
4.5
0.774
autocorrelation
4>i
0.8
0.88
0.924
0.72
0.235
goal threshold
*0
3.0
3.3
0.491
2.7
0.641
duration
T
120
132
0.535
108
0.604
Diagnosing Constraints to Sustainability
The potential value of the concept of sustainability lies in its ability to focus
research and intervention by identifying and ranking its constraints. Diagnosing
constraints entails a process of hypothesis formulation and testing using simulation of the
system model. Hypotheses should identify the current (or expected) value of a suspected
constraint, and a specific change that would relax the constraint. Sensitivity analysis then
provides the experimental tool for testing and ranking hypothesized constraints.


264
York, E.T., Jr. 1988. Improving sustainability with agricultural research. Environment
30(9): 18-20,36-40.
York, E.T., Jr. 1991. Agricultural sustainability and its implications to the horticulture
profession and the ability to meet global food needs. HortScience 26:1252-1256.
Zandstra, H. 1994. Sustainability and productivity growth: Issues, objectives, and
knowledge needs -- guidelines for working groups. In Reconciling sustainability
with productivity growth. Report of a workshop, Gainesville, Florida. May 1993.
Univ. of Florida and Cornell Univ.
Zunino, H., F. Borie, S. Aguilera, J.P. Martin, and K. Haider. 1982. Decomposition of
14C-labeled glucose, plant and microbial products and phenols in volcanic ash-
derived soil of Chile. Soil Biol. Biochem. 14:37-43.


10
such institutions threaten to dilute the concept of sustainable agriculture by co-opting it
while ignoring its more important and radical aspects.
Alternative values. Sustainable agriculture has been described as an umbrella term
encompassing several ideological approaches to agriculture (Gips, 1988) including organic
farming, biological agriculture, alternative agriculture, ecological agriculture, low-input
agriculture, biodynamic agriculture, regenerative agriculture, permaculture, and
agroecology (Carter, 1989; MacRae etal., 1989; Bidwell, 1986; OConnell, 1992;
Kirschenmann, 1991; Dahlberg, 1991).
Beus and Dunlap (1990) listed decentralization, independence, community,
harmony with nature, diversity, and restraint as key values of alternative agriculture.
Social values such as equity, the value of traditional agricultural systems, self-sufficiency,
preservation of agrarian culture, and preference for small, owner-operated farms have
been incorporated into definitions of sustainability (Weil, 1990; Keeney, 1989; Bidwell,
1986; Francis and Youngberg, 1990). The concept of equity is extended to include future
generations (Batie, 1989; Norgaard, 1991). Environmental values associated with
sustainability include mimicry of nature and an ecocentric ethic. Hauptli et al. (1990)
described mimicry of nature: . . sustainable agriculture attempts to mimic the key
characteristics of a natural ecosystem ... (p. 143). The ecocentric positionvaluing
ecosystems or species without regard to their impact on human welfareis illustrated by
Douglass (1984) who stated ecology-minded people . . define agricultural sustainability
in biophysical terms, and to allow its measurement to determine desirable population
levels (p. 5).


72
Inputs
Jones et al. (1995) cited a lack of emphasis on data standards as a barrier to
reusing farm models for different farms or applications. FSS input data structures and file
formats were designed to be flexible enough to be able to represent a range of farm types
and to accommodate possible future extensions of FSS. The general organization of data
and scheme for its use are being proposed as a starting point for developing data standards
for enterprise and farm-level systems analyses (Hansen etal., 1995). Detailed
presentation of input data requirements and formats in Appendix B supplements the
discussion in this section.
A farm scenario is the operation of a farm through a period of time with a given
set of initial conditions and rules for making decisions and scheduling activities. A
scenario may be replicated with stochastic sampling of input variables such as weather and
prices, but with the same initial conditions and decision rules for each replicate. At least
two filesa scenario file and a price fileare required to simulate a farm scenario. The
scenario file contains farm-level information and identifies the other input files. The price
file contains the parameters for models for generating sequences of prices. A scenario that
calls external IBSNAT crop simulation models also requires a minimum data set (MDS)
for each crop consisting of a crop management file, soil file, weather or climate file, and
genetic coefficient file (Tsuji et ah, 1994).


