<%BANNER%>

Renewable Water Flows and Wealth in Thailand

Permanent Link: http://ufdc.ufl.edu/UFE0024952/00001

Material Information

Title: Renewable Water Flows and Wealth in Thailand
Physical Description: 1 online resource (225 p.)
Language: english
Creator: Sweeney, Sharlynn
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: emergy, evapotranspiration, gis, income, thailand, water
Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Renewable energy flows are the basis of many production processes. As they vary over the landscape, they may play a significant role in differential economic development. This dissertation explored the relationship between renewable flows and measures of socioeconomic wealth in Thailand. Emergy evaluation was used to quantify renewable flows. A monthly water balance model was implemented using Geographic Information Systems (GIS) to quantify hydrologic flows for input data. Models of irrigation, river geopotential and tide provided additional input. Values were extracted at the provincial scale from coverages of water flows (mm/yr) and emergy flows (sej/yr), and analyzed as independent variables in regression models of wealth. Composite variables were created using principal components analysis (PCA) for additional independent variables. Exploratory data analysis for best fit regression models resulted in robust models for Gross Provincial Product (GPP) per capita, agricultural fraction of GPP per area, and total GPP per area, with adjusted-R2 values ranging from 0.81 to 0.91. According to squared part correlation coefficients, renewable water flows had significant, unique influence on these income measures when accounting for provincial membership in the Northeast region via an indicator variable (NE). The NE variable itself uniquely accounted for 20% to 76% of total variance in income. Of the renewable flows, actual evapotranspiration (AET) had the largest unique influence on income total variance (part-R2 = 0.29, GPP per area model). Squared partial correlations (partial-R2) indicated that after accounting for variance explained by NE, a 50% to 72% of remaining variance in income measures was attributed to AET. Rainfall and AET were the flows included most often as significant independent variables. Because emergy measures for AET and rainfall were directly proportional to these flows in mm, models using emergy flows were identical to models in mm. Using principal components (PCs) of water flows as independent variables increased model strength slightly, while PCs of emergy flows did not improve model strength. These research results, explicitly addressing the relationship between renewable resource flows and income, are an initial step toward the crucial goal of understanding environment-economic linkages.
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 Sharlynn Sweeney.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Brown, Mark T.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024952:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024952/00001

Material Information

Title: Renewable Water Flows and Wealth in Thailand
Physical Description: 1 online resource (225 p.)
Language: english
Creator: Sweeney, Sharlynn
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: emergy, evapotranspiration, gis, income, thailand, water
Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Renewable energy flows are the basis of many production processes. As they vary over the landscape, they may play a significant role in differential economic development. This dissertation explored the relationship between renewable flows and measures of socioeconomic wealth in Thailand. Emergy evaluation was used to quantify renewable flows. A monthly water balance model was implemented using Geographic Information Systems (GIS) to quantify hydrologic flows for input data. Models of irrigation, river geopotential and tide provided additional input. Values were extracted at the provincial scale from coverages of water flows (mm/yr) and emergy flows (sej/yr), and analyzed as independent variables in regression models of wealth. Composite variables were created using principal components analysis (PCA) for additional independent variables. Exploratory data analysis for best fit regression models resulted in robust models for Gross Provincial Product (GPP) per capita, agricultural fraction of GPP per area, and total GPP per area, with adjusted-R2 values ranging from 0.81 to 0.91. According to squared part correlation coefficients, renewable water flows had significant, unique influence on these income measures when accounting for provincial membership in the Northeast region via an indicator variable (NE). The NE variable itself uniquely accounted for 20% to 76% of total variance in income. Of the renewable flows, actual evapotranspiration (AET) had the largest unique influence on income total variance (part-R2 = 0.29, GPP per area model). Squared partial correlations (partial-R2) indicated that after accounting for variance explained by NE, a 50% to 72% of remaining variance in income measures was attributed to AET. Rainfall and AET were the flows included most often as significant independent variables. Because emergy measures for AET and rainfall were directly proportional to these flows in mm, models using emergy flows were identical to models in mm. Using principal components (PCs) of water flows as independent variables increased model strength slightly, while PCs of emergy flows did not improve model strength. These research results, explicitly addressing the relationship between renewable resource flows and income, are an initial step toward the crucial goal of understanding environment-economic linkages.
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 Sharlynn Sweeney.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Brown, Mark T.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2011-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024952:00001


This item has the following downloads:


Full Text

PAGE 1

1 RENEWABLE WATER FLOWS A ND WEALTH IN THAILAND By SHARLYNN DAWN SWEENEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Sharlynn Dawn Sweeney

PAGE 3

3 I dedicate this work to my parents.

PAGE 4

4 ACKNOWLEDGMENTS My graduate educational experience was a succes s due to the support of my fam ily, friends and colleagues, often overlapping categories. I thank my parents for their encouragement and contributions to my education, especially in my formative years. My brother is a continual source of inspiration with his boundless energy in blending educationa l endeavors, a fast-paced career and supporting a growing and beautiful family. My network of friends in Gainesville have supported me in direct emotional and intellectual ways, as well as indirectly by just being a joy to spend time with, and by providing a continual bounty of wonderful events and escapes. Special thanks go to my committee members. Dr. Mark Browns sound intellectual guidance, keen insight and good humor contribut ed significantly to my positive experiences during this research effort. Dr. Michael Binford played a major ro le in the early stages, providing logistical support, fruitful brainstorming sessions and engaging company during the Thailand field trip. Dr. Clay Montague has contributed greatly to my overa ll education via his input during the weekly systems seminar discussions over the y ears, and his superb co urses. Dr. Clyde Kiker also provided valuable guidance from an ec onomic viewpoint, through both his coursework and early brainstorming sessions. Financial support for a portion of my research was provided by the U.S. Department of Education and UF Environmental Engineering Sc iences through their Graduate Assistance in Areas of National Need (GAANN) Fellowshi p program. I thank the UF GAANN program directors, J.M.M. Anderson and J.M. Andino, for their assistance through the Fellowship. I also want to particularly thank two of my closest friends : May, whose ardent support and endless enthusiasm in all regards helped propel me forward; and John, for adding a new level of energy to my life when I needed it most, and for helping me to keep it all in perspective.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................9LIST OF FIGURES .......................................................................................................................11ABSTRACT ...................................................................................................................... .............15 CHAPTER 1 INTRODUCTION .................................................................................................................. 17Research Problem ...................................................................................................................17Review of Literature ...............................................................................................................18Environmental Basis of Economic Variability ................................................................ 18Geographic location ................................................................................................. 18Climate and hydrologic factors ................................................................................ 20System Evaluation and Emergy Analysis ........................................................................21Water Accounting ...................................................................................................................25Stream Networks ............................................................................................................... .....25Stream Order Concepts .................................................................................................... 25Streams from an Emergy Perspective .............................................................................. 26Description of Study Site ..................................................................................................... ...27Biophysical Characteristics ............................................................................................. 27Regions ....................................................................................................................... .....28Socioeconomic Characteristics ........................................................................................30Plan of Study ...........................................................................................................................322 METHODS ....................................................................................................................... ......36Emergy Analysis at National Scale ........................................................................................36Emergy Evaluation Protocols .......................................................................................... 36Emergy Evaluation Summary Indices ............................................................................. 37Total Renewable Flow Determination ............................................................................. 38Data Sources ....................................................................................................................39Unit Emergy Values ........................................................................................................ 40Provincial Scale Emergy Analysis .......................................................................................... 40Data Sources ....................................................................................................................40Irrigation Model ...............................................................................................................41Irrigation volume ......................................................................................................41Stream order transf ormity sub-model ...................................................................... 42Assigning stream order to irrigation water ............................................................... 45Geopotential Energy of Rain Runoff ...............................................................................45

PAGE 6

6 Inter-provincial River Flows: Geopoten tial and Chemical Potential Energy .................. 46Tidal Energy Allocation to Provinces .............................................................................47Tidal energy over the continental shelf .................................................................... 48Tidal energy within river channels ...........................................................................48Total Renewable Flow Determination ............................................................................. 50Statistical Analysis of Environment and Socio-Economic Condition .................................... 50Environmental Variables: Se lection and Processing .......................................................50Socio-Economic Variables: Sources and Processing ...................................................... 51Composite Variables via Prin cipal Components Analysis .............................................. 53Distance to Bangkok ........................................................................................................ 54Correlation Analysis ........................................................................................................ 54Regression Analysis ........................................................................................................ 543 RESULTS ....................................................................................................................... ........65National Emergy Analysis ......................................................................................................65Systems Diagram .............................................................................................................65National Emergy Table ....................................................................................................66Flow aggregation and summary diagram .................................................................67Summary emergy indices ......................................................................................... 68Provincial Emergy Analysis ................................................................................................... 69Systems Diagram .............................................................................................................69Renewable Emergy Flows ............................................................................................... 70Spatial surfaces of renewable emergy flows ............................................................ 71Stream order tran sformity model ............................................................................. 71Irrigation model ........................................................................................................73Provincial renewable emergy flows ......................................................................... 74Regional variation ....................................................................................................75Correlation Among Environmental Variables ........................................................................ 76Principal Components Analysis (PCA) of Environmental Variables .............................. 79PCA of standard unit water variables .......................................................................79PCA of renewable emergy measures ....................................................................... 80Socioeconomic Wealth Variables ........................................................................................... 82Variation Preand Post-Financial Crisis .........................................................................83Regional Variation in Socioeconomic Measures ............................................................ 83Correlation among Socioeconomic Measures .................................................................86Principal Components Analysis of Socioeconomic Variables ........................................ 87Time period B (1981-1993) ...................................................................................... 88Time period C (1996-2003) ...................................................................................... 89Bivariate Analysis of Environmen tal and Socioeconomic Variables .....................................90Standard Unit Water Flows and Wealth .......................................................................... 90Rainfall and wealth ...................................................................................................91Evapotranspired rainfall and wealth ......................................................................... 91Irrigation and wealth ................................................................................................92Soil moisture and wealth .......................................................................................... 92Emergy Flows and Wealth .............................................................................................. 93Terrestrial water emer gy flows and wealth ..............................................................93

