<%BANNER%>

Cropping Systems for Groundwater Security in India

Permanent Link: http://ufdc.ufl.edu/UFE0042709/00001

Material Information

Title: Cropping Systems for Groundwater Security in India Groundwater Responses to Agricultural Land Management
Physical Description: 1 online resource (204 p.)
Language: english
Creator: DOURTE,DANIEL RAY
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: GREEN -- GROUNDWATER -- INDIA -- SWAT -- WATER
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The total annual groundwater withdrawals in India (251 billion km3) are the highest of any nation. Depletion of groundwater resources is increasingly common in much of India, and farmers bear significant costs and greater vulnerability resulting from the loss or reduction of a reliable irrigation source. Three hypotheses were tested: (1) current rice cropland extent and management practices are depleting groundwater supplies, (2) tillage for water harvesting can significantly increase groundwater recharge in rainfed croplands, and (3) there are combinations of tillage, crop selection, and irrigation that are likely to increase groundwater recharge and reduce groundwater withdrawals. In order to test these hypotheses, there was the objective to evaluate improvements to the Green-Ampt infiltration routines of a hydrologic model, the Soil and Water Assessment Tool (SWAT), through the addition of a dynamic surface storage depth used for tillage parameterization. Also, the final objective was to assess the social and economic impacts of alternative agricultural land management. SWAT was used for simulating the groundwater balance (recharge ? irrigation pumping) of a 512 ha watershed to examine a variety of possible agricultural management options for groundwater sustainability. The best options for groundwater sustainability were evaluated based on predictions of groundwater recharge and withdrawals, evapotranspiration, and estimated household incomes. Reductions in rice cropland areas significantly improved the groundwater balance of the study area; water harvesting tillage simulated in all rainfed areas increased groundwater recharge by about 30 mm/year. Surface storage depth was shown to be the most important parameter for infiltration prediction in agricultural systems having 1.5 to 5.0 cm of surface storage capacity; surface storage depth was still important for infiltration prediction in systems having 0 to 1.0 cm of surface storage capacity. The vast extent of rice cropland areas and their highly negative groundwater balance suggest that irrigation from groundwater resources has caused much of the observed groundwater decline in India. Sensitivity analyses suggest that the addition of a variable surface storage depth head to the Green-Ampt infiltration routine can reduce uncertainty in infiltration simulations. Evidence of rainfall characterized by storms of greater intensity suggests that surface storage of runoff will become increasingly important for maintaining or improving current levels of groundwater recharge. Estimates of the economic impacts of selected management scenarios show promise that moderate management changes to improve the groundwater balance can still maintain or increase total watershed-scale income.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by DANIEL RAY DOURTE.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Haman, Dorota Z.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042709:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042709/00001

Material Information

Title: Cropping Systems for Groundwater Security in India Groundwater Responses to Agricultural Land Management
Physical Description: 1 online resource (204 p.)
Language: english
Creator: DOURTE,DANIEL RAY
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: GREEN -- GROUNDWATER -- INDIA -- SWAT -- WATER
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The total annual groundwater withdrawals in India (251 billion km3) are the highest of any nation. Depletion of groundwater resources is increasingly common in much of India, and farmers bear significant costs and greater vulnerability resulting from the loss or reduction of a reliable irrigation source. Three hypotheses were tested: (1) current rice cropland extent and management practices are depleting groundwater supplies, (2) tillage for water harvesting can significantly increase groundwater recharge in rainfed croplands, and (3) there are combinations of tillage, crop selection, and irrigation that are likely to increase groundwater recharge and reduce groundwater withdrawals. In order to test these hypotheses, there was the objective to evaluate improvements to the Green-Ampt infiltration routines of a hydrologic model, the Soil and Water Assessment Tool (SWAT), through the addition of a dynamic surface storage depth used for tillage parameterization. Also, the final objective was to assess the social and economic impacts of alternative agricultural land management. SWAT was used for simulating the groundwater balance (recharge ? irrigation pumping) of a 512 ha watershed to examine a variety of possible agricultural management options for groundwater sustainability. The best options for groundwater sustainability were evaluated based on predictions of groundwater recharge and withdrawals, evapotranspiration, and estimated household incomes. Reductions in rice cropland areas significantly improved the groundwater balance of the study area; water harvesting tillage simulated in all rainfed areas increased groundwater recharge by about 30 mm/year. Surface storage depth was shown to be the most important parameter for infiltration prediction in agricultural systems having 1.5 to 5.0 cm of surface storage capacity; surface storage depth was still important for infiltration prediction in systems having 0 to 1.0 cm of surface storage capacity. The vast extent of rice cropland areas and their highly negative groundwater balance suggest that irrigation from groundwater resources has caused much of the observed groundwater decline in India. Sensitivity analyses suggest that the addition of a variable surface storage depth head to the Green-Ampt infiltration routine can reduce uncertainty in infiltration simulations. Evidence of rainfall characterized by storms of greater intensity suggests that surface storage of runoff will become increasingly important for maintaining or improving current levels of groundwater recharge. Estimates of the economic impacts of selected management scenarios show promise that moderate management changes to improve the groundwater balance can still maintain or increase total watershed-scale income.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by DANIEL RAY DOURTE.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Haman, Dorota Z.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042709:00001


This item has the following downloads:


Full Text

PAGE 1

1 CROPPING SYSTEMS FOR GROUNDWATER SECURITY IN INDIA: GROUNDWATER RESPONSES TO AGRICULTURAL LAND MANAGEMENT By DANIEL R. DOURTE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFIL LMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 1

PAGE 2

2 201 1 Daniel R. Dourte

PAGE 3

3 ACKNOWLEDGMENTS I am deeply indebted to my advisor Dr. Dorota Haman for her substantial support and her consistent commitment to m y research, education, and professional development. I am very grateful for the help and wise counsel I have received from each person on my supervisory committee. I sincerely thank you: Dr. Rafael Muoz Carpena, Dr. Sanjay Shukla, Dr. Jim Jones, Dr. Lou is Motz, and Dr. Laila Racevskis.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ 3 LIST OF TABLES ................................ ................................ ................................ ........... 7 LIST OF FIGURES ................................ ................................ ................................ ...... 10 LIST OF ABBREVIATIONS ................................ ................................ .......................... 13 ABSTRACT ................................ ................................ ................................ .................. 16 CHAPTER 1 PROJECT OVERVIE W, GROUNDWATER MANAGEMENT, AND REASONS FOR GROUNDWATER DEPLETION ................................ ................................ .... 18 Introduction to Study Area ................................ ................................ ..................... 18 Climate of Wargal ................................ ................................ ............................ 19 Groundwater System of Wargal ................................ ................................ ...... 19 Goals of the Research ................................ ................................ ........................... 22 Hypotheses ................................ ................................ ................................ ..... 23 Objectives ................................ ................................ ................................ ....... 23 Groundwater in India: Evidence for and Causes of Depletion ................................ 24 Wargal Regiona l Groundwater Monitoring ................................ ....................... 24 Northwest India Large Region Groundwater Decline ................................ ....... 24 Groundwater Balance and Agricultural Management ................................ ...... 25 Historical Data: Rice Cropland Extent, Groundwater Irrigated Area, Rainfall ... 27 Water Balance Simulation Methods and the Soil and Water As sessment Tool ...... 30 Soil and Water Assessment Tool Overview ................................ ..................... 30 Summary of Water Balance Simulation Methods ................................ ............. 32 Hydrology in Ungauged Basins ................................ ................................ ....... 38 Factors Influencing Groundwater Recharge: Literature Review ............................. 39 Ratio nale and Significance of a Groundwater Recharge Literature Review ..... 39 Land Surface Environmental Factors and Groundwater Recharge .................. 41 La nd use and land cover ................................ ................................ ........... 43 Topography ................................ ................................ .............................. 45 Soil properties ................................ ................................ ........................... 47 Quantifying Gro undwater Recharge ................................ ................................ 48 Discussion and Conclusions on Factors Influencing Groundwater Recharge .. 49 Options for Managing Groundwater in Agr icultural Systems ................................ .. 50 Contributions of this Research ................................ ................................ ............... 51 2 RAINFALL INTENSITY DURATION FREQUENCY RELATIONSHIPS FOR ANDHRA PRADESH, INDIA : CHANGING RAINFALL PATTERNS AND IMPLICATIONS FOR GROUNDWATER RECHARGE ................................ .......... 63

PAGE 5

5 Rainfall Characterization and Water Resource Management ................................ 63 G roundwater Resources in India ................................ ................................ ........... 64 Objectives: Precipitation Characterization and Groundwater in India ..................... 65 Rainfall Characterization ................................ ................................ ........................ 67 Overview of IDF Analysis ................................ ................................ ................ 67 Methods for Rainfall IDF Development for Andhra Pradesh ............................ 68 Exploration of Trends in Occurrence of Rainfall Events of High and Low Intensity ................................ ................................ ................................ ....... 70 Results and Discussion: IDF Curves and Event Intensity Trends ........................... 71 Discussion: Precipitation Characterization and Groundwater in India .................... 72 Conclusions on Rainfall Intensity Trends and Groundwater Recharge ................... 73 3 IMPORTANCE OF SURFACE STORAGE PONDING DEPTH FOR PREDICTING INFILTRATION AND RUNOFF IN WATER CONSERVATION TILLAGE SYSTEMS ................................ ................................ .............................. 79 Modeling Infiltration and Tillage: Implications for Groundwater Recharge .............. 79 Infiltration, Surface Storage, and Changing Precipitation Character ................ 79 Green Ampt Infiltratio n and Depression Storage ................................ ............. 81 Groundwater Recharge and Tillage Management: Modeling Implications ....... 83 Methods for Analyzing the Importance of Surface Storage Depth .......................... 87 Site Description ................................ ................................ ............................... 87 Local Sensitivity Analysis ................................ ................................ ................ 88 Global Sensitivity Analysis ................................ ................................ .............. 88 Parameters for Green Ampt Infiltration Sensitivity Analysis ............................. 90 Development and Selection of Desig n Storms for the Analysis ....................... 93 Results and discussion of GAML Infiltration Sensitivity Analyses ........................... 93 Representative Design Storm for the A nalysis ................................ ................ 93 Tillage and MDS: GAML solution form ................................ ............................ 94 Local sensitivity analysis ................................ ................................ ................. 95 Global Sensitivity Analysis ................................ ................................ .............. 96 Conclusions on the Importance of Surface Storage Depth for Tillage Parameterization ................................ ................................ ................................ 97 4 EVALUAT ION OF AGRICULTURAL MANAGEMENT ALTERNATIVES FOR SUSTAINABLE GROUNDWATER IN INDIA: SIMULATED WATER BALANCE RESULTS ................................ ................................ ................................ ............ 106 Overview of Water Balance Simulation ................................ ................................ 106 Groundwater Management in India ................................ ............................... 106 Goals of Simulated Water Balance Experiments ................................ ........... 108 Water Balance Simulation Methods ................................ ................................ ..... 108 SWAT Preparation ................................ ................................ ........................ 108 Describing agricultural management ................................ ....................... 109 SWAT process modifications ................................ ................................ .. 110 SWAT Calibration and Evaluation ................................ ................................ 111 Parameter sensitivity and uncertainty analysis ................................ ........ 114

PAGE 6

6 Reservoir volume ................................ ................................ .................... 118 Groundwater recharge ................................ ................................ ............ 120 Results: SWAT Calibration, Evaluation, and Parameter Sensitivity ...................... 122 Uncertainty of Model Predictions ................................ ................................ ... 123 Groundwater Balance ................................ ................................ .................... 123 Results: Evaluation of Management Alternatives for Sustainable Groundwater ... 125 Significance of Tillage and MDS ................................ ................................ .... 126 Chang ing Extent and Irrigation Management of Rice Croplands ................... 128 Alternatives to Rice: Irrigated and Rainfed Crops ................................ .......... 129 Conclusions on Agricu ltural Management and Groundwater Supply in India ....... 130 5 SOCIAL AND ECONOMIC ASSESSMENT OF GROUNDWATER MANAGEMENT ................................ ................................ ................................ ... 149 Groundwater and I ndian Agricultural Economics ................................ ................. 149 Fieldwork on the Socioeconomics of Agricultural Management ........................... 150 Sustainability: Definitions ................................ ................................ .............. 151 Methods for Learning about Wargal Agricultural Management ...................... 152 Social and Biophysical Analysis Connections ................................ ................ 154 Results of Household Survey Fieldwork ................................ ........................ 154 Main concerns ................................ ................................ ........................ 155 Changes in groundwater and climate ................................ ...................... 156 Management and decision making ................................ .......................... 159 Valuing Groundwater ................................ ................................ ........................... 162 Economics of Alter native Agricultural Management ................................ ............. 163 Discussion and Summary of Fieldwork on the Socioeconomics of Agricultural Management ................................ ................................ ................................ .... 165 6 LIMITATI ONS, APPLICATIONS, AND CONCLUSIONS OF THE WARGAL STUDY OF GROUNDWATER DEPLETION AND AGRICULTURAL MANAGMENT ................................ ................................ ................................ ..... 174 Limitations ................................ ................................ ................................ ........... 174 Applica tions ................................ ................................ ................................ ......... 175 Conclusions ................................ ................................ ................................ ......... 176 APPENDIX A SWAT CODE MODIFICATIONS ................................ ................................ .......... 180 B INTERV IEW QUESTIONS FOR WARGAL FARMERS ................................ ........ 187 LIST OF REFERENCES ................................ ................................ ............................ 189 BIOGRAPHICAL SKETCH ................................ ................................ ......................... 204

PAGE 7

7 LIST OF TABLES Table page 1 1 Areas (ha) of cropland for three crop rotation in Wargal ................................ .... 58 1 2 Climate and hydrologic monitoring systems: obs ervation numbers, frequency, purpose ................................ ................................ ................................ ............. 60 1 3 Sources and factors included in analysis by Gogu and Dassargues; adapted from Gogu and Dassargues 2000 ................................ ................................ ...... 60 1 4 Summary of methods for quantifying groundwater recharge .............................. 61 1 5 Summary of the 9 recharge studies highlighting dominant factors ..................... 62 2 1 Annual maximum series rainfall intensity generated from hourly rainfall data from Hyderabad, Andhra Pradesh, India. Annual and Kharif season (June September) total rainfall and rain days. ................................ ............................. 75 2 2 Parameters for Weibull CDF: F(I) = 1 exp( where I is rainfall intensity, and parameters for IDF function: ................ 75 2 3 Rainfall intensity values from Weibull cumulative distribution functions (CDFs): hourly data 1993 2008 for Hyderabad ................................ .................. 76 2 4 Rainfall intensity (mm/hr) from Kothyari and Garde general formula fitted to recent intensity data (1993 2008) for Hydera bad 1 RMSE calculated based on CDF intensities. ................................ ................................ ............................ 77 2 5 Rainfall intensity (mm/hr) from Kothyari and Garde original formula fitted to older intensity data (1950 1980) for southern zone of India 1 RMSE calculated based on CDF intensities. ................................ ................................ 77 2 6 Differences (mm/hour) in predicted rainfall intensity values between Kothyari and Garde IDF formula using 1992 parameters 1 and using updated param eters fitted from this study 2 ................................ ................................ ...... 78 3 1 Parameter values needed for Green Ampt model: minimum, maximum, mean i ), s ), effective hydraulic conductivity (K se ), wetting f ), and maximum depression storage depth (MDS) for conventional tillage (CT) and tied ridge tillage (TR). ................................ ........ 100 3 2 IDF data (1993 2008 hourly rainfall), Weibull distribution used, values in the table are rainfa ll intensities in mm/hour ................................ ........................... 101

PAGE 8

8 3 3 4 hour design storms of 2, 5, 10 year return period used for rainfall input in GAML sensitivity analysis; 20 min time step ................................ .................... 102 3 4 GAML predicted runoff (RO) and infiltration (F) depths of conventional (CT) and tied ridge (TR) tillage for 2 year, 4 hour storm; standard and complete GAML solutions. Min, mean, and max MDS were 0, 5, 10 mm for CT and 15, 3 2.5, 50 mm for TR. ................................ ................................ ........................ 102 3 5 First order sensitivity indexes (S i ) for output F of the 5 GAML parameters for tied ridge (TR) and conventional tillage (CT) for both GAML solution forms, complete and standard. ................................ ................................ ................... 104 4 1 Descriptions of 25 alternative management scenarios with abbreviations used ................................ ................................ ................................ ................ 135 4 2 Maximum surface area and volume of the six reservoirs (tanks) in Wargal watershed ................................ ................................ ................................ ....... 138 4 3 Specific yield, natural recharge, and total recharge from DWTF method and total recharge from SWAT simulations in 2009 and 2010 ................................ 138 4 4 Nash Sutcliffe Efficiency (NSE), percent bias (PBIAS), and ratio of the root mean square error to the standard deviation of measured data (RSR) ............ 139 4 5 Parameter name, t statistic, p value, and range of the 13 parameters estimated during SWAT calibration ................................ ................................ 140 4 6 Summary of selected water balance components for existing and alter native agricultural management, sorted by groundwater balance (GW balance = recharge given specific yield of 0.013 from DWTF method. Simulations using generated weather 2000 2009, 792 mm ra infall. ................................ ............. 141 4 7 Summary of selected water balance components for existing and alternative agricultural management, sorted by groundwater balance (GW balance = recharge edicted groundwater level change, mm, given specific yield of 0.013 from DWTF method. Simulations using observed weather 2009, 667 mm rainfall. ................................ ........................ 142 4 8 Sensitivity of annual surface runoff and groundwater recharge to MDS at the basin scale: mm change in depth of recharge and runoff (from existing tillage, MDS = 0) per mm MDS ................................ ................................ ................... 146 4 9 Sensitivity of annual surface runoff and groundwater recharge in rainfed croplands (corn and cotton) in kharif season to MDS: mm change in depth of recharge and runoff (from existing tillage, MDS = 0) per mm MDS .................. 146

PAGE 9

9 5 1 Concerns of community member s ranked (for women) in order of decreasing frequency of reporting ................................ ................................ ..................... 168 5 2 Ways of observing groundwater depletion ................................ ....................... 168 5 3 Perceived c auses of local groundwater depletion ................................ ............ 168 5 4 Suggested reasons for recent reductions in local rainfall amounts .................. 169 5 5 Information use d to decide about crop selection: ranked (greatest to least ................................ ................... 169 5 6 Information used to decide about irrigation management: ranked (greatest to least ................................ ....... 169 5 7 Decision category and associated most important responses ......................... 171 5 8 Percentages of resp onses concerning who is responsible for irrigation pump control ................................ ................................ ................................ ............. 172 5 9 Percentages of responses concerning who is responsible for purchase farm inputs (seed and fertilizer) ................................ ................................ ............... 172 5 10 Literature review of yield increase for selected crops in response to tied ridge tillage; methods are measured yield (obs) or simulated yield (sim) .................. 172 5 11 Yields of crops commonly grown in Wargal based on household surveys and state agency data. Sample size indicates number of households that responded to surveys about yield data. Value ($/kg) based on 2009 2010 government of India minimum suppo rt prices (MSP) ................................ ....... 173 5 12 Groundwater balances (mm) and estimated values (USD) of the selected top seven management scenarios (Chapter 4) ................................ ...................... 173

PAGE 10

10 LIST OF FIGURES Fig ure page 1 1 Study area location in Wargal mandal, eastern Medak district, northwestern Andhra Pradesh ................................ ................................ ................................ 53 1 2 Groundwater sy stem diagram of Wargal (adapted from Dewandel et al., 2006 and Marechal et al., 2006). t is layer thickness, values are approximate. Water table height fluctuates between 15 and 35 m below surface. .................. 54 1 3 Water balance diagram showing connections between unsaturated and saturated zone water balances. I is irrigation, P is precipitation, SS is surface storage, Inf is infiltration, ET is evapotranspiration, RO is runoff, UF is unsaturated flow, C R is capillary rise, R is groundwater recharge, DP is deep percolation, SF is saturated flow. ................................ ................................ ....... 55 1 4 Total annual harvested areas of rice in India and Andhra Pradesh, 1961 2001 ................................ ................................ ................................ .................. 56 1 5 Area equipped for mechanized irrigation from groundwater source in India: 1961, 1971, 1981, 1986, 1993; population of India each year from 1961 to 2007. ................................ ................................ ................................ ................. 56 1 6 Annual rainfall in Telangana region (1960 2008): slight declining trend observed (about 3 mm/year reduction in annual rainfall if trend is assumed linear) ................................ ................................ ................................ ................ 57 1 7 Annual rainfall trends (1960 2008) in the 3 state region of the GRACE study of Rodell et al., 2009 ................................ ................................ ................ 57 1 8 Watershed boundary, reservoir locations (large dots), and digital elevation model of Wargal watershed, 2.2 m resolution ................................ .................... 58 1 9 Variogram plot of % clay in layer 1 of Wargal soil: used to evaluate spatial autocorrelation of soil properties and for kriging to generate continuous map of soil data ................................ ................................ ................................ ......... 59 2 1 Intensity duration frequency curves for 1, 2, 4, 8, and 24 hour durations developed from hourly rainfall data from 1993 2008 from Hyderabad, India ... 76 2 2 rainfall events (1974 2008) ................................ ................................ ................ 78 3 1 Illustration of infiltration modeled by Green s i are volum etric water content, saturated and initial, respectively, h p is depth of water ponded at the surface, L is distance from surface to the wetting front, f is wetting front suction. ................................ ................................ .................. 99

PAGE 11

11 3 2 Typical furrowed field beside tied ridged field following rain event (image from Jones and Baumhardt 2003) ................................ ................................ ..... 99 3 3 Empirical CDF for annual maximum series (AMS) rainfall intensity for 4 hour storms; Weibu ll, Gamma, and Gumbel CDF fitted to AMS data using maximum likelihood parameter estimation. ................................ ...................... 100 3 4 Infiltration (F) and runoff (RO) for a 4 hour 2 year design storm; complete and standard GAML s olutions ................................ ................................ ................ 101 3 5 Infiltration and runoff % changes from mean MDS values for 2 year, 4 hour storm for both GAML solution types; min, mean, max are values of MDS for conventional (CT: 0, 5, 10 mm) and tied ridge (TR: 15, 32.5, 50 mm) tillage. .. 103 3 6 Total effect sensitivity indices (outputs are F and RO; their sensitivities to each parameter are equal) for Green Ampt parameters for tied r idge and conventional tillage for 4 hour design storm of 2 year return period. ................ 105 4 1 Observed and simulated (for calibration and validation) watershed outlet reservoir volume and rainfall in Wa rgal watershed; 3/19/2010 to 12/3/2010 .... 137 4 2 Simulated and observed reservoir volumes: 3/19/2010 to 12/3/2010 ............... 137 4 3 Observed round water level change and rainfall during wet and dry periods .... 139 4 4 Basin scale response of groundwater recharge, groundwater balance, and runoff to MDS changes in rainfed croplands in kharif seas on and in both kharif and rabi seasons. MDS of 15, 32.5 or 50 mm. 30 to 50% less runoff with tillage for 15 mm < MDS < 50 mm; groundwater balance from 11 mm for existing management to 7 mm for 50 mm MDS in rainfed areas in both seasons. ................................ ................................ ................................ .......... 143 4 5 Field scale response of groundwater recharge and runoff to tillage changes (MDS depth) in rainfed croplands in kharif season. MDS of 15, 32.5 or 50 mm. 54 to 83% less runoff from rainfed croplands with tillage for 15 mm < MDS < 50 mm; groundwater balance from 11 mm for existing management to 6 mm for 50 mm MDS in rainfed areas in both seasons. ............................. 144 4 6 Field scale response of groundwater recha rge and runoff to tillage changes (MDS depth) in rainfed croplands in both kharif and rabi seasons ................... 145 4 7 Irrigation, recharge, ET, and groundwater balance for tied ridge tillage scenarios: 3 MD S depths in rainfed areas in kharif, rabi, and both seasons .... 147 4 8 Irrigation, recharge, ET, and groundwater balance for selected changes in extent and irrigation management of rice croplands ................................ ........ 148

PAGE 12

12 5 1 Some women take a break for lunch and to participate in discussions for this research ................................ ................................ ................................ .......... 170 5 2 One of six water harvesting struct ures (tanks) in the Wargal watershed used for increasing groundwater recharge ................................ ............................... 171

PAGE 13

13 LIST OF ABBREVIATION S ANGRAU Acharya N.G. Ranga Agricultural University: located in Hyderabad, India; partner university in this resea rch AQUASTAT This is the global web based information system on water resources and agriculture, developed by the Land and Water Division of the United Nations FAO ARS Agricultural Research Service of the USDA CDF Cumulative distribution function CGWB Cent ral Groundwater Board of India: a Ministry of Water Resources organization responsible for monitoring and maintaining sustainable groundwater resources in India CN Curve Number: an empirically based parameter used in the prediction of CN runoff and infiltr ation; developed by the Natural Resources Conservation Service (NRCS, formerly SCS) of the USDA CSIRO9 Commonwealth Scientific and Industrial Research Organization : climate model CT Conventional Tillage DEM Digital Elevation Model: a raster based represent ation of topography DRASTIC A groundwater vulnerability mapping method ET Evapotranspiration: the combination of plant transpiration and evaporation of water from soil, surface water, and vegetation surfaces FAO Food and Agriculture Organization of the Uni ted Nations FAOSTAT A database of time series information on food and agriculture: developed and supported by FAO GAML Green Ampt Mein Larson: physically based infiltration equations developed by Green and Ampt (1911) and improved by Mein and Larson (1973 ) GEV Generalized Extreme Value: a family of probability distributions that includes the Gumbel, Frchet and Weibull distributions as special cases

PAGE 14

14 GIS Geographic Information System: software tools used in the development and analysis of spatial data GLUE Generalized Likelihood Uncertainty Estimation: a simple, generic numerical method for use in calibration and uncertainty analyses of models of systems GRACE Gravity Recovery and Climate Experiment: twin orbiting satellite system used for observations abou t TWS GSA Global Sensitivity Analysis HYETOS A computer program used for stochastic disaggregation of rainfall to shorter time scales IAHS International Association of Hydrological Sciences ICRISAT International Crops Research Institute for the Semi Arid T ropics IDF Intensity, duration, frequency: relationships describing the rainfall pattern of a certain location IPCC Intergovernmental Panel on Climate Change LSA Local Sensitivity Analysis LULC Land use and land cover: describes the natural and human made landscapes MDS Maximum depression storage: the greatest area averaged depth of water that can be stored on the surface of a land area before runoff occurs MSP Minimum Support Price: an Indian national agriculture policy that sets the lowest prices for sele cted crops NASA National Aeronautics and Space Administration of the United States government; responsible for aerospace research NSE Nash Sutcliffe Efficiency: a model evaluation quantity NSS Near surface storage: depth of water stored in surface depressi ons, comparable to MDS PBIAS Percent bias: a model evaluation quantity PUB Predictions in Ungauged Basins: and IAHS initiative to reduce model prediction uncertainty

PAGE 15

15 RSR T he ratio of the root mean square error in model predictions to the standard deviation of measured data : a model evaluation quantity SCS Soil Conservation Service of the USDA; now called the Natural Resources Conservation Service (NRCS ) SINTACS A groundwater vulnerability mapping method adapted from DRASTIC for improved performance in Medit erranean areas SWAT Soil and Water Assessment Tool: a distributed parameter, landscape scale, open source hydrologic model TAW Total available water: depth of water in soil that is extractable by plants TR Tied Ridge tillage TWS Terrestrial water storage: total amount of water stored in a specified land area during some time UKHI United Kingdom Meteorological Office High Resolution General Circulation Model : a global climate model USDA United States Department of Agriculture WAVES A biophysically based wate r balance model

PAGE 16

16 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 CROPPING SYSTEMS FOR GROUNDWATER SECURITY IN INDIA : GROUNDWATER RESPONSES TO AGRICULTURAL LAND MANAGEMENT By Daniel R. Dourte May 201 1 Chair: Dorota Z. Haman Major: Agricultural and Biological Engineering The total annual groundwater withdrawals in India (251 billion km 3 ) are the highest of any natio n. Depletion of groundwater resources is increasingly common in much of India, and farmers bear significant costs and greater vulnerability resulting from the loss or reduction of a reliable irrigation source. Three hypotheses were tested: (1) current ri ce cropland extent and management practices are depleting groundwater supplies, (2) tillage for water harvesting can significantly increase groundwater recharge in rainfed croplands, and (3) there are combinations of tillage, crop selection, and irrigation that are likely to increase groundwater recharge and reduce groundwater withdrawals. In order to test these hypotheses there was the objective to evaluate improvements to the Green Ampt infiltration routines of a hydrologic model, the Soil a nd Water Ass essment Tool ( SWAT ) through the addition of a dynamic surface storage depth used for tillage parameterization. Also, the final objective was to assess the social and economic impacts of alternative agricultural land management. SWAT was used for simulat ing the groundwater balance (recharge irrigation pumping) of a 512 ha watershed to examine a variety of possible agricultural management options for groundwater

PAGE 17

17 sustainability. The best options for groundwater sustainability were evaluated based on pred ictions of groundwater recharge and withdrawals, evapotranspiration and estimated household incomes. Reductions in rice cropland areas significantly improved the groundwater balance of the study area; water harvesting tillage simulated in all rainfed are as increased groundwater recharge by about 30 mm/year. Surface storage depth was shown to be the most important parameter for infiltration prediction in agricultural systems having 1.5 to 5.0 cm of surface storage capacity; surface storage depth was still important for infiltration prediction in systems having 0 to 1.0 cm of surface storage capacity. The vast extent of rice cropland areas and their highly negative groundwater balance suggest that irrigation from groundwater resources has caused much of th e observed groundwater decline in India. Sensitivity analyses suggest that the addition of a variable surface storage depth head to the Green Ampt infiltration routine can reduce uncertainty in infiltration simulations. Evidence of rainfall characterized by storms of greater intensity suggests that surface storage of runoff will become increasingly important for maintaining or improving current levels of groundwater recharge. Estimates of the economic impacts of selected management scenarios show promise that moderate management changes to improve the groundwater balance can still maintain or increase total watershed scale income.

PAGE 18

18 CHAPTER 1 PROJECT OVERVIEW, GROUNDWATER MANAGEMENT, AND REASONS FOR GROUNDWATER DEPLETION Introduction to Study Area There is general agreement that water scarcity in India is severe (Alcamo et al. 2000; Yang et al., 2003). Total (761 km 3 ) and agricultural (688 km 3 ) water withdrawals for India in are the highest in the world (AQUASTAT, 2010). Groundwater in India is a highl y important resource for irrigation and household use, and its extensive use is resulting in widespread groundwater depletion (Shah et al., 2003; CGWB 2007; Rodell et al. 2009). More than half of the irrigation requirements of India are met from groundw ater (CGWB, 2002; Shah et al., 2003), and the number of mechanized borewells in India has increased from less than 1 million in 1960 to 26 28 million in 200 2 ( Mukherji and Shah 2005 ). India is characterized by substantial diversity in climate zones and l andscapes. One of the highest annual rainfall averages occurs in Cherrapunji in northeastern India: 11,430 mm. Western Rajasthan of northwestern India averages only 37 mm of annual rainfall; the n ational average a nnual rainfall is 1083 mm/year. India i s the second most populous and seventh largest country in the world (1.18 billion people, 328 Mha of total area and 180 Mha agricultural area; FAOSTAT 2010 ). The study area being considered here is in the Wargal mandal (a mandal is a local organization al unit, similar to a county) of Medak district, Andhra Pradesh, in the southern central part of India. This area was selected by Indian collaborators from Acharya N.G. Ranga Agricultural University (ANGRAU) in Hyderabad, Andhra Pradesh. Figure 1 1 illu strates the location of the study area. This 5 12 ha watershed was selected for analysis because of its relative proximity to ANGRAU and because it is

PAGE 19

19 illustrative of the common regional problem of groundwater decline, making farming systems increasingly v ulnerable. There are 157 households with landholdings in Wargal; 198 borewells are found in Wargal. Climate of Wargal The annual average rainfall in Wargal is 780 mm ; about 80% of rainfall is received from June September from the southwest monsoon syste m Average annual Potential evapotranspiration (PET) in Wargal is between 1600 and 2000 mm/year (UNEP, 1992). PET is the amount of water that would be evaporated and transpired from a landscape that had no soil moisture deficit. The year is divided into three seasons: kharif season is the rainy season and most important growing season from June September, rabi is the season from October February, summer is the very hot, dry season from March May. Groundwater System of Wargal Groundwater in the Wargal s tudy area occurs based on field observations (CGWB, 2007) and the work of Marechal et al. (2006), under largely unconfined conditions in consolidated formations (fractured granite ; transmissivity: 100 to 150 m 2 /day ; specific yield: 0.010 to 0.015 ) The s tudy by Marechal et al. (2006), in a watershed nearby the Wargal area, characterizes the hydrogeology as being Archean granite overlain by a clayey sandy regolith layer. This is consistent with the general hydrogeological characterization developed by Dew andel et al. (2006) for hard rock aquifers in India. The fissured granite layer 20 3 5 m thickness is typically where borewell casings are screened for groundwater withdrawal; this layer also is responsible for much of the horizontal flow (Dewandel et al ., 2006). The water table is 15 30 m below the ground surface, and there is very little interaction with surface water. The

PAGE 20

20 diagram of Figure 1 2 (adapted from Marechal et al., 2006) describes the aquifer system of Wargal. There are sometimes increased w ater quality problems as groundwater levels become reduced ; this is often to due naturally occurring arsenic and fluoride in the geology and soils of a region (Khan, 1994, concerning arsenic in Bangladesh; Subba Rao and Devadas, 2003 concerning fluoride i n India ) The concentrations of fluoride and arsenic in groundwater are sometimes greater deeper in an aquifer, and more water is withdrawn from deeper in the system as water levels decline There have been anecdotal reports of some people in Wargal show ing mild symptoms of fluoride contamination (Reddy, 2009) and the study of Subba Rao and Devadas ( 2003) in a district in western Andhra Pradesh documents a number of cases of fluoride contaminated groundwater. The Wargal watershed is located in the northw estern Andhra Pradesh district of Medak; groundwater resources for the Medak district (in 2005) were a total renewable groundwater volume of 813.19 Mm 3 and a net annual draft 704.07 Mm 3 leaving a balance of 109.12 Mm 3 of groundwater or a stage of ground w ater development of 87% (CGWB, 2007) ; these values were developed from large region simple water balance estimates and limited data on borewell withdrawals A decade of monitoring groundwater levels in borewells in Medak district (1996 2005) showed pre mo nsoon groundwater decline of 1.1 to 5.2 m. and post monsoon decline of 0.2 to 6.6 m during the 10 year period (CGWB, 2007). The monitoring co nsists of 26 wells being manually measured four times each year for depth to groundwater.