182
field preparation and harvest. Because beans and tomato are susceptible to nematodes and
soil-borne diseases (Chapter 5), no more than two bean and/or tomato crops were
simulated consecutively, and each ban or tomato crop required rotation with at-least one
season of maize, cassava or fallow. To maximize spatial diversity, the phases of each
multiple-year rotation were distributed among fields of equal size. Four additional
scenarios represented coffee monoculture with different assumed yields. Since no
process-level simulation model was available for coffee, its production was simulated by a
fixed schedule of field operations and fixed harvest amounts.
There is little a priori basis for hypothesizing that one particular crop contributes
more to farm sustainability than the others. However, the annual cropping scenarios were
used to test the hypothesis that diversification of crop enterprises leads to a more
sustainable farming system than any monoculture. They were also used to examine the
role of high value vegetables, represented by irrigated tomato.
Soil management. Soil processes can impact farm sustainability by modifying crop
yields. Soil management can also impose a cost on the farming system. The capabilities
of the crop models (Chapter 5) limit the soil-related determinants of sustainability that we
can test. This study includes scenarios designed to test the hypotheses that (a)
sustainability can be reduced by either excessive or insufficient N fertilizer inputs, and that
(b) soil erosion has an adverse impact on farm sustainability. Although the crop models
cannot predict soil erosion, Chapter 5 demonstrates a method for simulating the impact of
an assumed annual soil loss on crop production. The study includes erosion scenarios that
assume five constant levels (0, 25, 50, 100 and 150 Mg ha*1 yr*1) of annual soil loss.


160
strategy calls for a resource that is unavailable because of competing use by another
enterprise. The current model structure requires that each crop be simulated for an entire
season with a given set of resource requirements. There is no mechanism for feedback by
which resource constraints can alter crop performance. The appropriate way to deal with
the resource allocation problem is to resolve conflicts on a daily time step before
simulating crop and ecosystem processes (Fig. 5-27). This is not possible with the present
crop model structure.
For each simulation day
For each field
For each operation scheduled for today
Determine resources required.
For each resource
If there is a conflict (i.e., requirement > supply) then
Resolve the conflict through rescheduling or reallocation.
For each field
Simulate the adjusted operations.
Simulate crop and ecosystem processes,
end { for each field }.
end ( for each simulation day }.
Figure 5-27. Pseudocode representation of an algorithm for resolving resource conflicts
among crop enterprises.
The resource allocation problem could be solved by reorganizing the crop models
along hierarchical boundaries so that the ecosystem could function independently of crop
populations, then linking the models at code-level to a whole-farm model so that all
ecosystems on a farm could be simulated during any given simulation day. Caldwell and
Hansen (1993) demonstrated a hierarchical model structure. They reorganized a set of


134
observed yield variability can be attributed to two factors~P fertility and nematode
damageto which CROPGRO is not sensitive.
Maize. Simulated and observed maize grain yields are shown in Table 5-9 and Fig.
5-11. Simulated and observed harvest indices were reasonably consistent. Predicted time
to silking and physiological maturity was earlier than observed (Table 5-6). Although
yield predictions were not obviously biased, CERES clearly did not account for the range
of observed variability of yields (Fig. 5-11). It is more difficult with maize than with bean
to identify the determinants of yield variability that the model does not account for.
Table 5-9. Observed and predicted maize yields, Domingo and Trujillo farms, Cauca,
Colombia.
Farm
Planting
date
Plants
per ha
Canopy mass
(kg ha1)
obs. pred.
Grain yield
(kg ha1)
obs. pred.
Harvest index
obs. pred.
Domingo
Oct 14, 1993
50,309
12,210
12,382
4,664
4,683
0.382
0.378
Trujillo
Oct 14, 1993
55,556
21,016
13,423
6,746
4,877
0.321
0.363
Domingo
Mar 30,1994
47,778
7,600
11,788
2,310
5,034
0.304
0.427
Trujillo
Mar 28, 1994
62,500
14,179
14,381
5,274
5,584
0.372
0.388
Domingo
Oct 10, 1994
48,750
12,557
11,444
5,186
4,585
0.413
0.401
Trujillo
Oct 7, 1994
50,000
16,129
11,617
6,048
4,625
0.375
0.398
Cassava. The simulated cassava yield was close to the observed yield on the
Trujillo farm, but over predicted yield on the Domingo farm (Table 5-10). Because there
was only one treatment grown for one season, few inferences can be made about the
performance of CropSim CASSAVA in this environment. However, farmers report