PAGE 7

7 Coastal emergy flows and wealth ............................................................................. 94Total renewable emergy flow and wealth ................................................................ 94Emergy principal components and wealth ............................................................... 95Regression Analysis of Environment and Wealth .................................................................. 96Gross Provincial Product Per Capita ............................................................................... 96Water flow models for Gross Provincial Product per capita .................................... 97Emergy flow models for Gross Pr ovincial Product per capita ................................. 98Gross Provincial Product Per Area .................................................................................. 99Water flow models for Gross Provincial Product per area ....................................... 99Emergy flow models for Gross Pr ovincial Product per area .................................. 100Agricultural Gross Provincial Product Per Area ........................................................... 101Water flow models for agricultural Gross Provincial Product per area ................. 102Emergy flow models for agricultural Gross Provincial Product per area .............. 102Other Wealth Indices ..................................................................................................... 103Population Density ........................................................................................................ 105Wealth Model Comparisons .......................................................................................... 1054 DISCUSSION .................................................................................................................... ...146Summary ....................................................................................................................... ........146Principal Conclusions ...........................................................................................................147Discussion of Principal Conclusions ....................................................................................148Water Flow and Location .............................................................................................. 148Distinguishing the Unique Contribution of Water Flow to Wealth Variance ............... 149Use of Emergy Analysis to Characterize Environmental Flows ...................................153Total Renewable Emergy Flow ..................................................................................... 154Use of Principal Components to Repres ent Environmental and Economic Flows ....... 156Preand Post-crisis Time Periods ..................................................................................158Environmental Water Subsidy and Population Density ................................................ 159Suggestions for Further Research ......................................................................................... 159Temporal Window of Investigation ............................................................................... 159Other Ecosystem Subsidies ...........................................................................................160Water Balance Issues .....................................................................................................161Conclusion .................................................................................................................... ........162 APPENDIX A WATER BALANCE MODEL ............................................................................................. 165B ENERGY CIRCUIT DIAGRAMMING SYMBOLS USED IN SYSTEMS DIAGRAMS ...................................................................................................................... ...195C FOOTNOTES TO EMERGY EVALUATION TABLES .................................................... 196D PROVINCIAL DATA TABLES ..........................................................................................205E CORRELATION MATRICES .............................................................................................211

PAGE 8

8 LIST OF REFERENCES .............................................................................................................216BIOGRAPHICAL SKETCH .......................................................................................................225

PAGE 9

9 LIST OF TABLES Table page 2-1 Environmental variables for statistical analysis. ................................................................582-2 Socio-economic wealth variab les for statistical analysis. .................................................. 593-1 Emergy account for Thailand (c.2000). All flows are on an annual basis. ...................... 1073-2 Summary flows for Thailand (c. 2000). ...........................................................................1083-3 Summary emergy indices for Thailand (c.2000). ............................................................ 1093-4 Data for sample basins in the stream order transformity model. ..................................... 1103-5 Summary statistics for provincial emergy flows, mean in units of sej/m2/yr. ................1113-6 Eigenvalues and loadings for princi pal components for water flows (PCw). .................. 1113-7 Eigenvalues and loadings for principal components for emergy flows (PCem). ............ 1113-8 Summary statistics for socioeconomic wealth variables. ...............................................1123-9 Eigenvalues and loadings for principal co mponents of wealth measures in period B. .... 1133-10 Eigenvalues and loadings for principal co mponents of wealth measures in period C. .... 1133-11 Best fit models for Gross Provincial Product (GPP) per capita for five independent variable categories and 2 time periods, using standardized values for all variables. ....... 1143-12 Best fit models for GPP per area for five independent variable categories and 2 time periods. ...................................................................................................................... .......1153-13 Best fit models for agricultural GPP per area for five independent variable categories and 2 time periods. ........................................................................................................... 1163-14 Distance and water flow models fo r wealth PC1 for time period C, omitting provinces in the NE region. ............................................................................................. 1173-15 Best fit models for population density for five independent variable categories and 1 time period (period C). ..................................................................................................... 1173-16 Comparison of regression models using r oot mean square error (RMSE) differences and part-R2sums. .............................................................................................................. 118A-1 Equations used in the monthly Priestly -Taylor potential evapotranspiration (PET) model................................................................................................................................184

PAGE 10

10 A-2 Parameters used in the PET m odel equations, with data sources. ...................................184A-3 Thailand water balance calculations, ba sed on Thornthwaite and Mather (1955) as presented in Dunne & Leopold (1978). ........................................................................... 185A-4 Summary and difference stat istics for modeled provincial AET values versus Ahn and Tateishi AET estimation, and model wate rshed values versus TEI rainfall minus GRDC runoff. ..................................................................................................................189A-5 Water balance model results as aver age annual values, summarized by region. ............. 189A-6 Summary statistics for provincial water ba lance averages, perenni al stream density, and AET as a fraction of rainfall. .....................................................................................192C-1 Data sources and websites for na tional scale primary data inputs. ..................................197C-2 National emergy analysis footnotes containing energy and emergy conversion calculations. ................................................................................................................. ....198D-1 Provincial water flow valu es in units of mm/yr. .............................................................. 206D-2 Renewable emergy flows in units of sej/m2/yr. ............................................................... 207D-3 Selected provincial so cioeconomic variables. .................................................................209E-1 Pearson correlation matix for environmental variables, with P-values in italics. ............ 212E-2 Pearson correlation matix for socio-economic variables, with P-values in italics. ......... 213E-3 Pearson correlation matrix for water fl ows and socioeconomic variables, with Pvalues in italics. ............................................................................................................ ....214E-4 Correlation matrix of emergy flows and so cioeconomic variables, with P-values in italics. ...................................................................................................................... .........215

PAGE 11

11 LIST OF FIGURES Figure page 1-1 Map of Thailand showing regional location and major ro ads network (Intute, 2009). ..... 341-2 Satellite image of Thailand, taken Nove mber 2001, MODIS satellite (NASA, 2001). ..... 341-3 Map of Thailand with regional and provi ncial boundaries for the four-region study area. ....................................................................................................................................351-4 Average Gross Provincial Product (GPP) pe r capita for each region for the years 1981-2003, highlighting the regiona l disparity in wealth. ................................................. 352-1 Logic flowchart for determination of the renewable line item Total Water. ..................... 602-2 Global Map of Irrigated Areas (GMIA) raster coverage, providing percent area equipped for irrigation, and the Basi ns polygon coverage containing basin boundaries which have assigne d irrigation volumes. ........................................................ 602-3 Flowchart of the Geographic Information Systems (GIS) irrigation model. ..................... 612-4 Stream order, stream gauges and gauge watersheds used for the stream order transformity model. ........................................................................................................... .612-5 Schematic of transformity calculation, showing stream network, watershed outline and river discharge gauge. ................................................................................................. 622-6 Flowchart depicting the GIS procedures used to estimate the source stream of irrigation water and produce a coverage of the emergy of irrigation water evapotranspired from the land surface. ....................................................................................... 622-7 Map of Gulf of Thailand with locations and tidal ranges of point data obtained from the tidal prediction program WXTide32. ........................................................................... 632-8 Flowchart depicting the GIS procedures used to estimate the tidal energy dissipation in stream channels connected to the Gulf of Thailand. ...................................................... 632-9 Venn diagram illustrating part and partial correlations between a dependent variable Y and two independent variables, X1 and X2 (Stevens, 2003). ..........................................643-1 Systems diagram of material and energy flows in Thailand, with water flows shown in blue...............................................................................................................................1193-2 Aggregated systems diagram for the Th ailand economy with the standard national three-arm diagram (circa 2000). ...................................................................................... 1203-3 Bar graphs of selected emergy indices for 142 countri es, highlighting the relative placement of Thailand. ..................................................................................................... 121

PAGE 12

12 3-4 Systems diagrams representing three general provincial types in Thailand. ................... 1223-5 Raster coverages of average annual renewable emergy flows. ........................................ 1233-6 Typical Horton analysis (Horton, 19 45) relationships for the 19 evaluated watersheds. ................................................................................................................... ....1243-7 Stream transformities versus stream order for sample basins. ......................................... 1243-8 GIS coverages of estimated irrigation volume per year. ..................................................1253-9 Intermediary coverages generated by the emergy irrigation model which assigns stream order transformity to irrigated cells, and final coverage of the estimated emergy value of irrigation water used. ............................................................................ 1263-10 Provincial maps of re newable emergy flows. .................................................................. 1273-11 Provincial maps of irrigation emergy flow and selected emergy flow aggregates. .........1283-12 Box plots of provincial renewable emergy flows before transformation. .......................1293-13 Box plot of provincial renewable emergy flows after transformations detailed in Table 2-1. Note that for display purposes only, RO.chem and tide are distributions for only those provinces with non-zero values. ...............................................................1293-14 Box plots of emergy variables. ........................................................................................ 1303-15 Bivariate plots of AET and rainfall. ................................................................................. 1313-16 Matrix bivariate plot of aet.mm, ir g.mm, aetirg.mm, irg (e mergy), and aetirg (emergy). ..................................................................................................................... .....1313-17 Box plot of standardized environmental variables (z-scores) used in principal components analysis (PCA). ............................................................................................ 1323-18 Loading and score plots for principal components PC1w and PC2w of the water variable PCA. ...................................................................................................................1323-19 Loading and score plots for principal components PC1em and PC2em of the emergy variable PCA. ...................................................................................................................1333-20 Provincial maps of socioec onomic measures for period C. ............................................. 1343-21 Provincial maps of wealth indicators. .............................................................................. 1353-22 Box plots of socio-economic variables across time periods. ........................................... 1363-23 Box plots of transformed socio-econo mic variables by region and time period. ............ 137