PAGE 21

21 Artificial recharge of groundwater is an ancient and widespread practice in India. More than half a million artificial recharge structures (ponds and reservoirs from excavation and small dam construction) are scattered throughout the country (Sakthivadivel, 2007). There are s ix small reservoirs developed for groundwater recharge in Wargal. In field water harvesting through tillage practices has received little attention relative to the groundwater recharge schemes outside of cropland areas. In field water harvesting (or wate r harvesting tillage) research in India has been mostly concerned with residue management and runoff comparisons between conventionally tilled and no till systems (Rao et al., 1998; Bhattacharyya et al., 2006). The groundwater crisis can be addressed at b oth the supply and demand sides through water harvesting tillage: decreasing the runoff of rainfall from cropland areas reduces required irrigation withdrawals and also increases recharge of groundwater as stored water percolates beyond plant root zones. C onservation tillage is any tillage practice that reduces soil and water loss from a cropland area. Water harvesting tillage, a type of conservation tillage, describes the formation during primary cultivation of soil surface geometries that allow for poten tially substantial surface storage of rainfall or applied water. Soil surface microtopography has a significant impact on infiltration (Darboux and Huang, 2005), and in agricultural areas is largely influenced by tillage management. Numerous demonstratio ns of crop yield and soil moisture improvements have been demonstrated in response to water harvesting tillage (Twomlow and Breneau, 2000; Wiyo, 2000; Guzha, 2004; Tesfahunegn and Wortmann, 2008). However, the hydrologic benefits of water harvesting tilla ge at large spatial and time scales have received comparatively little

PAGE 22

22 research. An objective of this research was to demons trate the groundwater recharge responses to increases in rainfed croplands under water harvesting tillage and changes in crop selec tion and irrigation management. Additionally, the study produce d estimates of long term groundwater decline under present management practices. This analysis was completed using an established water balance model: the Soil and Water Assessment Tool (SWAT ), described by Arnold et al. ( 1998 ) Gassman et al. (2007), and Krysanova and Arnold ( 2008). Goals of the R esearch Agricultural water use makes up 80% of India more than half of the irrigation requirements of India are met by groundwater (CGWB, 2002; Shah et al. 2003); therefore, irrigation from groundwater can be credited with significant responsibility for the food grain self sufficiency achieved by India in the last three decades. Throughout Asia, irrigation from gr oundwater has become a major contributor to agricultural improvements in recent decades. The extensive survey of Shah et al. (2006) collected data from 2,629 well owners from 278 villages in India, Pakistan, Nepal, and Bangladesh for the purpose of assess ing the significance of irrigation from groundwater in the agricultural economies of South Asia. The ir data prompted the finding that for most farmers, irrigation from groundwater the ( Shah et al., 2006 ) S ome investigators have suggested that the groundwater socio ecology in Asia, and particularly in India, is at a critical point (Mukherji and Shah 2005; Shah et al., 2003). While from a resource management perspective, groundwater depletion could be said t o be self regulating ; meaning, as groundwater is depleted, extraction becomes more expensive and groundwater withdrawals are reduced. There is little evidence that this self regulation is

PAGE 23

2 3 happening in South Asia ( Shah et al., 2006), and if it does there are still severe consequences for households that lose the ability to irrigate from groundwater sources. The progression of groundwater use in agriculture has been organized into four stages (Shah et al., 2003): (1) expansion of borewell installations, (2) groundwater based agrarian boom, (3) onset of groundwater depletion concerns, and (4) collapse of groundwater based systems. Data from the Wargal study area (perceptions of farmers, numbers of failed borewells, and district scale ground water level monit oring) suggest that the study area is in stage three of the four stages: onset of groundwater depletion concerns. Broadly, the goal of this research is to find the best agricultural management solutions for reducing groundwater pumping and increasing grou ndwater recharge. The hypothese s being tested: Hypotheses The following hypotheses were tested in this research: Current rice cropland extent and management practices are depleting groundwater supplies Tillage for water harvesting can significantly increa se groundwater recharge in rainfed croplands There are combinations of tillage, crop selection, and irrigation that can improve groundwater recharge and reduce groundwater pumping while remaining sufficiently productive for household economies Objectives A ssess groundwater sustainability of agricultural management scenarios ( selected combinations of crop se lection, irrigation management, and tillage) at a watershed scale Evaluate impr ovements to a hydrologic model, Soil and Water Assessment Tool for simulat ing groundwater balance under various cropping systems

PAGE 24

24 Combine biophysical and social analyses to improve the relevance and likelihood of implementation of proposed agricultural management solu tions to groundwater depletion These objectives were designed t o test the three hypotheses about rice cropland extent, tillage for water harvesting, and combinations of management alternatives. Groundwater in India : Evidence f or and Causes of Depletion Wargal Regional Groundwater Monitoring The groundwater of Wargal m andal was estimated at 98 % stage of development in 2004, meaning that annual groundwater withdrawals were nearly equal to annual groundwater recharge (1018 ha m consumed of 1040 ha m recharged, CGWB, 2007). Rechar ge estimates were based on topographic, ge ologic, soils and weather data. Withdrawals estimates are based on total number of wells in the district (115,718 total, Groundwater Board (CGWB) observed a 2 4 m declin e in water level in 26 observation wells in Medak district during the 10 years from 1996 to 2005, suggesting an approximate annual decline of 20 cm (CGWB 2007). This decade of monitoring and the water balance simulations of Chapter 4 give some evidence that the groundwater of Wargal mandal is actually beyond a 100% stage of development, meaning withdrawals are exceeding recharge. Northwest India Large Region Groundwater Decline A recent groundwater depletion study that has received much attention is th e remote sensing analysis of Rodell et al. ( 2009 ) The study area was northwestern India, the states of Haryana, Punjab, and Rajasthan, which is distant from the study area being considered here. However, the analysis is illustrative of regional groundwa ter decline reports that are common in the literature. In this case, the nature of the

PAGE 25

25 observations required that the study area be very large. The NASA Gravity Recovery and Climate Experiment (GRACE) satellites were employed to measure changes in terres trial water storage (TWS). The GRACE system is different from typical remote sensing systems in that it does not use any sensing of electromagnetic waves (thermal, visible, or microwave), rather it uses changes in distance between the pair of GRACE satell ites as the y orbit the earth to estimate TWS changes. The changes in distance result from accelerations of the satellites in response to changes in the gravity field of Earth. The gravity field is altered by terrestrial landscapes, biomass, buildings, an d water. Information about topography, land use/land cover, vegetation, and surface water allows the e ffects of these to be accounted for, leaving groundwater to explain the remaining variation in gravity field. During the study period from August 2002 t o October 2008, 109 km 3 of groundwater loss was estimated; that is the equivalent of about 4 cm each year over the three state area. Rainfall was considered to be normal during the period. Groundwater Balance and Agricultural Management The groundwater ba lance is connected to the overall water balance at the land surface. In this project, where it is proposed that agricultural groundwater withdrawals are reducing groundwater storage volumes, a convincing connection should be shown between groundwater pump ing and groundwater level decline. Before looking at the relevant regional and national data, it is helpful to consider the simple water balances: In the unsaturated zone and at the soil surface: ET RO ( 1 1 ) where SW is change in soil water, P is precipitation, I is irrigation, ET is

PAGE 26

26 is the net chan ge in subsurface flow into and from the system. In the saturated zone (groundwater system): I ( 1 2 ) where S is change in groundwater storage, R is recharge (approx. = DP), I is irrigation pumped from groundwater, CR is capilla from the groundwater system. The wa ter balances are diagrammed in F igure 1 3 The diagram includes infiltration (Inf) as a water balance component in the unsaturated zone; Inf is an indirect flow, but it is the only component crossing the boundary at the surface. Inf can be expressed as: Inf = P + I RO ET boundaries of both zones are assumed to be the boundaries of the study area. All of the above quanti ties are manageable (except P) either directly or indirectly. If the goal is to improve or maintain S (groundwater storage), observing the diagram of F igure 1 3 indicates that irrigation and recharge remain as the manageable quantities in the groundwater balance, irrigation managed directly and recharge managed indirectly through management of SS, ET, and I. Irrigation and recharge appear in the surface water balance as I and DP. Therefore, the data presented about population, irrigated area, rice cropla nd, and others should be associated either with I or DP as these are the quantities at the surface that are directly influencing the groundwater balance. The widespread reports of groundwater decline in India suggest that it is safe to assume that changes in SF and CR are not responsible for the observed large regional groundwater decline There are large regions of northwestern and peninsular India where

PAGE 27

27 groundwater depletion is well documented (Foster and Chilton, 2003), and if depletion in these region s was caused by excessive groundwater withdrawals from neighboring areas, that would just mean that there is even more greater groundwater decline in those neighboring regions Historical Data: Rice Cropland Extent, Groundwater Irrigated Area, Rainfall T his section presents evidence from the literature and from national and state level data suggesting that rice cropland management and extent has resulted in the observed groundwater depletion in India. Indian groundwater depletion is generally attributed to the large observed increases in irrigation from groundwater that have occurred in India during the last 40 years (Rodell et al., 2009 ; Rao et al., 2001 ; Shah et al., 2003 ). Flooded rice is the dominant crop in India and is typically irrigated heavily t o maintain ponded surface water in fields. From 1961 to 2001, the national harvested rice area of India increased from 35 to 45 Mha and average yields jumped from 1500 to 3100 kg/ha (FAOSTAT 2010 ). T rends are similar, as shown in F igure 1 4 in the stud y domestic consumption; therefore, the huge increases in rice production from 1961 to he trends in rice production and population increase are similar: from 1961 to 200 7 rice production increased 277% and population increased 255 % (FAOSTAT 2010 ). Yield increases in the aforementioned period (214 %: 1961 to 2007) are partly the result of inc reases in irrigated area and improvements in irrigation management; plant variety and nutrient management improvements have also likely contributed to yield increases. Figure 1 5 shows a nearly fourfold increase in area irrigated from groundwater resource s between 1961 and 1993; also shown is the steady increase in

PAGE 28

28 of mechanized wells and borewells in India increased from less than 1 million in 1960 to more than 2 6 28 m illion in 200 2 ( Mukherji and Shah 2005 ). These data and figures show clear correlations between population, rice production, and groundwater irrigation. It could be argued that without a time series of water balance data, the causes of groundwater depl etion cannot be known. This is true; however, it can reasonably be inferred from the observed increases in area irrigated from groundwater, borewell installations, rice area, and from the literature on water balances of rice systems that irrigation wi thdrawals of groundwater are resulting in depletion of groundwater resources in India. The other possible explanation for reduced quantity of groundwater resources aside from management affecting recharge and/or groundwater withdrawal is declining ra infall trends. As mentioned above in the discussion of groundwater and surface water balances, recharge and irrigation are the manageable quantities of the groundwater balance. Recharge could be altered in numerous ways by management at the surface that affects surface storage, runoff, deep percolation, and/or evapotranspiration. If it is assumed, due to the increases in rice croplands which reduce runoff and increase surface storage, that the changes over time would not decrease the potential recharge o f a landscape, then groundwater depletion must result either from a reduction in water reaching the surface (rainfall) or an increase in groundwater pumping. Increased irrigation from groundwater of course increases water applied to the surface, but the p artitioning of the water generally about half the irrigation water leaves the system as evapotranspiration means the net change in the

PAGE 29

29 groundwa ter balance would be negative. Therefore, it should be examined whether precipitation changes could explain the observed groundwater depletion. Long term rainfall trends (a nnual rainfall depth data from Kothawale et al., 2008 ) are considered in F igures 1 6 and 1 7 for the regions of groundwater depletion mentioned above 20 cm/year in Medak district, Andhra Pra desh (CGWB 2007); 4 cm/year in Haryana, Punjab, and Rajasthan (Rodell et al., 2009). Small annual declines in total rainfall, if trends are assumed linear, suggest rainfall changes have little connection to groundwater depletion. Linear trends from 1960 were estimated to be: 1.66, 1.80, 0.98, and 0.22 mm/year in Punjab, East Rajasthan, Haryana, West Rajasthan, respectively. If a long term average annual declin e in rainfall is 3 mm and if 25 % of rainfall is assumed to contribute to groundwater recharg e, then groundwater decline associated with rainfall reduction would be at most 0.75 mm. Additionally, if the linear trends in annual rainfall are considered for the more recent time periods associated with the studies mentioned, then the rainfall trends become positive : 1.37 mm/year in Telangana (Medak study, 1996 2005) and 33, 41, 6, and 26 mm/year for the states in the northwest (Rodell study, 2002 2008). The Medak and GRACE studies are just two examples of numerous reports of groundwater depletion in India, and in these two regions based on the small increases in average annual rainfall totals during monitoring periods that demonstrated groundwater depletion, it seems unlikely that changes in rainfall are responsible for groundwater decline as rai nfall increased during the study periods. This suggests that groundwater pumping for irrigation is indeed the more probable cause of the reported groundwater depletion.

PAGE 30

30 Water Balance Simulation Methods and the Soil a nd Water Assessment Tool Soil and Wat er Assessment Tool Overview The Soil and Water Assessment Tool (SWAT ), Arnold et al., 1998; Gassman et al., 2007; Krysanova and Arnold, 2008), supported by the Agricultural Research Service (ARS) of the United States Department of Agriculture (USDA), was c hosen for modeling the water balance of the Wargal watershed for the purposes of estimating the groundwater balance. The model was used to estimate groundwater recharge and to upscale irrigation withdrawals. T he simplified groundwater balance, assuming n o net lateral sub surface flows, is recharge minus irrigation withdrawals. Hydrologic simulations have allowed the groundwater balance to be assessed under current and alternative farm management and have increased the evidence for the connections betwee n agricultural management and groundwater depletion SWAT is a continuous time, process based, distributed parameter hydrologic model that is used to estimate water quantity and quality at the landscape scale. The model is open source and supported rele ases are freely distributed. Geographic Information System (GIS) interfaces are available for preparation of landscape data. Water balance equations are solved for hydrologic response units (HRUs) that are developed based on combinations of soil type, sl ope, and land use/management. Watershed boundaries and subbasins are delineated from topographic data. Rainfall partitioning into runoff and infiltration is a highly important process description of any hydrologic model. SWAT has the advantage of allowi ng for the choice of the runoff/infiltration representation using the Curve Number (CN; SCS 1972) or Green Ampt Mein Larson (GAML; Green and Ampt 1911; Mein and Larson 1973) methods. It is suggested that it is preferable here to use the GAML method for the purpose of

PAGE 31

31 evaluating water harvesting tillage for groundwater recharge increases, as this sub daily time step method can respond to tillage management changes better than a daily time method given the episodic (high intensity) nature of rainfall in Wa rgal. Additionally, SWAT has been shown to perform well prior to calibration in numerous watersheds (Rosenthal et al., 1995; Bingner 1996 ) making possible the simulation of unga u ged or sparsely instrumented watersheds However, calibration does generall y yield substantial improvements in water balance predictions and was done in the study area using a time series of observed reservoir storage volume s Modifications have been made to SWAT code to include depression storage in the GAML infiltration routi ne for simulation of water harvesting tillage and rice croplands ; see C hapter 3 for details Literature values were used to parameterize tillage for maximum depression storage. To summarize, SWAT was chosen as the tool for hydrologic simulations based on : Physical basis for most process descriptions; spatial distribution of parameters Ability to simulate runoff and infiltration using GAML equations with sub daily precipitation data Incorporated, adjustable weather generator for long term simulations Demon strated performance in unga u ged watersheds (Rosenthal et al., 1995; Bingner 1996) GIS interface allow ing for fast development and input of spatial data O pen source code allowing for process modification Included sensitivity and uncertainty tools Demonstrat ed effectiveness at predicting recharge from land/water/atmosphere interactions at the surface: history of using SWAT output (recharge) as input to groundwater models (Kim et al., 2008)

PAGE 32

32 Summary of Water Balance Simulation Methods The methods employed here for the use of simulated water balance experiments closely follow the hydrologic modeling protocol proposed by Engel et al. (2007) for the purpose of improving the acceptability of modeling outcomes, remove bias of model users, developers, and increases th e repeatability of model application studies. Briefly, the procedure of the water balance experiments can be organized into six steps: 1. Collection and processing of landscape data Spatial data inputs to SWAT include a digital elevation model (DEM) for topograph ic representation, a land use/land cover (LULC) dataset for land cover and management description, and a soil map detailing all required soil properties. A DEM of the study area having 2.2 meter resolution was prepared using CartoSat IRSP5 remot e sensing stereo images (purchased from National Remote Sensing Centre, Department of Space, Gove rnment of India); see F igure 1 8 PCI Geomatics Orthoengine was used for image processing and DEM extraction. SWAT was used to delineate the watershed bounda ry based on topography, specifically the raster based flow direction grid that is developed from the DEM Two options were av ailable for LULC mapping: (1) u se only household survey data to create lumped LULC classes of arbitrary spatial distribution based on reported farm management practices or (2) use multispectral remote sensing (RS) images of CartoSat IRSP6 and supervised classification based on training areas from household surveys. It was decided to combine these options by using household survey dat a to obtain the extent of the major cropping sequences and using those areas to adjust the unsupervised classification of an RS Based on household surveys, the agricultural areas could be simplified to three crop ping systems. Based on 2008/2009 surveys of

PAGE 33

33 households in the watershed (N = 114), the dominant annual three crop rotations were : maize/potato/vegetable cotton/sunflower/vegetable and rice/rice/vegetable in seasons k harif/rabi/summer respectively Kharif season is from June to September, rabi season is from October to February, and summer season is from March to May. The e xtent of each crop in each rotation is presented in Table 1 1. Rice and vegetable croplands are the only irrigated areas Rice cropland areas were clearly distinguishable on the RS image, so there is little expected uncertainty in their extent and spatial distribution. Without an extensive time series of RS images and abundant groundtruth data, the rainfed crops of the watershed are remotely indistinguishable. However, th e extent of each crop can be assumed to be reliable because it is based on household surveys; it is only the spatial distribution of these areas that is uncertain. It is argued that uncertainty in location of these rainfed cropland areas is tolerable give n that there will be some interannual variability in spatial distribution of rainfed crops based on farmer decisions and there are small differences in hydrology (assuming uniform tillage) between these rainfed crops Rice cropland areas are more consist ent in their extent and spatial distribution because of the labor required to establish level, bund ed fields. It is arguably the rice areas that are of greatest hydrologic significance due to uniquely large irrigation, surface storage depth, evapotranspir ation, and deep percolation. Soil mapping was completed using soil data from 39 locations in the study area; sampling was done in 2 to 5 depth classes (to 1 meter maximum) at each point. Soil data includes saturated hydraulic conductivity ( K s ) % sand/si lt/clay, bulk density, and organic carbon. Two depth classes (0 28 cm and 28 94 cm) were extracted from the data for consistency at all locations. Weighted and harmonic means were used

PAGE 34

34 appropriately to aggregate data into the two uniform depth classes. Ordinary kriging using % clay in layer 1 of soil was used to interpolate point data to create a complete coverage. The choice of % clay was made over other soil properties because it showed the most spatial autocorrelation compared to that of the availabl e properties. This was decided based on visual inspection of semivariograms (plot of semivariance in property value in response to distance between observation locations); generally, small semivariance at low lag distances and larger semivariance at high lag distances suggests spatial autocorrelation. Semivariance is a spatial statistics quantity that describes the variance of some value (call it % clay) and % clay at some distance apart (lag distance). Semivariance is lag distance dependent as seen in F igure 1 9 See F igure 1 9 for the semivariogram of % clay. The result of kriging from point data was a continuous spatial distribution of % clay content throughout the watershed; 31 unique values of clay content resulted. To simplify the soil map k mean s clustering was used to group the 31 values of % clay into 6 groups ; associated soil properties were aggregated into the correct clay content classes using a nearest neighbor technique SWAT requir es a weather generator database for climate generation. C limate generation is required to allow for simulations to continue if there is a data gap or formatting error in weather input data; also, future scenario simulations require use of the weather generator database. The database can be adjusted to predict h ydrologic outcomes of climate change scenarios. This database of 168 climate parameters was developed using 34 years of daily climate data ( solar radiation, minimum and maximum temperatures rainfall) and 16 years of hourly rainfall data. Data used to ca lculate the required climate parameters were provided by the International Crops Research

PAGE 35

35 Institute for the Semi arid Tropics (ICRISAT; Singh, 2009 ). This was the nearest (45 km distant ) source of a long record of quality climate data, and inspection of a nnual average rainfall totals suggests there is little climate difference between ICRISAT and Wargal. Temperatures relative humidity and rainfall were measured in Wargal for input to SWAT for calibration and validation periods. Weather generator databas e location was listed in t he Wargal watershed as 17.76472 N, 78.62447 E and elevation of 600 m. 2. Design and installation of hydrologic/climate monitoring systems Climate data for SWAT input and watershed runoff data for model calibration and validation a re summarized in T able 1 2 Rainfall observations were collected manually and automatically. Automatic rainfall collection, using a tipping bucket gage that log ged each 0.25 mm of rainfall, allow ed for hourly precipitation data for GAML infiltration simu lation in SWAT. Sub daily data were generated from measured daily rainfall for wet days by HYETOS a program for disaggregation of daily rainfall data to hourly data ( Koutsoyiannis and Onof 2001 ) for periods when hourly rainfall data were unavailable ; th ere were 6 months in 2010 due to maintenance problems with the automated rain gage during which HYETOS was required to disaggregate daily rainfall observations into hourly data HYETOS uses the Bartlett Lewis pulse based method to generate storms from daily data based on 6 monthly parameters developed from a record of hourly data. Hourly data from nearby ICRISAT (Singh, 2009) were used to calculate the 6 Bartlett Lewis parameters. Disaggregated rainfall will not match actual hourly data, but it was t he best option available given the instrumentation challenges in the Wargal watershed. H YETOS disaggregated rainfall has been shown to closely match observations (Rodriguez Iturbe et al., 1988 ; Koutsoyiannis and Onof, 2001 ). Manual

PAGE 36

36 daily rainfall was co llected at 15 non evaporating gages d istributed throughout Wargal. Groundwater levels were monitored in 6 borewells and 3 open wells using an electronic depth sounder. One of the borewells was monitored continuously with a pressure transducer. Six rese rvoirs are found in the Wargal study area; these are water harvesting structures excavated long ago that serve to increase groundwater recharge. The largest of these reservoirs, Kothakunta, is found at the outlet of the watershed and creates a generally c losed watershed, meaning there is rarely outflow other than percolation of the tanks. Observations of water in Kothakunta reservoir were the data used for SWAT calibration and evaluation. The area and depth of water in Kothakunta reservoir, at the outlet of the study area, was monitored using GPS waypoint data and a staff gage that was manually read weekly or more frequently during rainy periods A GPS was used to weekly mark the perimeter of the flooded area of the reservoir by walking around the reserv oir at the water line and recording waypoints. Topographic survey data were used to develop the depth to volume and area to volume relationship s required to convert observations to storage volume. Increases in storage volume during a time interval are in terpreted as the outflow of the study area as this reservoir stores all outflow. available for data collection. It was decided that one target for calibration was sufficient gi ven the small size of the watershed; this is commonly done in hydrologic modeling. Also, recharge observations allowed an additional evaluation measure to further test the performance of SWAT.

PAGE 37

37 Irrigation pumping could be not be sampled in a large number o f fields, but 4 borewells were monitored with flowmeters to record irrigation pumping for a small sample of the most commonly irrigated crops: rice and vegetables. It is expected that for farmers with functional borewells there is little variability in ir rigation depths for a given crop due to the management of irrigation that is generally dependent only on electricity supply. Chapter 4 details the irrigation management setup in SWAT and presents the irrigation observations used. 3. Setup and initializati on of hydrologic model SWAT setup consisted of input of formatted landscape data (soils and topography and land use) and w eather data Cropping system management was parameterized through tillage specification, planting dates, crop selection, and irriga tion management. An initialization period of about a year and half (using some simulated and some measured weather data) was used to establish more suitable initial soil moisture and reservoir level conditions. 4. Model calibration and evaluation This s tep provided quantitative information on the uncertainty of simulated water balance components at the watershed scale using observed reservoir storage volume at the outlet of the watersh ed as the calibration target. 13 of the most important p arameters b ased on expert knowledge and sensitivity analyses of SWAT in the literature were adjusted to improve accuracy of prediction of outflow ( Chapter 4) Based on the recommendations of Moriasi et al ( 2007 ) the 3 quantitative statistics being used for model calibration and evaluation were the Nash Sutcliffe efficiency (NSE), percent bias (PBIAS), and the ratio of the root mean square error to the standard deviation of measured data (RSR).

PAGE 38

38 5. Water balance simulations of alternative cropping systems This st ep completed the simulation experiments by estimating the resulting water balance components of cropland management combinations in Wargal. Alternative farm management scenarios were created by combining tillage, crop selection, and irrigation management options, ba sed on accepted water conservation strategies and participation of community members; these alternative management options are detailed in Chapter 4. A 10 year simulation using generated weather data was used to develop water balance prediction s for existing management and also for all a lternative management options. 6. Evaluation of alternatives Groundwater recharge, evapotranspiration (ET) surface runoff, and irrigation withdrawals were among the outputs used to select management strategy c ombinations of greatest potential for groundwater quantity improvement. Chapter 4 provides details on the comparison of management alternatives. Hydrology in Ungauged Basins The International Association of Hydrological Sciences (IAHS) has launched a deca de long program (2003 2012) called Predictions in Ungauged Basins, or PUB (Sivapalan et al., 2003). The central goal of this initiative is to reduce predictive uncertainty of hydrologic models by improving hydrologic process understanding developing ways to generalize hydrology among similar watersheds, and increasing the available hydrologic and landscape data. This research contributes to the goals of PUB by improving a hydrologic model through increasing the physical basi s of the infiltration routines with the addition of a parameter which can be expected to have much lower uncertainty and variability than the other soil properties important for infiltration

PAGE 39

39 prediction : effective hydraulic conductivity and wetting front suction. Also, the use of small surface reservoirs as runoff gages increases the evidence for this being an effective strategy for acquiring data for model calibration and evaluation in the absence of streamflow measurements. Factors Influencing Groundwater Recharge: Literature Review O f the roughly 35 million km 3 of freshwater on Earth, 69% is frozen in glaciers and permafrost, and of the remaining 31% of global freshwater resources 96% occurs as groundwater the rest is divided between surface water and soil moisture (Shiklomanov 200 0 ). An important objective of this research is to understand how the groundwater balance can be managed through reductions in irrigation withdrawal s and increases in recharge through tillage management. The following sections review a variety of literatur e on groundwater recharge to assess which factors have the greatest influence on recharge. Groundwater recharge is primarily controlled by climate, topography, soil properties, land use/land cover, and geology (Rushton, 1988; Le Maitre et al., 1999; Delin et al., 2000; de Vries and Simmers, 2002; Lin et al., 2007). The following review of literature attempts to select the dominant factor( s) for a given climate/region. The strength of correlations between groundwater recharge and soil properties, topograp hy, geology, and land use/ land cover (LULC) are compared qualitatively, from a review of research on groundwater recharge estimation and prediction. Rationale and Significance of a Groundwater Recharge Literature Review The two main issues having manage ment implications associated with groundwater recharge are groundwater contamination (quality) and groundwater depletion or overdraft (quantity) (Lerner et al., 1990). Groundwater is the main source

PAGE 40

40 of water for agricultural, municipal, and industrial use s in some parts of the world; th erefore, the quantity and quality of groundwater is of great importance and is the focus of management efforts. Cities or villages that depend on groundwater for domestic supplies are vulnerable if groundwater supplies beco me of insufficient quantity or quality; water has to be imported or purified to meet domestic supplies. The same is true for farming systems and industries that depend on groundwater. Groundwater can be managed by changing water fluxes at the land surfa ce by altering soil properties (organic or mineral amendments), topography (water harvesting ponds, tillage), or LULC (impervious area, types of vegetation). Subsurface flows can also be altered by modifying geology through recharge and discharge structur es that penetrate confining or slow flowing formations; groundwater withdrawals for agricultural, municipal or industrial supply do of course influence subsurface flows. It is sometimes desirable to increase groundwater recharge to meet increased withdra wal demands, and it is sometimes the goal to decrease groundwater quantity to prepare for construction or relieve waterlogging and/or salinity problems in croplands. Managing anything well requires information about the quantity of interest and the releva nt processes; therefore, it is helpful to know which factors related to groundwater recharge are the most important to consider. To summarize, the objectives of this review are: Compare correlations of the four manageable factors soil properties, topogr aphy, geology, LULC to groundwater recharge Determine if there is consensus in the literature on the factor that dominates groundwater recharge (locally) Find methods of recharge measurement/estimation that are most regularly used

PAGE 41

41 Land Surface Environme ntal Factors and Groundwater Recharge With a focus on the four manageable, environmental factors being considered, three synthesis papers were selected to evaluate the prevalence of factors considered for groundwater recharge. Two of the reviews were sele cted to approach groundwater recharge from differ ent perspectives. The f irst perspective is that of vulnerability assessment, contamination, and water quality (Gogu and Dassargues 2000); the second perspective is that of recharge processes, estimation, a nd quantity (de Vries and Simmers 2002). The third review covers an extensive range of study areas and summarizes recharge rates and controls among them. Qualitative comparisons are made of these reviews to examine the frequency of consideration of the four factors in the studies included in each review. Additionally, nine studies of groundwater recharge relationships to the three most manageable of the four factors (soils, topography, LULC) were sampled and reviewed for comparison. These papers (publi shed in the last 10 years) were summarized and their methods and outcomes tabulated for comparison. Groundwater quantity and quality are well connected because the same factors that make a groundwater system vulnerable to quality degradation can improve groundwater recharge or quantity; areas of greater groundwater recharge are the areas that are most vuln erable to contamination. Ten methods of vulnerability assessment are reviewed by Gogu and Dassargues ( 2000 ) ; references and variables included for vul nerability assessment are given in Table 1 3 The outcome of a vulnerability analysis is a quantitative measure (often a map) of groundwater vulnerability to contamination; however, the measure of vulnerability is just a number some type of index that is relative to other vulnerabilities predicted with the same method. There is no absolute vulnerability quantification. The most thorough

PAGE 42

42 methods are the DRASTIC and SINTACS approaches which both use the following seven parameters for vulnerability class ification: depth to water, net recharge, aquifer media, soil media, topography, impact of the vadose zone, and hydraulic conductivity. The review by de Vries and Simmers ( 2002 ) suggests climate, geology, topography, soil condition, and vegetation are the factors determining groundwater recharge. Geology control of recharge is illustrated by the case of recharge differences in Botswana between the Kalahari sands and the adjacent Precambrian hard rock area. Rainfall in the two areas is similar (300 500 m m/year); tracer observations and modeling estimates show similar recharge in the Kalahari sands (1 10 mm) and hard rock areas (10 30 mm). However, spatial distribution of recharge is very different between the two areas. The hard rock area shows about 50 % of recharge occurring through preferential flow from fractured rock, creating small (5 10 km 2 ) groundwater basins that often coincide with morphologic (read: topographic) depressions. Groundwater recharge in the Kalahari sands is more spatially uniform, having less variability of soil moisture in the vadose zone and little geologic control of recharge. A very thorough review of 98 groundwater recharge studies reported in the literature was completed by Scanlon et al. (2006) focusing on recharge in arid and semi arid regions. Their conclusions are summarized in the following three sentences. The chloride mass balance technique is the most widely used method for estimating recharge. LULC change is the factor most regularly controlling recharge: in Niger LULC changes resulted in recharge responses that greatly exceeded those resulting from climate variability, including severe droughts; in Australia, deforestation has resulted in recharge increases of 2 orders of magnitude, causing severe groundwater sali nity

PAGE 43

43 problems due to the flushing of the large amounts of chlorides present in soils. The authors suggest that high sensitivity of groundwater recharge to LULC means that recharge can be managed in most cases by LULC changes. Land use and land cover Grou ndwater recharge in arid and semiarid regions is typically small relative to other water balance components. Rainfall distribution in these areas is characterized by marked wet and dry periods; the wet periods having storms of significant intensity. The result of these precipitation patterns is that groundwater recharge is largely episodic (de Vries and Simmers 2002). Episodic recharge describes groundwater recharge events of low frequency and large relative contributions. This type of recharge is high ly sensitive to topography and difficult to manage through vegetation changes (de Vries and Simmers 2000; Zhang et al., 1999). The conversion of native vegetation to annual cropping systems in southeastern Australia has increased groundwater recharge a nd caused salinity problems as a result of rising water tables. To assess the impact of agronomic practices (vegetation type) on groundwater recharge, Zhang et al. ( 1999 ) used a biophysically based water balance model (WAVES) at two sites in Australia to simulate water balance components in response to changes in crop rotation and type. Their simulations showed that perenializing croplands and growing crops having greater rooting depth did decrease occurrence of small recharge events and overall recharge Vegetation management was able to reduce annual recharge amounts by elimination of fallows and increased root depths. At one location, simulating two different rotations, a change to a deeper rooting variety (0.5 m depth to 1.0 m depth) resulted in redu ction in annual recharge of

PAGE 44

44 50%. The main conclusion was the recommendation that fallows be replaced with vigorously vegetative crops to reduce recharge. Water flows in the Dill Catchment, Germany were simulated to estimate sensitivity of flows to vegetat ion and soil property changes associated with vegetation changes (Huisman et al., 2004). Water flows considered were total runoff, actual evapotranspiration, surface runoff and ground water recharge; vegetation scenarios were pasture (perennial) and cropl and (annual): cropland was a common rotation of summer barley, winter rape, and winter wheat. A quantity called distinction level was used to compare hydrologic responses to vegetation change. The distinction level helps compare the sensitivity of the si mulated hydrologic fluxes to soil and vegetation parameter changes relative to the uncertainty in these fluxes due to parameter uncertainty. The results showed very little sensitivity of hydrologic fluxes to soil properties (saturated hydraulic conductivi ty, bulk density, available water content, depth of top soil layer). For the study area, groundwater recharge was little affected by changes from cropland to pasture because the opposite effects on runoff and evapotranspiration. Higher evapotranspiration and lower runoff amounts resulted from shifts of cropland to pasture (replacing 50% of cropland with pasture runoff: 596.9 to 536.0 mm/yr; ET: 292.9 to 352.8 mm/yr). Groundwater recharge and baseflow were estimated for the Upper Mississippi River basin (Arnold et al., 2000) using the Soil and Water Assessment Tool (SWAT) and two data analysis methods (recursive filter to separate baseflow from daily flow and hydrograph recession displacement to estimate groundwater recharge). Groundwater recharge, runof f, baseflow, and ET were most sensitive to (in order of decreasing

PAGE 45

45 sensitivity): Curve Number, soil available water capacity, evaporation compensation coefficient (adjusts the depth distribution of evapotranspiration from the soil). The greatest sensitivi ty to Curve Number is not surprising as this quantity contains considerable information about land use/land cover. Topography It has been shown by Delin et al. ( 2000 ) that even small differences in topography can substantially influence groundwater rechar ge. Their research site in the sand plains of central Minnesota was planted in maize under uniform tillage, eliminating most of the effects on recharge from land use and microtopography differences between sites. An upland and lowland site were selected, separated by 78 m of distance and 1.4 m of land surface elevation difference. Three methods for recharge estimation were used: well hydrograph analysis, unsaturated zone water balance, and chlorofluorocarbon tracer dating. Mean annual recharge at the si te of lower elevation was greater than recharge at the site of high er elevation by 30 % (estimate d from hydrograph analysis), 60 % (estima ted from water balance), and 80 % (estimated from chlorofluorocarbon dating). These results from a diversity of methods during a 4 year study suggest that small differences in topography (1.4 m) can be significant factors in estimation of groundwater. The resolution of the representation of topography can significantly impact predictions of land surface processes (Schoorl et al., 2000). Digital elevation models (DEMs) of the same terrain using different resolutions can result in result in different predictions of landscape properties or processes. Computational and sensing costs currently preclude very high resolution DE Ms, but the microtopographic relief that they represent does influence land surface hydrologic processes. The effect on hydrologic

PAGE 46

46 simulations of runoff in response to centimeter scale representations of a landscape were considered by Martin et al. ( 2008 ) ; interpolated at resolutions of 0.75 cm, 10 cm, 25 cm, 50 cm, 75 cm and 100 cm for vegetated and bare surfaces of a 20 m x 10 m slope in British Columbia, Canada. Elevations were obtained from a ground based laser scanner. Their results showed marked de creases in mean depression storage (MDS) for the landscape with increasing DEM grid size From 0.75 cm resolution to 25 cm resolution, MDS decreases from 2.4 cm to 1 cm, and is negligible at resolutions above 25 cm. They demonstrate the infiltration/runo ff partitioning changes in response to varying MDS, making clear the implications for groundwater recharge. The strength of topography as a predictor of groundwater recharge was quantified by Bakhsh and Kanwar ( 2008 ) in their study of landscape attributes and their correlation to subsurface drainage clusters. Flow data from measurements of subsurface drainage (pipes below agricultural fields) was used to find the landscape attributes (soil type, elevation, slope, aspect, curvature, flow length, flow direc tion, and flow accumulation) that best explained flow variations. Thirty six plots in a field each ha d their own drainage outlet and subsurface flow was measured from each plot. While subsurface flow was occurring laterally through drainage pipes, it was essentially deep drainage or potential groundwater recharge that was being intercepted. Stepwise discriminant analysis was used to select the attributes that made the biggest contribution to subsurface flow. Elevation and flow accumulation were both sel ected as the attributes that best explained the spatial variability of subsurface flows. Three soil types were classified for the site; saturated hydraulic conductivity was considerably