112
Table 5-2, Properties of soil layers, site near on-farm trials, Domingo farm.
Property
Ap
35 cm
55 cm
total N, percent
0.673
0.254
0.198
NH4+ N, mg kg'1 soil
46.2
16.9
12.9
N03' N, mg kg'1 soil
2.80
2.46
2.93
Bray II extractable P, mg kg'1 soil
1.9
1.1
1.5
total P, mg kg"1 soil
334
148
128
organic P, mg kg'1 soil
200
108
96
organic C, percent
5.3
3.4
2.4
C/N ratio
8
14
12
KC1 extractable K, cmolc kg'1 soil
0.18
0.05
0.05
KC1 extractable Ca, cmolc kg'1 soil
1.88
0.30
0.11
KC1 extractable Mg, cmolc kg'1 soil
0.43
0.07
0.05
KC1 extractable acidity (H + Al), cmolc kg'1 soil
0.79
0.23
0.17
ECEC, cmolc kg'1 soil
3.28
0.65
0.38
base saturation, percent
76
65
55
pH in water
5.4
5.6
5.6
pH in KC1
4.7
5.0
5.1
sand, percent
55.8
silt, percent
27.7
clay, percent
16.7
bulk density, g cm'3
0.42
0.44
0.45
water retention at 0.0 atm, percent volume
57.5
55.4
62.2
water retention at -0.3 atm, percent volume
40.7
40.9
45.9
water retention at -1.0 atm, percent volume
38.0
38.4
42.6
water retention at -15 atm, percent volume
32.5
33.0
37.5
water drained upper limit, percent volume
46.7
48.9


118
tomato models to respond to observed pest populations or damage (Batchelor et ah,
1993), the crop models do not simulate pest populations.
The crop models are limited in their ability to handle crop residues and organic
amendments. They do not account for standing or surface residue, but assume that all
residue is incorporated at the time of application or harvest. The crop models cannot
simulate application of organic soil amendments (eg., manure) after the first season while
they are running in sequential mode. Finally, when the models are run in sequence, they
always incorporate all stover into the soil; harvesting part of the stover or failing to
remove some of the harvest product results in a mass balance error.
Prior testing. The CERES maize model has undergone extensive testing in
temperate North America and Europe (Kiniiy & Jones, 1986) and in various regions of
tropical Africa (Keating et al., 1991; Singh et ah, 1993), Asia and the Pacific (Singh,
1985) with generally acceptable results. Simulated maize yields match observed yields
closely in a Hawaiian Andisol (Hydric Dystrandept by the 1975 version of Soil Taxonomy)
(Ritchie et al., 1990b). Although CROPGRO is relatively new and has not yet been
thoroughly tested, a predecessor, BEANGRO v.1.01, has been tested in Colombia (White
et ah, 1995). Predictions of yield response to population density and water stress were
generally good. Phenology predictions were poorer. Scholberg (1993) adapted
CROPGRO for tomato. Validation work is in progress. Matthews & Hunt (1994)
described an earlier Pascal version of CropSim CASSAVA. Predictions of response to
temperature and photoperiod in Australia and to water stress in Colombia were generally


76
The LINKAGES section identifies the possible interactions among resources. Resource
linkages are initialized from the LINKAGES section and identify what resources can be
exchanged with a particular resource in response to a deficit, surplus, sale, or fixed costs
of ownership. Variable cost linkages identify resources that may be used to replenish a
deficit or dispose of a surplus. Fixed cost linkages identify resources that are charged
fixed costs of ownership for a particular resource. Resource linkages also identify the
components of a debt resource and components in the denominator or numerator of an
aggregate resource. Each resource linkage has a price from the price file that determines
the rate of exchange between the pair of resources. The price need not have monetary
units; it may represent, for example, a barter rate between two commodities.
Operations. The OPERATIONS REQUIREMENTS section relates field
operations to particular resources by identifying combinations of resources such as labor
and equipment that are required to complete each operation. Operations are either
returned by the crop models or specified in the SCHEDULED OPERATIONS section.
The resources required for an operation should be timed resources or their descendants
(seasonal, capital, or machine). The linkages that operation requirements establish
between operations and resources are analogous to the resource linkages that link
resources to other resources.
The SCHEDULED OPERATIONS section provides a flexible means of
incorporating management operations that the crop models do not consider, such as post
harvest processing, marketing harvested products or purchasing supplies. They can also
represent management and production of livestock or crops for which no model is