PAGE 13

13 3-24 Box plot diagrams of the Index of Human Deprivation and its com ponents. .................. 1383-25 Box plot diagrams of the Human Ac hievement Index and its components. .................... 1393-26 Bivariate plots for provincial values of th ree alternative wealth indices, color coded by region. .........................................................................................................................1403-27 Regional box plots for the variable s income, expenditure and poverty. .......................... 1403-28 Bivariate plots of provincial values for GPP per capita versus th e wealth indices HDI and HAI for time period C. .............................................................................................. 1413-29 Box plot of standardized socio-economi c variables (z-scores) used in principal components analysis (PCA). ............................................................................................ 1413-30 Loading and score plots for prin cipal components PC1b and PC2b of the socioeconomic variable PCA for time period B. ............................................................. 1423-31 Loading and score plots for princi pal components PC1c and PC2c of the socioeconomic variable PCA for time period C. ............................................................. 1423-32 Bivariate plots of selected water flow variables (rain.mm, aet.mm, irg.mm, sm.mo) and socio-economic variables (popd.C, Gden.C, Gcap.C, pov.incid, HAI, PC1c), coded by region. ...............................................................................................................1433-33 Bivariate plots of GPP per km2 versus the emergy flows aet, ROgeo and (aet + ROgeo). ....................................................................................................................... .....1443-34 Bivariate plots of selected emergy flow variables (renew.W, renew.WI, PC1em, PC2em, PC3em) and socio-economic vari ables (popd.C, Gden.C, Gcap.C, pov.incid, HAI, PC1c), coded by region. ..........................................................................................145A-1 Flowchart for the Thailand water balance model, implemented on a monthly basis using Geographic Information Systems (GIS). ................................................................ 186A-2 Discharge gauges for the four region st udy area shown with the stream network. ......... 186A-3 Digitized watersheds, shown over the river network and elevation coverages that were used as a guide for the digitization process. ............................................................ 187A-4 Digitized watersheds used for comparing estimated runoff to discharge. Three of the larger watersheds (hatched lines ) encompass smaller watersheds. .................................. 187A-5 Average flow surfaces generate d by the water balance model. ....................................... 188A-6 Regional water budgets graphs derived from the monthly water balance model showing historical averages for rain, potential evapotranspiration (pet), actual evapotranspiration (aet), soil moisture defi cit (def) and runoff (ro) for four regions. ..... 190

PAGE 14

14 A-7 Provincial maps of water flows........................................................................................ 191A-8 Box plots of provincial wate r flows before transformation. ............................................ 192A-9 Box plots of transformed water flow variables for use in PCA and regression modeling. Units vary depending on the type of transformation. ..................................... 193A-10 Box plots of untransformed water flow variables. ........................................................... 194B-1 Standardized symbols used in energy circuit diagramming (adapted from Odum, 1996). ........................................................................................................................ .......195

PAGE 15

15 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RENEWABLE WATER FLOWS A ND WEALTH IN THAILAND By Sharlynn Dawn Sweeney August 2009 Chair: Mark T. Brown Major: Environmental Engineering Sciences Renewable energy flows are the basis of many pr oduction processes. As they vary over the landscape, they may play a significant role in differential economic development. This dissertation explored the relati onship between renewable flows a nd measures of socioeconomic wealth in Thailand. Emergy evaluation was used to quantify renewa ble flows. A monthly water balance model was implemented using Geographic Information Systems (GIS) to quantify hydrologic flows for input data. Models of irrigation, river geopotential and tide provided additional input. Values were extracted at the provincial scale from c overages of water flows (mm/yr) and emergy flows (sej/yr), and analyzed as independent variable s in regression models of wealth. Composite variables were created using principal components analysis (PCA) for additional independent variables. Exploratory data analysis for be st fit regression mode ls resulted in robust models for Gross Provincial Product (GPP) per capita, agricultural fraction of GPP per area, and total GPP per area, with adjusted-R2 values ranging from 0.81 to 0.91. Acco rding to squared part correlation coefficients, renewable water flows had significant, uni que influence on these income measures when accounting for provincial membership in th e Northeast region via an indicator variable

PAGE 16

16 (NE). The NE variable itself uniquely accounted for 20% to 76% of total variance in income. Of the renewable flows, actual evapotranspiration (A ET) had the largest unique influence on income total variance (part-R2 = 0.29, GPP per area model). Square d partial correlations (partial-R2) indicated that after accounting for variance e xplained by NE, a 50% to 72% of remaining variance in income measures was attributed to AET. Rainfall and AET were the flows included most often as significant independent variables. Because emergy measures for AET and rainfall were directly proportional to th ese flows in mm, models using emergy flows were identical to models in mm. Using principal components (PCs ) of water flows as independent variables increased model strength slightly, while PCs of emergy flows did not improve model strength. These research results, explicitly addressing the relationship between renewable resource flows and income, are an initial step toward the crucial goal of unde rstanding environmenteconomic linkages.

PAGE 17

17 CHAPTER 1 INTRODUCTION Research Problem A theoretical property of ecological systems is that they self-organize to m aximize power, or flow of useful energy (Odum, 1996), given spatially varied envi ronmental constraints, such as availability of nutrients, la nd, water, and energy. Similarly, human economies may operate as systems self-organizing toward maximum power given the possibilities an d limits established by available natural resource flows. Renewable ener gy flows in particular form the basis of all production processes on the planet including those of human soci eties. Production in excess of consumption by human society at a given scale can be exported to the larger system (e.g., region to nation) in exchange for energy in the form of goods and services that serve as feedbacks capable of catalyzing increased pr oduction from the region. This exch ange of energy is mediated by monetary currency, which flows counter-curre nt to the energy flows. If production is ultimately based on available renewable energy flow, and amplifying feedbacks are based on exported production, one might expect to see differences in production ability, and thus economic wealth, among regions of varying levels of free renewable resource flows. This research effort asked: Is wealth ba sed on renewable, indigenous energy flows? The main goal of this dissertation was to identif y relationships between flows of indigenous, renewable energy from the environment, and meas ures of economic activity and wealth indices in human-dominated systems. A dditionally, if a relationship is shown, a secondary goal was to determine if the strength of that relationship varies with other aspects of the system, such as geographic location or economic growth pha se (rapid growth versus recession). To address the relationship between econom y and environment, spatially defined ecological and economic conditions were investigated in Thailand as a case study. Thailand has

PAGE 18

18 a prevalent economic sector closely tied with e nvironmental resources, namely agriculture, and uneven economic development within the country. In addition, the financia l crisis in 1997, which involved the devaluation of the Th ai currency (baht) and subseque nt negative growth rates for several years, presents the opportunity to analyz e data from both a rapid growth phase and a recession period. Based on these aspects, Tha iland presents a good case study for exploring environment-economy relationships. Review of Literature Environmental Basis of Economic Variability The human economy de pends on ecosystem services as the basis for wealth, both directly and indirectly. Though many papers have been wr itten on the affect of humans and economic development on the environment (Broad, 1994; Glassman and Sneddon, 2003), it is rare to see detailed studies on the contribu tion of the environment (particularly free renewable resource flows) to economic wealth. Most economic theory implicitly assumes that all parts of the world have the same prospects for economic growth a nd long-term development, and that differences in performance are the result of differences in institutional factors (Sachs et al. 2001). Some researchers do attribute compar ative advantage to natural re source endowments, though datadriven studies usually consider only non-renewable, raw material subsidies with high market value such as fuels, minerals and metals (Lederman and Xu, 2006). Geographic location Though environmental flows of ma terials a nd energy are not usually accounted for explicitly, geographical features of the environment have been shown to affect economic development and growth (Olsson and Hibbs, 2005) The geographic variables investigated by Olsson and Hibbs (2005) were climate (as measur ed by the Kppen classification), latitude and the East-West orientation of c ontinental axis. They also look at biogeography, defined as the

PAGE 19

19 average numbers of locally av ailable wild plants and anim als suited to domestication 12,000 years ago. In an empirical study of the economic prosperity of nations, Hibbs and Olsson (2004), citing inspiration from Jared Diamond (1997), c oncluded that broad regions with richer biogeographic endowment transitione d earlier to settled agriculture and, thus, experienced earlier onset of accelerated technologica l development and economic growth. They also concluded that although politicalinstituti onal conditions are a powerful proxi mate source of the wealth of nations, geography and initial biogeography remain significant explanatory variables even after institutions are taken into account. Using regression models predic ting 1997 national Gross Domestic Product (GDP) per capita, they f ound that geography and initial biogeographic endowments alone explained between 50% and 60% of present-day international variation in log income per capita. When adding a measure of institutional quality, 80% of the variation in income was explained. An interesting addi tional regression showed that geography and biogeography accounted for 40% of the explained variance in instit utional quality, demonstrating the relevance of historical geogr aphy to institutional origins. Other than contributions to wealth from historical lega cies, geography may also drive wealth through more immediate effects on agricu ltural productivity, public health and transport costs (Gallup et al., 1999; Sachs et al, 2001; Sachs and Malaney, 2002). Sachs et al. (2001) emphasized temperate climate zones (with less in fectious disease and higher agricultural productivity) and the distance from sea trade as important geographic va riables that affect economic growth of nations. Araujo et al. ( 2002) found a positive relatio nship between economic growth in rural Mexico to road networks a nd to urban centers. Sica (2005) analysed the relationship between climate and growth in It alian provinces and concluded that maximum temperature and number of frost days affected the level of Italian provincial income. Minot et al.