PAGE 47

47 different among them (35 100 mm/hour), but the analysis only select ed topographic variables as having strong correlation with subsurface drainage flows. Soil properties Sensitivity analysis of groundwater recharge to numerous soil and vegetation parameters was completed using a simple one dimensional water balance with weather and vegetation information for England (Finch 1998). One at a time sensitivity analysis was used: all parameters being held at their mean values while the parameter of interest varies across its specified range. Numerous vegetation parameters we re considered for three vegetation types and mean annual recharge was estimated at 176.7, 290.5, and 96.4 mm for permanent short vegetation, annual short vegetation, and forest, respectively. Water balance simulations show ed greater sensitivity of ground water recharge to soil parameters (available water content, root zone depth) than to vegetation parameters (leaf area index, canopy height, growing season length) The simplified water balance model estimates runoff as 10% of rainfall if a rainfall event is greater than 5 mm; topography and a physically based description of infiltration/runoff processes were ignored for the purposes of the study. A method for estimating groundwater recharge in southern India was tested by Anuraga et al. ( 2006 ) using a soil water atmosphere plant model able to describe land use and soil types. Soil data for the study area of Bethamangala subwatershed was simplified into two classes: one sandy loam (Type 1) and one clay (Type 2). Annually, average recharge for the soils was 212 and 99 mm for the sandy and clayey soils, respectively; this difference is claimed to be due to the higher evaporative losses from the clay soil as a result of its higher water holding capacity. Annual water balances for three cropping intensities we re simulated. Type A: Finger millet, potato, and tomato are

PAGE 48

48 cropped during kharif, rabi, and summer seasons, respectively: total amount of irrigation is 410 mm/year. Type B: Finger millet is cropped during kharif season with supplemental irrigation of 15 0 mm/year. Type C: Finger millet is cropped during the kharif season without irrigation. Recharge for sandy and clayey soils, respectively, for cropping systems were 220 and 90 mm/year for system A, 232 and 124 mm/year for system B, and 184 and 84 mm/yea r for system C. This suggests that s oil texture influence on recharge is strong. A single layer soil water balance model, developed by Eilers et al. (2007) was used to estimate groundwater recharge during a 36 year period in semi arid northeast Nigeria. The model includes 11 soil and 13 vegetation parameters that are not spatially distributed. Simplicity and low data requirements of the mod el are touted as its strengths, and it is promoted for use in cropland areas for water management purposes. The si mulated water balance was used to find sensitivity of groundwater recharge to model inputs and parameters. Total available water (TAW), followed by rooting depth and then by crop coefficient, was found to be the parameter with the greatest effect on simul ated recharge. TAW depends both on soil properties and plant rooting depth. N ear surface storage (N SS ) depth was also shown to significantly influence recharge Quantifying Groundwater R echarge Methods for quantifying groundwater recharge can be grouped into 3 categories: physical, tracer, and numerical modeling (Scanlon et al., 2002). These methods are used differently in surface water based studies, unsaturated zone studies, and saturated zone studies. Table 1 4 summarizes the various methods for mea suring and estimating groundwater recharge. In this study in the Wargal watershed, unsaturated zone water balance modeling is used to estimate groundwater recharge using a

PAGE 49

49 distributed solution to the water balance equation at the surface and in the unsatu rated zone. Deep drainage, water percolating below the root zone is assumed equivalent to groundwater recharge. This equivalency assumption is valid based on the limited surface water/groundwater connections and the substantial depth to groundwater. Seep age meters, lysimetry, and tracers all only give point estimates, and sufficient point estimates to distribute groundwater recharge in the 500 ha study area would be too costly. The water table fluctuation method would not be effective because of the amou nt of groundwater pumping and number of wells (190); it would not be possible to have a sufficient number of groundwater wells that were uninfluenced by groundwater pumping. Discussion a nd Conclusions o n Factors Influencing Groundwater Recharge The studie s considered in this review span a range of locations and climates, making it difficult to make a rigorous comparison of correlations (groundwater recharge to soil properties, topography, geology, LULC). This review can, however, suggest environmental fac tors having the greatest correlation to recharge in a certain region. Table 1 5 summarizes the papers reviewed in this section : giving their location, the factors considered, and the factor explaining most of the variability in groundwater recharge. The dominance of simulated water balances in experimental methods is not uncommon in groundwater recharge studies, and it affirms the use of water balance simulations in the Wargal study area. Measured water balances are difficult as determination of boundary conditions of a groundwater system generally requires vast instrumentation for measurement of surface and subsurface quantities (including descriptions of geology) both in the study area and in surrounding areas. Tracer mass balances seem to be the most effective way of estimating local recharge; seepage

PAGE 50

50 meters can also provide accurate point estimates of recharge. The prevalence in the literature of simulation tools for estimating sensitivity of groundwater recharge to environmental factors supports the use of SWAT in this analysis to estimate groundwater recharge under current and modified management. This review and the extensive review by Scanlon et al. ( 2006 ) suggest that land use and land cover has the most control over recharge processes within a g iven climate. In 4 of the 9 studies tabulated above, a quantity related to LULC dominates groundwater recharge control. In 3 of the 9 studies, a topographic quantity was found to be most influential on groundwater recharge. Of the 6 studies that conside red LULC there were 2 which find a soil property having a greater influence on recharge. For agricultural areas, the sensitivity of recharge to LULC suggests that cropping system changes can have significant effects on groundwater recharge. Options for Ma naging Gro undwater i n Agricultural Systems Recall the water balance discussion in the earlier section Groundwater Balance and Agricultural Management Equation 1 1 described the water balance of the unsaturated zone and at the soil surface: I ET RO where SW is soil water, P is precipitation, I is irrigation, ET is evapotranspiration, RO is in small or large depressions, DP is subsurface flow into and from the system. Equation 1 2 described the water balance i n the saturated zone (groundwater system): I where S is groundwater storage, R is recharge (approx. = DP), net change in flow into and from the groundwater system. As discussed earlier, if the

PAGE 51

51 goal is to improve or maintain S (groundwater storage), then it is essentially irrigation and r echarge that remain as the manageable quantities in the groundwater balance. These quantities appear in the surface water balance as I and DP. Decreasing irrigation pumping (I) could be done through a variety of possible management changes: providing sup plemental rather than full irrigation, irrigating only at critical phenological times, changing crop planting date to increase effective rainfall, choosing crops with lower water requirements or improving application technology. At the farm scale, effort s to increase recharge (DP) include: excavating farm ponds, using zero or reduced tillage, mulching, and tillage for water harvesting (tied ridging, contour bunding, and others). All these techniques can increase groundwater recharge by increasing surface water storage depth which allows for longer infilt ration times of excess rainfall. The above review of groundwater recharge studies highlights the difficulties associated with measurement of groundwater recharge and the prevalence of water balance models for use in simulation of recharge. Contributions of this Research There are four major contributions that make this research significant. (1) The i mprovement of the GAML infiltration process description in SWAT to include surface storage depth as a tim e varying head and as a storage term at the surface. This allows for evaluation of water harvesting tillage and enables flooded croplands (rice) to be simulated with the use of lup.dat ) which allow for dynamic cropland ar eas to represent management changes and intensive cropping sequences This improved infiltration description has a stronger theoretical basis than the typical GAML formulation used in SWAT and similar hydrologic models, and sensitivity an alyses ( Chapter 3 ) have shown that infiltration predictions are sensitive

PAGE 52

52 (rank 2 of 5 parameters) to the parameter describing maximum depression storage in systems having non negligible surface storage depth. (2) The use of lup .dat in SWAT for sequences of multiple crops in a single year, common in semi arid and tropical ecosystems, and for imple menting cropping system changes is unique in its application here Typically this functionality is used for long term (decade scale) and broad land use change; for example, conve rsion of unmanaged forests to annual croplands. Here, it is used for the first time (according to lead SWAT model developer : Arnold, 2010) to model these intensive cropping sequences. (3) The use of reservoir storage volumes for calibration and evaluatio n of model predictions in a small watershed with little convergent surface flow is an innovative strategy for data acquisition and model a pplication in an ungauged basin ( Chapter 4; Liebe et al., 2009) Small reservoirs for irrigation use and groundwater recharge are common in India and in some areas of Africa; t his research demonstrates an effective inexpensive way to use these surface water bodies to calibrate and evaluate hydrologic models in watersheds without records of streamflow. (4) This research quantifies the impact of rice croplands on the groundwater balance at the landscape scale in a typical semi arid Indian watershed. In recent years there has been growing awareness, publicly and in the scientific literature, that groundwater depletion is becoming a problem in India ( CGWB 2007; Rodell et al., 2009 ; Shah 2007), but there is no information on the extent of irrigated croplands (largely rice) allowable for groundwater sustainability.

PAGE 53

53 Figure 1 1. Study area location in Wargal mandal, easte rn Medak district, northwestern Andhra Pradesh

PAGE 54

54 Figure 1 2 Groundwater system diagram of Wargal (adapted from Dewandel et al., 2006 and Marechal et al., 2006). t is layer thickness, values are approximate Water table height fluctuates between 15 and 35 m below surface.

PAGE 55

55 Figure 1 3 Water balance diagram showing connections between unsaturated and saturated zone water balances. I is irrigation, P is precipitation, SS is surface storage, Inf is infiltration, ET is evapotranspiration, RO is runoff, U F is unsaturated flow, CR is capillary rise, R is groundwater recharge, DP is deep percolation, SF is saturated flow

PAGE 56

56 Figure 1 4 Total annual harvested areas of rice in I ndia and Andhra Pradesh, 1961 2001 Figure 1 5 Area equipped for mechanized i rrigation from groundwater source in India: 1961, 1971, 1981, 1986, 1993; population of India each year from 1961 to 2007 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 5 10 15 20 25 30 35 40 45 50 1960 1970 1980 1990 2000 AP rice area, Mha India rice area, Mha India rice cropland area Andhra Pradesh rice cropland 0.46 1.16 7.30 11.90 17.70 20.60 26.54 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1960 1970 1980 1990 2000 Groundwater irrigated area, Mha Population of India, billion population of India groundwater irrigated area

PAGE 57

57 Figure 1 6 Annual rainfall in Telangana region (1960 2008): slight declining trend observed (about 3 mm/year reduction in annu al rainfall if trend is assumed linear) Figure 1 7 Annual rainfall trends (1960 2008) in the 3 state region of the GRACE study of Rodell et al., 2009 y = 2.8424x + 6552.6 0 200 400 600 800 1000 1200 1400 1960 1970 1980 1990 2000 2010 Annual rainfall depth, mm year Telangana annual rainfall 0 200 400 600 800 1000 1200 1960 1970 1980 1990 2000 2010 Annual rainfall depth, mm year Punjab annual rainfall Haryana annual rainfall E. Rajasthan annual rainfall W. Rajasthan annual rainfall

PAGE 58

58 Figure 1 8 Watershed boundary, reservoir locations (large dots), and d igit al elevation model o f Wargal watershed, 2.2 m resolution Table 1 1. Areas (ha) of cropland for three crop rotation in Wargal kharif rabi summer ha ha ha maize 65.96 potato 40.96 beans 10.30 cotton 65.14 sunflower 23.36 beans 10.30 rice 92.63 rice 60.60 beans 1 0.30 range 288.41 range 387.21 range 481.23

PAGE 59

59 Figure 1 9 Variogram plot of % clay in layer 1 of Wargal soil: used to evaluate spatial autocorrelation of soil properties and for kriging to generate continuous map of soil data

PAGE 60

60 Table 1 2. Clim ate and hydrologic monitoring systems: observation numbers, frequency, purpose Observation Collection method and number of observations Frequency Purpose Rainfall continuous, automated (1) and manual (15) bi weekly (continuous) and weekly (manual) input for hydrology model Weather Parameters (Solar radiation, wind speed, and temperature) manual (1) daily input for hydrology model Reservoir depth and area staff gage: manual reading (1) weekly (daily during rainy periods) calibration of hydrology model Drain flows to tanks water level recorder in flumes: continuous, automated (2) weekly calibration of hydrology model Groundwater levels pressure transducer: continuous, automated (1) and sounder: manual (9) bi weekly (continuous) and weekly (manual) calib ration of hydrology model Irrigation pumping flowmeters on borewells: manual reading (4) weekly input for hydrology model Table 1 3. Sources and factors included in analysis by Gogu and Dassargues; adapted from Gogu and Dassargues 2000 Reference and m odel name Parameters Topography Soils Geology LULC Albinet and Margat 1970 Goossens and Van Damme 1987 Carter and Palmer 1987 Foster 1987, GOD Aller et al., 1987, DRASTIC Navulur and Engel, SEEPAGE Civit a 1994, SINTACS Van Stempvoort et al., 1993, AVI Civita and De Regibus, 1995, ISIS Doerfliger and Zwahlen 1997, EPIK

PAGE 61

61 Table 1 4. Summary of methods for quantifying groundwater recharge Physical Tracer Numerical Modelin g Surface water Based Studies channel water budget, seepage meters heat, isotopic water balance simulation Unsaturated Zone Studies lysimetry, zero flux plane, Darcy's law applied, historical, environmental water balance simulation Saturated Zone Studie s water table fluctuation, Darcy's law historical, environmental water balance simulation

PAGE 62

62 Table 1 5. Summary of the 9 recharge studies highlighting dominant factors Study area; Reference Factors: *most important Experimental methods so uthern India; Anuraga et al., 2006 soil*, LULC (cropping system) distributed simulation Upper Mississippi River Basin; Arnold et al., 2000 soil, LULC* distributed simulation Nashua, Iowa; Bakhsh and Kanwar, 2008 soil, topography (elevation*, slope, aspec t, curvature, flow length, flow direction, flow accumulation*) drain flow measurement Central Minnesota corn field, Delin et al., 2000 topography(elevation*), soil simulation, tracers, well hydrographs Wallingford, UK; Finch, 1998 soil, LULC (LAI*, stom atal resistance, rooting depth, AWC) local simulation northeast Nigeria; Eilers et al., 2007 soil (total available water*), LULC (rooting depth, Kc) local simulation Dill catchment, Germany; Huisman et al., 2004 soil, LULC (cropping system*) distributed simulation British Colombia, Canada; Martin et al., 2008 topography (DEM resolution*) distributed simulation Mallee, SE Australia; Zhang et al., 1999 LULC (rooting depth*, fallowing) distributed simulation

PAGE 63

63 CHAPTER 2 RAINFALL INTENSITY DURATION FREQUEN CY RELATIONSHIPS FOR ANDHRA PRADESH, INDIA: CHANGING RAINFALL PATTERNS AND IMPLICATIONS FOR GROUNDWATER RECHARGE Rainfall Characterization and Water Resource Management The development and maintenance of rainfall intensity duration frequency (IDF) relation ships is important for a variety of water related design and management, including flood control, energy generation, water supply, agricultural drainage, and others. With growing competition for freshwater resources in much of the world, the role of update d IDF relationships for application in developing innovative water management strategies is becoming increasingly important. Improved characterization of rainfall, through updated IDF relationships, has been shown to improve water resource planning and ma nagement decisions (Karl et al. 1995; Angel and Huff 1997; Guo 2006). For example, a case study of Chicago urban drainage systems showed that drainage systems designed using updated IDF relationships, using shorter records of more recent data, performe d significantly better than those developed from older rainfall records (Guo 2006). There is mounting evidence from global and regional studies that precipitation patterns are shifting toward more common higher intensity storms and fewer light and moderat e events (Kunkel et al. 1999; Easterling et al. 2000; Trenberth et al. 2003; Goswami et al. 2006; Joshi and Rajeevan 2006). These observed changes in rainfall characteristics suggest that IDF analyses be regularly updated to include more recent and s horter records of rainfall time series and exclude older, less representative data. The analysis of Trenberth et al. (2003) notes that the Clausius Clapeyron equation relating vapor pressure and temperature suggests a 7% increase in atmospheric water

PAGE 64

64 cont ent for each C increase in average annual temperature. As a result of low level moisture convergence, local rainfall rates greatly exceed average regional or global evaporation rates; therefore, rainfall intensities could be expected to increase at a rat e at least as large as 7% / C. However, this differs from the accepted 1 2% / C increase in total annual precipitation depths (IPCC, 2001). To reconcile the differences in these predictions, it follows that low and moderate intensity precipitation ev ents will be less common, and precipitation would trend toward less frequent, higher intensity events (Trenberth et al. 2003). This argument is supported by global climate model predictions ( the UKHI and CSIRO9 general circulation models ; Hennessy et al. 1997) and by an investigation of rainfall records for the large region of central India (Goswami et al. 2006). Of course the spatial distribution of this change in precipitation character is uncertain, but it would mean greater risk of both dry spells a nd floods for some regions even though annual precipitation totals may increase slightly. The recent rainfall character study (Goswami et al. 2006) of a large central Indian region used analyses of daily rainfall data from 1803 stations (1951 2000), and found significant increases in frequency and magnitude of high intensity rain events (>100 mm/day) and significant decreases in frequency of light and moderate events (>5 and <100 mm/day). The authors observed that these trends are more difficult to noti ce based on analyses from individual station data due to the large variability in daily data. The regional analysis, however, having much larger sample size due to the large number of stations, is better able to detect long term trends in rainfall intensi ty. Groundwater Resources in India Groundwater in India is a highly important resource for irrigation and household use, and its extensive use is resulting in widespread groundwater depletion ( Chapter 1

PAGE 65

65 Shah et al. 2003; CGW B 2007; Rodell et al. 2009). There is general agreement that water scarcity in India is severe (Alcamo et al. 2000; Yang et al. 2003). Total (761 km 3 ) and agricultural (688 km 3 ) water withdrawals for India are the highest in the world, and nearly 90% of withdrawals are for agricultural use (AQUASTAT, 2010). More than half of the irrigation requirements of India are met from groundwater (CGWB, 2002; Shah et al. 2003), and the number of mechanized borewells in India has increased from less than 1 mill ion in 1960 to more than 20 million in 2000 (Shah, 2007). Groundwater in India is a highly important resource for irrigation and household use, and its extensive use is resulting in widespread groundwater depletion (Shah et al. 2003; CGWB 2007; Rodell e t al. 2009). A recent groundwater depletion study (Rodell et al. 2009) in the northwestern Indian states of Haryana, Punjab, and Rajasthan, is illustrative of common regional groundwater depletion problems in India. Using the Gravity Recovery and Clim ate Experiment (GRACE) satellites to measure changes in terrestrial water storage during the study period from August 2002 to October 2008, 109 km 3 of groundwater loss was estimated, or about 4 cm each year over the three state area. In the southern state of Andhra Pradesh, the Central Groundwater Board (CGWB) of India observed a 2 4 m decline in groundwater levels in 26 observation wells in the Medak district during the 10 years from 1996 to 2005, suggesting an approximate annual decline of about 30 cm (CGWB 2007). Objectives: Precipitation Characterization and Groundwater in India Considering the importance of groundwater resources in India, it is likely that if rainfall becomes characterized by more common events of high intensity, the result would be a higher fraction of rainfall contributing to runoff and a reduced fraction

PAGE 66

66 available for infiltration and groundwater recharge (de Vries and Simmers 2002; Gujja et al. 2009; Gupta et al. 2010, Rangan et al. 2010). This may put the already reduced strained groundwater resources of India at even greater risk of depletion. Several states in northwestern and peninsular India are experiencing groundwater depletion (CGWB 2007, Rodell et al. 2009). In Andhra Pradesh, for example, depleted groundwater supply, resulting from irrigation withdrawal expansions, has severe consequences for farmers that have invested in borewell infrastructure. Farmers bear substantial costs of groundwater depletion: greater yield variability, costs of failed borewells, and the expenses to develop new borewells. Updated IDF relationships that reflect the changes toward more episodic rainfall in Andhra Pradesh serve to emphasize the importance of surface storage depth as a management option for increasing infiltration and gro undwater recharge in agricultural areas An area in northern Andhra Pradesh, in the Medak district is studied here as a case study to demonstrate the advantages of updated IDF relationships for a region in India that is facing acute water shortages. The ob jective of this study is to develop updated IDF relationships for Hyderabad Andhra Pradesh, India. These updated rainfall descriptions are then used to demonstrate the changes in precipitation intensity and the need for regularly updated rainfall charact erizations using recently available sub daily rainfall records. The updated IDF relationships are compared to earlier analyses for the region, and the effects of changing rainfall character on groundwater resources in India are discussed. These IDF curve s can be useful tools for numerous types of water resource management projects in Hyderabad (a major city in India) and the surrounding region. For example, in a small agricultural watershed near Hyderabad,

PAGE 67

67 the causes and solutions of groundwater depletio n, a common regional problem, are be ing analyzed (Chapter 4 ), and these new IDF relationships are being used to improve predictions of infiltration (Chapter 3) and groundwater recharge. Rainfall Characterization Overview of IDF Analysis Generally, the de velopment of IDF curves follows four main steps. First, rainfall intensity data are organized into an annual maximum series. This is done for each duration of interest (1 hour, 4 hour, 8 hour, etc.) by finding the maximum rainfall intensity for the durat ion specified for each year. Second, a probability distribution is fitted to the annual maximum series using any choice of statistical techniques (maximum likelihood, l moments, or other). Third, the cumulative distribution function (CDF) chosen and para meterized in step two is used to calculate rainfall intensities from the frequencies (1/probability) desired (2 year, 5 year, 10 year, etc.) for each of the durations being considered. Fourth, the curves can be fitted to a parametric equation; this final step is optional and is useful if it is desirable to avoid using multiple CDFs for rainfall IDF prediction or if IDF curves are to be estimated for durations or frequencies not in the period of record. A common parametric equation form for IDF curves is ( Bernard, 1932): (2 1) where I t T is the rainfall intensity of a combination of T (frequency or return period, years) and t (duration, hours) and empirically calculated constants A1, A2, and A3. Limited IDF information is available for India. The study of Kothyari and Garde (1992), using rainfall data (1950 1980) from 80 recording gages grouped geographically

PAGE 68

68 based on rainfall characteristics, provided IDF curves for 5 large regions of India: northern, central, western, eastern, and southern. The Kothyari and Garde (1992) IDF relationships are the most current national IDF characterizations for India available in the literature, their IDF relationships will be used for comparison to those developed here. Their analysis found that the performance of E quation 2 1 (Bernard, 1932) could be improved by including some rainfall characteristic (Rchar) in the expression. Four rainfall properties were considered as candidates for inclusion in the equation: where I t T is the rainfall intensity of a combination of T (frequency or return period, years) and t (duration, hours), Rchar is some rainfall characteristic, and C1, C2, C3, and C4 are constants fitted to IDF data generated from the CDF w hich was fitted to observed data. Options considered for Rchar were mean annual rainfall (R), mean of the maximum monthly rainfall (Rmax), ratio (R/Rmax), and the 24 hr duration, two year return period rainfall depth (R 24 2 ) were used. It was found by com paring multiple regression correlation coefficients that R 24 2 was most effective at improving IDF curve fit to observed data, giving the IDF equation of the form (Kothyari and Garde 1992): ( 2 2 ) Methods for Rainfall IDF Development for Andhra Pradesh Two rainfall time series obtained ( Singh 2009 ) from recording gages at the International Crops Research Institute for the Semi arid Tropics (ICRISAT) near Hyderabad, Andhra Pradesh, India were used to devel op IDF relationships for this study. The first series consisted of 140,256 records of hourly rainfall (16 year period, 1993 2008); this data series was used for development of IDF relationships. The

PAGE 69

69 second rainfall time series consisted of 12,784 records of daily rainfall from (35 year period, 1974 2008); this data was used to analyze long term trends in occurrence of threshold based high intensity rainfall events. Sub daily rainfall records are important in rainfall IDF analyses for accurate determinati on of short duration (2, 4, 8 hour) storm intensities, and hourly rainfall records for India are sparse and are generally only 2007). Average annual rainfall in Hyderabad is 880 mm, and 75% of that rainfall a rrives during the rainy season from June to September called Kharif. The studies of Indian monsoon rainfall variability (May, 2004) and spatial coherence of tropical rainfall (Moron et al. 2007) suggest that the IDF relationships developed here, based on rainfall records from a recording station in Hyderabad are applicable for the Medak district. Using the hourly rainfall data, annual maximum series (AMS) for all durations considered were developed by calculating a moving average intensity for each durati on and then finding the maximum average intensity for each duration during a calendar year. This is step one of IDF curve development. The Weibull probability distribution was chosen based on graphical and log likelihood value comparisons of fit to the A MS data of Weibull, Generalized Extreme Value, Gamma, and Log Normal probability distributions. Maximum likelihood estimation of parameters for the Weibull CDF was completed independently for each of the five durations (1, 2, 4, 8, and 24 hour); see the r T able 2 2 (this is step two). The two parameter Weibull probability density and cumulative distribution functions are: ( 2 3 )

PAGE 70

70 ( 2 4 ) The five resulting CDFs were used to calculate rainfall intensity for the 5 durations at all the frequencies required; in this case 2, 5, 10, 15, 25, 50, 75, and 100 year return periods were considered (step 3). The IDF data were then fitted to a parametric equation of the form of E quation 2 2 for comparison to the Kothyari and Garde (1992) equation to examine the change in IDF relationships for the region in the last 3 decades. Exploration of Trends in Occurrence of Rainfall Events of High and Low Intensity Differences in IDF curves between those developed here and those develop ed using older rainfall records (Kothyari and Garde, 1992) can give evidence for a change in character of precipitation for the region near Hyderabad, India. However, trends in the total annual numbers of threshold based high and low/moderate intensity ra infall events for single recording station can add to the available evidence for changing rainfall mm/day; low to moderate days were those having 5 < I < 50 mm/day. These intensity classes are commonly used to group rainfall days into high and low/moderate classes (Angel and Huff 1997; Goswami et al. 2006). For each calendar year, days in each class were summed. A longer record was used for this trend analysis than was used for IDF curve development because of the lack of hourly rainfall data availability and the decision to use a more recent, shorter record for IDF analysis. Hourly rainfall data give more reliable IDF relationships, especially for short durations, and most of the available 2007).

PAGE 71

71 R esults and Discussion: IDF Curves and Event Intensity Trends Table 2 1 presents the 1, 2, 4, 8, and 24 hour annual maximum rainfall intensities for the 16 year record (1993 2008) from which IDF curves were develo ped. Annual and rainy season (k harif) rainf all totals and number of rain days are also presented. The IDF curves developed from the inversion of the fitted Weibull CDF at each duration, using the various probabilities (1/frequency) and solving for intensity, are plotted in F igure 2 1. These inten sities were used to fit coefficients C1, C2, C3, and C4 to equation 3 2. Fitting was done separately for each duration and then also for all durations combined (see r esulting constants in T able 2 2); coefficients were fitted using an automated iterative r outine with the goal of minimizing the average root mean squared error (RMSE) between intensiti es from the CDF and those from E quation 2 2. Using the coefficients of Kothy ari and Garde (1992) listed in T able 2 2 the RMSE of rainfall intensities from IDF relationships developed using recent data (1993 2008) and those calculated from the original Kothyari and Garde formula (1950 1980 rainfall data) was 28.98 mm/hour. Mean difference (intensities from original Kothyari and Garde formula minus intensities f rom newly parameterized W eibull CDF) was 25.43 mm/hour (T able 3 4), and percent difference from the original 1992 formula was greater than 100% for many return period duration combinations. This suggests significant increases in intensity of rainfall eve nts during the last 3 decades for the region near Hyderabad, India. Analyzing the daily rainfall data from Hyderabad (1974 2008), noticeable trends decreasing numbers of low and moderate intensity events (5 < I <50 mm/day) were observed (F igure 2 2). These trends have low to moderate statistical significance

PAGE 72

72 based on t tests of the hypothesis of no trend; p values of 0.27 and 0.082 for slopes of increasing trend of high in tensity events and decreasing trend of low and moderate events, respectively. The non parametric Mann Kendall test of trend significance rainfall events, but no signific ant trend for annual numbers of low and moderate intensity events. The moderate strength of the trends is consistent with the observations of Goswami et al. (2006) that the high variability of single gage records makes it difficult to observe trends of hi gh statistical significance, but the directions of the trends are consistent with their analysis for all of central India. Discussion: Precipitation Characterization and Groundwater in India Changing rainfall characteristics generally have important impact s on the hydrology of a region. In semi arid regions, like that of the Hyderabad region, the annual potential evapotranspiration is much greater than precipitation, runoff generation is generally Hortonian, and groundwater recharge in these regions depend s largely on high intensity storms and the storage of excess rainfall in surface depressions (de Vries and Simmers 2002). Channelized flow is highly seasonal, meaning most groundwater recharge results from areal infiltration and percolation of surface st orage of excess rainfall. Higher intensity rainfall patterns lead to greater runoff and a lower proportion of rainfall being infiltrated and available for groundwater recharge. For the agricultural regions of semi arid India, groundwater recharge is of m ajor concern: groundwater depletion is common as a result of substantial irrigation withdrawals. Given the scarcity and seasonality of surface water resources in the region, increased rainfall variability makes irrigation from groundwater resources even m ore important. The evidence provided by this study suggests rainfall in peninsular India is becoming increasingly

PAGE 73

73 characterized by higher intensity events and fewer low and moderate intensity events. This result has a variety of management applications, but for agricultural areas dependent on groundwater, one application would be to increase investments in reservoirs, farm ponds, and water harvesting tillage to increase the infiltration of rainfall. The value of groundwater resources can be expected to i ncrease when there is more variable rainfall. Conclusions on Rainfall Intensity Trends and Groundwater Recharge Rainfall IDF relationships are useful tools for various hydrologic analyses, and the updating and maintenance of these relationships is importan t for decision making that requires information about the character of rainfall. The recent evidence of more common high intensity rainfall events was extended in this study to illustrate IDF differences between old and new records and to recommend that I DF relationships be regularly updated using more recent rainfall records. The newly developed IDF relationship for the Hyderabad region of southern India, using the formula of Kothyari and Garde with updated parameters based on 1993 2008 rainfall data, in cluding the 2 year, 24 hour rainfall of 103.2 mm, is These new IDF curves show noticeably greater rainfall intensity patterns on average 25 mm/hour or 123% greater for 1 to 24 hour durations compared t o the previously available rainfall characterization (Kothyari and Garde 1992) for the southern region of India. Based on this large IDF difference, it is recommended that IDF relationships be regularly updated to improve decision s and design s that utiliz e these relationships. These results are consistent with the growing consensus that precipitation patterns are shifting, especially in lower latitudes, toward higher intensity rainfall events and a decrease in moderate

PAGE 74

74 and low intensity rainfall (Owor et al. 2009; Pall et al. 2006; Trenberth et al. 2003; Allen and Ingram, 2002). Similar studies of changing IDF relationships using shorter, recent rainfall records in other regions of India and in other parts of the world could give increased evidence for ch anging rainfall characteristics.

PAGE 75

75 Table 2 1. Annual maximum series rainfall intensity generated from hourly rainfall data from Hyderabad, Andhra Pradesh, India. Annual and Kharif season (June September) total rainfall and rain days Annual maximum se ries rainfall intensity, mm/hr Rainfall totals, mm K harif 1 hour 2 hour 4 hour 8 hour 24 hour A nnual K harif rain days 1993 29.5 17.1 10.0 5.0 2.4 831 588 49 1994 34.8 23.4 14.1 8.9 6.0 848 550 62 1995 54.1 29.9 15.7 7.9 4.0 1266 747 54 1996 50.3 27. 8 16.2 11.1 4.7 1063 911 64 1997 29.2 17.8 11.8 7.7 3.9 743 433 48 1998 34.1 21.2 12.0 7.9 2.8 1181 887 64 1999 30.0 16.6 8.7 4.4 1.9 580 455 63 2000 154.9 136.4 105.6 66.0 36.9 2016 1797 66 2001 37.4 20.7 10.5 6.4 2.6 688 514 53 2002 18.5 14.1 7.1 3 .8 2.2 623 473 48 2003 50.6 41.2 25.4 12.8 4.3 926 789 66 2004 29.4 26.5 14.3 7.7 3.2 783 546 47 2005 46.7 32.3 18.3 9.3 4.4 1192 850 68 2006 33.8 24.6 17.2 11.9 4.6 889 636 75 2007 30.2 28.4 16.1 8.8 2.9 717 571 77 2008 44.0 25.8 15.6 9.9 7.0 1109 7 58 65 Table 2 2. Parameters for Weibull CDF: F(I) = 1 exp( where I is rainfall intensity, and parameters for IDF function: Weibull CDF parameters IDF function parameters Storm Duration shape C1 C2 C3 C4 1 hour 50.1056 1.6863 7.1068 0.4018 0.2259 0.7100 2 hour 35.2153 1.4214 7.1059 0.4303 0.2614 0.6885 4 hour 21.5984 1.2173 7.1035 0.4264 0.2976 0.6734 8 hour 12.6868 1.1703 7.1025 0.4113 0.3075 0.6686 24 hour 6.0630 1.0670 7.1022 0.40 50 0.3318 0.6560 fitted to all 5 durations 7.1052 0.2532 0.4786 0.3889 Kothyari and Garde parameters 7.1000 0.2000 0.7100 0.3300

PAGE 76

76 Figure 2 1. Intensity duration frequency curves for 1, 2, 4, 8, and 24 hour durations developed from hour ly rainfall data from 1993 2008 from Hyderabad, India Table 2 3 Rainfall intensity values from Weibull cumulative distribution functions (CDFs) : hourly data 1993 2008 for Hyderabad t (hours) T, return period (years) duration 2 5 10 15 25 50 75 100 1 40.32 66.44 82.16 90.36 100.22 112.51 119.70 123.94 2 27.21 49.22 63.32 70.89 80.15 91.94 98.95 103.12 4 15.98 31.93 42.85 48.89 56.43 66.23 72.17 75.73 8 9.28 19.05 25.87 29.67 34.45 40.70 44.50 46.78 24 4.30 9.47 13.25 15.40 18.13 21.77 24.01 25.37 0 20 40 60 80 100 120 140 0 5 10 15 20 25 Intensity, mm/hour Duration, hours T= 2 year T= 5 year T = 10 year T = 15 year T = 25 year T = 50 year T = 75 year T = 100 year

PAGE 77

77 Table 2 4. Rainfall intensity (mm/hr) from Kothyari and Garde general formula fitted to recent intensity data (1993 2008) for Hyderabad 1 RMSE calculated based on CDF intensities t (hours) T, return period (years) duration 2 5 10 15 25 50 75 100 1 51.40 64.82 77.25 85.60 97.42 116.11 128.66 138.38 2 36.89 46.52 55.44 61.43 69.92 83.33 92.33 99.31 4 26.47 33.39 39.79 44.09 50.18 59.80 66.26 71.27 8 19.00 23.96 28.55 31.64 36.01 42.92 47.56 51.15 24 11.23 14.16 16.88 18.70 21.28 25.37 28.11 30.23 RMSE 6.27 1 : Table 2 5 Rainfall intensity (mm/hr) from Kothyari and Garde original formula fitted to older intensity data (1950 1980) for southern zone of India 1 RMSE calcul ated based on CDF intensities t (hours) T, return period (years) duration 2 5 10 15 25 50 75 100 1 37.67 45.24 51.97 56.36 62.43 71.71 77.77 82.37 2 23.03 27.66 31.77 34.46 38.16 43.84 47.54 50.35 4 14.08 16.91 19.42 21.06 23.33 26.80 29.06 30.7 8 8 8.61 10.34 11.87 12.88 14.26 16.38 17.77 18.82 24 3.94 4.74 5.44 5.90 6.54 7.51 8.14 8.63 RMSE 28.98 1 : RMSE

PAGE 78

78 Table 2 6 Differences (mm/hour) in predicted rainfall intensity values betwee n Kothyari and Garde IDF formula using 1992 parameters 1 and using updated parameters fitted from this study 2 t (hours) T, return period (years) duration 2 5 10 15 25 50 75 100 1 13.73 19.58 25.28 29.24 35.00 44.40 50.89 56.01 2 13.86 18. 86 23.67 26.98 31.75 39.49 44.79 48.95 4 12.40 16.48 20.37 23.03 26.85 33.00 37.20 40.49 8 10.39 13.62 16.68 18.77 21.75 26.53 29.79 32.33 24 7.28 9.42 11.44 12.80 14.75 17.86 19.96 21.61 Mean Difference 25.43 1 : and 2 : Figure 2 2. < I < 50 mm/day) rainfall events (1974 2008) 0 10 20 30 40 50 60 0 1 2 3 4 5 6 7 8 9 1970 1975 1980 1985 1990 1995 2000 2005 2010 N(5
PAGE 79