136
typical yields of about 10 Mg ha'1 of dry root grown for the starch market (Knapp,
personal communication). CropSim CASSAVA does not simulate plant response to N
stress and therefore is expected to over predict yields under the low level of soil fertility
management typical for the region.
Tomato. Simulated fresh fruit yields were 24,833 1729 kg ha'1 (SD, n = 10)
based on 6% dry matter for tomato planted April 1, with 75 kg applied N ha'1 and
automatic irrigation. Tomato was not included in any of CIATs on-farm trials. However,
a farmer in Siberia, located about 2 km from the Domingo farm, reported obtaining yields
for irrigated tomato ranging from 7200 to 8400 kg fresh fruit ha'1. The farmers reported
yields were lower than expected; FEDCAFE (1986) reported that the average tomato
yield in the Cauca region was 27,000 kg ha'1 in 1983.
Tomato is susceptible to a range of diseases and pests. I observed severe wilting
and foliar necrosis in a field of tomatoes near the Domingo farm. These symptoms were
not consistent with physiological maturity. Furthermore, many tomato cultivars are
susceptible to the same root knot nematodes (Meloidogyne spp.) that apparently reduced
bean yields (Overman, 1991).
Weather variability
Table 5-11 and Figure 5-12 present distributions of simulated yields in response to
actual and generated weather variability. With the exception of October-planted bean,
yield distributions were not significantly affected by the source of weather data (P = 0.05).
Although standard deviations of simulated time to maturity was consistently lower with


198
Sources of risk
Eliminating either weather or price risk reduced overall farm risk after 6 or 9 years
(Fig. 6-15c and d). Removing price risk improved sustainability at nine years while the
effect of removing weather risk was not significant (P = 0.05) (Fig. 6-16, Table 6-11).
The effect of eliminating price risk was much greater than the effect of eliminating weather
risk, supporting the hypothesis that price variability contributes more to farm risk than
weather. This suggests that farmers in the Cauca region are more vulnerable to the
uncertainties of their economic environment than to their physical environment.
Figure 6-13. Sustainability time plot of base and resource scenarios.


193
reducing crop yields. The level of applied N affected immediate crop yields, but did not
result in discemable long-term trends due to buildup or depletion of soil N (Chapter 5).
The reduction of sustainability with more applied N supports the hypothesis that excessive
N decreases sustainability by increasing costs.
Simulation results support the hypothesis that increasing soil erosion decreases
farm sustainability (Fig. 6-10, Table 6-9). Although the fixed amounts of soil loss applied
to the farm do not represent the erosion process realistically, this procedure does
demonstrate the influence of soil loss on farm sustainability, and the potential value of
Figure 6-9. Sustainability time plot of base and nitrogen management scenarios.