PAGE 20

20 (2003) found that the agro-climatic variables sl ope and soil type, along with market access, explained 74% of variation in ru ral poverty rates in Vietnam, i ndicating that poverty in remote areas is linked to low agricultural po tential and lack of market access. Some research efforts have provided evid ence for location-related environment-economy links in Thailand, as well. Felkner and Townse nd (2007) found that at both the national and village level, wealth and entrepreneurship frequency varied positively with proximity to major rivers, with higher soil fertility, with lower el evation and with lower variation in rainfall. Furthermore, their Geographic Information System s (GIS) suitability model identifying locations in four provinces that had "favorable" envi ronmental and geographic conditions for economic activity revealed that many of these locations had actually developed significant economic clusters. Since presumably geography does not eff ectively change, they suggest that these results hint at a causality from geography to the clustering of economic activity and wealth. Climate and hydrologic factors There are many non-quantitative references to the role of renewable water flows in economic wealth generation. Falkenma rk (1989) re fers to a poor understanding on the part of society of the fact that the amount of water made available through the hydrologic cycle may indeed involve a distinct constrai nt to the development potential of a semi-arid or arid country. Kovacs (1989) also emphasizes that hydr ologic conditions influence socio-economic development. A suite of studies by Mendelsohn (1994, 2004, 2007) does quantitatively address links between environment and income, including hydrologic factors, using a Ricardian model. The method examines how land values shift with c limate and other control variables, with land values determined by a net revenue per hectare e quation that includes market prices, crop output, crop inputs and input prices. Th e crop output term is a functi on of purchased inputs, labor

PAGE 21

21 characteristics, farm characteristics and climate variables (temperature a nd precipitation or soil moisture). Ricardian models completed in Brazil, India and the United States all indicate that land value, or agricultural net revenue, depends upon soil moisture, temperature, soils and economic conditions (Mendelsohn et al., 2004). Clim ate variables alone explained 67% of the variance in land value in the United Sates in 1982, and in combination with soil and socioeconomic variables, 78% of the variance wa s explained (Mendelsohn et al., 1994). Expanding the research focus from net la nd revenues to total income leve ls, Mendelsohn et al. (2007) present results that in the United States and Brazil, the same variables that explain farm performance also explain why some rural di stricts and counties have higher total (both agricultural and non-agricu ltural) income per capita than othe rs. Climate variables, along with flooding, erosion and slope variable s explained 23% of the variati on in rural income in the US. Climate variables and soil type explai ned 46% of the variation in Brazil. System Evaluation and Emergy Analysis Because of myriad cross-disciplinary factors affecting incom e levels and income disparity, a systems perspective is useful for both organi zing a research approach, and attempting to address the multitude of energy flows that may be affecting wealth. Taking a whole-systems view, both ecological and human-dominated system s self-organize to tran sform available energy into new energy and material flows (Odum, 1994). As energy transformations occur, some of the available energy is lost, and the remaining en ergy after transformation is in a new form, generally more concentrated with the ability to do types of work the previous form could not. A sequence of these energy transf ormations results in material s and flows with decreasing quantities of energy, but increasing quality (O dum, 1996). These transformation chains form networks of transformations comprising energy hierarchies where the higher levels feed back to interact with the contributing en ergy flows. Energy hierarchy is an organizational principal that

PAGE 22

22 is readily observed in many systems (e.g. food we bs, stream network convergence in watersheds, government organization). According to Odum (1996), based on principl es first put forth by Lotka (1922), during self-organization, system de signs develop and prevail that maximize power inflow, energy transformation and those transformations that feed back to reinforce production and efficiency. Based on this maximum power principle, a system that can draw more resources and use these flows to maintain and enhance syst em structure and assets will outcompete systems that have fewer resources to drive their activities. Systems analysis methods that ignore energy quality and only account for energy content undervalue the contributions of concentrated, lo wer energy flows such as human work and fossil fuels, relative to diffuse, plentiful energy fl ows like sunlight and wind. The simple energy content of a material or flow does not necessarily reflect the type of work it is able to do, or its concentration, transportability, and amplifying feedback ability, all qualities that shape the patterns and processes in systems. Emergy analysis is a technique to account for the value of a flow beyond the value of its remaining energy cont ent. Its fundamental assumption is that the value of a resource is proportiona l to the total energy required in all transformations that produce the resource. Emergy is defined as the fundame ntal source energy requi red, both directly and indirectly, to create a product or service (Odum, 1996). Th is puts all products of nature, technology, and the economy on a common basis of the prior work required to produce them. Emergy is measured in units of so lar emjoules (sej). Emjoules are defined as the previous energy required to make something (sometimes referred to as embodied energy), and the word solar indicates that all forms of cont ributing energy are expressed in un its of solar en ergy that would be required to generate all inputs.

PAGE 23

23 Emergy analysis uses sunlight, the basic en ergy source for all global processes, as the common currency for all global processes. Several publications present deta ils of the standard emergy evaluation methodology (Odum, 1996; Odum et al., 2000; Brown and Ulgiati, 2004). The quality of energy in any give n product is expressed as the transformity, and measured as the ratio of the emergy input in solar emjoules (s ej) to energy output, or available energy (J). Transformities can be used to convert flows from energy units to emergy units. If flows are in units of mass or dollars, specific emergy (sej/g ) or emergy per dollar (sej/$) may be used for conversions. An umbrella term to encompass all types of emergy conversion factors, unit emergy value (UEV), is seen in more recent emergy literature, and can be used in place of the terms transformity, specific emergy and emergy per doll ar, effectively accounting for any units in the denominator of the emergy conversion factor. Once all flows are on the same unit basis, summary indices can be calculate d, enabling comparison of differe nt systems and assessment of various aspects of system condition (Brown and Ulgiati, 1997). In this study, emergy contributions of renewable resource flows in Th ailand, with a focus on water, are evaluated and compared to wealth measures. It is assumed that water flows will be prom inent when looking for relationships between environment and economy in Thailand, particularly in rural areas which are more agricultural. Different aspects of water have been evaluate d in previous emergy studies (Odum et al., 1987; Odum and Arding, 1991; Odum, 1996; Brown and McClanahan, 1996; Romitelli, 1997; Buenfil, 2001; Huang et al., 2007), including chemical poten tial energy of water, geopotential energy of water, the capacity of water to assimilate wast es, solid and solute transport by water, and the relative values of differe nt sources of water.

PAGE 24

24 The chemical potential energy of water was us ed to measure the emergy input of fresh water to fisheries and/or aqu aculture in Odum et al. (1987), Odum and Arding (1991) and Brown et al. (1991). Brown and McClanah an (1996) evaluated the water us ed in the production of lowenergy rice in Thailand, finding a va lue of 760 E12 solar emjoules pe r year (sej/yr) per hectare of cultivated land, which was nearly 37% of the total emergy yield of the rice produced. Odum et al. (1987) also evaluated ecosystem services pr ovided by the Mississippi Basin, finding that large values in wetland service and sediment deposition were diverted into the sea by diking and channelizing. Romitelli (1997) evaluated work d one by water energy in six Brazilian watersheds. That study indicated that river water accumulate d energy through the river network, and in the lower reaches of the watershed, the physical energy of the stream water spreads sediments and nutrients onto the floodplains, affecting lowl and productivity. Huang et al. (2007) also investigated the converg ence of river energy, focusing on its relation to the energetic hierarchy of land use in a watershed in Taiwan. That study calculated stream transformities based on geopotential energy, finding that stream order corre sponds to transformity, and that the hierarchy of land use in the basin is correlated to the hierarchy of streams and runoff. Buenfil (2001) evaluated the emergy value and transformity of many sources of water from the global (e.g., glaciers and global rainfall) to local scale (e.g., lakes and potable water utilities), finding that transformities for potable water are equivale nt in magnitude to gasoline and electricity. Past emergy research on renewa ble emergy flow values and th eir contribution to systems processes serves as an info rmation and methodological basis for the emergy accounting portion of this study. The lack of prev ious explicit investigations of the role renewable emergy flows play in wealth disparity in one of the aspects prompting this research.

PAGE 25

25 Water Accounting Water availability and transfer are criti cal to both the ecology and economy of huma n systems. In addition, from an emergy perspectiv e, water flow is often the dominant renewable flow in many countries (Sweeney et al., 2006). Maps of components of the hydrologic cycle, such as precipitation isopleths a nd stream networks, reveal the sp atial heterogeneity inherent in this phenomenon. In order to investigate spa tially variable environmental energy flow contributions to Thailands human systems, it was critical to formulate a water balance model to capture the major flows of wate r that enter and exit the regions of interest, particularly evapotranspiration. As a major outflow of wa ter from drainage basins, evapotranspiration dominates the water balance and controls such hydrologic phenomena as soil moisture content, groundwater recharge, and stream flow. It is al so one of the more difficult parameters to measure, as opposed to rainfall and stream flow, for which direct measures can be made fairly easily. Appendix A contains the lite rature review performed to he lp formulate the water balance implemented in this study, as well as the methodological details and the primary output of water flow values. Stream Networks In addition to the water flowing through vegeta tion from direct rainfall, transport and use of stream water is a m ajor component of overall water use in Thailand. A brief review of stream ordering concepts is given, followed by previous research on the value of stream water as determined by emergy analysis. Stream Order Concepts Stream order is a measure of the position of a stream in the hierarchy of tributaries (Leopold et al., 1964). Horton (1945) was the first to classify the hierarchical organization of streams, proposing an ordering scheme in which the chosen order of a stream extends upstream

PAGE 26

26 to the headward tip of the longest tributary it dr ains. Horton showed that stream order is related to number of streams, channel length and drai nage area by simple geometric relationships. Whenever streams of the same order meet, they cr eate a stream of the next order. Strahler (1957) suggested a different way of designating stream or der to overcome the difficulty of renumbering one headwater tributary. His method is to restrict the designation of order to stream segments, and thus not re-number lower orders by tracing b ack a higher order to its headwaters. When two first-order streams come together, they form a second-order stream. When two second-order streams come together, they form a third-order stream. Streams of lower order joining a higher order stream do not change the or der of the higher stream. The hi erarchical numbering method of Strahler was the ordering method used in this study. Usefulness of the stream-order system depe nds on the premise that, on average, if a sufficiently large sample is treated, order num ber is directly proportional to size of the contributing watershed, to channel dimensions a nd to stream discharge at that place in the system. Because order number is dimensionless, two drainage networks differing greatly in linear scale can be compared w ith respect to corresponding points in their geometry through the use of order number (Strahler, 1964). Streams from an Emergy Perspective Odum (1996) identified a stream network as a typical exam ple of an energy hierarchy, where higher numbers of smaller streams converge to form fewer numbers of larger rivers. The potential energy of water elevated by rain falling at higher elevations is used up in transformation steps, as small streams generate larger ones. The flow of geopotential energy of elevated water is converted into kinetic ener gy, friction and work on ge ological and biological systems. The chemical potential of fresh water, formerly rain, is concentrated in volume and delivered to areas further downstream. Due to the work done on the water as it moves along the