79 CHAPTER 3 IMPORTANCE OF SURFACE STORAGE POND ING DEPTH FOR PREDIC TING INFILTRATION AND RUN OFF IN WATER CONSERV ATION TILLAGE SYSTEM S Modeling Infiltration and Tillage : Implications for Groundwater Recharge Infiltration Surface Storage and Cha nging Precipitation Character Infilt ration is the quantity of water that enters the soil at the intersection of the soil surface and the atmosphere. The rate of infiltration is controlled by the rate of water application (precipitation and irrigation), by the depth of surface water storage, and by the properties of the soil. In agricultural areas, tillage management arguably has the greatest influence on infiltration (Blevins et al., 1990; Knapen et al., 2008; Zhang et al., 2007). Its effects on infiltration result from greater surface rou ghness, which increases depression storage, and from lower bulk density, which increases hydraulic conductivity. Disruption of surface sealing and crusts through tillage also increases infiltration capacity of a soil. While there is extensive literature on tillage management effects on surface storage and roughness (Kamphorst et al., 2000; Planchon et al., 2002; Guzha, 2004), there has been comparatively little work done to quantify the importance of surface storage in simulation of infiltration and runof f. Excess rainfall (total precipitation minus interception, surface storage, and infiltration) is the amount of water available for direct runoff; this is generally greater than runoff, which is the amount of that converges into surface streamflow. For s implicity, excess rainfall is hereafter referred to as runoff (RO). In semi arid regions, potential evapotranspiration is much greater than p recipitation; the ratio of precipitation to potential evapotranspiration for semi arid zones is 0.20 0.50 (UNEP, 19 92). Annual precipitation for semi arid zones varies from 300 800 mm, in areas with summer rains and from 200 500 mm in areas with winter rains

PAGE 80

80 (UNEP, 1992). Groundwater in semi arid regions is an important resource for mitigation of droughts and dry spe lls (Shah et al., 2006). Channelized flow is uncommon and very seasonal, and groundwater recharge in these regions depends largely on high intensity storms and the areal infiltration and deep percolation of runoff stored in surface depressions (de Vries a nd Simmers, 2002). One such semi arid region is the Wargal agricultural watershed, Medak district, Andhra Pradesh, India, where groundwater depletion is a growing concern. More than 80% of annual rainfall (780 mm) in Wargal falls during the 4 months fro m June to September. Much of it is episodic, arriving in short, high intensity events. There is growing consensus that precipitation patterns are shifting, especially in lower latitudes, toward more common higher intensity rainfall events and fewer moder ate and low intensity rainfall events ( Chapter 2; Owor et al., 2009; Pall et al., 2006; Trenberth et al., 2003; Allen and Ingram, 2002). These trends toward increasingly episodic rainfall highlight the value of surface storage for management of infiltrat ion and groundwater recharge and the potential importance of water harvesting tillage methods for dry spell mitigation. For an episodic rainfall pattern having high intensity storms, daily time step models of infiltration and runoff can be expected to ove restimate infiltration and thus underestimate runoff. The hourly or sub hourly time step solutions of Green Ampt have been shown to simulate infiltration better than the daily time step Curve Number (SCS, 1972) method under a variety of conditions (Rawls and Brakensiek, 1986; Wilcox et al., 1990; Stone and Sadler, 1991; Van Mullem, 1991). Curve Number infiltration has established guidelines

PAGE 81

81 for parameterization of tillage management; however, describing tillage for Green Ampt infiltration is generally not considered. Green Ampt Infiltration and Depression Storage The Green Ampt infiltration model (Green and Ampt, 1911) can be described as an approximate, physically Water is assumed to infiltrate w ith a distinctly defined wetting front as illustrated in Figure 3 1. The development of the Green Ampt equations is a direct application of constant, ponded infiltration, ( 3 1) where q is water flux [L/t] K s is sa turated hydraulic conductivity [L/t] and dH/dz is the hydraulic head gradien t in the direction of flow [L/L] Adding up heads in Figure 3 1 (dH = h p f ), substituting infiltration rate q by dF/dt, where we call F cumulative s i )*L), replacing K s with K se (effective hydraulic conductivity, accounting for air in the soil matrix of unsaturated soils), and substituting the soi l s i or ( 3 2) Although in reality, 0 < h p (t) < MDS, in the classical use of derivation and use of the GAML eq uations, the ponding depth, h p is assumed negligible and is removed from the expressions in equation 3 2. Separation of variables and integration results in the familiar Green Ampt infiltration equation: ( 3 3)

PAGE 82

82 M ein and Larson (1973), Chu (1978) and Skaggs and Khaleel (1982) improved the applicability of the Green Ampt equation to allow for calculation under conditions of variable natural rainfall intensity (with periods of non ponding and ponding conditions). Th e time to ponding, t p the cumulative infiltration at the time of ponding, F p p the time to infiltrate Fp if ponding started from the beginning of the rainfall event, were introduced to allow for the following solution system. For t < t p f = i (r ainfall intensity) and for t = t p f = i. Time to ponding can be calculated from t p = F p /i, where, ( 3 4) Consideration of the ponding time adjustments finally result into, ( 3 5) Using this form of the equation, h p appears only in the surface water balance, calculated after runoff (RO) predicted by GAML, as dRO = dR dF dh p ( 3 6) where dRO [L] is the increment of excess rainfall produc ed for an in cremental rain, dR [L], dF [L] is the increment in cumulative infiltration for the same time step, and dh p [L] is the change in surface storage depth for the same period. Equations 3 4, 3 5, and 3 s form of the Green Ampt infiltration model now allows for infiltration to be simulated before and after ponding and if ponding ends and begins again. The MDS=0 solution, shown in the results to illustrate tillage and MDS importance, represents a negligibl e h p both as a head and in a surface water balance. This solution form (MDS=0) is used to predict infiltration and runoff in a variety of field scale and landscape scale hydrologic models, Soil and Water

PAGE 83

83 Assessment Tool (SWAT, Arnold et al., 1998; Gassman et al., 2007; Krysanova and Arnold, 2008) is one example. It is proposed here that for agricultural systems under some form of water harvesting tillage, surface storage h p should be included as a time varying head (0 < h p (t) < MDS) in the GAML equations i n addition to the water balance at the surface (eq. 3 6). To simplify the handling of the time dependent boundary condition in equation ( 3 2) we can assume that h p is constant within each time step but changed for each time step ( This simplifi cation is acceptable in water harvesting tillage systems where dh p for each dt is small relative to the total MDS if the time step is kept sufficiently small. This allows integration of equation (2) to yield, ( 3 7) Equations 3 4, 3 6, and 3 7, solved concurrently for each time step will here be for water cons ervation tillage systems against the standard formulation is examined by repeating local and global sensitivity analyses with both forms of the GAML equations: with h p only included in surface water balance for the conventional Green Ampt method (standard GAML) and with h p as a time varying head in the GAML equation and also in the surface water balance (complete GAML). Sensitivity analyses were used to measure the strength of influence of h p on infiltration predictions and to compare GAML infiltration sen sitivity to all the remaining parameters. Groundwater Recharge a nd Tillage Management : Modeling Implications In this chapter the focus is on improving the performance of infiltration predictions from the GAML equations through the addition of the typical ly neglected surface storage

PAGE 84

84 depth term, h p ; the importance of all five GAML parameters under conventional and tied ridge tillage is compared. Conventional tillage describes disking, harrowing, moldboard plowing, or chisel plowing with no consistent orien tation to elevation contours. Water harvesting tillage is a type of conservation tillage characterized by significant increases in microrelief or depression storage through the creation of surface geometries allowing for substantial water storage (i.e. ti ed ridge tillage, pit tillage, contour bunding). Essentially, small basins are formed in croplands which can store excess rainfall and increase infiltration. The agronomic benefits of water harvesting tillage are generally realized in the form of increas ed crop yields resulting from increased plant available water (Twomlow and Breneau, 2000; Wiyo, 2000; Guzha, 2004; Tesfahunegn and Wortmann, 2008). In addition, increased infiltration can result in the potential hydrologic benefit of increased groundwater recharge (Foster and Chilton, 2003), an important benefit in many intensive agricultural areas of the world where aquifer levels are decreasing (Konikow and Kendy, 2005). Rainfed agricultural systems, especially those in resource poor and generally ungaug ed watersheds, are increasingly using water harvesting tillage to increase infiltration depths (Twomlow and Breneau, 2000; Wiyo et al., 2000; Tesfahunegn and Wortmann, 2008; Derib et al., 2009; Araya and Stroosnijder, 2010), and it may be in these systems, especially in the tropical regions having vulnerable soils, that alternative tillage management is the most economical means of increasing production (Rockstrm, 2001). Also, irrigated rice croplands, having substantial and well identified surface stora ge depths, are extensive in many typically ungauged watersheds of Southeast Asia. Including surface storage depth in models using GAML predicted

PAGE 85

85 infiltration is suggested as a possible means of improving model process descriptions for predictions in ungau ged basins (i.e. PUB project, see Sivapalan et al., 2003). In addition to the growing use of water harvesting tillage in rainfed systems, there is evidence in the scientific literature and in recent agricultural applications that there is renewed interest in water harvesting tillage even in humid, irrigated areas (Truman and Nuti, 2009 and 2010). In Mississippi, USA, for example, water harvesting tillage in the form of furrow diking (synonymous with tied ridging) has been approved for cost sharing under t implementing furrow diking, half of the costs are paid by the state. In an analysis of the hydrology and economics of water harvesting tillage for irrigated cotton in Georgia, USA, it was found that a grower would recover investment costs for furrow diking tillage after the first 16 ha mm of irrigation water savings. This was based on a cost of $1.17 / ha mm to pump irrigation water and $18.50 / ha for implementing furrow diking (Trum an and Nuti, 2009). Additionally, furrow diking resulted in a 2.6 times reduction in soil loss and 25% lower runoff compared to conventional tillage. There are numerous process based hydrology and water quality models using GAML infiltration with fixed or negligible MDS used in a surface water balance and assuming negligible h p in GAML equations (MDS=0 or standard GAML solution form): SWMM (Storm Water Management Model; Huber and Dickinson, 1992), PRZM (Pesticide Root Zone Model; Carsel et al., 1998), WEPP (Water Erosion Prediction Project; Laflen et al., 1997), SWAT (Soil and Water Assessment Tool; Arnold et al., 1998), VFSMOD (Vegetative Filter Strip Modeling System, Muoz Carpena et al., 1999), WaSiM ETH (Water Flow and Balance Simulation Model; Schulla and Jasper,

PAGE 86

86 2000), and others. It is argued here that using the complete GAML form in models using the Green Ampt method for infiltration prediction, there is no real loss of model parsimony. If MDS values for various tillage management can be assumed to have less uncertainty than other infiltration parameters, a reasonable assumption given the simplicity of measurement, the lower range of values, and the strength of literature values of MDS (Kamphorst et al., 2000; Planchon et al., 2002; Jones and Baumha rdt, 2003; Guzha, 2004), then there may be a reduction in model equifinality concerns, meaning that a model user may have a better chance of getting the right answer for the right reasons (Beven, 2006; Gupta et al., 2008; Hughes, 2010). Parameter estimat ion completed during GAML model calibration to match observed data can produce models of similar performance regardless of the solution form being used. For example in a watershed having extensive areas under water harvesting tillage, even if the MDS=0 GA ML solution is used, streamflow could still be predicted well after calibration by increasing the values for K se f to achieve the same runoff reductions that actually result from the MDS of the areas under water harvesting tillage. The use of addit ional measured input factors (as opposed to simply using calibration procedures) is suggested as a way to avoid this problem (Ritter et al., 2003). Here it is proposed that including h p (and MDS) in the GAML equations and in a surface water balance would provide more certain predictions of infiltration and runoff, compared to MDS=0 and standard GAML solutions, for evaluating hydrologic effects of water harvesting tillage. The objectives of this chapter were to evaluate the importance of surface storage d epth for Green Ampt predicted infiltration in agricultural systems

PAGE 87

87 under conventional and water harvesting tillage, and to test the proposal that for systems with water harvesting tillage the surface storage depth could be used to improve the process descr iption of GAML infiltration. Methods for Analyzing the Importance of Surface Storage Depth S ite D escription The study area is a 512 ha watershed, called Wargal, in a rural, agricultural region of northwestern Andhra Pradesh, India (Figure 1 1 ). Soils in t he area are mostly sandy loam and sandy clay loam with some small areas of clay loam and sandy clay (Soil Survey Staff, 1999). Rice, maize, cotton, potato, and sunflower are the most commonly grown crops. Conventional tillage, the typical practice in the area, is either moldboard or chisel plough drawn by animals and sometimes tractors. Tied ridge tillage, also called furrow diking, basin tillage, or reservoir tillage, is the practice of creating earthen ridges perpendicular to direction of primary tilla ge (Figure 3 2 ). Tied ridging was selected as the example of water harvesting tillage because it is sometimes used for potato and other vegetable crops planted late in the rainy season, and equipment is commercially available for mechanization. The ridge s and ties can be created with hand tools or with a variety of animal drawn or tractor drawn implements. Typically, furrow spacing is 75 to 100 cm with ties every 2 to 6 meters (Jones and Baumhardt, 2003). Ridge and tie heights are approximately 20 cm. A nalysis of the surface geometry suggests a maximum depression storage of 100 mm. Profile measurements of tied ridge tilled systems show 50 to 60 mm maximum depression storage (Jones and Clark, 1987). Groundwater depletion is a growing concern in the stud y region as a result of substantial rice cropland areas, high irrigation withdrawals, density of borewells, and

PAGE 88

88 unmetered electrical supply. This is a common scenario in much of India (Rodell et al., 2009). The broad goal of work in the Wargal watershed is to evaluate combinations of agricultural management options for improving groundwater recharge and reducing irrigation withdrawals. These evaluations are being made using a simulated water balance from the semi distributed hydrology model, SWAT (Arnold et al., 1998; Gassman et al., 2007; Krysanova and Arnold, 2008). One of the management options being investigated for increasing groundwater recharge is to expand the extent of cropland areas under water harvesting tillage. It is expected that the infil tration increases from water harvesting tillage would result in groundwater recharge improvements. Local Sensitivity Analysis A simple, local sensitivity analysis (LSA), changing one parameter at a time, was completed to find the changes in runoff (RO) and infiltration (F) depths obtained for a 4 hour design storm of 2 year return period in the Wargal area; results from all 3 GAML solution forms were compared (MDS=0, standard, and complete). Percent changes in F and RO were calculated based on the differen ce in RO and F from solutions using mean MDS for each tillage treatment, conventional (CT) and tied ridge (TR). Here, only MDS was varied from its minimum, mean, and maximum values. The purpose of the local sensitivity analysis was to give a simple, easil y interpretable measure of MDS importance in GAML solutions. Global Sensitivity Analysis A global analysis of sensitivity (GSA) was included here to provide a more complete examination of the importance of MDS in the context of all the GAML parameters. A dditional summaries of GSA methods (Saltelli et al., 2000; Saltelli, 2005;

PAGE 89

89 Saltelli et al., 2008) and recent applications to hydro ecological models (Muoz Carpena et al., 2007; 2010; Muoz Carpena and Muller, 2009; Fox et al., 2010; Chu Agor et al., 2011) are available. For a given model input factor X i the first order variance based global sensitivity index S i gives the ratio of the output variance if the true value of the input was known to the total output variance (Y). Formally, a first order S i is d efined as S i = V[ E(Y|X i ) ] / V(Y), where X i is a parameter or input of a model, Y is the output of interest, E is expected value, and V is variance. If a model if perfectly additive, meaning there are no parameter interactions, then the sum of all S i quals 1. This in practice represents a quantitative measure of how much of the total output variance is explained by the variation of the single parameter X i alone (direct effect). First order indexes are used to select the input factor(s) of the model t hat would result in the greatest reduction in variance, V(Y), if the true value of the factor was known; hence they are typically used to prioritize the importance of model inputs (Saltelli, 2005). Total effect global sensitivity indexes, S Ti are calcula ted to account for the presence of interactions (higher order effects). S Ti gives the ratio of the expected amount of variance that would remain if all other input factors were known (fixed) and only the input factor of interest was varied over its ran ge to the total output variance (Y), i.e. S Ti = E[ V(Y|X i ) ] / V(Y), and are computed similarly to first order S i They are generally used to fix a factor having little influence on V(Y) or to exclude unimportant input factors during model development ( Saltelli, 2005). If a model is non additive, meaning there are parameter interactions, then the sum of all S Ti is greater than 1.

PAGE 90

90 A program was developed in Matlab (Mathworks Inc., Natick, MA) for Green Ampt infiltration of unsteady rainfall and was used to read in rainfall intensities, GAML parameters, and solve for cumulative infiltration (F), surface runoff (RO), surface storage (h pi ), and f (infiltration rate). Equation 7, an implicit function in terms of F, was solved for F using a combination of bi section, secant, and inverse quadratic interpolation methods (fzero routine in Matlab). h pi was calculated at each time step i based on rainfall intensity, infiltration rate, previous time step h p 1 and MDS. Variance based first order sensitivity indexe s (S i ) were obtained for all the GAML parameters for various storm durations and return periods using the extended Fourier amplitude sensitivity test (FAST) (Saltelli et al., 1999). These were computed for target outputs total storm RO and F using the GAM L Matlab program linked to the sensitivity and uncertainty analysis package SimLab 2.2 (Saltelli, 2005) that was used to sample input parameter sets and for post processing of outputs to generate sensitivity indexes. Parameter sets of size 5000 were gener ated from the ranges and probability distributions selected for all five Green Ampt parameters. Parameters for Green Ampt Infiltration Sensitivity Analysis The first step in a good sensitivity analysis is a clear understanding of the objective; here, the f irst objective of the sensitivity analysis was to rank the relative importance of surface storage depth against all other model input factors for two tillage strategies: conventional (CT) and tied ridge (TR). The second step is to develop ranges and proba bility distributions for all parameters of the model being used. Table 3 1 d isplays parameter values needed to use the Green Ampt model for unsteady rainfall at the Wargal site. Ranges of the Green Ampt parameters for the study area were based

PAGE 91

91 on observat ions and estimations from literature values; probability distributions were assigned to each of the parameters based on the range and values found. Laboratory measurements of saturated hydraulic conductivity were performed on undisturbed soil cores coll ected from the experimental site (Figure 1 1 ) by partners at Acharya N. G. Ranga Agricultural University (ANGRAU) in variable depth increments from 0 1 m (136 samples from N = 37 locations); there were typically samples from 4 soil layers at each locatio n Values of saturated hydraulic conductivity (K s ) ranged from 0.131 to 165 mm/hr, based on constant head permeameter K s measurements. For use in Green Ampt, K s values only in the top soil layer (0 30 cm) were used; the geometric mean K s in this layer was 31.8 mm/hour. Hydraulic conductivity (K) of soil is dependent on water content; generally a drier soil has a lower K than a wetter soil. The effective hydraulic conductivity (K se ) in the Green Ampt equation accounts for this air entrapment making K se < K s The recommendation of Bouwer (1966) is to approximate K se using K se = 0.5*K s ; this approximation was used here in the absence of field measurements of K se A log normal probability distribution was fitted to the K se data; Weibull, log normal, and Generalized Extreme Value (GEV) distributions were compared graphically. The GEV distribution fit the data most closely, but the log normal distribution, also a good fit, was selected because it is what is typically used for K s distribution (Carsel and Pa rish, 1988; Meyer et al., 1997) and the GEV distribution is unavailable in the SimLab (Saltelli, 2005) sensitivity/uncertainty analysis package that was used here. f ) values for soils in Wargal watershed. The soil properties tables of Rawls et al. (1983) were f based on observed soil texture data. The soil texture used here

PAGE 92

92 was sandy loam; it was one of the two dominant textures based on observations and measured K s values were more consistent with those associated with sandy loam than with sandy clay loam. Soil texture analysis for the experiment site was made using the Bouyco us Hydrometer method (Kalra and Maynard, 1991). Without additional data for f was assumed uniform with parameters given in Table 3 1. i ) and saturated s ) f using tabular data for sandy loam soil (Rawls et al., 1983; Fangmeier et al., 2005). Spatial variability is typically high for these parameters and i Literature value s based on soil texture were the best estimates available for the study area. The maximum surface storage depth (MDS) represents the maximum depth of surface storage of water over an area. It is the equivalent depth of water stored on the surface of the e ntire area if it was covered with an equal depth of water everywhere. There is not sufficient data in the literature to recommend changes in the other parameters for different tillage. Tillage is expected to influence infiltration through soil surface ge ometry changes, not through changes in soil structure. MDS is changed because significant differences in this parameter have been found for different kinds of tillage (Table 3 1). Based on a summary of literature values, MDS will be given a mean value of 5 mm for conventional tillage. With no data about distribution of MDS measurements, distribution will be assumed uniform: minimum and maximum MDS of 0 and 10 mm, respectively, for conventional tillage. MDS for tied ridging will be assigned a mean of 32. 5 based on the assumed uniform distribution and using the minimum

PAGE 93

93 value of Guzha et al. (2004) of 15 mm depression storage and a maximum value of 50 mm based on a conservative assessment of surface geometry and the work of Jones and Clark (1987). Developme nt and Selection of Design Storms for the Analysis Hourly data of rainfall from 1993 2008 (16 years; Singh, 2009) from recording gages at the International Crops Research Institute for the Semi arid Tropics (ICRISAT) near Hyderabad, Andhra Pradesh, India ( 20 km from study area) were used to generate rainfall intensity duration frequency (IDF) curves that were used for synthetic storm generation of design storms. IDF relationships were developed by fitting a probability distribution function to the annual m aximum series rainfall intensity data (separately for selected durations: 2, 4, 8, and 24 hours). Parameters for distribution functions were estimated using maximum likelihood estimation. Three distribution functions commonly used for rainfall IDF repre sentation, Gumbel, Gamma, and Weibull ( Koutsoyiannis et al., 1998; Mohymont et al., 2004), were compared graphically with the empirical cumulative distribution functions (CDF) of annual maximum series rainfall for the selected durations. Results and discus sion of GAML Infiltration Sensitivity Analyses Representative Design Storm for t he Analysis Plotting the Gumbel, Gamma, and Weibull CDF with the empirical CDF of 4 hour rainfall annual maximums showed similar fit to the data for Weibull and Gamma distribut ions (Figure 3 3) The Weibull distribution function was selected to be used here based on simplicity of form and frequency of use in the literature (Wilks, 1989; D. Koutsoyiannis et al., 1998; Madsen et al., 2009). The 16 year hourly rainfall record cou ld be considered too brief for prediction of intensities and durations of low frequency

PAGE 94

94 storms (75 and 100 years; Table 3 2), but the unusually heavy rainfall events of August 2000 (92 year daily high rainfall for Hyderabad; Geological Survey of India, 200 1) result in rainfall IDF relationships that can be considered reliable for low frequency storms. As only 2, 5, and 10 year storms of 4 hour duration were used here, this record was deemed adequate for generating design storms of those frequencies. In the 16 year period of record, there were seventy 4 hour storms, thirty three 6 hour storms, and nine 8 hour storms; these were analyzed to determine storm type for the area using mass distribution curves. Analyses of rainfall mass distribution curves suggest storms should be described as Type II (SCS, 1986), meaning total depths of rainfall before and after half of the storm duration are about equal, and peak rainfall intensity occurs at half the storm duration; this is sometimes referred to as an intermediat e storm. The low number of long duration storms for the record period (nine 8 hour storms) is evidence of the episodic nature of rainfall in the region; storms are typically short and of high intensity. Therefore, it was decided to consider 4 hour storm s for this analysis given that storms of this duration are much more common than 8 hour or 24 hour storms). The rainfall intensity duration frequency (IDF) data used to develop design storms are presented in Table 3 2. The rainfall intensity data for 4 h our design storms of the three return periods (2, 5, 10 year return period) that were used for this analysis are presented in Table 3 3. Tillage and MDS: GAML solution form The differences in infiltration and runoff depths between the standard and compl ete GAML solutions are demonstrated in Figure 3 4 For the MDS range of conventional tillage (CT; 0 10 mm), there are small differences in predicted RO and F. On average across the MDS range for CT, using the complete GAML form, predicted F

PAGE 95

95 was 1.2% grea ter and predicted RO was 2.5% lower than the F and RO predicted using the standard GAML form. These differences increase substantially for the MDS range of tied ridge tillage (TR; 15 50 mm). On average across the MDS range for TR, using the complete GAML form, predicted F was 5.8% greater and predicted RO was 40.8% lower than the F and RO predicted using the standard GAML form. This shows that the solution form should be considered for GAML predictions in agricultural areas with significant MDS (> 15 mm) Also, Figure 3 4 demonstrates the importance of the MDS parameter for describing tillage, for either GAML solution form. The global sensitivity analysis was required to make quantitative comparisons between the importance of MDS and the other 4 GAML pa rameters. Local sensitivity analysis Results of the local sensitivity analysis (LSA; Table 3 4 and Fig ure 3 5 ) illustrate substantial infiltration and runoff depth differences between CT and TR tillage treatments, meaning that MDS can be an important param eter for improving descriptions of tillage management. Under the complete GAML solution with mean MDS values of 32.5 mm and 5.0 mm for tied ridge (TR) and conventional tillage (CT), respectively, total runoff depth was 65% lower for TR than for CT, and pr edicted total storm infiltration was 31% greater for TR than for CT. Differences in RO and F within tillage treatments and between solution forms show that the manner in which h p is included in GAML solutions is important for tillage treatments with signi ficant MDS (> 15 mm). For TR with mean MDS using the complete GAML solution form, predicted RO decreased by 31% and F increased by 6% compared to the standard GAML form. GAML solution form is less important in conventionally tilled areas with low MDS (< 10

PAGE 96

96 mm): for CT with mean MDS and the complete GAML solution form, predicted RO decreased by 3% and F increased by 6% compared to the standard GAML form. Global Sensitivity Analysis The first order sensitivity indices (S i ) for each of the GAML parameters for both tillage treatments and both GAML solution forms are reported in Table 3 5 and Figure 3 6 Design storms of 4 hour duration (T = 2, 5, and 10 years) were used. The S i of Table 3 5 show that for a tied ridged system, MDS is the second most importa nt parameter for GAML predicted infiltration, meaning that for tillage with non negligible surface storage, the parameter MDS is shown to explain more output variability in F and RO than all other parameters except K se ; parameters ranked by importance are: K se f s i Of the five GAML parameters, K se and MDS are the two parameters that are the least likely candidates for elimination from the solution or for fixing at a constant value for uncertainty analysis. Also differences in S i of MDS between t he two forms of GAML solutions gives more evidence that it does matter how h p is included in the GAML equations, as indicated by the percent differences in MDS S i between complete and standard GAML solutions (Table 3 5). U nder tied ridging, using the sta ndard GAML solution form resulted in decreases in the S i for MDS of around 23%, making it of importance comparable to f but MDS remained the second most important parameter using the standard GAML form. For a conventionally tilled system, sensitivity of infiltration to MDS is expectedly smaller but is still significant and is greater by importance are: K se f s i There were significant changes in MDS S i between solution forms for simulations using CT, but the lower overall S i of MDS compared to those under TR, mean that the manner in which H is included in the GAML solution is of little importance for conventionally tilled systems.

PAGE 97

97 Plotting S i for the GAML parameters, Figure 3 6 illustrates the strength of influence of parameters on the variability of the outputs of interest (RO and F). Given the parameter ranges and distri butions for sandy loam soils and tillage systems typical of this study area, it is evident that MDS becomes a significantly more important parameter if there is tillage having surface storage ranging from 15 50 mm. However, in systems having less surface storage (0 10 mm), MDS is one of the least important parameters. Figure 3 6 suggests that in studies of infiltration and runoff in conventionally tilled agricultural systems, most efforts should be directed toward improving estimates of K se f It also suggests that in systems employing some type of tillage to reduce runoff, most efforts should be directed toward improving estimates of K se and MDS. The sum of all S i was about 0.98, meaning that the model is almost perfectly additive; therefore, any refinement to a parameter will result in direct reductions in output uncertainty. Conclusions on the Importance of Surface Storage Depth for Tillage Parameterization One outcome of sensitivity analysis is to guide investigators about which uncertain inpu t factors contribute the most to the model output uncertainty. This study shows that K se and MDS are the most important factors controlling total infiltration (F) and runoff (RO) in systems having tillage with significant depression storage. As a result, the most attention and resources spent on parameterization of GAML predicted infiltration should focus on the parameters to which infiltration is most sensitive. Substantial differences in predicted runoff and infiltration depths between conventional (CT) and tied ridge tilled (TR) systems were observed; therefore, MDS was shown to be a useful parameter for describing water harvesting tillage. The results demonstrated here show small differences in F and RO for CT between the two solution

PAGE 98

98 forms, standard a nd complete, suggesting that the typical simplification of GAML to the standard form is an adequate solution for GAML predictions under conventionally tilled systems. However, the marked differences in first order variance based global sensitivity indexes S i for F and RO depths between the two solutions for TR suggest that h p should be included both in the water balance and in the Green Ampt equations. The main advantage of the complete GAML solution over the MDS=0 and standard GAML solutions is that it represents a simple way to describe water harvesting tillage (using the MDS parameter and the accompanying time varying h p ) and to reduce the uncertainty of infiltration and runoff predictions. Thus, hydrologic models equipped with the complete GAML soluti on could become useful tools for estimating the landscape scale impacts of water harvesting tillage (of different MDS and areal extent) on infiltration, evapotranspiration, and groundwater recharge. Based on the predicted infiltration increases from wate r harvesting tillage and also on the growing evidence of changing precipitation character, it could be expected that surface storage of excess rainfall will become an increasingly important management concern in agricultural areas and that there may be an increase in extent of agricultural systems having some form of water harvesting tillage, including tied ridging, contour bunding, or pit tillage. Therefore, the advantages of the complete GAML solution demonstrated here could provide improved predictions of infiltration and runoff for field and landscape scale hydrologic models employing standard or MDS=0 GAML solutions.

PAGE 99

99 Figure 3 1. Illustration of infiltration modeled by Green s i are volumetric water content, saturated and initial, respectively, h p is depth of f is wetting front suc tion Fig ure 3 2 Typical furrowed field beside tied ridged field following rain event (image from Jones and Baumhardt 2003)

PAGE 100

100 Table 3 1. Parameter values needed for Green Ampt model: minimum, maximum, mean values and estimated probability distribution i ), s ), effective hydraulic conductivity (K se ), wetting f ), and maximum depression storage depth (MDS) for conventional tillage (CT) and tied ridge tillage (TR) Parameter Min Max Mean Units Distribution 1 0.1 0.2 0.1 mm/mm uniform 1 0.4 0.6 0.5 mm/mm uniform 1 26.7 194.0 110.4 mm uniform Ke 2 1.1 44.1 15.9 mm/hr log MDS (CT) 3 0.0 10.0 5.0 mm uniform MDS (TR) 4 15.0 5 0 .0 32.5 mm uniform 1 Rawl s et al., 1983; Fangmeier et al., 2005 2 laboratory analysis of local soils in Wargal study area 3 Kamphorst et al., 2000; Planchon et al., 2002; Guzha et al., 2004 4 Guzha et al., 2004; Jones and Clark, 1987 Figure 3 3. Empirical CDF for annual m aximum series (AMS) rainfall intensity for 4 hour storms; Weibull, Gamma, and Gumbel CDF fitted to AMS data using maximum likelihood parameter estimation 0 0.2 0.4 0.6 0.8 1 0 20 40 60 80 100 120 Cumulative probability Annual maximum series rainfall intensity (4 hour storm; mm/hour) AMS data CDF Weibull CDF Gamma CDF Gumble CDF

PAGE 101

101 Table 3 2 IDF data (1993 2008 hourly rainfall), Weibull distribution used, values in the table ar e rainfall intensities in mm/hour Duration t, (hours) T, return period (years) 2 5 10 15 25 50 75 100 1 42.3 62.4 73.6 79.3 85.9 94.0 98.3 101.3 2 29.0 45.8 55.5 60.6 66.6 74.0 78.0 80.8 4 17.3 29.3 36.8 40.7 45.4 51.4 54.7 56.9 8 10.0 17.3 21.9 24.3 27.2 30.9 32.9 34.3 24 4.7 8.6 11.2 12.6 14.2 16.4 17.6 18.5 Figure 3 4. Infiltration (F) and runoff (RO) for a 4 hour 2 year design storm; complete and standard GAML solutions 0 20 40 60 80 100 120 140 160 180 0 5 10 15 20 25 30 35 40 45 50 F, RO depths (mm) MDS (mm) F, complete GAML F, standard GAML RO, complete GAML RO, standard GAML CT TR

PAGE 102

102 Table 3 3 4 hour design storms of 2, 5, 10 year return period used for rainfall input in GAML sensitivity analysis; 20 min time step Duration Design storm intensity, mm/hr hours 2 year 5 year 10 year 0.00 0.0 0.0 0.0 0.33 24.8 29.8 34.2 0.67 27.2 32.7 37.6 1.00 30.5 36.7 42.2 1.33 35.5 42.6 49.0 1.67 43.7 52.6 60. 4 2.00 62.6 75.3 86.5 2.33 89.6 107.7 123.8 2.67 50.8 61.0 70.1 3.00 39.0 46.8 53.8 3.33 32.7 39.3 45.2 3.67 28.7 34.5 39.7 4.00 25.9 31.1 35.8 Table 3 4 GAML predicted runoff (RO) and infiltration (F) depths of conventional (CT) and tied ridge (TR) tillage for 2 year, 4 hour storm; standard and complete GAML solutions. Min, mean, and max MDS were 0, 5, 10 mm for CT and 15, 32.5, 50 mm for TR Runoff Infiltration Complete GAML solution MDS TR CT MDS = 0 TR CT MDS = 0 Min 40.7 59.7 59.7 123 .0 104.0 104.0 Mean 18.9 53.3 59.7 144.8 110.3 104.0 Max 0.0 47.0 59.7 163.7 116.7 104.0 Runoff Infiltration Standard GAML solution MDS TR CT MDS = 0 TR CT MDS = 0 Min 44.7 59.7 59.7 119.0 104.0 104.0 Mean 27.2 54.7 59.7 136.5 109.0 104.0 Max 9.7 49.7 59.7 154.0 114.0 104.0

PAGE 103

103 Figure 3 5 Infiltration and runoff % changes from mean MDS values for 2 year, 4 hour storm for both GAML solution types; min, mean, max are values of MDS for conventional (CT: 0, 5, 10 mm) and tied ridge (TR: 15 32.5, 50 mm) tillage 15 10 5 0 5 10 15 min mean max % Change in infiltration from mean MDS infiltration CT: complete GAML TR: complete GAML CT: standard GAML TR: standard GAML 150 100 50 0 50 100 150 min mean max % Change in runoff from mean MDS infiltration CT: complete GAML TR: complete GAML CT: standard GAML TR: standard GAML

PAGE 104

104 Table 3 5 First order sensitivity indexes (S i ) for output F of the 5 GAML parameters for tied ridge (TR) and conventional tillage (CT) for both GAML solution forms, complete and standard Tillage type S i of GAML paramete rs solution form K se f s i MDS storm t, T %change TR 0.8210 0.0343 0.0117 0.0069 0.1105 4 hour, 2 year complete 0.8384 0.0358 0.0115 0.0062 0.0891 4 hour, 5 year GAML 0.8425 0.0375 0.0131 0.0062 0.0815 4 hour, 10 year TR 0.827 3 0.0578 0.0099 0.0053 0.0858 4 hour, 2 year 28.8% standard 0.8378 0.0571 0.0091 0.0048 0.0685 4 hour, 5 year 30.1% GAML 0.8390 0.0588 0.0106 0.0049 0.0623 4 hour, 10 year 30.8% CT 0.8982 0.0588 0.0115 0.0063 0.0115 4 hour, 2 year com plete 0.8951 0.0573 0.0106 0.0055 0.0089 4 hour, 5 year GAML 0.8925 0.0581 0.0118 0.0054 0.0081 4 hour, 10 year CT 0.8965 0.0641 0.0110 0.0060 0.0079 4 hour, 2 year 45.6% Standard 0.8934 0.0620 0.0101 0.0053 0.0059 4 hour, 5 year 50.8 % GAML 0.8905 0.0627 0.0113 0.0052 0.0054 4 hour, 10 year 50.0% 1 % change in MDS S i is the increase in MDS S i when changing from standard to complete GAML solution form

PAGE 105

105 Figure 3 6 Total effect sensitivity indices (outputs are F and RO; their sen sitivities to each parameter are equal) for Green Ampt parameters for tied ridge and conven tional tillage for 4 hour design storm of 2 year return period 0 0.2 0.4 0.6 0.8 1 First order sensitivity indices TR: complete GAML TR: standard GAML CT: complete GAML CT: standard GAML K se f s i MDS

PAGE 106

106 CHAPTER 4 EVALUATION OF AGRICU LTURAL MANAGEMENT AL TERNATIVES FOR SUSTAINABLE GROUNDWA TER IN INDIA : SIMULATED WATER BA LANCE RESULTS Overview of Water Balance Simulation Groundwater Management in India There are 170 Mha of cultivated land in Indi a more than half the total land area of 328 Mha (FAOSTAT 2010 ). Irrigated areas are 45% (77 Mha) of total cultivated area (AQUASTAT 2010 ), and more than half of the irrigation requirements are from groundwater resources ( CGWB 2002; Shah et al., 2003 ). Total groundwater withdrawals of 251 billion km 3 are the highest of any nation (AQUASTAT 2010 ) The harv ested area of rice croplands in India (44 Mha) and the irrigation diversions for rice areas are greater than those of all other crops, and rice production at the national scale is the most valuable agricultural activity valued at $30 billion/year (FAOSTA T 2010 ). Field scale water balance experiments in the Wargal study area have shown a groundwater balance of about 690 mm (1235 mm or irrigation; 542 mm of return flow recharge) for a rainy season rice cropping system (Reddy, 2009) All these data shoul d lead to the conclusion: rice cropland areas in India are extensive, highly important economically, have large consumptive water use, and depend heavily on groundwater resources. For farming households which make up about half of the population of Indi a (FAOSTAT, 2010; 2009 population of India ) groundwater depletion is concerning because it is expensive and increases vulnerability A study of the costs of groundwater depletion in Andhra Pradesh has shown that increasing groundwater recharge is the mos t economical way to manage groundwater depletion (Reddy, 2005) The w ater

PAGE 107

107 balance simulations of this research in the Wargal watershed have served to quantify the impacts on the groundwater balance of recharge structures (small reservoir s), current farm m anagement practices, and alternative farm management options. The costs of groundwater depletion refer to capital losses resulting from borewell infrastructure being ineffective due to lower groundwater levels, new required investments to deepen borewells and lower pumps, higher energy costs for groundwater withdrawal, and the indirect costs of lower net incomes from reduced areas of flooded rice croplands. It was shown in the study by Reddy (2005) of three villages in northern Andhra Pradesh that groundw ater depletion costs are most damaging to farmers with smaller land holdings, and disparities in groundwater access and control generally develop those having larger land holdings are able to adapt to groundwater depletion and retain access to groundwate r resources. The options for managing groundwater depletion are numerous, but they can be simplified into the following categories: increasing surface storage and infiltration of excess rainfall (to increase groundwater recharge) reducing groundwater with drawals, and decreasing evapotranspiration (ET) in rainfed and irrigated croplands. This is explained in the conceptual model and discussion of surface and groundwater balances in Chapter 1. The simulated water balance experiments presented in this chapt er serve to examine some of the possible agricultural management options for groundwater sustainability ; the best options for groundwater sustainability are evaluated based on predictions of groundwater recharge and withdrawals, ET, a nd estimated household incomes.