251
Caldwell, R.M. and J.W. Hansen. 1993. Simulation of multiple cropping systems with
CropSys. p. 397-412. In F.W.T. Penning de Vries, P. Teng & K. Metselaar (ed.)
Systems Approaches for Agricultural Development, Vol. 2. Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Carter, H.O. 1989. Agricultural sustainability: an overview and research assessment. Calif.
Agrie. 43(3): 16-18,37.
Casagrande, J.T. and M.C. Pike. 1978. An improved approximate formula for calculating
sample sizes for comparing two binomial distributions. Biometrics 34:483-486.
Castillo, W.C. 1990, 1993. Series estadsticas de precios promedios corrientes de frijol a
nivel de consumidor en Cali y Palmira Valle del Cauca, Centro Internacional de
Agricultura Tropical (CIAT). (unpublished report)
Castro, D.M., C.P. Roa and R.L. Rivera. 1992. Genesis y taxonoma de los Andisoles
Colombianos. Suelos Ecuatoriales 22:50-68.
Corporacin de Abastecimentos del Valle del Cauca S.A. (CAVASA). 1994. Series
histricas de precios por kilogramo registradas en el mercado mayorista de
CAVASA. CAVASA, Cali, Colombia, (unpublished report)
Conway, G.R. 1985. Agroecosystems analysis. Agrie. Admin. 20:31-55.
Conway, G.R. 1994. Sustainability in agricultural development: trade-offs with
productivity, stability and equitibility. J. Farming Systems Research Extension
4(2): 1-14.
Conway, G.R. and E.B. Barbier. 1990. After the green revolution: sustainable agriculture
for development. Earthscan Publications, London.
Cordray, S.M., L.S. Lev, R.P. Dick, and H. Murray. 1993. Sustainability of Pacific
Northwest horticultural producers. J. Prod. Agrie. 6:121-135.
Cramb, R.A. 1993. Shifting cultivation and sustainable agriculture in East Malaysia: a
longitudinal case study. Agrie. Systems 42:209-266.
Dagpunar, J. 1988. Principles of Random Variate Generation. Oxford Univ. Press, New
York.
Dahlberg, K.A. 1991. Sustainable agriculture -- fad or harbinger? BioScience
41(5):337-340.


22
that balances productivity, environmental soundness and socioeconomic viability goals.
Lai (1991) proposed a sustainability coefficient as a function of output per unit of input at
optimal per capita productivity or profit, output per unit of decline in the most limiting or
least renewable resource, and the minimum assured output level. Sands and Podmore
(1993) proposed an environmental sustainability index as an aggregation of sub-indices of
soil productivity, ecosystem stability, and potential to degrade the environment. Selection
of components of the sub-indices and the form of aggregation functions were indicated as
important research topics. Stockle et al. (1994) proposed a framework for evaluating
sustainability based on nine system attributes: profitability, productivity, quality of soil,
water, and air, energy efficiency, fish and wildlife habitat, quality of life, and social
acceptance. Production system sustainability is determined by scoring attributes as
weighted functions of quantifiable, long-term constraints, then combining weighted
attribute scores into an integrated measure.
The consistent inability to specify aggregation functions in these studies points to
the weakness of interpreting sustainability as the ability to fulfill diverse sets of goals as a
conceptual foundation for characterization. Diagnosis is limited by need to decide a-priori
the relative importance of different types of constraints to sustainability. Stockle et al.
(1994) acknowledged and defended the subjectivity needed to aggregate diverse system
attributes into an integrated measure of sustainability.


222
bundle of resources required to complete the operation. The resources should be timed
resources or descendants (i.e., capital, machine, or seasonal resources). A line contains
repeating blocks of resource requirements which represent a combination of mutually
dependent resources. A subsequent line with the same index and combination of
operation, crop and method represents an alternate resource bundle. An operation uses
resource bundles sequentially in the specified proportions until the operation is complete
or until the resource bundles are exhausted. While the operation date is within the time
window, any unfinished portion of the operation is deferred to the next day.
Eight operation type codes-PLNT, IRRT FERT, RESD, CHEM,
TILL, HARV and MRKTare predefined. The farm model will recognize any
additional operation codes found in the OPERATION REQUIREMENTS section. Jones
et al. (1994, Appendix B) lists method codes. Planting (PLNT) uses the planting
material codes (eg., PM001 indicates dry seeding). Method codes for application of
fertilizer (FERT), organic material (RESD) and pesticides (CHEM) use the
chemical applications codes. Tillage methods (TILL) are specified by implement codes.
Harvest (HARV) methods are HA001" for harvest product, HA002" for leaves, and
HA003 for canopy.
The example in Fig. B-6 specifies the requirements for marketing (MRKT) maize
(MZ). This operation has a priority of seven, can be spread across up to 14 days, and
' requires either two hours of both JOSE and HIRED TRUCK, or 2.5 hours of both
HIRED LABOR and the truck per Mg of grain.


52
Sustainability Applied to Farming Systems
Although sus