PAGE 27

27 network, an increase in ener gy quality follows the river system organization downstream. Diamond (1984) investigated this stream hierarchical pattern by calculating the geopotential transformities for stream orders in the Mississippi River basin. He calculated the transformity for each order by dividing the emergy of rainfall over a watershed by the geopotential energy of each order. Romitelli (1997) applied a similar me thod to watersheds in both Brazil and North Carolina, with the addition of calculating a stre am chemical potential transformity for each stream order by dividing stream outflow chemical energy into rainfall emergy inputs to the watershed. Description of Study Site Biophysical Characteristics Thailand is centrally located in m ainland Sout h East Asia between 5 and 21 latitude north and 97 and 106 longit ude east. Total area is 513,115 km2 and elevation ranges from sea level to 2,590 m, with 80% of the country below 500 m altitude, and only 5% above 1000 m (Donner, 1978). Figure 1-1 displays a map of Th ailand showing the general regional location and topography and Figure 1-2 contains a satellite image of Tha iland, excluding the Southern peninsular region. Thailand lies within the area be tween two mountain systems, w ith one system in the North and far West, and the other to the east of Th ailand, on the boundary with Laos. The wide depression between these two mountain systems contai ns the alluvial plains of the Chao Phraya, also referred to as the Central region, and the Northeast Region, also known as the Khorat Plateau (Figure 1-2). The Chao Phra ya river system feeds into the de lta at the head of the Gulf of Thailand, and drains about one-third of the nations territory. The Northeast region drains into the Mekong River through a separate river network.

PAGE 28

28 Thailand has a tropical monsoon climate with a distinct wet season and a hot dry season, for which the length varies among regions. The monsoons alternately blow southwesterly and northeasterly over Thailand. The surrounding seas as well as the physiographic terrain, contribute much to modifying the monsoon eff ects on various localitie s of the country. The majority of the country receives over 80% of annual rainfall from May to October during the southwestern monsoon, which carries moisture from the Indian Ocean. Most areas of the country receive adequate amounts of annual rainfall, but the duration of the rainy season and the amount of rain vary substantia lly from region to region. The drier, more seasonal areas are in the center of the Khorat plateau, in the countrys northeast ern region, which is exposed to a shorter season of the southwest monsoon, and in some parts of the west, which lie in the rain shadow of the mountains along the Burmese border (Land Devel opment Department, 2001).Two distinct types of climate are recognized: tropical rain forest climate and tropical savannah climate (UNEP/EAPAP, 1997). The tropical rainforest climate is characterized by uniformly high temperature and heavy rainfall without possessing a ny distinct dry season, mostly confined to the southern peninsular region, and areas of the sout heast with heavy rainfa ll. The tropical savannah climate is characterized by less pr ecipitation with three distinct seasons. In general, a rainy season extends from May to October, a hot dry season from March to April, and a cold dry season from November to February. Regions Landforms and drainage divi de the country into four main natural regions: North, Northeast, Central and South. For reporting of so cio-economic data, the government considers a separate East region, the Bangkok Metropolitan Area (BKK) and oc casionally a West region, all of which are part of the Central region when th ey are not considered separately (Figure 1-3).

PAGE 29

29 Although Bangkok geographically is part of the central plain, as the capital and largest city, this area may be considered in other respects a separate re gion. Additionally, the Bangkok province and the heavily urbani zed surrounding provinces are some times considered a region for statistical purposes the Ba ngkok Metropolitan Area. Each of the four major geographical regions differs from the others in population, basic resources, natura l features, and level of social and economic development. The North has mountains incised by steep rive r valleys and upland areas that border the central plain. A series of rivers unite in the lo wlands to form the Chao Phraya watershed. These natural features make possible several differe nt types of agriculture, including wet-rice cultivation in the valleys and shifting cultivation in the uplands. The Northeast with its poor soils is not favored agriculturally and many areas contain laterite soils, with poor water retention (Parnwell, 1988). The regi on consists mainly of the dry Khorat Plateau and a few low hills. The mons oon season is short and intense here, bringing heavy flooding in the river valleys in the wet season, and a prolonged dry season. Mountains ring the plateau on the west and the south, and th e Mekong delineates much of the eastern rim. As measured by income levels, poverty incidence an d other measures of social welfare, the Northeast is the poores t region in Thailand. The Central region is a natural self-contained basin often termed the rice bowl of Asia. The complex irrigation system developed for wet -rice agriculture in this region provided the necessary economic support to sustain the developm ent of the Thai state from the thirteenthcentury kingdom of Sukothai to contempor ary Bangkok (Library of Congress, 2003). The relatively flat landscape facilitated inland water and road transport. This fertile area was able to sustain a dense population, 422 persons per square kilometer in 1987, compar ed with an average

PAGE 30

30 of 98 for the country as a whole. The terrain of the region is dominated by the Chao Phraya and its tributaries and by the cultivated paddy fields Metropolitan Bangkok, the fo cal point of trade, transport, and industrial ac tivity, is situated on the southern edge of the re gion at the head of the Gulf of Thailand and includes part of the delta of the Chao Phraya system. The South, a narrow peninsula, is distinctive in climate, terrain, and resources. Its economy is based on rice cultivation for subsistence and rubber production for industry. Other sources of income include coconut plantations, tin mining, a nd tourism, which is particularly lucrative on Phuket Island. Rolling and mountainous terrain and the absence of large rivers are conspicuous features of the South. North-sout h mountain barriers and impenetrab le tropical forest caused the early isolation and separate political development of this regi on. International access through the Andaman Sea and the Gulf of Thailand made the South a crossroads for both Theravada Buddhism, centered at Nakhon Si Thammarat, and Is lam, especially in th e former sultanate of Pattani on the border with Malaysia (Library of Congress, 2003). Socioeconomic Characteristics Thailands population in creased from 23 million people in 1961 to 65 million by 2006 (Population Reference Bureau, 2006). During this period, a successful National Family Planning Program reduced the countrys annual population gr owth rate from 3.1 percent in 1960 to 0.7 per cent in 2005, the lowest rate of population gr owth among countries in the Mekong sub region (Population Reference Bureau, 2006). The country s current population density is around 120 people per square km, with the highest concentration is in ur ban areas. Roughly 50% of the population was engaged in agriculture in the year 2000 (Krongkaew and Kakwani, 2003). When the First National Development Plan wa s launched in 1961, Thailand was a typical agricultural economy with over 80% of the population engaged in agricultural activities and with rice as a major crop for both export a nd domestic consumption (Falvey, 2000). From the first

PAGE 31

31 plan onward, the relative importance of agriculture declined, and the industrial and manufacturing sectors became much more establ ished in the 1980s and 1990s. Agriculture as a proportion of gross domestic product (GDP) fell fr om over 30% in the 1970s to around 10% in the 1990s. Despite the low proportion of total GDP, Thailand is th e worlds largest rice exporter and a high-ranking exporter of other crops and food products, feeding around 4 times its population in the year 2000 (FAO, 2000). Agriculture is the main natural resource export sector and in term of providing livelihoods, is overwhelmingly the most important sector. During the last four decades of Thailands de velopment, economic gr owth has been rapid, with double-digit growth percentage s in the late 1980s. The financia l crisis resulted in negative growth in 1997 and 1998, recovering to around 4. 5% in 2000. There are a variety of accounts and speculation on the causes of the crisis (UNDP 1999; Siamwalla, 2005). It is believed to have started with the financial collapse of the Thai baht, which followed severe financial overextension by Thai banks, partly in the real estate sector. The overall growth of the Thai econo my, averaging 6.8% from 1961 through 2000, has brought about an overall reduction in poverty incidence, 57% to 16% in the same time period, but income inequality has rise n (Krongkaew and Kakwani, 2003). Th e inequality as measured by the Gini coefficient was 0.451 in 1988, in creasing to 0.481 in 1990 and 0.490 in 2000. Regionally skewed distributions of economic activities contribute to Thailands high and rising national income disparities (Me dhi et al., 1992). Figure 1-4 cont ains a plot of average Gross Provincial Product (GPP) per capita for each region for the years 1981-2003, highlighting the regional disparity. Values are from the Nati onal Economic Social and Development Board (NESDB, 2006).