PAGE 108

108 Goals of Simulated Water Balance Experiments In consideration of the significance of groundwater resources in India, t here are three hypotheses being tested in this chapter by simulating the water balance of the Wargal watershed : (1) current rice cropland extent and management practices are depleting groundwater supplies, (2) tillage for water harvesting can significantly increase groundwater recharge in rainfed croplands, and (3) there are combinations of tillage, crop selection, and irrigation t hat can improve groundwater recharge and reduce groundwater withdrawals T he objectives of the simulation experiments : Assess groundwater sustainability of agricultural management practices (crop selection, irrigation management, tillage) at a watershed s cale quantify the groundwater balance (recharge irrigation pumping) under current and alternative agricultural management Evaluate improvements to the Green Ampt infiltration routines of a hydrologic model (Soil and Water Assessment Tool SWAT; Arnold et al., 1998 ; Krysanova and Arnold, 2008 ) for simulating groundwater balance under various cropping systems Water Balance Simulation Methods SWAT Preparation Preparing SWAT for hydrologic simulations required spatial datasets for soils, topography, climat e, and LULC as described in Chapter 1: Summary of Water Balance Simulation Methods Detailed management of agricultural areas was described using planting dates, harvest dates, and irrigation information based on data from the Wargal area. Code modific ations were made (Appendix A ) to accommodate a surface storage depth for parameterization of water harvesting tillage. Agricultural management alternatives were prepared by developing text files to change cropland areas, irrigation depths, and tillage ind uced MDS.

PAGE 109

109 Describing agricultural management Dates for planting and harvesting were those typical of kharif (June September rainy season), rabi (October February) and summer seasons (March May). Flowmeters on borewells were used to estimate daily irriga tion applications for rice in k harif (N = 6 where N is the number of seasonal flowmeter records) and rabi (N = 5) seasons and for vegetables in summer (N = 4 ). Flow volumes were monitored by research partners from ANGRAU, measured field sizes allowed for conversion to application depth. Based on flowmeter data, daily irrigation applications in k harif and rabi seasons for rice were 8 mm/day and were 4 mm/day for vegetables in summer season. Modeling c ropping system sequences ( in kharif, rabi, and summer) required adjusting cropland areas dynamically 3 times each year. This was accomplished in SWAT using a lup.dat (land use update) file and 3 text files which described changing cropland areas that were read in by SWAT at specified dates each year. The se asonal cropland area files and lup.dat were also used for describing changed cropland areas for alternative management options. To compare the water balance results of alternative agricultural management, existing management and 25 alternative management s cenarios were simulated (Table 4 1) both for observed weather data (2009 2010) and for a 10 year simulation using generated weather data (2000 2009) from WXGEN (Sharpley and Williams, 1990) included in SWAT. WXGEN weather data were generated based on a da tabase of 168 climate param eters for the W argal watershed. Simulations were completed for both measured and generated weather data because the years of 2009 and 2010, for which there are observations of weather data, are not necessarily good representatio ns of average annual rainfall. In 2009, annual rainfall was 667 mm (about 1 1 0 mm below

PAGE 110

110 normal); in 2010, annual rainfall was 1022 mm (about 2 4 0 mm above normal). The use of simulated weather data also resulted in some dry and some wet years, but the aver age annual rainfall during the 10 years (792 mm) was very close to the long term average in Wargal of 7 8 0 mm. Most of the results presented are from the 2000 2009 generated weather dataset because it was decided that the rainfall data of the generated dat aset better represented average conditions. The trends in results were nearly identical for both weather datasets. SWAT process modifications To test the significance of tillage management for increasing groundwater recharge the GAML infiltration routines of SWAT were modified to allow for parameterization of tillage using a maximum depression storage (MDS) value see Chapter 3 and E quation 3 2. In order for the modified infiltration routine to work proper ly, 2 2 subroutines of SWAT required alterations ( Appendix A ). The subroutines requiring the most extensive changes were those for GAML infiltration (surqgreenampt.f), actual evapotranspiration (etact.f), irrigation from aquifer (irrsub.f), and tillage des cription (newtillmix.f and readtill.f). The i nfiltration routine was altered to allow for accumulation of surface storage of excess rainfall which created a time varying head that influence d infiltration rates (Chapter 3) After modifications, s urface st orage accumulate d up to the MDS value ; additional excess rainfall b ecame surface runoff ET routine changes were made in order for stored surface water in cropland areas to be evaporated; evaporation rates were dependent on leaf area index as modeled by S WAT. The irrigation routine was changed so that irrigation applications could be stored on the surface if they were in excess of infiltration rates and there was a non zero MDS value The tillage description

PAGE 111

111 routines were changed to add the MDS parameter to the till.dat tillage database and to create the correct MDS in an area after a tillage operation with non zero MDS was completed. These changes have allowed for ponded rice to be simulated more accurately and in combination with dynamically varying cr opland areas; this was not possible with previous releases of SWAT. SWAT Calibration and Evaluation Model calibration is the estimation of parameters of model to improve its predictions of observed data. Model evaluation is the quantitative assessment o f the performance of a model in predicting observations which were not used for model development or parameter estimation (calibration). The target for SWAT calibration and evaluation was a time series of 5 0 reservoir volumes f rom 3 / 19 /20 10 to 12/3/2010 ( Figure 4 1); this was the best available runoff gauge given that there is limited channelized flow in the watershed. The reservoir at the outlet of the Wargal watershed creates a generally closed watershed, and the volume of water stored in this reservoir was considered a good substitute for streamflow data which are typically used to calibrate and evaluate hydrologic models. This time series of reservoir volumes begins in a dry portion of the year and extends beyond the rainy kharif season. As discussed in Chapter 1, these observations were obtained from GPS waypoint data and tank level gage data to measure flooded area and depth of water, respectively. The 13 most important parameters of SWAT were varied across a specified range based on field data an d literature values. Varying these 13 parameters required re writing 838 text files for input to SWAT; therefore, automated calibration software wa s required. SWAT CUP3 ( Abbaspour, 2008 ) a program developed by the systems analysis and modeling group of Eawag ( Swiss Federal Institute of Aquatic Science and

PAGE 112

112 Technology ), was used for SWAT calibration, evaluation, and sensitivity and uncertainty analysis. Generalized likelihood uncertainty estimation (GLUE) was implemented (Beven and Binley, 1992) in SWAT C UP3. GLUE requires few assumptions about parameter distributions and is a simple model independent method for parameter estimation. GLUE allows for the non uniqueness of model realizations, meaning that numerous parameter sets can result in a model havin g similar performance. Model uncertainty comes only from parameter uncertainty under GLUE methodology. There is flexibility in the choice of performance measure or objective function. Here the Nash Sutcliffe efficiency (NSE) was used; it is one of the 3 quantitative model evaluation techniques recommended in the model evaluation guidelines work of Moriasi et al. (2007). NSE is expressed by : NSE = ( 4 1 ) where Y i obs is the i th observation for the output of interest, Y i sim is i th simulated value, and Y mean is the mean of all the observations us ed (n total observations). The range of NSE is > 0.5 is said to describe satisfactory model performance (Moriasi et al., 2007). Model evaluation was by g raphical comparison NSE, percent bias (PBIAS), and the root mea n squared error (RMSE) observations standard deviation ratio (RSR) were the 4 methods used to evaluate model performance, following the recommendations of Moriasi et al. ( 2007). PBIAS and RSR are expressed by: PBIAS = (4 2)

PAGE 113

113 RSR = (4 3) NSE > 0.5 RSR < 0.70, and PBIAS < |25% | can be said to describe satisfactory model performance (Moriasi et al., 2007) for monthly data of streamflow. These guidelines should of course be adjusted based on prediction variable (sediment, phosphorus, or others) observation time step, and measurement uncerta inty (Moriasi et al., 2007). These guidelines are generally more stringent for streamflow predictions than for other variables (water quality) and model performance is usually worse as observation time step is reduced ( Engel et al., 2007; Yuan et al., 200 1) ; that is, model performance on a monthly basis can be expected to be greater than that on a weekly or daily basis The average observation time step of the reservoir volumes used for model calibration and evaluation in the Wargal watershed was 5 days (range of 1 to 28 days) There was one year of reservoir volumes available for use for model calibration and evaluation; therefore, some strategy was required to be able to use all the data for both model evaluation and calibration. K fold cross validatio n is one strategy that allows all the data to be used f or parameter estimation and for model evaluation (Wallach et al., 2006) Cross validation is similar to the method of splitting a dataset into 2 parts and using one part for parameter estimation and o ne part for evaluation of the model. In cross validation, an observation or group of observations is removed from the dataset and parameter estimation proceeds. Since the observations removed were not used for parameter estimation, the model can be safel y evaluated for its ability to simulate the removed observations. Next, the observation or group of observations is replaced, and the next observation or group of observations is removed and the model

PAGE 114

114 calibrated again. This continues until all data have been removed and replaced. In the Wargal area, groups of 10 observations from the record of 50 were removed, meaning 5 fold cross validation was used. The result is 5 different estimates of the parameter vector which are each used separately to predict t he reservoir volume to evaluate the model. One of the advantages of cross validation is that all the data are used in the same way for both model calibration and evaluation (Wallach et al., 2006). Deciding which parameters to vary (and by how much) is a n important part of model calibration. The literature on calibration of SWAT was used to assist with choices of parameters for calibration (Bosch et al., 2004; Van Liew et al., 2005 and 2007). A departure from typical practice was that saturated hydrauli c conductivity ( K s ) for soils was varied separately for rice croplands and other areas due to the uniquely low K s of puddled rice soils. Puddling describes the tillage of rice fields when they are flooded; the resulting particle settling in order of parti cle size (largest to smallest) results in a soil of very low hydraulic conductivity ( Kukal and Aggarwal 2002; Marechal et al., 2006). More than 15 GLUE calibration/sensitivity runs (each of N=5000 or N=2500) were completed using anywhere from 5 to 14 par ameters for calibration; these preliminary calibration efforts were used to decide on the final 13 parameters to calibrate and how much they should vary. Parameter ranges were carefully selected (based on observations in Wargal and literature values) to e nsure values did not depart from realistic ranges for the soils and land use of Wargal. Parameter sensitivity and uncertainty analysis A complete global sensitivity analysis of SWAT i s not an objective of this work; this has been done well by others (van Greinsven et al., 2006) but it is important to have a meaningful ranking of parameters by order of importance. A sensitivity analysis

PAGE 115

115 was completed through SWAT CUP3 to determine parameter sensitivities using the multiple regression system: ( Muleta and Nicklow 2006 ; Abbas p our et al., 2007) This system regresses the Monte Carlo sample d parameters from GLUE against the predicted reservoir volumes g A t test and accompanying p value are used to compare the relative sens itivities of each parameter b i ; p value is the probability that the absolute value of the regression coefficient i would be as large or larger if there was no relationship between that parameter b i and the output, g (Helton and Davis, 2000) A low p valu e corresponds to an important parameter (low probability of the parameter not making significant changes to predicted reservoir volume ) The sensitivities calculated are based on linear approximations of the sensitivity of the model ou t put to changes in a parameter while all other parameters are changing ; therefore, it is a type of global sensitivity ( Muleta and Nicklow, 2006 ) ; this method requires a large number of simulations and is dependent on choice of parameter ranges (Abbasour et al., 2007) SWAT CU P3 (Abbaspour, 2008) uses a flexible format to assist with generation of parameter sets for calibration and model sensitivity analysis. The following system defines the range of a parameter: x__.__________ . X is the type of change applied to the parameter (x = v, a, or r) : v__ means the original parameter value is to be replaced by a value in the specified range, a__ means a value from the specified range is added to the original parameter value, and r__ means the original parameter value is multiplied by (1+ a value in the specified range ). is the SWAT parameter name, is the SWAT file extension, is the soil

PAGE 116

116 hydrologic group (A, B, C, D; optional), is the soil tex ture (optional), is the landuse category (in this study it is just crop type, optional), is the number of the subbasin in the watershed (in this study it is 1 to 6, optional), and is the class of slope range (optional). The 13 parameters varied for sensitivity and uncertainty analyses were in order of importance based on the step wise regression based global sensitivity analysis described earlier : SOL_AWC(layer#): available water holding capacity of the soil layer in mm water / mm soil. This was varied in soil layers 1 and 2 for Wargal. SOL_K(layer#): saturated hydraulic conductivity of the soil layer in mm/h r. This is actually 2 parameters as t his was adjusted separately in layer 1 of rice croplands. RES_K: hydraulic conduct ivity (mm/hr) of a reservoir bottom. CN2: the initial SCS curve number for moisture condition II; it is a function of soil permeability, land use, operations, and soil moisture EVRSV: reservoir evaporation coefficient (0 1) used to adjust for shading and vegetation cover. GWQMN: the threshold depth of water (mm) in the shallow aquifer required for baseflow to occur. EPCO: plant water uptake compensation factor; this is another calibration parameter similar to ESCO. EPCO ranges from 0.01 to 1, and the clo ser to 1 EPCO is the more plant water uptake is allowed to come from deeper soil layers. ESCO: soil evaporation compensation factor; this is a calibration parameter that adjusts the depth in the soil profile from which evaporation can occur. ESCO ranges f rom 0.01 to 1, and values closer to 1 increase the depth at which evaporation from soil can occur. ALPHA_BF: is an index of the groundwater flow response to recharge changes; it is close to 1 for land with fast baseflow response to recharge and it is aroun d 0.1 0.3 for land with slow baseflow response to recharge. RCHRG_DP: the fraction of percolation from the root zone which reaches the deep aquifer. EVLAI: leaf area index at which no evaporation occurs from surface water beneath a canopy; this is used in wetlands and in this study is used for rice croplands

PAGE 117

117 GW_DELAY : the number of days required for water th at drains out of the root zone to reach the shallow aquifer where it is stored, becomes baseflow, or becomes recharge to deep aquifer. It may be surpri sing to notice that CN2 is the fourth most important parameter even though Green Ampt is used to predict runoff and infiltration in place of the SCS Curve Number method This is because CN2 is used by SWAT in the equations to update the Green Ampt effecti ve hydraulic conductivity (K se ) each day based on soil moisture ; CN2 is adjusted daily for soil moisture The routine to calculate K se was modified to remove CN2, but the subsequent calibration attempts showed that this code modification resulted in poor model performance; therefore, the original SWAT algorithm for K se which includes CN2, was retained. An important objective of this research was to evaluate the inclusion of surface storage term in Green Ampt infiltration to parameterize water harvesting t illage. The maximum surface storage term (MDS) was not included in the automated calibration and sensitivity analysis because the only areas under existing management with non negligible MDS were rice croplands, and the depth of MDS (150 mm) in these area s is not really uncertain. The s ensitivity of groundwater recharge and surface runoff to MDS in SWAT was evaluated after the overall sensitivity analysis by using the calibrated model with other parameters fixed while varying MDS depth in rainfed cropland s in different seasons. This local sensitivity gives easily understandable information about the response of recharge and other water balance components at both basin and field scales to MDS changes Uncertainty analysis is a quantitative assessment of mo del prediction variability for the purpose of checking that model output behavior is acceptable for the given variability in model input variable and parameters. In this analysis, the variability in

PAGE 118

118 parameters was assumed to be the only source of input va riability. The ranges of values for the 13 input parameters (same as those used for final model calibration) were developed based on measured data from the study area, on values from the literature, and on numerous GLUE parameter estimation runs. A graph ical evaluation of uncertainty was used here by plotting the 95% prediction uncertainty range (95 PPU; Abbasour et al., 2007) together with the observed and simulated reservoir volumes Given the ranges of parameters included in parameter estimation, simu lated values can be expected to fall in the 95 PPU with probability 0.95. The 95 PPU was developed by calculating the 2.5 th and 97.5 th percentiles of the cumulative distribution for each simulated point (Abbaspour, 2008) Ideally, all the observations sh ould fall in the 95 PPU and it should be as narrower as possible; though the narrowness depends on the variability assigned to each input parameter Reservoir volume Before initiation of this project, t he Wargal watershed was an ungauged basin, meaning tha t there were hydrologic observations of insufficient quantity or quality for many practical applications (Sivapalan et al., 2003). Some target for model calibration and evaluation is required; therefore, it was decided to use reservoir volume as an easily measurable runoff gauge. There are six small reservoirs in the 512 ha Wargal watershed. These were constructed several decades ago according to farmers in the area, by excav ating earthen dams (about 2 5 m high) at locations of seasonal surface flow for the purpose of storing rainfall runoff to increase groundwater recharge. The maximum storage volumes and surface areas of these reservoirs are listed in Table 4 2 Reservoir characteristics were determined from a combination of topographic survey ing and GPS tracking. The largest reservoir, called Kothakunta and found in

PAGE 119

119 subbasin one, is at the outlet of the watershed and this is the reservoir that was monitored. Detailed surveying in the reservoir was used to make geometric simplifications to develop an area volume relationship. The equation relating reservoir area to volume that was used here is of the form C = a A b ( Sawunyama et al., 2006) observations. This general form has been shown to perform very well in predicting reservoir volumes based on measured areas in west and southern African ( Sawunyama et al., 2006 ; Liebe et al., 2009). From measurements of reservoir geometry for reservoir number one in Wargal (this i s at the outlet of the watershed), the area volume relationship developed was V = 0.00857 A 1.428 where volume is in m 3 and area is in m 2 This is very similar to that developed by Liebe et al. (2005). Reservoir depth and area were monitored weekly or sub weekly. Depth was measured by reading the water level on a graduated staff that was installed in the reservoir. Area was measured by traversing the reservoir at the waterline with a handheld differential GPS unit and recording waypoints every few met ers The waypoint data were processed in a GIS to construct polygons of reservoir water surface area from which area could be calculated Area and depth data on concurrent days were used to develop gauge depth to area relationships of A = 9874.6 D 1.600 5 for A <= 105,000 m 2 and A = 7564.9 D 2.0064 for A > 105,000 m 2 Two methods were used for reservoir monitoring because of delays in installation of the level gauge; during this time, the GPS method was effective but more time consuming. The resulting time series of rese rvoir volumes ( Figure 4 1) was used, in conjunction with recharge observations, for calibration and evaluation of SWAT.

PAGE 120

120 Groundwater recharge Reliable groundwater recharge predictions are probably the most important outcome of hydrologic modeling in the Wargal study area. Irrigation withdrawals are generally well known based on flowmeter data in Wargal and confirmed in the literature for a nearby watershed (Mare chal et al., 2006). Groundwater recharge, however, is very difficult to meas ure; it is highly spatially variable due to the variety of land management in Wargal. Flooded rice croplands, water harvesting reservoirs, rainfed croplands, and unmanaged rangelands all have very different contributions to groundwater recharge. While be ing only a surface water model and not capable of simulating groundwater flow, SWAT is able to estimate recharge from all these different areas. Knowing recharge and irrigation withdrawals gives an evaluation of the simplified groundwater balance (see Cha pter 1): groundwater recharge minus irrigation withdrawals recharge was to use annual observations of recharge based on changes in groundwater levels during wet and dry per iods. A method for quantifying recharge from water table changes, called the double water table fluctuation method (DWTF), allows both specific yield (Sy) and recharge to be estimated (Marechal et al., 2006). The DWTF method is useful only in areas where there are very clear wet and dry parts of the year, with negligible natural recharge occurring during dry parts of the year. The water table fluctuation method (WTF), discussed in Chapter 1 as the leading physical recharge measurement method, connects ch anges in groundwater table elevations to aquifer storage changes ; it is only applicable in unconfined aquifers These quantities are related by the aquifer c haracteristic, specific yield, Sy sometimes called drainable or

PAGE 121

121 effective porosity. Sy is the vo lume of water drained from a n aquifer divided by the total bulk aquifer volume: Sy = Vwd / Vt. The main weakness of the WTF method is that specific yield is generally not known or it is not known at the required scale; for a large aquifer, there can be si gnificant spatial variability in Sy. The groundwater balance equation from Chapter 1 can be written as (Marechal et al., 2006): R + RF + Qon = ET + IP + Qoff + S ( 4 2 ) where R is natural recharge, RF is return flow recharge, Qon and Qoff are subsu rface flows in and out of the system, ET is evapotranspiration from groundwater, IP is irrigation pumping, and is change in groundwater storage. As discussed in Chapter 1, it is proposed that Qon = Qoff because of the similarity of the surrounding land scape and farming system management and ET from groundwater is negligible because of the large depth of groundwater (15 m) can be written in terms of specific yield and change in groundwater level: = Sy h. The groundwater balance can then be wr itten as: R + RF = IP + ( 4 3 ) This equation can then be used separately for wet and dry periods; for dry periods, R is zero and is negative. The dry period can be used to calculate Sy; irrigation pumping (IP) is known from a sample of f lowmeters on borewells and return flow recharge (RF) is known from upscaling plot scale water balances of irrigated rice and vegetables: Sy = (RF dry IP dry ) / dry ( 4 4 ) The wet period is then used to calculate natural recharge: R = IP wet + Sy wet RF wet ( 4 5 )

PAGE 122

122 Results : SWAT Calibration, Evaluation, and Parameter Sensitivity SWAT performance in predicting outlet reservoir volumes was very good for the calibration period (NSE = 0.954) and was satisfactory for the cross validation period (NSE = 0.729). Graphical comparison s (Figures 4 1 and 4 2) of simulated and observed reservoir volumes also show acceptable simulation performance. Good agreement was found between annual groundwater recharge simulated by SWAT and recharge observed from DWTF method (Table 4 3). Due to the fewer data available for parameter estimation for cross validation, model performance suffers as shown by the higher PBIAS and RSR values (14.7 and 0.525, respectively; Table 4 4 ). Cross validation model performance ca n still be called satisfactory based on the Moriasi et al. ( 2007) recommendations: NSE > 0.5, RSR < 0.70, and PBIAS < |25%| To increase confidence that SWAT was indeed routing runoff (about 40 mm for whole watershed for the monitoring period) from the wat ershed to the monitored reservoir at the outlet of Wargal, some simple water balance calculations were completed. Direct rainfall input to the reservoir, was about 116,000 m 3 based on the observed reservoir areas and daily rainfall values. Maximum obser ved reservoir volume exceeded 350,000 m 3 ; therefore, considering evaporation and percolation of stored reservoir volume, it can be concluded that the majority of flow to the reservoir is from surface runoff from the watershed. The regression based sensi tivity analysis from the GLUE generated parameter sets found that soil available water content (AWC) and saturated hydraulic conductivity in rice croplands were the most sensitive parameters. Having p values less than 0.1, the 6 most important parameters of the 13 included in parameter estimation and sensitivity analysis were, in order of importance (Table 4 5): soil available water content

PAGE 123

123 in soil layers 1 and 2 (r__SOL_AWC(1,2).sol), saturated hydraulic conductivity of soil in layer 1 of rice croplands ( v__SOL_K(1).sol______RICE), seepage rate of reservoirs (v__RES_K.res), SCS Curve Number for moisture condition 2 (r__CN2.mgt), reservoir evaporation coefficient (v__EVRSV.res), and saturated hydraulic conductivity of soil in layer 1 of all non rice areas ( r__SOL_K(1).sol). The best estimates of all parameters are given in Table 4 5 also. Uncertainty of Model Predictions Uncertainty of model predictions is represented by t he 95% predic tion uncertainty (95 PPU), the range of predicted reservoirs between th e 2.5 and 97.5 percentiles of all predicted values as the 13 calibration parameters vary across their ranges. This is presented graphically in Figure 4 1: 64% of observations are contained by the 95 PPU. Typically, around 80% of observations can be expec ted to be contained in the 95 PPU ( Abbaspour et al., 2007 ), the lower value obtained here (64%) can still be considered as an acceptable amount of uncertainty given the other measures of model evaluation that were quite good. The prediction difficulty inc reases as observation time step decreases ( Moriasi et al., 2007 ), and it can be expected to be more difficult to be predict a cumulative output (reservoir volume) than a rate [L 3 /t]. Therefore, the 64% of observations captured by the 95 PPU can be conside red adequate. Groundwater Balance An important criterion used to compare results of different agricultural management was the groundwater balance: simplified in this case to mean groundwater recharge minus irrigation withdrawals (subsurface flows into and from the aquifer system were assumed equal). Recharge observations based on the DWTF are presented in Table 4 3; these matched well with recharge predictions from SWAT. The

PAGE 124

124 distinct wet and dry periods and the large changes in groundwater level observati ons are shown by Figure 4 3. Based on the DWTF method described above, specific yield (Sy) of the aquifer in Wargal was found to be 0.013 based on 2009 and 2010 groundwater levels (Table 4 3). This Sy value is typical of hard rock aquifers and is consist ent with that found by Marechal et al. (2006), Sy = 0.014, using the DWTF method in Maheshwaram watershed very near to the Wargal watershed. A small Sy means that even very small changes in the groundwater balance ( ) result in very large changes in the depth to the water table ( = / Sy); for example, under existing management and using generated weather data for 2000 2009, the average annual groundwater balance was 10.9 mm. This would result in water table d ecline of 835 mm for an aquifer having Sy of 0.013. Using measured weather data the groundwater balances were simulated to be 44.1 mm and 155.4 mm for 2009 and 2010, respectively, under existing management. The groundwater balance defined as recharge minus irrigation pumping is a useful quantity for evaluating agricultural management impacts on groundwater supply, as it considers the subsurface water balance components that are managed more directly. However, it is a simplification of the actual subs urface water balance in which horizontal flows would be indirectly managed. For example, in Table 4 6, the management scenarios with large positive groundwater balances would not really be expected to realize the large increases in water table height ( GW ) shown in the table. These values would be realistic only if the same management changes were made in the areas surrounding the watershed. So if the management changes that increased the groundwater balance were made only in the Wargal watershed, the resulting rise in

PAGE 125

125 water table height would be lower than that shown by Table 4 6 due to increased subsurface flows out of the Wargal groundwater system. Results: Evaluation of Management Alternatives for Sustainable Groundwater Detailed results of the effe cts of tillage and rice cropland management on the groundwater balance are given in the next 3 sections; the remainder of this paragraph is a brief summary of these results. SWAT simulations have predicted an average annual groundwater balance (recharge irrigation = 535.1 mm 546.0 mm) of 10.9 mm for the period from 2000 2009 using generated weather data. For years 2009 and 2010, SWAT predicted a groundwater balance of 44.1 mm and 155.4 mm, respectively, using observed weather data. The 6 water harv esting structures have a significant groundwater recharge impact; the predicted contribution to groundwater recharge was about 12% of the annual total (2009 2010 annual average: 75 mm recharge from tanks, 602 mm total groundwater recharge). Changing tilla ge in rainfed areas in kharif (June September) season from conventional to tied ridge (MDS = 15, 32.5, or 50 mm) resulted in an estimated 2 3.5% increase in basin scale groundwater recharge. Tied ridge tillage in rabi (October February) season rainfed areas resulted in negligible basin scale recharge increase due to the small amount of rainfall (around 20% of the annual rainfall total) and the reduce extent of rainfed croplands in the rabi season. As expected, changing the extent and irrigation manage ment of rice croplands resulted in significant changes in the groundwater balance. The predicted changes in groundwater balances for alternative management are expectedly small (relative to total recharge and irrigation depths) Ideally, additional obser vations to evaluate the uncertainty of SWAT would be used to improve confidence in the water balance predictions. Without more observational evidence, the options are (1) to trust the quality of input data, process

PAGE 126

126 descriptions of SWAT, and the SWAT evalu ations from observed reservoir volumes (NSE = 0.954) and groundwater recharge or (2) reject the quality of the predictions based on insufficient observations for evaluation. This is of course the typical quandary in hydrologic modeling (Kirchner, 2006 ): t he challenge of knowing if a model user is getting the right answers for the right reasons. Given t he long history of successful application of SWAT in a variety of systems (Gassman et al., 2007) and the good performance (based on guidelines of Moriasi et al., 2007) of SWAT demonstrated in Wargal, it is recommended that option (1) is chosen here. Significance of Tillage and MDS Changing tillage in rainfed areas from conventional to tied ridge tillage (or another water harvesting tillage ) through the introd uction of a non negligible surface storage term (MDS, maximum depression storage) does make significant impacts on the groundwater balance of the watershed. Here, existing tillage (mouldboard or chisel plow) is assumed to have negligible MDS (Chapter 3). The rainfed crops in kharif season are largely corn and cotton: areas of 105.7 and 104.4 ha respectively, or 20.6% and 20.4% of the watershed area. The rainfed crops in rabi season are largely potato and sunflower: areas of 65.6 and 37.4 ha respective ly, or 12.8% and 7.3% of the watershed area. Basin scale sensitivity of groundwater recharge to MDS depth (Figure 4 4 ) is not striking based on percent changes in recharge from existing management; groundwater recharge increase d 1.8 to 3.5% compared to re charge under existing management. However, that amount ed to a recharge increase of 9.5 to 18.4 mm/year, and adding a 3 2.5 mm MDS in rainfed crops in only in k harif season was enough to change the groundwater balance from negative to positive: 10.9 mm/yea r to 2.4 mm/year (Table 4

PAGE 127

127 6 ) For simulations using measured data (2009) ad ding a 32.5 mm MDS in rainfed crops only in k harif (rainy) season changed the groundwater balance from 44.1 mm/year to 34.2 mm/year (Table 4 7 ). Runoff depths at the basin scal e are much smaller (around 30 to 60 mm/year, depending on management scenario) than recharge depths ; therefore, the percent change in basin scale runoff in response to tillage change is much greater (Figure 4 4 ). Compared to existing tillage, runoff is re duced by 48% for MDS depth of 50 mm in k harif corn and cotton, and runoff is reduced by 31% for MDS depth of 15 mm in k harif corn and cotton. The extent of rainfed croplands (around 41% in kharif season and 20% in rabi season) suggests that in watersheds h aving a higher proportion of rainfed areas, tillage or water harvesting would have an even greater impact on the groundwater balance of a basin. Comparing runoff and groundwater recharge only from rainfed croplands where MDS changes were simulated gives a more direct assessment of MDS effects on water balance components. Figures 4 5 and 4 6 show the sensitivities of groundwater recharge and runoff to MDS depth in rainfed croplands for kharif season only (Figure 4 5 ) and for MDS in both kharif and rabi sea sons (Figure 4 6 ). Compared to annual recharge from corn and cotton areas under existing tillage, groundwater recharge from these areas increased 14.2, 19.9, and 24.3% for MDS depth s of 15, 32.5, and 50 mm respectively, in kharif corn and cotton. F or MD S depth s of 15, 32.5, and 50 mm in both kharif and rabi rainfed areas, groundwater recharge from these areas increased by 6.7, 9.4, 12.3%, respectively. The increased groundwater recharge in response to changes in MDS provides evidence that there is great er average soil moisture (deep percolation and recharge are predicted in SWAT in response to soil water content that exceeds field

PAGE 128

128 capacity) resulting from tillage induced MDS. Changes in ET in response to tillage were small but noticeable at the basin sc ale, and ET changes from tied ridge tilled areas of kharif crops increase d by as much as 5% compared to conventionally tilled rainfed kharif crops. Sensitivities of recharge and runoff to MDS have been presented in the form of percent change in water balan ce component for different MDS. It was decided to add an additional local measure of sensitivity: a very basic measure, the simple derivative, X, where represents a change in some output variable Y and X represents a change in some input factor X. Tables 4 8 and 4 9 give s ensitivity of annual surface runoff and groundwater recharge to MDS ( RO MDS and RCHG MDS, respectively) at the basin and field scales in mm change in depth (from existing tillage, MDS = 0) of recharge or runoff per mm d epth of MDS F or MDS from 15 to 50 mm in kharif rainfed areas the v alues of ranged from 1 to 1.9 in kharif rainfed croplands and from 0.3 to 0.6 at the whole basin scale. This suggests that in rainfed croplands having MDS > 15 mm, there is an increase in annual recharge of greater than 1 mm for each mm of MDS. Values for DS ranged from 0.6 to 3.3 for MDS from 15 to 50 mm in kharif rainfed areas. In summary, results showing sensitivity of simulated recharge and runoff to tillage man agement (varying MDS ) can be found in Tables 4 6, 4 7, 4 8, and 4 9, Figures 4 4, 4 5, 4 6, and 4 7. Chang ing Extent and Irrigation Management of Rice Croplands The basin scale groundwater balance under existing management was simulated to be 10.9 mm/year (546.0 mm irrigation, 535.1 mm recharge). Using the observed weather data for 2009 and 2010, t he basin scale groundwater balance under existing management was simulated to be 44.1 mm/year for 2009 (546.0 mm irrigation, 501.9