PAGE 32

32 Glassmen and Sneddon (2003) present some sali ent statistics regard ing the dominance of Bangkok, and on a slightly lesser scale, the prom inence of the Central Region as well. They report that in 1960, Bangkok contained 8.1 percent of the countrys inhab itants and 23.8 of its gross domestic product. By 2000, it had increased its share to 10.4 percen t of the population and 35.2 percent of the gross domestic product. The Ce ntral region in general also exhibits economic dominance relative to the rest of the country. Between 1981 and 2000, the Central Region and the Bangkok Metropolitan Region (Bangkok a nd its surrounding 5 provinces) combined maintained nearly 90 percent of national ma nufacturing value added and increased their combined share of the gross domestic product from 64 to 71 percent, in spite of having only onethird of the national popul ation. This wealth disparity of grea t concern to Thai policymakers, and has prompted calls to intensify efforts to unde rstand both the causes of income disparity, and possible means to achieve more equity and further raise standard s of living (UNDP, 1999). Plan of Study There are three broad goals for this study: Quantify hydrologic flows and renewable emergy flows using continuous raster coverages in a GIS env ironment. Demarcate the environmental flows that correl ate to wealth measures at the provincial scale in Thailand, including alte rnative measures of renewable flows that account for energy quality, and develop re gression models relating these measures to income and other criteria of wealth. Assess the relative benefit of alternative m easures of environment, such as emergy and principal components of water flows or emergy flows, in explaining wealth variance, and explore whether wealth models differ among re gions and between preand post-financial crisis time periods. Each goal has specific essential tasks. For the first component, a monthly water budget was performed for three regions in Thailand usi ng average monthly climate normals. This study considers long-term average climate conditions fo r estimating water flows with the intent of

PAGE 33

33 representing historical average conditions wi thin which economic development evolved. In addition, raster coverages were derived for the standard set of renewable emergy flows, with original GIS models developed specifically for irrigation flow s, runoff geopotential emergy, and stream order transformities. For the second objective, environm ental variables were extracted at the provincial level for subsequent regression analysis to assess the relationship be tween environmental flows and various measures of wealth. Distance to Ba ngkok and regional dummy variables were also considered. Wealth measures were generally annual averages for two time periods: 1981-1995 (pre-crisis) and 1998-2003 (post-crisis). After ex tensive data explora tion of correlation and scatter plot matrices, best-fit models were determined for each wealth measure and several categories of independent variables: dist ance to Bangkok only, water flows, principal components of water flows, renewable emergy fl ows and principal components of those emergy flows. Finally, part and partia l correlation were used to asse ss the unique contribution of each significant renewable flow to wealth variance in the presence of location variables. For the third objective, adjusted-R2 values and root mean square error values were utilized as comparators of model strength among independent variable categories and between preand post-economic crisis time periods. Regional affect s on wealth models were considered through the examination scatter plots coded by region and the consideration of re gional dummy variables and distance to Bangkok when exploring best-fit models.

PAGE 34

34 Figure 1-1. Map of Thailand showing regional location and major road s network (Intute, 2009). Khorat Plateau Central/Chao Phraya Valley Inner Chao Phraya Delta Figure 1-2. Satellite image of Thailand, take n November 2001, MODIS satellite (NASA, 2001), with approx imate areas demarcated for the Khorat Plateau and the Central Valley and Inner Delta of the Chao Phraya River basin.

PAGE 35

35 Figure 1-3. Map of Thailand with regional and provincial boundaries fo r the four-region study area. 0.0E+00 2.0E+04 4.0E+04 6.0E+04 8.0E+04 1.0E+05 1.2E+05 1.4E+05 1.6E+05 GPP capita by region, 1981-2003 (1988 baht) BBK E C NE Figure 1-4. Average Gross Provincial Product (GPP) per capita for each region for the years 1981-2003, highlighting the regional di sparity in wealth. (BKK = Bangkok metropolitan region, E = East, C = Central, NE = Northeast).

PAGE 36

36 CHAPTER 2 METHODS A systems-based accounting approach was used to investigate relationships between environm ental energy and material flows and economic development. A comprehensive emergy analysis was first performed for Thailand at the national scale. Next, a more detailed renewable emergy analysis was implemented for each of the provinces of the Central, East, and Northeast regions. In order to capture th e spatial variability of environmental flows across Thailands landscapes, a number of Geographic Information Systems (GIS) models were implemented to obtain the estimated values needed for input in to the emergy accounts. At the provincial level, socio-economic statistics reported by Thai statistical offices were compared to the spatially corresponding environmental flows. The emergy analysis process for the national an d provincial scale is de scribed first. Next the main models used to estimate the data require d for the provincial analysis are described (with the water balance described in Appendix A). Finall y, the statistical methods used to investigate possible relationships between environment and economics are given. Emergy Analysis at National Scale In order to permit holistic assessment, em ergy analysis requires detailed conceptualization of the system, including its boundaries, primary inflows from imports and internal stocks, transformations, and exports. Analyses were don e at both the national an d provincial scales (45 provinces), with values for the material and en ergy flows obtained from the literature, web-based data sources, and from models described within this work. Emergy Evaluation Protocols The framew ork for emergy analysis is well-defined, utilizing with systems diagrams, tables of quantified system inputs, and standardized calculations of aggregate flows and indices to

PAGE 37

37 summarize condition (Odum, 1996). Emergy evaluation at national scale started with a systems diagram for identifying the major flows of ener gy and materials across the national boundary, as well as the main internal stocks and transformati ons, at the time scale of one year. The lateral boundaries were defined as the po litical border, and included th e continental shelf along areas with a coastline. The upper boundary was defined as 1000 m above the earth and water surfaces, and the lower boundary 2 m below the earth surfa ce or floor of the lakes or seas (Odum, 1996). Definitions for symbols used in systems diagramming are given in Appendix B. A line item table was then made of a ll known flows across th e boundary, including dispersed environmental flows, concentrated raw material flows from mining, imported goods and services, exported goods and services, and money flows. Raw data on flows were compiled and converted into emergy units using unit emergy values (UEV). Unit emergy values have been computed for many hundreds of products, permitt ing relatively straightforward tabulation of emergy values once the physical flows (mass or energy) driving each process is known. Emergy is the product of the physical fl ow (units of mass or energy) a nd UEV (emergy per physical unit). Many of the methods for estimating UEVs are su mmarized in Odum (1996). A notes section was created to accompany the main emergy table deta iling the calculations wh ich convert raw data inputs into the energy, mass or dollar valu es to which the UEVs are applied. Emergy Evaluation Summary Indices After converting physical flows to emergy, the various flows comprising a national resource bas e may be added together. The synthesis of flows permitted by conversion to common emergy units not only permits their addition, but also their fractional contribution to the economy. Formal rules for aggregating flows into categories permit the computation of numerous summary indices which allow comparison between systems (Brown and Ulgiati, 2004). The emergy line item flows were summed to provide a value for total emergy use in the

PAGE 38

38 system, and additional aggregated summary flows and indices were calcul ated, integrating the major inputs from the human economy and inputs coming free from the environment. Emergy indices included emergy use per capita and pe r area, total renewable emergy use, emergy to money ratio, percent of use derived from renewa ble flows, and investment ratio, among others. Calculations for deriving summary indices are gi ven in the standardized tables that accompany an emergy evaluation. Total Renewable Flow Determination Due to its role in many of the emergy indices, and its importance to the aims of this research, one of more important areas for st andardization in the emergy accounting method is determination of the total renewable summary fl ow. The standard proced ure, though exceptions can be found in the emergy literature, is to list al l major renewable flows as line items, but to use only the largest value for Total Renewable Flow (R), thus avoi ding double-counting of the flows from the three external biospheric inputs: gravitational energy, deep heat flow energy, and solar energy (Odum, 1996). In recent practice, both the chemical potential of rain (or evapotranspiration) and the geopoten tial of runoff have been listed as separate line items, though summing them is not considered double-counting, a nd they may be used together as the largest renewable flow (Odum, 1996). To try to reduc e confusion for the audience unfamiliar with emergy, as well as codify whether rainfall or eva potranspiration is to be used, all of the water calculations are performed prior to insertion into the main line item table, and are detailed in the notes section accompanying the main table (Appendix C). The resulting line item emergy value is called Total Water and considers chemical potential of rain, chemical potential of evapotranspiration, chemical and ge opotential of rain runoff, ch emical and geopotential of river inflow, and chemical and geopotential of river ou tflows. Figure 2-1 contains a logic flowchart

PAGE 39

39 (adapted from Sweeney et al., 2006) documenting the steps used to determine the Total Water line item flow, which is often equal to the to tal renewable flow for many regions/countries. With annual renewable water flow now represen ted by one line item, the largest flow from the main table may be chosen for total renewa ble flow. Another option is to sum the largest terrestrial renewable flow and the tidal fl ow, as discussed by Campbell (2000). Campbell proposed that adding rain and tide will not doublecount any of the three independent biospheric energy inputs (solar radiation, deep heat, gravit ational attraction) becaus e, relative to the timescale of national economies, radiation is a negligible input to tide, deep heat is a long term input for tide, gravitational attraction is a negligible input for rain, and d eep heat is a negligible input for rain. Accordingly, the renewa ble emergy base of Thailand and each of its provinces might be calculated by adding the tidal flow (if present) to the largest terrestrial renewable flow, which happens to be water for every province. For the provincial-scale emergy anal ysis, 3 variations of total renewable will be considered: renewable water (renew.W), renewable water plus irrigation (renew.WI) and renewable water plus ir rigation water plus tide (renew.WTI). Data Sources Data compiled from a diverse set of publis hed international da ta sources and GIS coverages were used as input for the tabulati on of the emergy table line items for the national account. The data were compiled and transfor med within the National Emergy Accounts Database (NEAD), created by Sweeney et al. (20 06), which generates standard national emergy accounts and summary flows for 134 countries for the year 2000. Sp atial coverages of geographically referenced data were chosen for deriving renewable flows to allow for calculations within a GIS environment and subseq uent analyses at the provincial scale. The consideration of spatial varia tion avoids estimating annual flows based on one or only a few point estimates biased to certain regions of th e country. Data sources for line item flows for the

PAGE 40

40 national analyses can be found in Appendix C (Table C-1), and details on th e calculation used to convert this data to emergy un its can be found in Table C-2. Unit Emergy Values Unit emergy values (UEV ) are the crucial link between the energy, mass, or dollar value of a flow, and the prior emergy it took to produce and convey the flow to its present location. Emergy values were calculated within the na tional accounts by multiplying the mass, energy, or money content of flow (grams, joules, $) by the UEV (sej/gram, sej/J, sej/$) assigned to that flow. In the absence of a comprehensive set of location specific UEVs for every product and process in Thailand, previously computed UEVs were adopted (see Table C-2 for UEVs used in this study). Provincial Scale Emergy Analysis The next scale of evaluation considered the 45 provinces of the Central (C), East (E), Bangkok (BKK) and Northeast regions (NE). At the provincial scale, particular attention is required to estimate renewable flows with enough re solution to capture variability at this spatial scale. A m onthly water balance was performed using IDRISI GIS software, which produced spatial coverages of water flow s (see Appendix A). These coverages provided the average annual evapotranspiration and runoff variables used as two of the inputs to the provincial emergy analysis. In addition, in order to capture variation across provinces, models were implemented for irrigation, runoff geopotential cross-border river flows and tidal energy. These models are described in this section. Data Sources The renewable emergy flows of solar radiat ion, rainfall, evapotra nspiration (AET) and wind were obtained by performi ng the energy an d emergy calculations using map algebra on coverages used in the water ba lance model (Appendix A). Values were then extracted using a