PAGE 129

129 mm recharge) and 155.4 mm/ye ar for 2010 (546.0 mm irrigation, 701.4 mm recharge). Comparing the water balance components (irrigation, recharge, ET, and groundwater balance) in Figures 4 7 and 4 8 (and Table 4 6) shows a much greater difference in water balance components in response to changes in rice cropland extent and irrigation management (Figure 4 8 ) than to changes in MDS from tillage (Figure 4 7 ). While the reduction of rice cropland areas by 50 and 75% in k harif and rabi seasons may be unrealistic, these scenarios do demonst rate the obviously substantial influence that the extent and management of rice croplands have on the groundwater balance of this watershed. However, even modest changes in irrigation or extent of rice croplands resulted in significant changes in the wate r balance. Decreasing daily irrigation depths by 25% in k harif season only resulted in a change in the basin scale groundwater balance from 10.9 mm to 18 mm ( 472.9 mm irrigation, 492.9 mm recharge) Decreasing the extent of rice croplands in both kharif and rabi seasons by 25% (rep lacing with corn and cotton in k harif and potato and sunflower in rabi) resulted in the groundwater balance increasing to 34.3 mm ( Table 4 6 and Figure 4 8 ): 418.3 m m irrigation, 452.6 mm recharge Alternatives to Rice: Irrigat ed and Rainfed Crops The yield variability of rainfed crops is one of numerous objections by farmers in the area to replacing parts of rice croplands with alternative crops (Chapter 5) One of the ways to weaken this objection is by replacing rice croplan ds with other irrigated crops. In some of the alternative management scenarios simulated here, rice croplands in rabi season were replaced by irrigated corn. Irrigation depths for corn would be lower (4 mm/day compared to 8 mm/day for rice) and the absen ce of ponded irrigation water means ET depths would be lower. Replacing rice with irrigated corn would not

PAGE 130

130 necessarily be acceptable for farmers in Wargal, but it is expected that this replacement would have less yield variability than a rainfed rice repl acement crop. Therefore, irrigated crops replacing rice might be preferred over rainfed crops replacing rice areas. A 25% reduction in k harif and rabi rice areas, with replac ement crops of rainfed corn in k harif and irrigated corn in rabi (scenario code: red_rice_k corn,r irr corn,25 ; Table 4 1 and 4 6 ) seemed a good compromise between the risk of increased yield variability and the need to address declining groundwater quantity. This scenario produced a groundwater balance of 18.5 mm (446.4 mm irrigati on, 465.0 mm recharge ; Table 4 6 ). The addition of water harvesting tillage in the rainfed kharif areas ( red_rice_k corn,r irr corn,25: mds33_kharif ) improves the groundwater balance of this scenarios to 36.1 mm (446.4 mm irrigation, 482.5 mm recharge ; Ta ble 4 6 ). Conclusions on Agricultural Management and Groundwater Supply in India The objective to improve the widely used dist ributed hydrologic model, SWAT, was achieved. The process modifications to SWAT to include the complete representation of Green Ampt infiltration by modeling a variable ponding depth, has been shown to be effective in parameterizing tillage for water harvesting. The changes are supported by the th eoretical basis of Green Ampt (Chapter 3), and they have resulted in the new ability of SWAT to use Green Ampt methods to predict changes in runoff and infiltration in response to tillage management with varying MDS. This process modification allows rice croplands to be modeled as typical cropland hydrologic response units (HRUs) in SWAT; previously rice croplands had to be modeled at potholes, an HRU in which there was rarely surface runoff and the majority of outflow was by ET and deep percolation. Rice areas were a special HRU that was parameterized with a maximum conical storage volum e, and this volume parameter meant that dynamically updating rice

PAGE 131

131 cropland areas (to model crop rotations) during the year using the lup.dat routine in SWAT was not possible. With the changes to SWAT, rice area MDS is easily adjustable, and rice croplands can be correctly and easily included in dynamic landuse updates. These improvements in SWAT are significant and can be expected to reduce prediction uncertainty as MDS estimates will likely be more certain than other important parameters like soil AWC or K s Choosing the best management practices for su stainable groundwater in Wargal, or a similar watershed is of course a complicated decision with s ome required subjectivity. Understanding the behavior of stakeholders is critical for management alternati ves to be effective. The choice in this study is based on the predicted groundwater balances and on estimates of the agronomic and economic impacts of the management scenarios being compared. The results presented here give quantitative predictions of th e groundwater balance (and other water balance components) for a multitude of management scenarios. The following 7 alternatives were chosen as the best management options for increasing groundwater supply b ased on (1) participation of growers in the area ( constraints on labor, preferences for rice cultivation, little interes t in alternative rainfed crops ( Chapter 5), (2) the groundwater balance results (Chapter 4), and (3) consideration of yield variability and risk. red_rice_irr,k,r,25 (GW balance 38.2 mm/year): Irrigation depths for rice in both kharif and rabi seasons are reduced by 25% (from 8 to 6 mm/day). While there would not likely be any water stress for rice with these reduced irrigation depths, there may be small declines in yield and increas es in labor requirements due to the intermittent absence of flooded conditions and the non aquatic weeds that might result. red_rice_k corn,r irr corn,25: mds33_kharif (GW balance 36.1 mm/year): The extent of rice croplands is reduced by 25%, being replace d with corn in kharif season and irrigated corn (4 mm/day) in rabi season. Water harvesting tillage

PAGE 132

132 (MDS = 32.5 mm) is simulated for kharif rainfed crops. Moderate increases in yield variability can be expected due to the replacement of some areas of irr igated rice with rainfed corn. Yield and returns from irrigated corn in rabi would be comparable to those of rice. Small improvements in kharif rainfed crop yield may result from the water harvesting tillage. red_rice_k,r,25 ( GW balance 36.1 mm/year 34.3 mm/year) : The extent of rice croplands is reduced by 25%, being replaced with corn and cotton in kharif season and potato and sunflower in rabi season. Moderate increases in yield variability can be expected due to the replacement of some areas of irriga ted rice with rainfed crops. red_rice_irr,k,25: mds33_kharif ( GW balance 31.3 mm/year ) : Irrigation depths for rice in kharif season only are reduced by 25% (from 8 to 6 mm/day). Water harvesting tillage (MDS = 32.5 mm) is simulated for kharif rainfed crop s. Small to negligible declines in yield may result from the irrigation reductions. Small improvements in kharif rainfed crop yield may result from the water harvesting tillage. red_rice_k corn,r irr corn,25 (GW balance 18.5 mm/year) : The extent of rice croplands is reduced by 25%, being replaced with corn in kharif season and irrigated corn (4 mm/day) in rabi season Moderate increases in yield variability can be expected due to the replacement of some areas of irrigated rice with rainfed corn. Yield a nd returns from irrigated corn in rabi would be comparable to those of rice. red_rice_irr,k,25 (GW balance 18.0 mm/year) : Irrigation depths for rice in kharif season only are reduced by 25% (from 8 to 6 mm/day). Small to negligible declines in yield may r esult from the irrigation reductions. mds50_kharif (GW balance 6.0 mm/year) : Water harvesting tillage (MDS = 50.0 mm) is simulated for kharif rainfed crops. Small improvements in kharif rainfed crop yield may result from the water harvesting tillage. The first hypothesis given at the beginning of Chapter 4, that current rice cropland extent and management practices are depleting groundwater supplies, can be considered to have sufficient evidence to be supported based on observed field scale water balances of rice croplands (Reddy, 2009), the water balance simulations (Chapter 4 Figure 4 8 ) of rice croplands and the whole watershed, and the regional data in Chapter 1. The second hypothesis, that tillage for water harvesting can significantly increase groun dwater recharge in rainfed croplands is also accepted based on the

PAGE 133

133 results of this chapter A 2 3 .5 % increase is groundwater recharge at the basin scale and a 14 24% increase in recharge at the field scale gives sufficient evidence that tillage for water harvesting can be in important part of agricultural management strategies for improving groundwater supply. The third hypothesis, that there are combinations of tillage, crop selection, and irrigation that can improve groundwater recharge and reduce grou ndwater withdrawals is also supported by the evidence from the water balance simulations The third hypothesis was likely the easiest to test and is probably the most obvious result. It hardly requires water balance estimates from a distributed hydrolog ic model to know that there are combinations of tillage, crop selection, and irrigation that can improve the groundwater balance of the watershed, but the evidence from the simulation experiments does give reliable estimates on the extent of management cha nges required (i.e. how much rice cropland area reduction is needed to change the groundwater balance from negative to positive) and their quantitative impact on various water balance components. As discussed in Chapter 1, it is essentially irrigation and recharge that are the directly manageable quantities in the groundwater balance. Simulations showed that irrigation changes ( from reduced areas of irrigated crops and/or reduced depths of irrigation of irrigated crops) in rice cropland areas greatly influ enced the groundwater balance of the Wargal area. Reduced irrigation and extent of rice croplands decreased groundwater recharge (less return flow recharge from flooded croplands) but the reduction in irrigation at the basin scale was always greater than the reduction in recharge, resulting in consistent groundwater balance improvements. This is because of the lower evapotranspiration that resulted from the irrigation reductions.

PAGE 134

1 34 The other major influence on groundwater recharge explored here was water harvesting tillage in rainfed croplands. Recharge from rainfed areas was significantly increased when water harvesting tillage (MDS of 15 50 mm) replaced conventional tillage. With growing evidence for rainfall patterns to be more commonly characterized by greater storms of high intensity (and fewer low/moderate intensity storms ; Chapter 2 ), i t is strongly recommended that tillage for increasing surface storage of excess rainfall be considered as an important management option for managing droughts and d ry spells. The increased recharge associated with water har vesting tillage can result in more reliable groundwater resource s for supplemental irrigation during droughts and dry spells; the increased infiltration associated with water harvesting tillage ca n result in soil moisture increases having direct agronomic benefits during droughts and dry spells.

PAGE 135

135 Table 4 1. Descriptions of 25 alternative management scenarios with abbreviations used Management scenario: abbreviation Description existing current agricultural management in Wargal watershed mds50_kharif tillage change: MDS of 50 mm in kharif season corn and cotton mds33_kharif tillage change: MDS of 32.5 mm in kharif season corn and cotton mds15_kharif tillage change: MDS of 15 mm in kha rif season corn and cotton mds50_rabi tillage change: MDS of 50 mm in rabi season potato and sunflower mds33_rabi tillage change: MDS of 32.5 mm in rabi season potato and sunflower mds15_rabi tillage change: MDS of 15 mm in rabi season potato a nd sunflower mds50_k,r tillage change: MDS of 50 mm in kharif and rabi mds33_k,r tillage change: MDS of 32.5 mm in kharif and rabi mds15_k,r tillage change: MDS of 15 mm in kharif and rabi red_rice_k,r,25 reduce rice cropl ands in kharif and rabi by 25%; r eplace with equal parts corn and cotton in kharif, potato and sunflower in rabi red_rice_k,r,50 reduce rice cropl ands in kharif and rabi by 50 %; replace with equal parts corn and cotton in kharif, potato and sunflower in rabi red_ric e_k,r,75 reduce rice cropl ands in kharif and rabi by 7 5%; replace with equal parts corn and cotton in kharif, potato and sunflower in rabi red_rice_k corn,r irr corn,25 reduce rice crop lands in kharif and rabi by 25%; replace with rainfed corn in khari f, irrigated corn in rabi red_rice_k corn,r irr corn,50 reduce rice crop lands in kharif and rabi by 50 %; replace with rainfed corn in kharif, irrigated corn in rabi

PAGE 136

136 Table 4 1. Continued Management scenario: abbreviation Description red_rice_k irr corn,r irr corn,25 reduce rice crop lands in kharif and rabi by 25%; replace with irrigated corn in kharif, irrigated corn in rabi red_rice_k irr corn,r irr corn,50 reduce rice croplands in kharif and rabi by 50 %; replace with irrigated corn in khar if, irrigated corn in rabi red_rice_irr,k,25 reduce rice daily irrigation depths by 25% in kharif to 6 mm/day red_rice_irr,k,50 reduce rice daily irrigation depths by 50 % in kharif to 4 mm/day red_rice_irr,k,r,25 reduce rice daily irrigation de pths b y 25% in kharif and rabi to 6 mm/day red_rice_irr,k,r,50 reduce rice daily irrigation depths by 50 % in kharif and rabi to 4 mm/day red_rice_r irr corn,25 reduce rice croplands in rabi seaso n only by 25%; replace with irrigated corn in rabi red_rice_r irr corn,50 reduce rice croplands in rabi season only by 50 %; replace with irrigated corn in rabi red_rice_k corn,r irr corn,25: mds33_kharif reduce rice croplands in kharif and rabi by 25%; replace with rainfed corn in kharif, irrigated c orn in rabi; MDS of 32.5 mm in kharif corn and cotton red_rice_k irr corn,r irr corn,25: mds33_kharif reduce rice croplands in kharif and rabi by 25%; replace with irrigated corn in kharif, irrigated corn in rabi; MDS of 32.5 mm in kharif corn and cott on red_rice_irr,k,25: mds33_kharif reduce rice daily irr igation depths by 25% in kharif to 6 mm/day; MDS of 32.5 mm in kharif corn and cotton

PAGE 137

137 Figure 4 1. Observed and simulated (for calibration and validation) watershed outlet reservoir volume and rainfall in Wargal watershed; 3 / 19 /20 10 to 12/3/2010 Figure 4 2. Simulated and observed reservoir volumes: 3/19/2010 to 12/3/2010 0 10 20 30 40 50 60 70 80 90 100 0 50000 100000 150000 200000 250000 300000 350000 400000 3/19/2010 5/19/2010 7/19/2010 9/19/2010 11/19/2010 Daily rainfall, mm Reservoir volume at watershed outlet, m3 95 PPU Rainfall Observed reservoir volume Simulated reservoir volume; NSE = 0.95 cross validation simulation; NSE = 0.73 R = 0.947 0 50000 100000 150000 200000 250000 300000 350000 400000 0 100000 200000 300000 400000 Simulated reservoir volume, m3 Observed reservoir volume, m3

PAGE 138

138 Table 4 2 Maximum surface area and volume of the six reservoirs (tanks) in Wargal watershed Tank subbasin Area, ha Volume, 10 4 m 3 1 21.40 35.03 2 0.96 0.42 3 0.59 0.21 4 0.77 0.30 5 0.32 0.09 6 0.75 0.29 Table 4 3 Specific yield, natural recharge, and total recharge from DWTF method and total recharge from SWAT simulations in 2009 and 2010 2009 2010 DWTF SWAT DWTF SWAT Sy 0.0139 0.0123 R, mm 264.2 419.9 R + RF, mm 496.1 501.9 674.7 701 .4 Rainfall, mm 667.0 667.0 1022.0 1022.0

PAGE 139

139 Figure 4 3 Observed g roundwater level change and rainfall during wet and dry periods Table 4 4 Nash Sutclif fe Efficiency (NSE), percent bias (PBIAS), and ratio of the root mean square error to the standard deviation of measured data (RSR) Calibration describes the parameter estimation using all observations, and evaluation describes the 5 fold cross validatio n. NSE PBIAS RSR Calibration 0.954 3.185 0.215 Evaluation 0.729 14.657 0.520 0 50 100 150 200 250 555 560 565 570 575 580 585 Rainfall, mm Groundwater elevation, m. above sea level Groundwater level, basin average rainfall

PAGE 140

140 Table 4 5 Parameter name, t statistic, p value, and range of the 13 parameters estimated during SWAT calibration r__ indicates that the initial parameter is multi plied by (1+[value in range]); v__ indicates that the initial parameter is replaced by the value in the range. Sensitivity Range Parameter Name t Stat P Value Min Max Best estimate Units r__SOL_AWC(1,2).sol 37.304 0.000 0. 25 0.75 0.47 mm/mm v__SOL_K(1).sol______RICE 18.190 0.000 0.50 1.25 0.99 mm/hr v__RES_K.res 18.167 0.000 0.50 1.00 0.56 mm/hr r__CN2.mgt 13.658 0.000 0.10 0.05 0.09 none v__EVRSV.res 3.775 0.000 0.50 0.95 0.78 none r__SOL_K(1).sol 3.136 0.002 0.25 0.75 0.26 mm/hr r__GWQMN.gw 1.570 0.116 950.00 1100.00 1073.78 mm v__EPCO.bsn 0.932 0.352 0.50 0.95 0.78 none v__ESCO.bsn 0.900 0.368 0.50 0.95 0.80 none r__ALPHA_BF.gw 0.801 0.423 0.25 0.50 0.27 days r__RCHRG_DP.gw 0.548 0.584 0.70 0.80 0.76 none v__EVLAI.bsn 0.236 0.814 3.50 5.00 3.84 m 2 /m 2 r__GW_DELAY.gw 0.213 0.832 30.00 50.00 30.70 days

PAGE 141

141 T able 4 6 Summary of selected water balance components for existing and alternative agricultural management sorted by groundwater balance (GW balance = recharge irrigation) is predicted groundwater level change, mm, given specific yield of 0.01 3 from DWTF method. Simulations using g enerated weather 2000 2009 792 mm rainfall Average annual values, mm Management scenario GW balance Irrigation Recharge ET red_rice_k,r,75 10335.0 124.0 163.6 287.6 534.1 red_rice_k,r,50 6596.2 79.2 2 90.9 370.1 553.4 red_rice_irr,k,r,50 6477.1 77.7 290.4 368.1 585.4 red_rice_k corn,r irr corn,50 3712.4 44.5 347.3 391.8 570.9 red_rice_irr,k,r,25 3183.8 38.2 418.0 456.2 591.8 red_rice_k corn,r irr corn,25: mds33_kharif 3008.3 36.1 446.4 482.5 583.0 red_rice_k,r,25 2858.2 34.3 418.3 452.6 572.7 red_rice_irr,k,25: mds33_kharif 2610.4 31.3 474.9 506.3 593.9 red_rice_irr,k,50 2243.9 26.9 404.2 431.1 583.9 red_rice_k corn,r irr corn,25 1543.2 18.5 446.4 465.0 583.0 red_rice_irr,k,25 1499.5 18.0 474.9 492.9 593.9 mds50_k,r 627.9 7.5 546.0 553.5 595.1 mds50_kharif 496.0 6.0 546.0 552.0 593.9 mds33_k,r 281.3 3.4 546.0 549.4 591.4 mds33_kharif 201.7 2.4 546.0 548.4 592.1 mds15_k,r 60.7 0.7 546.0 545.3 591.3 mds15_kharif 117.5 1.4 546.0 544.6 592. 0 red_rice_r irr corn,50 580.9 7.0 488.7 481.8 587.1 red_rice_r irr corn,25 603.9 7.2 517.2 509.9 591.1 mds50_rabi 778.0 9.3 546.0 536.7 593.3 mds33_rabi 829.6 10.0 546.0 536.0 591.4 mds15_rabi 852.4 10.2 546.0 535.8 591.3 existing 909.2 10.9 546.0 535.1 592.1 red_rice_k irr corn,r irr corn,25: mds33_kharif 1508.6 18.1 548.5 530.4 595.8 red_rice_k irr corn,r irr corn,50 2099.1 25.2 475.9 450.7 587.2 red_rice_k irr corn,r irr corn,25 3197.4 38.4 548.5 510.1 595.9

PAGE 142

142 Table 4 7 Su mmary of selected water balance components for existing and alternative agricultural management, sorted by groundwater balance (GW ba lance = recharge irrigation). given specific yield of 0.013 from DWTF method. Simulations using observed weather 2009 667 mm rainfall Average annual values, mm Management scenario GW balance Irrigation Recharge ET red_ric e_k,r,75 8967.2 107.6 163.6 271.2 493.0 red_rice_k,r,50 4813.1 57.8 290.9 348.7 531.7 red_rice_irr,k,r,50 3250.7 39.0 290.4 329.4 602.2 red_rice_k corn,r irr corn,50 1288.3 15.5 347.3 362.7 558.3 red_rice_k,r,25 601.4 7.2 418.3 425.5 570.5 red_rice_k corn,r irr corn,25: mds33_kharif 162.7 2.0 446.4 448.4 607.2 red_rice_irr,k,50 210.9 2.5 404.2 401.7 606.1 red_rice_irr,k,r,25 331.0 4.0 418.0 414.1 602.2 red_rice_irr,k,25: mds33_kharif 1023.6 12.3 474.9 462.7 582.0 red_rice_k corn,r irr corn,25 1100.4 13.2 446.4 433.2 584.9 red_rice_irr,k,25 1848.3 22.2 474.9 452.8 606.8 mds33_k,r 2843.4 34.1 546.0 511.9 645.4 mds33_kharif 2851.4 34.2 546.0 511.8 636.9 mds15_kharif 2854.1 34.2 546.0 511.8 633.1 mds50_k,r 2879.5 34.6 546.0 511.4 632.5 mds50_kharif 3052.1 36.6 546.0 509.4 625.6 mds15_k,r 3404.1 40.8 546.0 505.2 669.2 red_rice_r irr corn,50 3434.6 41.2 488.7 447.5 600.3 red_rice_r irr corn,25 3476.2 41.7 517.2 475.5 605.9 mds50_rabi 3499.9 42.0 546.0 504.0 616.1 mds3 3_rabi 3665.7 44.0 546.0 502.0 617.7 existing 3672.7 44.1 546.0 501.9 609.3 mds15_rabi 4225.6 50.7 546.0 495.3 592.1 red_rice_k irr corn,r irr corn,25: mds33_kharif 4405.2 52.9 548.5 495.6 643.6 red_rice_k irr corn,r irr corn,50 5806.7 69.7 4 75.9 406.2 602.7 red_rice_k irr corn,r irr corn,25 6874.0 82.5 548.5 466.0 620.9

PAGE 143

143 Figure 4 4 Basin scale response of groundwater recharge, groundwater balance, and runoff to MDS changes in rainfed croplands in khari f season and in both kharif and rabi seasons MDS of 15, 32.5 or 50 mm. 30 to 50% less runoff with tillage for 15 mm < MDS < 50 mm; groundwater balance from 11 mm for existing management to 7 mm for 50 mm MDS in rainfed areas in both seasons. 60 50 40 30 20 10 0 15 10 5 0 5 10 15 20 25 Percent change in runoff (in rainfed croplands) from existing (MDS = 0) management Percent change in groudwater recharge from existing (MDS = 0) management; Basin scale groundwater balance, mm Percent change in recharge (basin scale) in response to MDS management Groundwater balance: basin scale, mm Percent change in runoff (basin scale) in response to MDS management

PAGE 144

144 Figure 4 5 Field scale response of groundwater recharge and runoff to tillage changes ( MDS depth) in rainfed croplands in kharif season MDS of 15, 32.5 or 50 mm. 54 to 83 % less runoff from rainfed croplands with tillage for 15 mm < MDS < 50 mm; groundwater balance from 11 mm for existin g management to 6 mm for 50 mm MDS in rainfed areas in both seasons. 90 80 70 60 50 40 30 20 10 0 15 10 5 0 5 10 15 20 25 Percent change in runoff (in rainfed croplands) from existing (MDS = 0) management Percent change in groudwater recharge (in rainfed croplands) from existing (MDS = 0) management; basin scale groundwater balance, mm Percent change in recharge for three MDS depths in kharif Groundwater balance: basin scale, mm Percent change in runoff for three MDS depths in kharif

PAGE 145

145 Figure 4 6 Field scale response of groundwater recharge and runoff to tillage changes ( MDS depth) in rainfed croplands in both kharif and rabi season s 90 80 70 60 50 40 30 20 10 0 15 10 5 0 5 10 15 20 25 Percent change in runoff (in rainfed croplands) from existing (MDS = 0) management Percent change in groudwater recharge (in rainfed croplands) from existing (MDS = 0) management; basin scale groundwater balance, mm Percent change in recharge for three MDS depths in kharif and rabi Groundwater balance: basin scale, mm Percent change in runoff for three MDS depths in kharif and rabi

PAGE 146

146 Table 4 8 Sensiti vity of annual surface runoff and groundwater recharge to MDS at the basin scale : mm change in depth of recharge and runoff (from existing tillage, MDS = 0) per mm MDS Management Scenario MDS depth Increase in recharge, mm Increase in recharge per mm MDS Decrease in runoff, mm Decrease in runoff per mm MDS mds50_kharif 50 .0 49. 9 1.0 76. 1 1.5 mds33_kharif 32.5 40.9 1. 3 69.0 2.1 mds15_kharif 15 .0 29.1 1.9 49. 2 3. 3 existing 0 0 0 0 0 Table 4 9 Sensitivity of annual surface runoff and groundwater recha rge in rainfed croplands (corn and cotton) in kharif season to MDS: mm change in depth of recharge and runoff (from existing tillage, MDS = 0) per mm MDS Management Scenario MDS depth Increase in recharge, mm Increase in recharge per mm MDS Decrease in ru noff, mm Decrease in runoff per mm MDS mds50_kharif 50 .0 16.9 0.3 30.1 0.6 mds33_kharif 32.5 13.3 0.4 27.3 0.8 mds15_kharif 15 .0 9.5 0.6 19.5 1.3 existing 0 0 0 0 0

PAGE 147

147 Figure 4 7 Irrigation, recharge, ET, and groundwater balance for tied ridge t illage scenarios: 3 MDS depths in rainfed areas in kharif, rabi, and both seasons 100 0 100 200 300 400 500 600 700 Depth of water balance componenet, mm Management scenarios: tillage changes Irrigation Recharge ET GW balance

PAGE 148

148 Figure 4 8 Irrigation, recharge, ET, and groundwater balan ce for selected changes in extent and irrigation management of rice croplands 100 0 100 200 300 400 500 600 700 Depth of water balance componenet, mm Management scenarios: rice extent and irrigation changes Irrigation Recharge ET GW balance

PAGE 149

149 CHAPTER 5 SOCIAL AND ECONOM IC ASSESSMENT OF GROUNDWATER MANAG EMENT Groundwater and Indian Agricultural Economi c s The g roundwater irrigated areas of India nearly tripled in extent from 11.9 million hectares in 1971 to 33.1 million ha in 1999 (Mukherji and Shah 200 5 ), and i t has been shown through analysis of over 240 districts across India that the productivity ($/ha) of groundwater irrigated areas are about 35% greater than those of surface water irrigated areas (Mukherji and Shah 200 5 ). Throughout Asia, and especially in India, i rrigation from groundwater has become a major contributor to agricultural improvements in recent decades. However, as discussed in earlier chapters, there are mounting concerns about groundwater depletion in India (CGWB 2007; Rodell et al., 2009 ), and th ere are also energy efficiency concerns associated with groundwater depletion that have resulted from increa sed groundwater irrigated areas. The decline of groundwater resources in some regions has decreased net energy ratios in agriculture (energy output / energy input) as a result of greater pumping and fertilizer inputs (Gurunathan and Palanisami 2008). Groundwater scarcity in India is increasingly becoming a problem for livelihood security and the research presented in this chapter focuses on (1) in tegrating gender social and economic analyses and to improve the relevance of science based recommendations related to groundwater management regimes in rural India (2) estimating the economic consequences of changing agricultural management, and (3) di scussing application possibilities and policy options for alternative agricultural management

PAGE 150

150 Fieldwork on the Socioeconomics of Agricultural Management In contribution to the analysis of social and economic effects of agricultural management changes, f ieldwork was done to find the decision making processes involved in determining farm management practices in the Wargal watershed, including the nature of information used by stakeholders to decide about farm management. A thorough analysis require d that the differences between decision making by men and women (involvement, power relations, access to and use of information) be understood. Fieldwork in the study area aimed to begin this process. The related biophysical goal in Wargal was to find cropping system management options that can realize sustainable water resources and that are sufficiently productive. This was done through simulation experiments using a distributed water balance model Watershed scale analyses of groundwater responses to crop selection, tillage strategy, and irrigation management helped select the best management practices for improving groundwater supply for the people of the watershed ( Chapter 4) Information from people of the study area has served three purposes for the bi ophysical experimentation aspect of this project: Validation of problem confirms that the problem (groundwater depletion) is actually a concern to people in the watershed. Interactions with farmers in the area showed that groundwater depletion was obser ved by nearly everyone in the area, and this depletion was one of the major concerns of households in the area. Simulation of appropriate management options participation from the community to recommend management options (alternative crops, tillage, and irrigation) that they would prefer, that they consider economically viable, and that they expect would reduce irrigation withdrawals. This information increases the relevance of hydrologic simulation experiments. Five weeks of preliminary fieldwork has provided some information on crop preferences, available tillage implements, and typical irrigation practices. This information is being used to choose implementable management scenarios for simulation experiments.

PAGE 151

151 Implementation and motivation to change field trials and actual management changes by farmers in the area are expected to be easier to initiate as a result of the involvement of the community. An improved understanding of how certain farm management decisions are made (what information is use d to make them) helps expose possible motivations for changes in decisions about resource management. From interactions with farmers in the area, it was learned that water availability was the most dominant information used in decisions about crop selecti on. The interaction of research generated management recommendations and the implementation of actual, local management changes is nuanced and requires significant understanding of local information systems (Roncoli et al., 2002; Shah et al., 2003). As st ated in their study about local rainfall forecasting in rural Africa (Roncoli et al., research For r esearch models generated in the highly developed parts of the world to have practical transferability to help improve rural livelihoods, the ir management recommendations must be able to fit into local knowledge systems Sustainability: Definitions Broadl y, this project focuses on sustainable water resource management in India; in the Wargal watershed the specific concern is groundwater sustainability. The definition of sustainability (or security) of a biophysical quantity is generally straightforward. It is the maintenance of a state or process at a desired level or rate into the future. T he subj ect of interest in this project is groundwater quantity; so sustainability can be more narrowly defined as some amount of groundwater (distance below ground l evel or aquifer storage volume) to be m aintained. This chosen storage amount is informed by past and current groundwater levels and expert judgment. The groundwater balance has an obvious dependence on decisions by people, which is why

PAGE 152

152 it was decided tha t an improved understanding of how the people of Wargal make decisions would contribute to the overall goal of finding ways to sustainably manage water resources in the area. Methods for Learning about Wargal Agricultural Management The plan for fieldwork in the Wargal watershed was focused on meeting the following objectives: Explore the perceptions of men and women concerning groundwater resources and changing climate Identify gender differences in constraints to changing farm management practices (tillag e techniques, crop se lection, irrigation management), and explore the local farm management decision making processes The intention was to meet with members of households having dependence on the study area for their livelihood sources. It was planned to meet with men and women both together and separately to discuss their decision making processes, cropping system preferences and constraints, responsibility divisions, and perceptions of water resources and climate change. Meetings with community level go vernment were organized to understand the regulation of water and land use management. The local governing body, Panchayat, consisted of five men who are charged with tax collection, data acquisition (crop types, yield estimates, crop condition, rainfall, population), and settling local disputes. Farmers who participated in interviews were selected partly by partners from ANGRAU (they knew some farmers who had agreed to metered borewells and other instrumentation in their fields) and partly by just appro aching people who were working or resting in the fields. Most participants were selected by the latter method

PAGE 153

153 T ranslators from ICRISAT and ANGRAU (interviews were in Telugu) assisted with interviews. The fieldwork tools consisted of informal interview s of open ended questioning (Norman et al., 1995) Typically, interviews began with introductions and small talk and continued with informal discussions about main concerns in the community, changes in groundwater supply and climate, and how decisions wer e made and who was responsible for certain farm operations. Questions were grouped into these three themes during question preparation, but no mention of the themes was made during interviews. Questions were not read from a questionnaire; they were asked from memory in order for interviews to be more conversational. Note takers recorded all responses. O bservations were regularly made of who was doing farm operations (tillage, weeding, planting, transplanti ng, controlling borewell pump). Some responden ts were direct owners of the farm we visited and these respondents were either alone or with family members (both men and women) who were working on the farm. Hired day laborers also participated in the discussions summarized here. For much of the result s presented below, sample sizes of women are slightly larger than those of men (these numbers are totals, including responses in gender separated and mixed groups). The larger numbers of women reflects the special efforts made to include women in discussi ons; also, due to the nature of many of the field operations for which they were responsible (for example, weeding by hand), women were more likely to be found working in groups. Interviews generally happened in groups, but responses were only recorded fo r people who gave answers. All members of a group were given the opportunity to respond to each question. Interviews

PAGE 154

154 lasted from 10 to 30 minutes The informality of survey administration and method of participant selection resulted in some incomplete s urveys; not all questions were answered by all participants. This is a common weakness of informal surveys, making them less suited to quantitative analysis (Norman et al., 1995). Social and Biophysical Analysis Connections During the design of simulati on experiments, decisions have to be made by the user of the water balance model about the types of management pract ice scenarios to simulate. It was hypothesized that management scenarios developed from farmer generated information would have higher prod uctivity and g reater chance of implementation This hypothesis could not be tested by survey data collection, but would require farmers to implement management changes and to evaluate their effectiveness. Participation of farmers from the study are a was expected to result in a more realistic range of management scenarios (Chapter 4). Results of Household Survey Fieldwork Wargal mandal is one of 45 mandals of Medak district, a Telugu language region. Medak district has a population of 2.8 million, 80% of which is rural. There are about 150 households working in agricultural systems in Wargal. There are about 190 active borewells; total land area is 512 ha. Cropping systems are dominated by irrigated rice which represents 41% and 49% of cropland area in kharif and rabi seasons, respectively Questions are f indings from interviews with community members were grouped into the following three themes: main concerns, changes in groundwater and climate, and management and decision making. Main concerns were responses about the most pressing issues in the community. Changes in both groundwater and climate captured

PAGE 155

155 responses about direction of changes observed, how they are measured, and what the causes of the changes are. Management and decision making exp lores the information used to make farm management decisions and who is responsible for selected management activities. These three themes were identified for organizational purposes during survey development. Main c oncerns One topic of interactions with farmers was their main concerns in the community. This was approached in a broad, open ended manner to include information related to agriculture or other aspects of the livelihood system. Percentages of various responses are presented in Table 5 1. A f ew key themes stand out and warra nt attention. The data suggest, based on numbers of responses for each concern mentioned by participants, that women are more concerned with social livelihood issues like education and adequate health care for their childr en, bus routes to town for easier access to markets, and sufficient, clean water for domestic use. Public transportation (bus route public transportation. T he most prevale nt concerns of women tend to coalesce around their domestic roles of providing care and education for their children and their roles of going to market As an example of the sometimes divergent responses of men and women, 59% of female participants mentio ned access to higher education as a leading concern while 0% of male participants mentioned this. 41% of women and 3% of men mentioned access to healthcare services as a major concern. Some women did express concern about lack of groundwater and electric ity for powering irrigation pumps. Men were mostly concerned with issues that directly impact their ability to be successful farmers and wage earners : such as affordable fertilizer, reliable and inexpensive seed,

PAGE 156

156 adequate electric power for running irriga tion pumps, and availability of local non farm employment Participants of both genders expressed concern about there being less groundwater and an increasing rate of failing borewells and both were concerned with power shortages that impact the function ing of irrigation pumps. Generally, there is a discrepancy in the concerns between the men and women we interviewed. This is evidenced by noticing that four of the five top concerns for women are not mentioned as concerns by men except for one male respon dent who did mention concern about quality of and acc ess to water for domestic use. However women seemed to share several of the most mentioned concerns of the men This is most likely because the overall economic prosperity of the family depends on agr icultural quantities like seed, fertilizer, electricity, and water availability and those are things men and women generally discuss together. Some women have also taken out help groups and this money is often used for fa rm inputs so that women have an added stakeholder concern in the success of their family farm. It should be noted that these questions were left open so that although men did not respo nd that they are concerned with education and health care this does not mean that these issues are not actual concerns they also have. Rather, their more immediate concerns were lth related. Changes in groundwater and c limate Water r esource sustainability is an important objective of this work; t herefore, it was decided that the perceptions of farmers about groundwater and climate changes

PAGE 157

157 should be explored. Farmers were asked if they had noticed any changes in groundwater levels in recent years; the question was always open, not assuming any direction of resource changes. Have you noticed any If groundwater depletion was observed, p articipants were asked to consider methods of noticing groundwater depletion, expected causes of groundwater depletion, and expected cause of reduced rainfall. Tables 5 2, 5 3, and 5 4 summarize their responses. Table 5 2 summari zes the groundwater sensing or measurement options; without instrumentation changes in groundwater levels are still noticed by men and women in the research area. More women cited failed drilling attempts as a method of observing groundwater decline. Thi s method likely has the most severe economic consequences for families in the research area. A failed borewell must still be paid for and this usually is done by taking out a loan but which requires a bountiful harvest so the loan may be paid back. Only a small percentage of responses indicated that groundwater decline was not observed; this gives more support (in addition to published groundwater decline data) to the effort to find farm management options for improving groundwater levels. Those that res ponded that they had noticed groundwater decline were asked what they thought was causing the decline. A few different responses were given and it sometimes took about one minute for participants to choose a cause. Table 5 3 summarizes the responses conce rning cause of groundwater depletion. A clear majority of research participants, whether female or male, have perceived a negative change in groundwater quantity An equal number of both men and women claim to have noticed this change by visually noting t he rate and strength of flow coming

PAGE 158

158 from borewell pumps. This shows that both men and women have an interest in observing flow rate. A clear majority of women have correlated changes in groundwater because of failed drilling attempts. Similarly, more wo men associate groundwater depletion with the proliferation of borewells over the last ten to fifteen years: of the 23 people who were asked about reasons for changes in groundwater, 15 associated this with more borewells being drilled and 12 of these were women. What is more relevant is that this recognition of correlation between borewell development and groundwater depletion shows that there is community knowledge about current farming systems and management decisions and how these may be contributing to groundwater depletion. This suggests that growers in the area would be somewhat receptive to farm management scenarios which show promise of improving groundwater levels. Lastly, all research participants in the watershed, both men (N = 12) and women (N Board suggests noticeable rainfall declines in the Wargal study area These data show a decline in rainfall from the long term mean (873 mm annually) of about 30% over the per iod from 2001 2006 (CGWB 2007). When discussing past rainfall and water levels in the area, a few farmers mentioned that they remember when there used to be puddles and little ponds all over the fields in the past but in recent years the fields dry up e arlier because there is less rain. Participants who suggested reduced rainfall as a cause of groundwater decline were asked why rainfall amounts were less than average. Similar to the responses about cause of groundwater decline, there was little variabi lity among responses. Shoulder shruggin g was common and some suggested such a thing could not be known. As seen in Table 5 4, local knowledge, sometimes explains

PAGE 159

159 changes in rainfall as being associated with the work of the gods and the misdeeds of people Management and decision m aking Management of farming systems in the study area is adaptive, considering diverse information from which decisions are made. Questions were asked of farmers about what information is used to make decisions about crop select ion and irrigation management (Appendix B ) Tables 5 5 and 5 6 pr esent participant responses and the associated percentages of total responses to inquiries about how crop selection and irrigation management decisions are made Agreement in the responses of men and women is illustrated in Table 5 5 Responses for each question are ranked in the table from highest to lowest frequencies of occurrence. Both men and women reported that observations of water availability and conversations within the family a nd community are the most important sources of information for decisions about crop selection. One response about information for crop selection for which responses was about expected produce pri ce. In the sample interviewed (26 women, 20 men), a much higher proportion of men than women indicated that produce price would be considered for deciding on crop selection and extent In discussions of timing and depth of irrigation, fewer women than me n gave the response: electricity is on; therefore, irrigation happens (Table 5 6) That response was the most frequently occurring response by men. This supply side management method may be partly the cause of the local groundwater depletion. Our initia l fieldwork suggests that irrigation management is generally the responsibility of men, but for some growers women did have some irrigation responsibility.

PAGE 160

160 Considering the biophysical goal of improving groundwater levels in this watershed, the prevalenc e of water availability as an important influence on crop selection decisions suggests that people in the community would be responsive to management scenarios that are designed to do well under limited water availability. This suggests that water balance simulations be designed to find management options that broaden the range of crops under consideration when there is low water availability. Some tillage techniques which result in greater rainfall infiltration (and greater estimated yield) could achieve this. However, observations in the field and participation of community members have revealed that there is limited labor availability and very little mechanization available for field operations. Therefore, management scenarios for water balance simula tions were designed with consideration of these constraints on labor and mechanization T he sources of infor mation used for decision making, based on conversations with growers ( Tables 5 5 and 5 6), can be summarized into the following 6 categories ranke d in order of decreasing frequency of occurrence: Decision support information Water availability Observation of plants and soil Family communication Discussion with neighbors Expected market price of produce Personal experience This information is used t o shape decisions about crop selection, irrigation, and fertility management Decision making information that could be categorized as water availability was reported by nearly all farmers involved, suggesting that groundwater decline and rainfall variabi lity are significant concerns. Water availability is observed by

PAGE 161

161 impressions of rainfall of the previous year, water tank levels, soil moisture, plant appearance, and flow rate (intermittency) of borewells. Figure 5 3 displays a reservoir (one of six in the study area) after a recent rain. These reservoirs were excavated decades ago and are maintained to delay runoff of monsoon rains and increase groundwater recharge. In areas where agriculture is the dominant livelihood activity, attention to water avai lability and other limiting factors is nearly universal among stake holders (Roncoli et al., 2002). Tools used to predict risk of agricultural activities typically have some measure of water availability being the most sensitive input (Kazianga and Udry 2 006), so it is not surprising that growers indicated that water availability is the most important factor in decisions about crop selection. Table 5 7 illustrates the consistency between men and women in the most frequently observed response s concerning f arm management Table 5 7 lists the responses mentioned by the greatest number of male and female participants for four farm management questions : what is information used to decide about (1) crop selection and (2) irrigation management (3) what is the p referred rice cropland replacement, and (4) who is responsible for borewell pump operation Only in the replacement of rice crop did responses diverge; the men suggesting growing cotton and the women opting for a cover crop or fallow, which would require much less labor and management but would provide no direct cash income. In a report on the changing roles of men and women in response to desertification, Gurung et al. ( 2006) suggest that women are more responsible for food crop production and land conse rvation and men are typically more responsible for cash crop management. The limited data presented here is consistent with that finding : women responding to a

PAGE 162

162 question about crop selection if there was insufficient irrigation water for rice cultivation s aid that they would prefer a cover crop or fallow (which would likely improve productivity of future plantings of food crops) and men suggested cotton (which would increase short term household income). Discussions with farmers provided information about who turns on and off the borewell pumps used for irrigation withdrawals and who goes to the market to buy seeds and fertilizer. Tables 5 8 and 5 9 summarize responses to these two questions. en; women o nly if man is unavailable borewell pumps and were responsible for seed and fertilizer acquisition, but women would occasionally do these jobs if needed. This arrangement seemed to wor k well and appeared satisfactory to all involved. However, some investigators have suggested that and Piana 2006; Zwarteveen 2008). Considering what a productive resource afford them less power than men (Pangare 1998). Women usually reported that they were afraid of the electrical box where the pump switch (or wire connectors) was located. Similarly, the purchasing and delivery of seed and fertilizer was typically the responsibility of the men ( Table 5 9), possibly increasing their access to information about marketing and crop suggestions from vendors. Valuing Groundwater Much of consider ation of agricultural management alternatives for Wargal is focused on groundwater sustainability, meaning the assurance of sufficient groundwater quantity indefinitely. For a more complete analysis of the possible management

PAGE 163

163 options, beyond hydrology (Ch apter 4) and simple economic estimates (Chapter 5), t he p resent and future value s of groundwater resources should be estimated to assist in decision making about groundwater management. While uncertainties in policy, climate change, and national and globa l economies make future value estimates highly uncertain, it is plausible that groundwater in the Wargal area will be much more valuable in the future. Groundwater valuation must always be based on the context (NRC, 1997) meaning that the value of the se rvice provided by groundwater should be quantified for each situation. There are both in situ and extractive services provided by groundwater. In situ services are generally more difficult to quantify in terms of economic value; these services include su bsidence avoidance, saltwater intrusion prevention, recreational flow maintenance, and provision of water supply buffer. Extractive services include municipal, industrial, and agricultural supply ; these are usually easier to value. In contexts like that of Wargal, where groundwater services are almost entirely irrigation of agricultural areas valuation is based largely on incomes from sale of irrigated crops. No quantitative estimate of present or future value of the groundwater resource was made here d ue to the large uncertainties in aquifer thickness and irrigated crop values. However, groundwater valuation in Wargal would be an important extension of the research presented here. Economics of Alternative Agricultural Management There are important eco nomic impacts at the h ousehold level and at the regional and national levels of changing agricultural management. Household level impacts are estimated here based on predicted yield changes and expected prices for different management scenarios Whil e there are predicted improvements in the groundwater balance resulting from shifting some rice croplands to rainfed crops, there is likely much

PAGE 164

164 greater variability in yield (and income) from rainfed areas. This is a substantial obstacle for growers and w ater managers concerned with the future of groundwater resources in India. There are 3 policy options that could contribute to the implementation of management changes that show promise for improving the groundwater balance One option would be to change the minimum support price ( MSP ) structure; lowering the MSP of rice and increasing the MSPs of popularly grown rainfed crops could lead to the required rice cropland reductions. MSP is an Indian national agricultural policy to set the lowest prices that prod ucers receive for farm yields. The MSP system is administered by a variety of commodity specific, public organizations. Similarly to the MSP adjustments if rainfed or less irrigated grains were System, to low income families instead of rice (or instead of a portion of the rice), this could increase the demand for rainfed or less irrigated grains. A third option may be to subsidize equipment for TR tillage. Plows that are tractor or animal draw n can create the ridges and ties for water harvesting tillage in rainfed croplands. To evaluate the expected economic impacts of selected management alternatives, e stimates of the total watershed scale value of crop yields were made for existing management and for the leading 7 management scenarios using average observed yields, cropland a reas, MSPs for 2009 2010 (Tables 5 1 0 and 5 1 1 ). As described in Chapter 4, the leading 7 management scenarios were selected from the 25 alternatives based on (1) partic ipation of growers in the area (constraints on labor, preferences for rice cultivation, little interest in alternative rainfed crops (Chapter 5), (2) the groundwater balance results (Chapter 4), and (3) consideration of yield variability and risk. A

PAGE 165

165 liter ature review of yield response to tied ridge (TR) tillage was completed (Table 5 10) to estimate the expected yield increase in rainfed areas where TR was simulated (MDS: 15 50 mm). Based on the review of 9 studies of yield response to TR excluding yield increases greater than 100%, the average increase in yield was 27.4%. To be conservative in estimates of yield increase in Wargal, a 13% increase was used as the estimated change in yield for rainfed crops under TR. For simulated scenarios having 25% re duced rice irrigation depths in kharif and rabi seasons, rice yields were estimated to decrease by 25%. For simulated scenarios having 25% reduced rice irrigation depths only in rabi season rice yields were estimated to decrease by 1 5%. For simulated sc enarios having irrigated corn in rabi season, corn yields were estimated to increase by 25%. With these estimates of yield changes, the cropland areas from each management scenario, and the crop values from MSP data, the watershed scale value s of existing and leading 7 management sc enarios were calculated (Table 5 12). Surprisingly, the mds50_kharif scenario, TR tillage of MDS=50 mm in kharif corn and cotton, was found to be the most valuable ($351,021), followed closely by existing management ($342,234). While there is an unknown amount of uncertainty in the estimates of yield changes these estimates do seem reasonable and are consistent with the literature. Water harvesting tillage was shown to have positive impacts both on the groundwater balance and on the economic yields of the Wargal watershed. Discussion and Summary of Fieldwork on the Socioeconomics of Agricultural Management Some investigators have suggested that the groundwater socio ecology in Asia, and particularly in India, is at a critical point. The progression of groundwater use in agriculture has been organized into four stages (Shah et al., 2003): expansion of

PAGE 166

166 borewell installations, groundwater based agrarian boom, onset of groundwater depletion concerns, and collapse of groundwater ba sed systems. Observations (from groundwater level measurements and local household perceptions) in Wargal suggest that this area is in stage three of the progression: groundwater depletion is an observed and growing concern. The data from interviews with farmers of Wargal suggest s that farm management decisions in Wargal do consider this, but farmers are understandably still trying to maximize their groundwater irrigated areas. For Wargal residents to have sufficient groundwater resources for domestic and agricultural needs in the future, they will likely need to implement changes in the way farming systems are managed. The types and extent of changes can begin to be answered from the results of the water balance simulations experiments (Chapter 4). Cons idering this context, the main findings of this initial assessment of social and gender issues related to farm management decision making in the area are summarized: Groundwater availability is a concern for many people in the study area; of all their conc erns in the community it ranks in the middle (for both men of women) in terms of frequency of responses from participants. If the four stages of the progression of groundwater based agrarian societies are correct, the stage that follows the current one (o bservation of depleted groundwater) is a rapid decline in productivity and livelihood security. It seems now may be an important time for finding the best ways to increase productivity, yet require the least groundwater based irrigation. Households have observed groundwater depletion by noticing more failed drilling attempts and more intermittent, lower pressure flow from borewells. If substantial management changes are achieved which reduce groundwater withdrawals, these means of measurement will be im portant indicators to farmers of the increased availability of groundwater. Without any preemptive management changes, local methods of groundwater measurement will require management changes to reduce groundwater use, but there would likely be a period o f high vulnerability as local people adjust to the very limited availability of groundwater. Data from farmer participation suggests more women than men attribute the cause of

PAGE 167

167 groundwater depletion to increased borewells, so it may be that women are more amenable than men to adopt management alternatives which reduce dependence on borewells. For both men and women, water availability was the information most commonly used to make decisions about crop selection. Concerning irrigation management: a common response was that irrigation timing and depth depended mostly on electricity supply (if power is on, irrigation happens); this supply side management perspective may be part of the reason for groundwater depletion in the area. It was almost exclusively t he responsibility of men to control borewell pumps, but participants indicated that both men and women were involved in deciding when to turn pumps on or off. Efforts to find combinations of management options that are most promising for improving groun dwater levels, while still maintaining economic viability of farms, have focused on the improvement and use of water balance simulation tools (Chapter 4). Water balance simulations have been used to estimate changes in the groundwater balance (recharge estimated irrigation withdrawals) in response to changes in rice cropland extent, tillage, and irrigation management. It is hoped that field trials of the management options most likely to improve groundwater quantity (based on simulations) will be initia ted. The participation of local farmers in development of management scenarios has resulted in more relevant and sustainable management scenarios being considered, and therefore, a greater chance for successful on farm implementation of management recomme ndations.

PAGE 168

168 Table 5 1. Concerns of community members ranked (for women) in order of decreasing frequency of reporting Women Men Women Men Issues most concerning to participants % of total responses rank of concern No local higher education 59 0 1 10 Bus routes are limited into Wargal 55 0 2 10 Limited power for irrigation 41 50 3 2 Water quality for drinking in homes 41 3 3 9 Limited service at healthcare clinic; no staffed doctor 41 3 3 10 Less groundwater, failed bores, water availability 31 28 4 5 Vulnerability due to irregular rains 28 9 4 7 Seed is expensive and not reliable 14 34 5 3 No available non farm work 14 31 5 4 Lack of affordable fertilizer 10 59 6 1 Labor availability/rising cost of labor 10 22 6 6 Less land, more children 10 0 6 10 Total number of participants 29 32 Table 5 2. Ways of observing groundwater depletion Women Men Methods of observing groundwater decline % of total responses Failed drilling attempts 71 20 Time of continuous flow, pressure of flow vi sually noted 57 80 Not observed 10 20 Total number of participants 21 15 Table 5 3. Perceived causes of local groundwater depletion Women Men Suggested causes of groundwater decline % of total responses Less rain 58 75 More borewells 83 50 Total number of participants 24 12

PAGE 169

169 Table 5 4. Suggested reasons for recent reductions in local rainfall amounts Women Men Reasons for observed changes in rainfall % of total responses gods, more population, people doing bad things 86 63 Less greenery (fo rest plantations logged; trees around croplands cut) 3 26 Unknown 10 32 Total number of participants 29 12 Table 5 5 Information used to decide about crop selection: ranked (greatest to least Women Men Dec ision information: crop selection % of total responses Observe water availability 81 80 Talk with family 65 50 Talk with neighbors 38 40 Market price information 8 30 Personal experience 0 45 Total number of participants 26 20 Table 5 6. Informati on used to decide about irrigation management: ranked (greatest to least Women Men Decision information: irrigation timing and depth % of total responses Observing plants, soil 85 77 Irrigate when power is on ; occasional shutoff if sufficient rain 40 62 Rainfall depth and timing 30 23 Ke eping 3 inches of water in rice fields 0 15 Total number of participants 20 13

PAGE 170

170 Figure 5 1 Some women take a break for lunch and to participate in discussions for this research

PAGE 171

171 Figure 5 2 One of six water harvesting structures (tanks) in the Wargal watershed used for increasing groundwater recharge Table 5 7. Decision category and associated most important responses Decision type or responsibility Dominant respons e Women Men crop selection observe water availability observe water availability irrigation management observe plants, soil observe plants, soil; irrigation if electricity is on rice cropland extent 1/8 to 1/4 of cropland 1/8 to 1/4 of cropland ri ce crop replacement if insufficient water cover crop, fallow cotton pump responsibility men men

PAGE 172

172 Table 5 8. Percentages of responses concerning who is responsible for irrigation pump control Women Men Responsibility for borewell pump control % of tot al responses Men ; women only if man is unavailable 100 100 Women 0 0 Total number of participants 19 8 Table 5 9. Percentages of responses concerning who is responsible for purchase farm inputs (seed and fertilizer) Women Men Responsibility for pur chasing of farm inputs % of total responses Men; women only if man is unavailable 100 100 Women 0 0 Total number of participants 10 6 Table 5 10 Literature review of yield increase for selected crops in response to tied ridge tillage; methods are me asured yield (obs) or simulated yield (sim) Crop Location % yield increase Methods Reference corn Nebraska 12 obs Duley, 1960 sorghum Kansas 17 obs Luebs, 1962 sorghum Kansas 22 obs Musick, 1981 cotton Texas 32 obs Gerard et al., 1983a sorghum Texas 33 obs Gerard et al., 1984 cotton Texas 36 obs Clark, 1983 sorghum Texas 40 sim Krishna et al., 1987 sorghum Texas 104 obs Gerard et al., 1983b sorghum Texas 176 obs Jones and Clark, 1987

PAGE 173

173 Table 5 11. Yields of crops commonly grown in Wargal bas ed on household surveys and state agency data. Sample size indicates number of households that responded to surveys about yield data. Value ($/kg) based on 2009 2010 government of India minimum support prices (MSP) Corn Cotton Rice Potato Sunflower Gr een bean yield: kg/ha 1936 398 2866 5389 917 2047 sample size 19 5 11 9 3 11 value: $/kg 0.188 0.617 0.213 0.186 0.497 0.448 Table 5 12. Groundwater balance s (mm) and estimated values (USD) of the selected top seven management scenarios (Chapter 4) Annual values, watershed scale Management scenario GW balance, mm USD, $ red_rice_irr,k,r,25 38.2 304 770 red_rice_k corn,r irr corn,25: mds33_kharif 36.1 339 998 red_rice_k,r,25 34.3 335 593 red_rice_irr,k,25: mds33_kharif 31.3 337 433 red_rice_k c orn,r irr corn,25 18.5 329 358 red_rice_irr,k,25 18.0 328 646 mds50_kharif 6.0 351 021 existing 10.9 342 234

PAGE 174

174 CHAPTER 6 LIMITATIONS APPLICATIONS, AND CONCLUSIONS OF THE WARGAL STUDY OF GROUNDWATER DEPLETIO N AND AGRICULTURAL M ANAGMENT Limitations The simplified groundwater balance is one weakness of this study; extending this research to more explicitly represent groundwater flow in three dimensions would certainly add valuable information. This would allow changes in water table elevation to be predicted with more certainty than the one dimensional estimates used here based on observed specific yield. Increased groundwater monitoring would be required to model flow across the watershed boundaries, and more certain information on the hydraulic pr operties of the aquifer system would need to be observed. Increased field scale water balance data would have improved confidence in watershed having no previous init iatives or instrumentation. It was a limitation of this research to have few hydrologic observations available to calibrate and evaluate SWAT. Observations of soil moisture, field scale runoff, and reservoir areas/volumes of the other five reservoirs wou ld have added to the quality of the water balance modeling. The preliminary fieldwork to learn about decision making and management of farmers in Wargal was not designed and completed in a way that allowed for any quantitative or any rigorous qualitative f indings. While there was a lot of important information that was learned (this enriched the project and made the modeled scenarios more realistic), an improvement in the depth of the socio economic analysis would add a lot to this work. A better understa about changes in policy, (2) the ability/willingness of farmers to change tillage strategy (detailed cost and labor information), and (3) the preferred alternatives to rice croplands.

PAGE 175

175 Applications This r esearch should add to a growing body of evidence that extent of rice croplands in some parts of India is beyond what is sustainable for long term supply of groundwater resources. A gricultural policy changes to reduce support for rice cultivation or to fav or rainfed crops may be required to manage groundwater depletion. It is not meant to suggest here that there is something inherently unsustainable about flooded rice cultivation, only that the extent should be carefully managed. This research provides an improved method of describing tillage management using an expanded form of Green Ampt infiltration that includes a time varying surface storage depth. This change is supported by the theoretical foundations of Green Ampt and by detailed sensitivity analys es (Chapter 3). SWAT was modified to include the modified infiltration description and was used to simulate management alternatives for the Wargal area, demonstrating the significance of tillage management in a way that was previously unavailable in SWAT (and other landscape hydrology models). SWAT is a very widely used water balance simulation tool, and the modified form developed and used here can be useful in numerous other applications. Another application of this research is to increase the considera tion of tillage management for increasing groundwater recharge. Water harvesting tillage is not effective in all systems, particularly in soils having very low hydraulic conductivity, but this research has demonstrated that water harvesting tillage can ma ke moderate increases in groundwater recharge. Surface water storage through tillage may become a more important management factor, especially given evidence of increasing rainfall variability, in areas where groundwater depletion is a concern.

PAGE 176

176 Conclusi ons This research of the Wargal watershed in peninsular India is a case study of the causes and possible solutions to groundwater depletion in India. In Chapter 1, the national and state level data on rice production in India and the field scale water ba lances of rice croplands gave strong indications that the extent and management of rice croplands is likely a significant contributor to groundwater depletion. Chapter 1 also highlighted the importance of improving simulation tools for making predictions about the hydrologic impacts of management changes in ungauged basins. In Chapter 2, the rainfall IDF analysis, together with the literature on changing rainfall character, showed that rainfall patterns are likely becoming more episodic in the Wargal reg ion, leading to a lower proportion of rainfall contributing to infiltration and recharge. The application being that surface storage will become a more important management concern for agronomic and groundwater recharge improvements. The sensitivity ana lyses of Chapter 3 compared the importance of the 5 parameters used for Green Ampt infiltration predictions, finding that the parameter describing tillage induced surface storage (MDS) was important (2 nd only to effective hydraulic conductivity) for infilt ration predictions in areas under water harvesting tillage. MDS is usually neglected or only included in a surface water balance for Green Ampt infiltration, and it was shown in Chapter 3 that MDS should be included both in the water balance and in the Gr een Ampt equations to improve infiltration predictions for areas having water harvesting tillage. In Chapter 4, the improved infiltration equations were implemented in SWAT to evaluate tillage for increasing groundwater recharge in Wargal. It was found that tillage was indeed effective at increasing recharge, changing the annual groundwater balance

PAGE 177

177 from 11 mm (existing management, 535 mm recharge) to 6 mm (MDS=50 mm in kharif rainfed crops, 552 mm recharge) with water harvesting tillage applied in the rainfed croplands during kharif/rainy season (41% of the watershed area) For watersheds with more extensive rainfed areas, this suggests that water harvesting tillage could have even more significant improvements to the groundwater balance. Also in Chap ter 4, small reductions (25%) in rice cropland irrigation or areal extent shifted the groundwater balance from negative to positive, increasing the evidence that the extent and management of rice croplands is a probable contributor to groundwater depletion Chapter 5 recounted the participation of growers in Wargal, reporting on the information used in decision making, preferred agricultural management, and major concerns in the community. Groundwater depletion was a common concern of participants. Esti mates of watershed scale economic yield were made to compare existing and the most promising alternative management scenarios. These estimates showed the possible economic gains associated with water harvesting tillage and the probable losses associated with reduced rice cropland extent or irrigation. The participation of farmers from the study area during the early stages of the design of this research has substantially improved this research. The problem of groundwater depletion was verified by farmers in Wargal. The management scenarios (labor, capital) and management preferences (rainfed and irrigated crops that work best for them). This combination of hydrologic and socio economic sciences has the potential to improve the adoptability of management recommendations. The cooperation of farmers in learning about water resource availability and decision making

PAGE 178

178 for farm management improves the relationship between sc ientists and stakeholders; this may mean that farmers would be more open to on farm trials of alternative management. It could be said that groundwater depletion in India is a self regulating problem: declining water tables will naturally lead to less gro undwater withdrawal as it becomes more expensive and less predictable to pump groundwater from greater depths. This is of course true, but without any early management changes there are serious costs associated with allowing the groundwater resource to re ach the stage of self regulation. Depleted hard rock aquifers can have serious fluoride and arsenic contamination, and the loss of a reliable water source for irrigation during droughts and dry spells can lead to much more vulnerable agricultural systems. So what is the solution to groundwater depletion in agricultural areas of India? The silver bullet is an idiomatic expression describing something that is completely effective at resolving some kind of problem, and as is often the case in any type of sea rch for the best management practices, there is no silver bullet agricultural management for groundwater depletion in Wargal. There is no technology or single management option that is clear leader in reducing or reversing groundwater depletion. It coul d be argued that groundwater depletion in Wargal (and in numerous regions like Wargal) has socio economic roots ; agricultural systems have become established that depend heavily on irrigation from groundwater sources. Groundwater has supported booms in th e agricultural economies of Asia, and it will continue to be an important resource for irrigation. However, the risks associated with reduced irrigation from groundwater (more yield variability) should somehow be compared to the risks of a

PAGE 179

179 severely deplet ed groundwater system (water quality, possible complete loss of irrigation source). While groundwater depletion has socio economic causes and solutions, part of those solutions will be technological and managerial. It was shown in Chapter 4 that are sev eral combinations of alternative management options that show promise for improving groundwater supply. Expanding the use of water harvesting tillage and reducing rice cropland extent or irrigation depths are two options that considerably improved the gro undwater balance. The required extents of these changes, in order to make substantial groundwater balance improvements (Chapter 4), can be useful quantities for evaluating these management options in other regions where groundwater depletion is a problem

PAGE 180

180 APPENDIX A SWAT CODE MODIFICATI ONS Changes to the following 2 2 subroutines of SWAT were required to correctly operationalize the modified GAML infiltration. Please contact the author for the source code or for details about the code changes. It is expected that the next official SWAT release will include these modified routines (Arnold, 2010) : surq_greenampt.f etact.f irrsub.f newtillmix.f readtill.f surface.f volq.f simulate.f operaten.f readhru.f rchday.f rchmon.f readmgt.f subbasin.f resetlu.f readlup.f watbal.f zero0.f header.f allocate_parms.f modparm.f parm.mod The complete, modified subroutine for SWAT GAML infiltration is included below. Other subroutines requiring major changes were those for evapotranspiration (etact.f), irrigation (irr sub.f), and tillage (newtillmix.f and readtill.f).

PAGE 181

181 subroutine surq_greenampt !! ~ ~ ~ PURPOSE ~ ~ ~ !! Predicts daily runoff given breakpoint precipitation and snow melt !! using the Green & Ampt technique !! ~ ~ ~ INCOMING VARIABLES ~ ~ ~ !! name |units |definition !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! cnday(:) |none |curve number for current day, HRU and at !! |current soil moist ure !! idt |minutes |length of time step used to report !! |precipitation data for sub daily modeling !! ihru |none |HRU number !! iyr |year |year being simulated (eg 198 0) !! mds(:) |mm H2O |maximum depression storage for HRU !! nstep |none |max number of time steps per day !! newrti(:) |mm/hr |infiltration rate for last time step from the !! |pre vious day !! nstep |none |number of rainfall time steps for day !! precipdt(:) |mm H2O |precipitation for the time step during day !! sol_k(1,:) |mm/hr |saturated hydraulic conductivity of 1st soil !! |layer !! sol_por(:,:)|none |total porosity of soil layer expressed as a !! |fraction of the total volume !! sol_sumfc(:)|mm H2O |amount of water held in the soil profile !! |at field capacity !! sol_sw(:) |mm H2O |amount of water stored in soil profile on !! |any given day !! swtrg(:) |none |rainfall event flag: !! | 0: no r ainfall event over midnight !! | 1: rainfall event over midnight !! wfsh(:) |mm |average capillary suction at wetting front !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! ~ ~ ~ OUTGOING VARIABLES ~ ~ ~ !! name |units |definition !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! h_depth(:,:)|mm H2O |surface storage depth for time step !! h_day(:) |mm H2O | end of day or daily surface storage depth !! |in HRU !! hhqday(:) |mm H2O |surface runoff generated each hour of day !! |in HRU !! newrti(:) |mm/hr |infiltration rate for last time step from the !! |previous day !! surfq(:) |mm H2O |surface runoff for the day in HRU

PAGE 182

182 !! swtrg(:) |none |rainfall event flag: !! | 0: no rainfall event over midnight !! | 1: rainfall event over midnight !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! ~ ~ ~ LOCAL DEFINITIONS ~ ~ ~ !! name |units |definition !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! adj_hc |mm/hr |adjusted hydraulic conductivity !! cuminf(:) |mm H2O |cumulative infiltration for day !! cumr(:) |mm H2O |cumulative rainfall for day !! d f(:) |mm H2O |incremental infiltration for time step !! dthet |mm/mm |initial moisture deficit !! excum(:) |mm H2O |cumulative runoff for day !! exinc(:) |mm H2O |runoff for time step !! incro(:) |mm H2O |incremental actual runoff for time step !! |accounting for surface storage for time step !! f1 |mm H2O |test value for cumulative infiltration !! j |none |HRU number !! k |none |counter !! kk |hour |hour of day in which runoff is generated !! psidt |mm |suction at wetting front*initial moisture !! |deficit !! rateinf(:) |mm/hr |infiltration rate for time step !! rintns(:) |mm/hr |rainfall intensity !! soilw |mm H2O |amount of water in soil profile !! tst |mm H2O |test value for cumulative infiltration !! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ !! ~ ~ ~ SUBROUTINES/FUNCTIONS CALLED ~ ~ ~ !! Intrinsic: Sum, Exp, Real, Mod !! ~ ~ ~ ~ ~ ~ END SPECIFICATIONS ~ ~ ~ ~ ~ ~ use parm integer :: j, k, kk real :: adj_hc, dthet, soi lw, psidt, tst, f1 real, dimension (nstep+1) :: cumr, cuminf, excum, exinc, rateinf real, dimension (nstep+1) :: rintns, df, incro !! array location #1 is for last time step of prev day j = 0

PAGE 183

183 j = ihru !! reset values for day cumr = 0. cuminf = 0. excum = 0. exinc = 0. rateinf = 0. rintns = 0. ro = 0. incro = 0. !! h_depth from last time step of previous day is !! is already defined as initia l h for current day !! at the end of this routine if (h_day(j) >= 0.) then h_depth(j,1) = h_day(j) end if !! calculate effective hydraulic conductivity adj_hc = 0. !! calculate effective hydraulic conductiv ity as half of saturated !! conductivity to remove CN from calculations for sensitivity analysis adj_hc = (56.82 sol_k(1,j) ** 0.286) / & (1. + 0.051 Exp(0.062 cnday(j))) 2. if (adj_hc <= 0.) adj_hc = 0.001 !! adj_hc = sol_k(1,j) 0.5 !! if (adj_hc <= 0.) adj_hc = 0.001 dthet = 0. if (swtrg(j) == 1) then swtrg(j) = 0 dthet = 0.001 sol_por(1,j) 0.95 rateinf(1) = newrti(j) newrti(j) = 0. else soilw = 0. if (sol_sw(j) >= sol_sumfc(j)) then soilw = 0.999 sol_sumfc(j) else soilw = sol_sw(j) end if

PAGE 184

184 dthet = (1. soilw / sol_sumfc(j)) sol_p or(1,j) 0.95 rateinf(1) = 2000. end if psidt = 0. psidt = dthet wfsh(j) k = 1 rintns(1) = 60. precipdt(2) / Real(idt) if (j == 1) write(126,5002) k, precipdt(1), cumr(1), rintns(1), & & rateinf(1), cuminf(1), excum(1), exinc(1) do k = 2, nstep+1 !! calculate total amount of rainfall during day for time step !! and rainfall intensity for time step cumr(k) = cumr(k 1) + precipdt(k) rintns(k) = 60. precipdt(k+1) / Real(idt) !! if rainfall intensity is less than infiltration rate !! everything will infiltrate if (rateinf(k 1) >= rintns(k 1).and.h_depth(j, k 1)==0.) then cuminf(k) = cu minf(k 1) + rintns(k 1) Real(idt) / 60. if (excum(k 1) > 0.) then excum(k) = excum(k 1) exinc(k) = 0. else excum(k) = 0. exinc(k) = 0. end if else !! if rainfall intensity is greater than infiltration rate !! or if h_depth > 0 !! find cumulative infiltration for time step by successive !! substitution tst = 0. tst = adj_hc Real(idt) / 60. do f1 = 0. f1 = cuminf(k 1) + adj_hc Real(idt) / 60. + (psidt + dthet*h_depth(j, k 1)) Log((tst + psidt + dthet*h_depth(j, k 1)) / (cuminf(k 1) + psidt + dthet*h_d epth(j, k 1))) if (Abs(f1 tst) <= 0.001) then cuminf(k) = f1

PAGE 185

185 !! Calculate the actual surface storage, h_depth, !! according to a parameterized mds; limit h_depth to mds !! Calculate the actual runoff for th e time step as !! incro = h_depth (before mds constraint) mds df(k) = cuminf(k) cuminf(k 1) h_depth(j, k) = h_depth(j, k 1) + precipdt(k) df(k) if (h_depth(j, k) < 1.e 6) then h_depth(j, k) = 0. end if if (h_depth(j, k) > mds(j)) then if (h_depth(j, k) > mds(j)) then incro(k) = h_depth(j, k) mds(j) incro(k) = h_depth(j, k) mds(j) h_depth(j, k) = mds(j) end if !! Calculate cumulative runoff, excum(k), for the day as the sum !! of previous step RO, excum(k 1), and RO for current time step, incro(k) excum(k) = excum(k 1) + incro(k) exinc(k) = excum(k) excum(k 1) if (exinc(k) < 0.) exinc(k) = 0. kk = 0 kk = (k 1) idt if (Mod(kk,60) == 0) then kk = kk / 60 else kk = 1 + kk / 60 end if hhqday(kk) = hhqday(kk) + exinc(k) surfq(j) = surfq( j) + exinc(k) exit else tst = 0. tst = f1 end if end do end if !! calculate new rate of infiltration rateinf(k) = adj_hc ((psidt + h_dep th(j, k)*dthet) / & (cuminf(k) + 1.e 6) + 1) if (j == 1) write(126,5002) k, precipdt(k, cumr(k), rintns(k),&

PAGE 186

186 & rateinf(k), cuminf(k), excum(k), exinc(k) end do !! define h_depth for first time st ep of following day and define daily h_depth, h_day !! as the same as h for last time step of previous day h_depth(j, 1) = h_depth(j, nstep+1) h_day(j) = h_depth(j, nstep+1) if (Sum(precipdt) > 12.) then swtrg(j) = 1 new rti(j) = rateinf(nstep) end if return 5000 format(//,'Excess rainfall calculation for day ',i3,' of year ', & & i4,' for sub basin',i4,'.',/) 5001 format(t2,'Time',t9,'Incremental',t22,'Cumulative',t35,'Rain fall',& & t45,'Infiltration',t59,'Cumulative',t71,'Cumulative',t82, & & 'Incremental',/,t2,'Step',t10,'Rainfall',t23,'Rainfall', & & t35,'Intensity',t49,'Rate',t58,'Infiltration',t73,'Runoff',& & t84,'Runoff', /,t12,'(mm)',t25,'(mm)',t36,'(mm/h)',t48, & & '(mm/h)',t62,'(mm)',t74,'(mm)',t85,'(mm)',/) 5002 format(i5,t12,f5.2,t24,f6.2,t36,f6.2,t47,f7.2,t61,f6.2,t73,f6.2, & & t84,f6.2) end

PAGE 187

187 APPENDIX B INTERVIEW QUESTIONS FOR WAR GAL FARMERS The following questions were used as a general guide for interviews with farmers in Wargal. Questions were sometimes asked differently, as the interviews became better adapted to the phrasing of questions that would be most easily understood b y participants Grouping of questions into themes of main concerns, groundwater and rainfall changes, management and decision making, and general questions was done simply for organizational purposes. Questions were asked to individuals and to groups. I n group settings, all participants were asked to respond separat ely. While responses in groups could be expected to differ from those of individuals, no distinction in responses (between individual alone or in a group) was made in the results shown (Chapt er 5). Main concerns What problems are most concerning about life in Wargal? What things would you most like to see improved about your life? Perceptions on groundwater and rainfall changes Have gro undwater levels been changing? If yes, how is this obse rved? What is the cause of groundwater decline? Is anything done to increase groundwater supply? Have you noticed changes in rainfall? If yes, what kinds of changes? W hat do you think caused the changes in rainfall? How is rainfall measured /observed? Ma nagement and decision making

PAGE 188

188 How do you decide what to grow on your farm? Who makes the decisions? What information helps with this decision? How much of your cropland is in paddy each year? Why do you grow this? What would you grow in place of paddy if there was not enough water for this crop? Have you considered crops ICRISAT promotes ? (millet, sorghum, pigeon pea) How are fields tilled, planted, weeded, harvested? Who does these operations? How do droughts and dry spells affect crop selection, till age, irrigation? Who makes decisions about drought mitigation strategies? How do you decide when to irrigate and how much to irrigate? Who irrigates or who turns pump on and off? What changes in management would increase water availability? Who buys farm inputs like seed and fertilizer? General questions How often does the government ag. Dept./extension come to talk to you? Do you visit the local progressive farmer/s? Are you on subsidized rations? Are you a member of a SHG or is someone in your household a member?