PAGE 41

41 vector coverage of provinc ial boundaries obtained from Thailand on a Disc (TEI, 1996). Deep earth heat flow point data from the Internati onal Heat Flow Commissi on (Pollack et al., 1991) were inverse-distance interpol ated using Idrisi software, producing a heat flow raster from which a sum was extracted for each province. More de tailed information was processed for irrigation, runoff geopotential energy, energy in cross-boundary water flows and tidal energy as described in the following sections. Irrigation Model The vast majority of irrigation water in Th ailand com es from rivers and surface storage reservoirs. The portion of irrigation from groundwater in 1995 was only 0.2% (FAO, 1999). Using IDRISI software, a GIS model was develo ped to estimate the volume of water extracted from surface water sources that is applied to paddies and fields. Estimating irrigation per province was limited by both a lack of actual consumption values a nd a lack of measured input data for these area units. The best data available were for ir rigation demand reported for the major water basins. The provincia l irrigation volume estimation wa s based on the Global Map of Irrigated Areas, v4.0 (GMIA) (Siebert et al., 2006), and the Thai National Committee on Irrigation and Drainage (THAICID) table of irrigation volumes pe r basin and irrigated area per basin (Sethaputra et al., 2001). The basin irrigation volumes we re reported only for the year 1993. The GMIA had a resolution of 5 minutes (0 .0833 degrees), and cell va lues were irrigated area expressed as a percentage of total ar ea equipped for irrigation for the year 2000. Irrigation volume Due to lack of spatial data on irrigation volumes at a resolution finer than the bas in scale, the volume reported per basin, divided by the i rrigated area per basin, was considered the average volume per area and applie d to all irrigated areas within the respective basins in order to generate a spatial map of irrigation volumes. Figure 2-2 shows the percent area equipped for

PAGE 42

42 irrigation from GMIA, as well as the basin boundaries which defi ne the watersheds for which volume irrigation is reported. Figu re 2-3 contains a flow chart of the estimation process. The result is an irrigation map from which volume can be extracted using the provincial boundary coverage. The derived sums of irrigation volume applied per province were then multiplied by 0.5 to generate a conservative estimate of the fraction of irriga tion that is actually evapotranspired in paddies and fields. This frac tion is based on a several studies which report the following general irrigation effici encies: 40% 60% in the Chao Phraya area of Central Thailand (Vudhivanich et al., 2002), 38% 67% in Northeast Thailand (Mekong River Commission, 2007) and 44% 83% in Northeast Thailand (Kangrang and Chaleeraktrakoon, 2007). Stream order transformity sub-model In order to include irrigati on in the provincial emergy accoun ts, and to approxima te energy quality differences of irrigation water pulled from different levels of the watershed hierarchy, a model based on stream network position was developed for estimating transformities for water. The transformity for rainfall represents th e average amount of biospheric emergy inputs necessary to deliver the rainfall water to the po int at which it lands on the earths surface. Because irrigation water from streams has been s ubject to more work done by the landscape, i.e. runoff and convergence through stream networ k patterns and elevation change, a more representative transformity than that of rainfa ll is required. A stream order transformity model was developed to estimate a suite of transform ities based on location of water in the stream hierarchy. This study employed a method similar to Romi tellis (1997) stream chemical potential transformity. Transformity of water at a give n point (discharge sta tion) was calculated by dividing the emergy of the rainfall in the contributing watershed by the chemical potential energy of the water at the given point. This is based on the assumption that it takes the full amount of

PAGE 43

43 rainfall volume (and energy and emergy), to generate the discharge (and available energy and emergy) at any point in the stream segment. Data availability defined th e points used (discharge data at gauges), and points were aggregated by stream order in order to use an average transformity for stream order anywhere in Thaila nd in order to transform irrigation volumes to irrigation emergy. Building the stream water transformity mode l based on stream order required a stream network coverage coded for stream order. The stream network was obtained from Thailand on a Disc (TEI, 1996), and a thorough inspection was perf ormed of all segments. A Strahler ordering algorithm was used to generate the original datase t, which can cause errors in areas of braided channels and dense canal networks, thus corre ctions were made manually where necessary by tracking stream order from the first order streams. One area in need of repair was the lower valley of the Chao Phraya. Four major tributaries join to form the Chao Phraya at the northern edge of the Central Region, at which point the st ream order of seven was correctly assigned in the original dataset. Just downstream of this juncture, many branches split from the Chao Phraya into a dense, non-dendritic network of channels a nd canals. This area was coded as stream order zero in the original coverage. Because this study is concerned with the va lue of the water as it related to the energy required to generate it a nd place it at that location, an order of seven was assigned to the portions of the netw ork connected to the Chao Phraya to reflect the source of the water. Because discharge data is essential for a stream water transformity estimate, one constraint on sample size is the availability of river gauge data. Secondar ily, in order to consider the minimum inputs required to generate a stream order, stations needed to be in close proximity to the areas of the stream network where two segments of the same order meet to generate a higher

PAGE 44

44 order. Applying the above criteria to the gauge data from the Global River Discharge Commission (GRDC, 2004) resulte d in a set of 19 gauges and th eir associated watersheds. Watersheds were generated by onscreen digitization using the Thailand Environmental Institute (TEI) stream network coverage (TEI, 1996) and the ETOPO5 digital elevation model (NOAA, 1988) as guides. Resulting areas were crosscheck ed with reported areas for the gauges and corresponded to within 10%. Figure 2-4 displays the stream network with corrected stream order values, and the set of gauges and watersheds used to estimate stream water chemical potential emergy for each stream order. A basic schematic of a watershed and associ ated gauge, along with the stream chemical potential transformity calculation, can be seen in Figure 2-5. Average annu al rainfall, extracted from the interpolated TEI rainfall data using the watershed boundaries, was multiplied by the GRDC reported watershed area to obtain an annual estimate of rainfall volume over the watershed. This volume was converted to chemical energy (rainfall volume*water density*Gibbs free energy of rain water), and th en to emergy (rainfall chemical energy*rainfall chemical energy transformity) to obtain the em ergy required to generate the runoff chemical energy observed at the gauge (runoff volume*water density*Gibbs free energy of stream water). The watershed emergy was divided by the runo ff energy to obtain estimates for stream transformity categorized by stream order. The calculated stream tran sformities were then averaged by stream order, resulting in an es timated average stream water chemical potential transformity for each stream order. The highest stream order of the selected gauges was order 7, and the Mekong was not represented. Based on th e relationships between order and discharge and order and basin area, the poi nt at which the Mekong starts to border Thailand was assumed to be order 8, allowing for a transformity estimate for that order. The estimated transformities

PAGE 45

45 (average per order) were then applied to m odeling irrigation evapotranspiration by assigning stream orders to irrigation volume on a spatial basis. Assigning stream order to irrigation water The emergy value of irrigation water was calcula ted using both the estimates of irrigation volum e per cell, and stream order transformities described in the preceding sections. In the absence of specific information about the spatial origin of irrigation water at each cell, it is assumed that the stream water applied to a land cell originates from the nearest perennial stream. The flowchart in Figure 2-6 illustrates the GIS procedures used to deri ve the final irrigation chemical potential emergy map. A map of the distance of each land cell from the nearest perennial stream network cell was produced w ith the IDRISI Distance module. Next, the Allocate module assigned to the land cells the st ream order identifier of the nearest stream network feature from which the distance was cal culated. Using the Assign module, the stream order values were replaced with the stream order transformities derived in the stream order model. The map of transformities was then multiplied by the map of the chemical potential energy of the irrigation water evapo-tran spired (cubic meters of irrigation*0.5*1000kg/m3*4940 J/kg), resulting in a raster map of the emergy va lue of the irrigation water evapo-transpired on land. Geopotential Energy of Rain Runoff The standard runoff geopoten tial calculation was perfor med (Odum 1996), however, rather than applying the averag e elevation of the province as the hydraulic head for the total runoff volume for the province, the calculation was performed via GIS map algebra in IDRISI on a cell by cell basis. Each cell was assigned its el evation as the hydraulic head. This elevation raster, the raster for annual runoff volume from the water balance model, and the necessary equation constants were then used in the geopotential em ergy calculation: rain runoff

PAGE 46

46 geopotoential emergy = runoff volume (m3/yr)*elevation(m)*1000 kg/m3*9.8 m/sec2*4.7E+04 sej/J. Inter-provincial River Flows: Geopotential and Chemical Potential Energy Water entry via rivers al ong borders was comp iled using long-term average annual discharge data (point source ga uge stations) from the Global R unoff Data Centre (GRDC, 2004). Data were not available for ev ery cross-border river flow, but most of the major flows have gauges, and the gauge nearest the border was us ed. The calculation of th e geopotential energy of the river flow requires elevation change and discharge volume. El evation change from entry to exit was determined from the elevation GIS c overage and average discharge at entry was obtained from the GRDC gauge coverage. Because the bulk of irrigation withdrawals are from major rivers, modeled irrigation volumes per province were subtra cted from the river discharge volumes at border entry before performing the river geopotential energy calculation. The calculation was performed with th e following equation: river geoptoential emergy = river volume at entry to province (m3/yr)*elevation change (m)*1000 kg/m3*9.8 m/sec2*4.7E+04 sej/J. Other than withdrawals for irrigation, the chemical potential energy imported from upstream provinces is only relevant in the coastal provinces, where freshwater flows are deposited in the marine waters of the continental shelf, making use of the chemical potential. In the study region, there are 4 major rivers that ente r the Gulf of Thailand, the water of which was considered as an upstream input via cross-border flows. All othe r freshwater inputs to the Gulf were assumed to be generated by rainfall runoff with in the coastal provinces. In the inner Gulf of Thailand, the discharge volumes of the 4 major rivers are of great magnitude and the coastlines are relatively small, thus all provinces with coastlines along the inner Gulf share in the chemical potential energy generated by the river discharge. In order to account for the shared energy, the inner Gulf estuary was assumed to extend 20-30 km from the shore, which resulted in the