PAGE 189

189 LIST OF REFERENCES Abbaspour K.C. 2008. SWAT CUP user manual. Federal Institute of Aquatic Science and Technology (Eawag). Available at: http://www.eawag.ch/forschung/sia m/software/swat/index. Accessed 3 December 201 0 Abbaspour, K.C., J. Yang, I. Maximov, R. Siber, K. Bogner, J. Mieleitner, J. Zobrist, and R. Srinivasan. 2007. Modelling hydrology and water quality in the pre alpine/alpine Thur watershed using SWAT. Jo urnal of Hydrology 333: 413 430 Alcamo, J., T. Henrichs and T. Rosch 2000. World water in 2025: global mod eling and scenario analysis. In World Water Scenarios Analyses F.R. Rijsberman, ed. Marseille, France: World Water Council. Allen, M.R. and W.J Ingram. 2002. Constraints on future changes in climate and the hydrologic cycle. Nature 419 : 224 232. Angel J.R. and H.A. Huff 1997 Changes in heavy rainfall in the Midwestern United States. Journal of Water Resources Planning and Management 121: 2 46 249. Anuraga, T.S., L. Ruiz, M.S. Mohan Kumar, M. Sekhar, and A. Leijnse. 2006. Estimating groundwater recharge using land use and soil data: a case study in South India. Agricultural Water Management 84(1 2): 65 76. AQUASTAT 2010 Water Resources Development and Management Service. Rome, Italy: Food and Agriculture Organization of the United Nations Available at: http://www.fao.org/nr/water/aquastat/main/index.stm Accessed 12 O ctober 2010. Araya, A. and L. Stroosnijder 2010. Effects of tied ridges and mulch on barley (Hordeum vulgare) rainwater use efficiency and production in Northern Ethiopia. Agricultural Water Management 97 : 841 847. Arnold J.G 2010. Personal communi cation. Arnold, J. G., R. Sr inivasan, R.S. Muttiah, and J. R. Williams. 1998. Large area hydrologic modeling and assessment: Part I. Model development. Journal of the American Water Resources Association 34(1): 73 89. Arnold J.G., R.S. Muttiah, R. Srinivasan, and P.M. Allen. 2000. Regional estimation of base flow and groundwater rech arge in the Upper Mississippi River Basin. Journal of Hydrology 227: 21 40.

PAGE 190

190 Bakhsh, A B. and R. S. Kanwar. 2008. Soil and landscape attributes interpret subsurface drainage clusters. Australian Journal of Soil Research 46: 735 744. Bernard, M.M. 1932 Formulas for rainfall intensities of long durations. Transactions of the American Society of Civil Engineers 96: 592 624. Beven, K. 2006. A manifesto for the equifinality thesis. Journal of Hydrology 320 : 18 36. Beven, K. and A. Binley. 1992. The F uture of Distributed Models Model Calibration and Uncertainty Prediction. Hydrological Processes 6(3): 279 298. Bhattacharyya, R V Prakash, S. Kundu and H. S. Gupta. 2006. Effect of tillage and crop rotations on pore size distribution and soil hydr aulic conductivity in sandy clay loam soil of the Indian Himalayas. Soil and Tillage Research 86(2): 129 140. Bingner, R.L. 1996. Runoff Simulated from Goodwin Creek Watershed Using SWAT. Transactions of the ASAE 39(1): 85 90. Blevins, R.L., W.W. Frye, P.L. Baldwin, and S.D. Robertson 1990. Tillage effects on sediment and soluble nutrient losses from a Maury silt loam. Journal of Environmental Quality 19 : 683 686. Bosch, D.D., J.M. Sheridan, H.L. Batten, and J. G. Arnold. 2004. Evaluation of the SWA T model on a coastal plain agricultural watershed. Transactions of the ASAE 47(5): 1493 1506. Bouwer, H. 1966. Rapid field measurement of air entry value and hydraulic conductivity of soil as significant parameters in flow system analysis. Water Resourc es Research 2 : 729 738. Carsel R.F. and R.S. Parrish 1988. Developing Joint Probability Distributions of Soil Water Retention Characteristics. Water Resources Research 24, 755 769. Carsel, R.F., J.C. Imhoff, P.R. Hummel, J.M. Cheplick, and A.S. Doni gian Jr. 1998. PRZM 3, Model for predicting pesticide and nitrogen fate in the crop root and unsaturated zones: Users manual for release 3.0 Athens, GA: United States Environmental Protection Agency CGWB 2002. Master Plan for Artificial Recharge to Groundwater in India. Faridibad, India: Central Groundwater Board. Available at: http://cgwb.gov.in/index.html Accessed 8 April 20 09

PAGE 191

191 CGWB 2007. Groundwater Information, Medak District, Andhra Pradesh. Fa ridibad, India: Central Groundwater Board. Available at: http://cgwb.gov.in/documents/SR/Medak.pdf Accessed 8 April 20 09 Chu, S.T. 1978. Infiltration during an unsteady rain. Water Resources Re search 14: 461 466. Chu Agor, M.L., R. Muoz Carpenaa, G. Kiker, A. Emanuelsson, and I. Linkov. 2011. Exploring sea level rise vulnerability of coastal habitats through global sensitivity and uncertainty analysis. Environmental Modeling and Software 26: 593 604. Clark, L.E. 1983. Response of cotton to cultural practices. Texas Agricultural Experiment Station Progress Report 4174 4176: 17 29. Darboux, F. and C. Huang. 2005. Does soil surface roughness increase or decrease water and particle transfer? Soil Science Society of America Journal 69: 748 756. Das D G. Samanta B.K. Mandal, R.T. Chowdhury, C.R. Chanda, P. Chowdhury, B.K. Basu, and D. Chakraborti 1996. Arsenic in groundwater in six districts of West Bengal, India. Environmental Geochemi cal Health 18:5 15. de Vries, J.J. and I. Simmers. 2002. Groundwater recharge: an overview of processes and challenges. Hydrogeology Journal 10: 5 17. Delin, G.N., R. W. Healy, M. K. Landon, and J. K. Bhlke. 2000. Effects of topography and soil prope rties on recharge at two sites in an agricultural field. Journal of the American Water Resources Association 36: 1401 1416. Derib, S.D., T. Assefa B. Berhanu, and G. Zeleke 2009. Impacts of micro basin water harvesting structures in improving vegetativ e cover in degraded hillslope areas of north east Ethiopia. The Rangeland Journal 31 : 259 265. Dewandel, B., P. Lachassagne, R. Wyns, J.C. Marechal, and N.S. Krishnamurthy 2006. A generalized 3 D geological and hydrogeological conceptual model of gran ite aquifers controlled by single or multiphase weathering. Journal of Hydrology 330: 260 284. Duley, F.L. 1960. Yields of different cropping systems and fertilizer tests under stubble mulching and plowing in eastern Nebraska. Nebraska Agricultural Exp eriment Station Research Bulletin: 190.

PAGE 192

192 Easterling, D R G.A. Meehl, C. Parmesan S.A. Changnon, T.R. Karl and L.O. Mearns. 2000. Climate extremes, observations, modeling, and impacts. Science 298 : 2068 2074. Eilers, V. H. M., R. C. Carter, and K. R. Rushton. 2007. A single layer soil water balance model for estimating deep drainage (potential recharge): An application to cropped land in semi arid North east Nigeria. Geoderma 140: 119 131. Engel, B., D. Storm, M. White, J. Arnold, and M. Arabi. 200 7. A hydrologic/water quality model application protocol. Journal of the American Water Resources Association 43(5): 1223 1236. Fangmeier, D.D., Elliot, W.J., Workman, S.R., Huffman, R.L., Schwab, G.O. 2005. Soil and Water Conservation Engineering, fif th ed. Thomson Delmar, New York. FAOSTAT. 2010. Database on Agriculture Rome, Italy: Food and Agriculture Organization of the United Nations Available at: http://faostat.fao.org/default.aspx Accesse d 5 October 2010. Finch, J. W. 1998. Estimating direct groundwater recharge using a simple water balance model sensitivity to land surface parameters. Journal of Hydrology 211: 112 125. Foster, S. S.D. and P.J. Chilton, P.J. 2003. Groundwater: the proc esses and Philosophical Transactions Royal Society London B 358: 1935 1955. Fox, G.A., R. Muoz Carpena, and G.J. Sabbagh, G.J. 2010. Influence of flow concentration on parameter importance and prediction uncerta inty of pesticide trapping by vegetative filter strips. Journal of Hydrology 384: 164 173. Gassman, P. W., M. R. Reyes, C. H. Green, and J. G. Arnold. 2007. The soil and water assessment tool: historical development, applications and future research dir ections. Transactions of the ASABE 50 (4): 1211 1250. Geological Survey of India. 2001. August 2000 Flood in Hyderabad City: Causative Factors and Suggestions to avoid Recurrence. Available at: http://www.portal.gsi.gov.in/gsiDoc/pub/cs_hydflood_aug00.pdf Accessed 9 March 201 1 Gerard C.J., P.D. Sexton, and D.M Matus. 1983a. Furrow diking for cotton production in the Rolling Plains. Texas Agricultural Experiment Station Repo rt 4174 4176: 1 16.

PAGE 193

193 Gerard C.J., P.D. Sexton, L.E. Clark, and E.C. Gilmore. 1983b. Sorghum for grain: Production strategies in the Rolling Plains. Texas Agricultural Experiment Station Bulletin 1428: 1 12. Gerard, C.J., P.D. Sexton, and D.M. Conover. 1 984. Effect of furrow diking, subsoiling and slope position on crop yields. Agronomy Journal 76: 945 950. Gogu, R.C. and A. Dassargues. 2000. Current trends and future challenges on groundwater vulnerability assessment using overlay and index methods. Environmental Geology 39: 549 559. Goswami B N V. Venugopal, D. Sengupta, M.S. Madhusoodanan, and P.K. Xavier 2006. Increasing trends of extreme rain events over India in a warming environment. Science 314: 1442 1445. Green, W. H. and G. A. Ampt. 1911 Studies on Soil Physics. 1. The flow of air and water through soils. Journal of Agricultural Science 4(1):1 24. Gujja B S. Dalai, H. Sha ik and V. Goud. 2009. Adapting to climate change in the Godavari River basin of India by restoring traditiona l water storage systems. Climate and Development 1, 229 240. Guo, Y. Updating rainfall IDF relationships to maintain urban dr ainage design standards. 2006 Journal of Hydrologic Engineering 11(5): 506 509. Gupta, H.V., T. Wagener, and Y. Liu. 2008. Reconciling theory with observations, elements of a diagnostic approach to model evaluation. Hydrological Processes 22 : 3802 3813. Gupta, P K S. Panigrahy, and J.S. Parihar. 2010 Modeling the spatiotemporal changes in the runoff of major basins in In dia under the future climate change scenario. Bulletin of the National Natural Resources Management System 35 : 57 62. Guzha, A.C. 2004. Effects of tillage on soil microrelief, surface depression storage and soil water storage. Soil and Tillage Research 76:105 114. Haan C T B.J. Barfield, and J.C. Hayes. 1994 Design Hydrology and Sedimentology for Small Catchments Academic Pr ess, San Diego. Helton, J.C. and F.J. Davis. 2000. Sampling based methods In Sensitivity Analysis. A. Saltel li, K. Ch an, and E.M Scott, eds. New York: Wiley Hennessy, K.J., J.M.Gregory, and J.F.B Mitchell 1997. Changes in daily precipitation under enhanced greenhouse conditions. Climate Dynamics 13: 667 680.

PAGE 194

194 Huber, W.C. and R.E. Dickinson 1992. Storm Water Manage ment Model, Athens, GA: Environmental Research Laboratory Office of Research and Development United States Environmental Protection Agency. Huff, F .A. 1990 Time distributions of heavy rainstorms in Illinois. Illinois State Wat er Survey Circular, 173. Hughes, D.A. 2010. Hydrologic models, mathematics or science? Hydrological Processes 24 : 2199 2201. Huisman, J. A., L. Breuer, and H.G. Frede. 2004. Sensitivity of simulated hydrological fluxes towards changes in soil propertie s in response to land use change. Physics and Chemistry of the Earth 29: 749 758. IPCC (Intergovernmental Panel on Climate Change). 2001. Climate Change 2001, The Scientific Basis. Cambridge England: Cambridge University Press. Jain S K P.K. Agarwa l, and V.P. Singh. 2007 Hydrology and water resources of India. In Water Science Technology Library 57 Dordrecht, The Netherlands : Springer. Jones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Batchelor, L.A. Hunt, P.W. Wilkens, U. Singh, A. J. Gijsman, and J. T. Ritchie. 2003. The DSSAT cropping system model. European Journal of Agronomy 18(3 4): 235 265. Jones, O.R., and R.L. Baumhardt. 2003. Furrow Dikes. In Encyclopedia of Water Science 317 320. B.A. Stewart and T.A. Howell, eds New York, NY: Marcel Dekker. Jones, O.R., and R.N. Clark 1987. Effects of furrow dikes on water conservation and dryland crop yields. Soil Science Society of America Journal 51 : 1307 1314. Joshi U and M. Rajeevan 2006. Trends i n precipitation extremes over India. Pune, India: National Climate Centre Available at: http://www.imdpune.gov.in/ncc_rept/RESEARCH%20%20REPORT%203. pdf Accessed 15 December 2010. Kalra, Y.P. and D. G. Maynard. 1991. Methods Manual for Forest Soil and Plant Analysis. Inf. Rep. NOR X 319. For. Can., Northwest Reg., North. Four. Cent. Edmonton, AB. Kamphorst, E.C., V. Jetten, J. Gurif, J. Pitknen, B.V. Iversen, J.T. Douglas, and A. Paz 2000. Predicting depressional storage from soil surface roughness. Soil Science Society of America Journal 64 : 1749 1758.

PAGE 195

195 Karl T R., R.W. Knight, and N. Plummer. 1995 Trends in high frequency climate variability in the twentieth century. Nature 3 77: 217 220. Kim, N.W., M. Chung, Y.S. Won, and J.G. Arnold. 2008. Development and application of the integrated SWAT MODFLOW model. Journal of Hydrology 356: 1 16. Kirchner, J.W. 2006. Getting the right answers for the right reasons: Linking measure ments, analyses, and models to advance the science of hydrology Water Resour ces Research 42: 1 5. Klute, A., Dirksen, C. 1986. Hydraulic conductivity and diffusivity: laboratory methods. In Methods of Soil Analysis, Part 1. Physical and Mineralogical Methods, 2nd ed. Agron. Monogr. 9, 687 734. A. Klute, ed. Madison, WI: ASA. Knapen, A., J. Poesen, G. Govers, and S. De Baets 2008. The effect of conservation tillage on runoff erosivity and soil erodibility during concentrated flow. Hydrologic Proc esses 22 : 1497 1508. Konikow, L.F. and E. Kendy 2005. Groundwater depletion: A global problem. Hydrogeology Journal 13: 317 320. Kothawale, D.R., K. Rupa Kumar, D.A. Mooley, A.A. Munot, B. Parthasarathy, and N.A. Sontakke. 2008. IITM Indian regional/ subdivisional Monthly Rainfall data set (IITM IMR). Pune, India: Indian In stitute of Tropical Meteorology. Available at: http://www.tropmet.res.in Accessed 5 October 2010. Kothyari U C and R.J. Garde. 1992 Rainfall intensity duration frequency formula for India. Journal of Hydraulic Engineering 118: 323 336. Koutsoyiannis, D., C. Onof. 2001 Rainfall disaggregation using adjusting procedures on a Poisson clu ster model. Journal of Hydrology 246: 109 122 Koutsoyiannis, D., D. Kozonis, and A. Manetas. 1 998. A mathematical framework for studying rainfall intensity duration frequency relationships. Journal of Hydrology 206 : 118 135. Krishna, J.H., G.F. Arkin, J.R. Williams, and J.R. Mulkey. 1987. Simulat ing furrow dike impacts on runoffand sorghum yields. Transactions ASAE 30: 143 147. Krysanvoa, V. and J.G. Arnold. 2008. Advances in ecohydrological modelling with SWAT a review. Hydrological Sciences 53(5): 939 947.

PAGE 196

196 Kukal, S.S. and G.C. Aggarwal. 20 02. Percolation losses of water in relation to puddling intensity and depth in a sandy loam rice (Oryza sativa) field. Agricultural Water Management 57: 49 59 Kunkel, K.E., K. Andsager, and D.R. Easterling. 1999. Long term trends in extreme precipitati on events over the conterminous United States and Canada. Journal of Climate 12: 2515 2527 Laflen, J.M., W.J. Elliot, D.C. Flanagan, C.R. Meyer, and M.A. Nearing 1997. WEPP predicting water erosion using a process based model. Journal of Soil Water Co nservation 52: 96 102. Le Maitre, D.C., D.F. Scott, and C. Colvin. 1999 A review of information on interactions between veg etation and groundwater. WaterSA 25 (2): 137 152. Lerner, D.N., A.S Issar, and I. Simmers. 1990. Groundwater Recharge: a Guide to Understanding and Estimating Natural Recharg e Hannover, Germany: Verlag Heinz Heise. Liebe, J., N. van de Giesen, and M.S. Andreini. 2005. Estimation of small reservoir storage capacities in a semi arid environment: a case study in the upper e ast r egion of Ghana. Physics and Chemistry of the Earth 30 : 448 454. Liebe, J., N. van de Giesen, M. S. Andreini, M.T. Walter, and T. Steenhuis. 2009. Determining watershed response in data poor environments with remotely sensed small reservoirs as runoff ga u ges. Water Resources Research 45: W07410 Lin, Y.S., Y.W. Lin, Y. Wang, Y.G. Chen, M.L. Hsu, S.H. Chiang, and Z.S. Chen. 2007. Relationships between topography and spatial variations in groundwater and soil morphology within the Taoyuan Hukou tableland, northwestern Taiwan. Geomorphology 90(1 2): 36 54. Luebs, R.E. 1962. Investigations of cropping systems, tillage methods, and cultural practices for dryland farming at the Fort Hays (Kansas) Branch Experiment Station. Kansas Agricultural Experiment Sta tion Bulletin 449: 32 33. Madsen H K. Arnbjerg Nielsen P.S. Mikkelsen. 2009 Update of regional intensity duration frequency curves in Denmark: Tendency towards increased storm intensities. Atmospheric Research 92: 343 349. Marechal, J.C., B. Dewand el, S. Ahmed, L. Galeazzi, and F.K. Zaidi. 2006. Combined estimation of specific yield and natural recharge in a semi arid groundwater basin with irrigated agriculture Journal of Hydrology 329 : 281 293

PAGE 197

197 Martin, Y., C. Valeo, and M. Tait. 2008. Centim etre scale digital representations of terrain and impacts on depression storage and runoff. Catena 75: 223 233. May W. 2004 Variability and extremes of daily rainfall during the Indian summer monsoon in the period 1901 1989. Global and Planetary Chan ge 44: 83 105. Mein, R.G. and C. L. Larson. 1973. Modeling infiltration during a steady rain. Water Resources Research 9(2): 384 394. Meyer, P.D., M.L. Rockhold, and G.W. Gee. 1997. Uncertainty analyses of infiltration and subsurface flow and transport for SDMP sites NUREG/CR 6565. Washington, DC. : U.S. Nuclear Regulatory Commission Mohymont, B., G.R. Demare and D.N. Faka. 2004. Establishment of IDF curves for precipitation in the tropical area of central Africa comparison of tech niques and res ults. Natural Hazards and Earth System Science s 4 (3) 375 387. Moriasi, D.N., J.G. Arnold, M.W. Van Liew, R.L. Bingner, R.D. Harmel, and T.L. Veith. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Tr ansactions of the ASABE 50(3): 885 900. Moron V A W Robertson M.N. Ward, and P. Camberlin. 2007 Spatial coherence of tropical rainfall at the regional scale. Journal of Climate 20 : 5244 5263. Mukherji, A., and T. Shah. 2005. Groundwater socio eco logy and governance: a review of institutions and policies in selected countries Hydrogeology Journal 13: 328 345. Muleta, M.K. and J.W. Nicklow. 2006. Sensitivity and uncertainty analysis coupled with automatic calibration for a distributed watershed model Journal of Hydrology 306: 127 145 Muoz Carpena, R., and S.J. Muller 2009. Formal exploration of the complexity and relevance of biogeochemical models through global sensitivity and uncertainty analysis, opportunities and challenges. Estudios en la Zona no Saturada del Suelo 9 : 1 12. Muoz Carpena, R., G.A. Fox, and G.J. Sabbagh 2010. Parameter Importance and Uncertainty in Predicting Runoff Pesticide Reduction with Filter Strips. Journal of Environmental Quality 39: 630 641.

PAGE 198

198 Muoz Carpena, R ., J.E. Parsons, and J.W. Gilliam 1999. Modeling hydrology and sediment transport in vegetative filter strips. Journal of Hydrology 214:111 129. Muoz Carpena, R., Z. Zajac, and Y. Kuo. 2007. Global sensitivity and uncertainty analyses of the water q uality model VFSMOD W. Transactions of the ASABE 50: 1719 1732. Musick, J.T. 1981. Precipitation management techniques new and old. In Annual Groundwater Management District Conference, 8th Smith, D.D., ed. Lubbock, TX: High Plains Underground Water Co nservation District. Norman, D.W, F.W. Worman, J.D. Siebe rt, and E. Modiakgotla. 1995 The Farming Systems Approach to Development and Appropriate Technology Generation. FAO Farm Syst ems Management Series 10. AGSP. Rome, Italy : Food and Agricultural Or ganization of the United Nations NRC (National Research Council). 1997. Valuing Ground Water: Economic Concepts and Approaches. Washington D.C.: National Academy Press. O'Brien, K., R. Leichenko, U. Kelkar, H. Venema, G. Aandahl, H. Tompkins, A. Javed, S. Bhadwal, S. Barg, L. Nygaard, and J. West. 2004. Mapping vulnerability to multiple stressors: climate ch ange and globalization in India. Global Environmental Change 14 ( 4 ): 303 313. Owor, M., R.G. Taylor, C. Tindimugaya, and D. Mwesigwa, D. 2009. R ainfall intensity and groundwater recharge, evidence from the Upper Nile Basin. Environmental Research Letters 4: 1 6. Pall, P., M.R. Allen, and D.A. Stone. 2006. Testing the Clausius Clapeyron constraint on changes in extreme precipitation under CO2 war ming. Climate Dynamics 28: 351 363. Philip J R. 1957. The theory of infiltration. 4. Sorptivity and algebraic infiltration equations. Soil Science 84: 257 264. Planchon, O., M. Esteves, N. Silvera, and J.M. Lapetite, 2002. Microrelief induced by t illage measurement and modeling of surface storage capacity. Catena 46: 141 157. Rangan H C.A. Kull and L. Alexander 2010 Forest plantations, water availability, and regional climate change: controversies surrounding Acacia mearnsii plantation s in the upper Palnis Hills, southern India Regional Environmental Change 10, 103 117.

PAGE 199

199 Rao, K.P. C., T. S. Steenhuis, A .L. Cogle, S.T. Srinivasan, D.F. Yule and G. D. Smith. 1998. Rainfall infiltration and runoff from an Alfisol in semi arid tropical I ndi a. I and II. No till systems. Soil and Tillage Research 48(1 2): 51 69. Rao, P.B K. Subrahmanyam, and R.L. Dhar. 2001. Geoenvironmental effects of groundwater regime in Andhra Pradesh, India. Environmental Geology 40: 4 5. Rawls, W.J., and D.L. Brake nsiek. 1986. Comparison between Green Ampt and curve number runoff predictions. Transactions of the ASAE 29: 1597 1599. Rawls, W.J., D.L. Brakensiek, and N. Miller. 1983. Green Ampt infiltration parameters from soil data. Proceedings of the American S ociety of Civil Engineers 109: 62 70. Reddy, D 2009. Personal communication. Water Technology Centre, ANGRAU, Hyderabad, India. Reddy, V.R. 2005. Costs of resource depletion externalities: a study of groundwater overexploitation in Andhra Pradesh, In dia. Environment and Development Economics 10: 533 556. Ritter, A., F. Hupet, R. Muoz Carpena, S. Lambot, and M. Vanclooster 2003. Using inverse methods for estimating soil hydraulic properties from field data as an alternative to direct methods. Ag ricultural Water Management 59: 77 96. Rodell, M., I. Velicogna, and J.S. Famiglietti. 2009. Satellite based estimates of groundwater depletion in India. Nature 460: 999 1002. Rodriguez Iturbe, I., Cox, D.R., Isham, V. 1988. A point process model for r ainfall: Further developments. Proceedings of the Royal Society of London A 417: 283 298. Rosenthal, W. D., R. Srinivasan, and J. G. Arnold. 1995. Alternative river management using a linked GIS hydrology model. Transactions of the ASAE 38 (3): 783 790. R ushton, K.R. 1988. Numerical and conceptual models for recharge estimation in arid and semi arid zones. In: Estimation of Natural Ground Water Recharge 223 238. I. Simmers, ed Dordrecht, The Netherlands: Reidel Sakthivadivel, R 2007. The Groundwa ter Recharge Movement in India. In: The Agricultural Groundwater Revolution: Opportunities and Threats to Development 195 210. M. Giordano and K.G. Villholth e ds. Cambridge, MA: CABI.

PAGE 200

200 Saltelli, A. 2005. Global sensiti vity analysis an introduction. In : Sensitivity Analysis of Model Output 2 7 43. K.M. Hanson and F.M. Hemez, eds Los Alamos, NM: Los Alamos N ational Laboratory. Saltelli, A., M. Ratto, T. Andres, F. Campolongo, J. Cariboni, D. Gatelli, M. Saisana, and S. Tarantola 2008. Global Se nsitivity Analysis: The Primer Chichester England: Wiley Interscience Saltelli A S. Tarantola and F. Campolongo 2000. Sensitivity analysis as an ingredient of modeling. Statistical Science 15: 377 395. Saltelli, A., S. Tarantola, and K. Chan, K. 1999. A quantitative, model independent method for global sensitivity analysis of model output. Technometrics 41 : 39 56. Sawunyama, T., A. Senzanje, and A. Mhizha. 2006. Estimation of small reservoir storage capacities in Limpopo River Basin using ge ographical information systems (GIS) and remotely sensed surface areas: Case of Mzingwane catchment. Physics and Chemistry of the Earth 31 : 935 943. Scanlon, B.R., K.E. Keese, A.L. Flint, L.E. Flint, C.B. Gaye, W.M. Edmunds, and I. Simmers. 2006. Global synthesis of groundwater recharge in semiarid and arid regions. Hydrological Processes 20: 3335 3370. Scanlon, B.R., R.W. Healy, and P.G. Cook. 2002. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeology Journal 10: 18 39. Schulla, J. and K. Jasper. 2000. Model Description WaSiM ETH. Zu rich: Institute of Geography, ETH. SCS (Soil Conservation Service). 1972 National Engineering Handbook, Section 4 Hydrology. Washington, D C. : U .S. Government printing Office. SCS ( S oil Conservation Service). 1986. Technical Release 55, Urban H ydrology for Small Watersheds. Washington, D.C.: U.S. Government printing Office Shah, T. 2007 The groundwater economy of south Asia : an assessment of size, significance and th e socio eco logical impacts. In: The Agricultural Groundwater Revolution: Opportunities and Threats to Development M. Giordano and K.G. Villholt h, eds Battaramulla, Sri Lanka: I nternational Water Management Institute Shah, T A.D. Roy A.S. Qureshi, and J. Wang 2003 Sustaining Asia's groundwater boom: An overview of issues and evidence. Natural Resources Forum 27: 103 141.

PAGE 201

201 S hah T., O.P. Singh, and A. Mukherji 2006. groundwater irrigation economy: analyses from a survey in Ind ia, P akistan, Nepal Terai and Bangladesh Hydrogeology Journal 14: 286 309 Sharpley A N and J.R. Williams. 1990 EPIC Erosion Productivity Impact Calculator: 1. Model documentation. Technical Bulletin 1768. Washington, D.C.: U SDA Agricultural Re search Service. UNESCO. Paris, France: United Nations Educational, Scientific, and Cultural Organization. Av ailable at: http://webworld.unesco.org/water/ihp/db/shiklomanov/summary/html/figure _1.html Accessed 22 June 2010. Singh, P 2009. Personal communication. Inte rnational Crops Research Institute for the Semiarid Tropics, Patancheru, India. Sivapalan, M., K. Takeuchi, S.W. Franks, V.K. Gupta, H. Karambiri, V. Lakshim, X. Liang, J. J. McDonnell, E. M. Mendiondo, O. Connell, T. Oki, J.W. Pomeroy, D. Scher tzer, S. Uhle nbrook, and E. Zehe. 2003. IAHS Decade on Predictions in Ungauged Basins (PUB), 2003 2012: Shaping an exciting future for the hydrological sciences Hydrological Sciences Journal 48 (6): 857 880. Skaggs, R. W. and R. Khaleel. 1982. Infiltration. In Hydr ologic Modeling of Small Watershed s Haan, C.T., H.P. Johnson, and D.L. Brakensiek, D.L. eds. St. Joseph : ASAE. Soil Survey Staff. 1999. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys (2nd ed.). Washington DC: US Department of Agriculture Soil Conservation Service. Stone, K.C. and E.J. Sadler. 1991. Runoff using Green Ampt and SCS curve number procedures and its effect on the CERES Maize model. ASAE Paper No. 91 2612. St. Joseph, MI : ASAE. Subba Rao, N. and D.J. Devadas. 2003. Fluoride incidence in groundwater in a n area of Pen insula India. Environmental Geology 45: 243 251. Tesfahunegn, G.B. and C. S. Wortmann. 2008. Tie r idge tillage for high altitud e pulse production in northern E thiopia. Agronom y Journal 100(2): 447 453. Trenberth, K.E., D. Aigu, R.M. Rasmussen, and D.B. Parsons 2003. The changing character of precipitation Bulletin of the American Meteorological Society 84: 1205 1217.

PAGE 202

202 Truman, C.C. and R.C. Nuti. 2009. Improved water capt ure and erosion reduction through furrow diking. Agricultural Water Management 96: 1071 1077. Truman, C.C. and R.C. Nuti. 2010. Furrow diking in conservation tillage. Agricultural Water Management 97: 835 840. Twomlow, S.J. and P.M.C. Breneau 2000. The influence of tillage on semi arid soil water regimes in Zimbabwe. Geoderma 95: 33 51. UNEP (United Nations Environment Programme). 1992. World Atlas of Desertification London: Edward Arnold van Griensven, A., T. Meixner, S. Grunwald, T. Bishop, M Di Luzio, and R. Srinivasan. 2006. A global sensitivity analysis tool for the parameters of multi variable catchment models. Journal of Hydrol ogy 324(1 4): 10 23. Van Liew, M.W., J.G. Arnold, and D. D. Bosch. 2005. Problems and potential of autocalibr ating a hydrologic model. Transactions of the ASAE 48(3): 1025 1040. Van Liew, M.W., T.L. Veith, D.D. Bosch, and J. G. Arnold. 2007. Suitability of SWAT for the Conservation Effects Assessment Project: comparison on USDA ARS watersheds. Journal of Hydrolo gic Engineering 12(2): 173 189. Van Mullem, J.A. 1991. Runoff and Peak Discharges Using Green Ampt Infiltration Model. Journal of Hydraulic Engineering 117: 354 370. Wallach, D., D. Makowski, and J. W. Jones. 2006. Working with Dynamic Crop Models: Eva luation, Analysis, Parameterization and Application Amsterdam, The Netherlands: Elsevier. Wilcox, B.P., W.J Rawls D.L. Brakensiek, and J.R. Wright 1990. Predicting runoff from rangeland catchments: a comparison of two models. Water Resources Researc h 26 : 2401 2410. Wilks, D.S. 1989. Rainfall intensity, the Weibull distribution, and estimation of daily surface run off. Journal of Applied Meteorology 28 : 52 58. Wiyo, K.A., Z.M. Kasomekera, and J. Feyen. 2000. Effect of tied ridging on soil water s tatus of a maize crop under Malawi conditions. Agricultural Water Management 45: 101 125. Yang H P. Reichert K.C. Abbaspour and A.J.B. Zehnder 2003. A water resources threshold and its implications for food security. Environmental Science and Tec hnology 37 : 3048 3054.

PAGE 203

203 Yang, J., P. Reichert, K.C. Abbaspour KC, J. Xia, and H. Yang. 2008. Comparing uncertainty analysis techniques for a SWAT application to Chaohe Basin in China. Journal of Hydrology 358: 1 23 Yuan, Y., R.L. Bingner, and R. A. Rebich. 2001. Evaluation of AnnAGNPS on Mississippi Delta MSEA watersheds. Trans actions of the ASAE 44(5): 1183 1190. C. Mit chell, P. L. Milthorpe, and M. Yee. 1999. Estimating episodic recharge under different crop/pasture rotations in the Mallee region, Part 2: Recharge control by agronomic practices. Agricultural Water Management 42: 237 249. Zhang, G.S., K.Y. Chan, A. Oates, D.P. Heenan, and G.B. Huang 2007. Relationship between soil struct ure and runoff/soil loss after 24 years of conservation tillage. Soil and Tillage Research 92: 122 128.

PAGE 204

204 BIOGRAPHICAL SKETCH Daniel Dourte was born in Harrisburg, Pennsylvania. Pursu ing an undergraduate degree in mechanical e ngineering, he studied at R ochester Institute of Technology and Messiah College. He graduated 2004 with a B.S. in mechanical e ngineering from Messiah College. Working on assistive devices for persons with disabilities resulted in two trips to Burkina Faso in 2004 and 2005. Intera ctions with farmers there inspired an interest in agriculture and water management. To test this interest and start the process of learning about farming systems, Daniel worked for vegetable farm in Hustontown, Pennsylvania as an irrigation and crop manage r. In the spring of 2006, he started a Master of Engineering program in agricultural and b iological e ngineering at the University of Florida. In this program he performed water balance experiments to measure crop water use of mature southern highbush blu eberries. A planning trip to several areas in India in January 2008 began his work on groundwater depletion in India. The results of his work in India have provided an improved hydrology model (SWAT), a demonstration of the importance of surface storage for infiltration predictions, and an assessment of agricultural management alternatives for groundwater sustainability. Daniel has been married to Natalie Dourte for 3 years, and they have a 1 year old daughter, Aubrey. Daniel volunteers as a vegetable ga rden instructor at Howard Bishop Middle School and is an active vegetable gardener at home. He rides his bike instead of driving as often as possible and enjoys all kinds of exercise. Daniel enjoys international travel and it is a goal of his to be invol ved in agricultural development and water management in the global South.