PAGE 47

47 inclusion of 7 provinces in the area of influe nce for the 4 major river outlets. The chemical potential of the 4 major rivers of the inner Gulf was allocated to these 7 provinces in proportion to their coastline lengths. Discharge volumes were obtained from estuary budgets produced by LOICZ (Dupra et al., 2000). The 4 coastal provinces contai ning the 4 rivers within thei r borders were assigned their respective shares of the combined river discharge volume, while the internally generated runoff from rain was not included due to the assumption th at this is already in the river runoff discharge value. The remaining coastal provinces had ad ditional runoff volume added according to the internally generated rainfall runo ff within their borders. To account for the rain runoff chemical potential generated within coastal provinces, rain runoff volume for coastal provinces was converted to chemical potential energy using standard calculations, and was then converted to emergy using the transformity for rain chemical potential: rain runoff chemical potential emergy = runoff volume (m3/yr)*1000 kg/m3*4940J/kg*3.1E+04 sej/J. Because the vast majority of the river volume entering the estu aries was generated farther up stream within the country, a transformity for river water was needed to conve rt the major river discharge chemical potential energy to emergy. For the major river runo ff chemical potential emergy, a volume-based weighted average of stream order transformity was used, resulting in the following equation for this component: river runoff chemical potential emergy = river runoff volume (m3/yr)*1000 kg/m3*4940J/kg*1.6E+05 sej/J. Tidal Energy Allocation to Provinces The tidal energy contribution to provin ces involves two com ponents, the energy dissipation over the continental shelf, and the energy dissipation in the river channels that experience tidal intrusion and river fl ow reversal during high tide events.

PAGE 48

48 Tidal energy over the continental shelf Based on the relative size of the Thai a nd Cambodian coastlines along the Gulf of Thailand, it is assumed that Thai land receives about 2/3 of th e tidal energy in the gulf. A predictive tide program WXTide32, provided tidal amplitude and tid al frequency predictions for 9 points along and near the coastline of the study area. Figure 2-7 shows a map of the Gulf of Thailand, with locations and tidal ranges of poi nt data obtained from the tidal prediction program. The dotted line approximates the boundary of the Gulf of Thailand and the South China Sea, while the solid line approximates the demarcation between the gulf areas allocated to Thailand and Cambodia separately. For the 11 coas tal provinces, tidal data was assigned as follows: if one or more points fell along the coastl ine, the data from the single point or average of the points was used; if no points were located al ong a provincial coastline, then the data from the nearest point was used. The continental sh elf area assigned to each province was the portion of the total shelf area that was proportional to the provi ncial coastline length. Then, the standard tidal emergy calculation was perf ormed: tidal emergy = area (m2)*# tides/yr*(0.5*range2) *1025kg/m3*9.8 m/sec2*7.4E+04 sej/J. Tidal energy within river channels Because the four major Thai rivers emptyi ng into the gulf ex perience significant tidal influx, it is possible that several of the inland provinces receive tidal energy via dissipation in their river channels. Using IDRISI, A GIS model was devised to estimate the energy received in tidally influenced stream channels. Figure 28 contains a flowchart that details the GIS operations performed. The average upper limits for tidal intrusion in the 4 rivers for both the rainy season and dry season were obtained from estuary budgets compile d by Dupra et al. (2000). A value of 2 meters was used for the tidal amplitude at the rive r mouths (Dupra et al., 2000). Using LINERAS, a

PAGE 49

49 coverage was generated that consisted of the 4 major rivers with a width of one pixel (500m). The COSTGROW algorithm was used to generate a spread of increasing values from the coastline, moving upstream through the 4 major river channels, to a limit of 1000 pixels. Various friction surfaces were trialed for the dry s eason until a costgrow surface was achieved that resulted in similar values at the 4 tidal in trusion limit points (COS TGROW 1000). The friction surface chosen was the square root of elev ation. The COSTGROW 1000 pixel values that geographically corresponded to the 4 tidal intrusion points pe r season (wet and dry) were averaged for each season. These averages were then rounded to the nearest 100 and those values (500 for the dry season and 200 for the wet season) were then used as the costgrow limits to generate the final costgrow surfaces that represen t areas of tidal intrusion. For the final costgrow surfaces, COSTGROW was run for the full stream network with an adjustment in the friction surface to account for varying stream sizes: the sq uare root of elevation was multiplied by stream class (1=major river, 2=perennial stream, 3=intermittent stream). Next, the costgrow surfaces were converted from unitless pixel values (1 through 200 and 1 through 500) to tidal amplitude. Assuming 2 mete rs at the river mouths and 0 meters at the limit of tidal intrusion, a linea r interpolation of 0 to 2 was performed over 200 and 500 unit lengths. These values then replaced the costgrow values. To derive area of the stream channel, width values were assumed (via map inspection) corresponding to stream class (stream class 1 = 300m, 2 = 25m and 3=10m). Once tidal amplitude and stream area were derived, the standard tidal energy and emergy conversion calculation was performed, as done for tidal flow on the continental shelf. With pixel va lues in emergy units, provincial bounda ries could then be used to extract total tidal emergy values per province.

PAGE 50

50 Total Renewable Flow Determination Total Renewable Flow is often considered to be the largest renewabl e flow among the suite of renewable em ergy flows. As mentioned in th e description of Total Renewable Flow for the national scale, the renewable emergy base each of the provinces might also be calculated by adding the tidal flow (if present) to the largest terrestrial renewable flow which happens to be water for every province. A further consideration was the irrigation flow, which is predominately renewable. When inputs to two irrigation dams in Thailand were analyzed in an emergy synthesis study in 1996 (Brown and McClanahan), the emergy of irrigation water was 85 to 90% renewable. In order to analyze se veral options of Total Renewable Flow at the provincial scale, 3 variations of total renewable will be consid ered: renewable water (renew.W), renewable water plus irrigation (renew.WI) and renewable water plus irrigation water plus tide (renew.WTI). Statistical Analysis of Environm ent and Socio-Economic Condition The suite of measures from the environm en tal and emergy models described previously, along with a group of socio-economic variables from statistical abstracts and the literature, were subjected to a variety of screeni ng measures to identify representative sets of variables for entry into multivariate models to assess the degree of relationship between the Thai environment and socio-economic condition at the provincial scale. Minitab software was used for variable selection, correlation analysis, principal componen ts analysis (PCA) and least-squares multiple regression analysis. Environmental Variables: Selection and Processing The environmental and em ergy models developed for this study produced a multitude of variables, some of which were highly correlate d. Correlation matrices an d scatter plots were analyzed in order to reduce the number of variab les to a representative set that limited excessive multi-collinearity. The final variables chosen co rrespond to 4 categories: location, water flows,

PAGE 51

51 emergy line item flows, and aggregated emer gy summary flows (Table 2-1). The location variables were meant to capture regional variat ion and distance to the Bangkok market and ports. The water flows were from the water balance m odel and represent long-term average conditions (circa 1960-1990) and were in units of water dept h (mm), with the exception of stream density, reported as km/km2. The emergy signature flows were the renewable flows from the environment, reported in average emergy density units for each province, sej/m2/yr. The emergy aggregates represent several options for calcula ting total renewable emergy entering the system each year. Because the final environmental variables selected varied greatly in measurement scale, skew, and outlier presence, tran sformation to approximate norma l distributions was performed before proceeding with the correlation and re gression analyses. An iterative process was employed, involving investigations of histogram s, probability plots and the Anderson-Darling (AD) goodness-of-fit statistic, with a p-value targ et of 0.05. In order to set a limit on the vast number of transformations possibl e, the choices were limited to ta king the log (base ten) or using the following as exponents for each variable: -1, -0.5, 0.5, 1 and 2. The transformation for each variable producing the distributi on with the highest p-value was c hosen. In some cases, a p-value of 0.05 was not met, but the transformation was accepted if visual inspection of the histogram and probability plot showed a distribution appro aching normal. Table 2-1 lists the final set of environmental variables for entry into exploratory data anaylses, and the transformations applied. Socio-Economic Variables: Sources and Processing Socio-economi c measures for each province were obtained from the National Economic Social and Development Board (NESDB) online statistical tables (www.nesdb.go.th) and two Thailand Human Development Report public ations produced by the United Nations Development Programme (UNDP, 1999; UNDP, 2003). The socio-economic variables include

PAGE 52

52 both Gross Provincial Product (GPP, in 1988 consta nt currency) and population for the years 1981-2003 (from NESDB), the Index of Human De privation (IHD) and its components (UNDP, 1999), the Human Achievement Index (HAI) and its components (UNDP, 2003), the Human Development Index (UNDP, 2003), poverty inci dence in 2000 and personal income and expenditure in 2000 (UNDP, 2003). The IHD is a co mposite index developed for Thailand using data from roughly 1990 through 1996, employing 48 indi cators that cover 8 components: health, education, employment, income, housing and environment, transport and communication, consumer goods, and a womens rights index. The HAI is a composite index developed for Thailand using data from 1998 through 2002, employi ng 40 indicators that c over eight aspects of human development: health, education, employme nt, income, housing and living environment, family and community life, transport and communi cation, and participation. The HDI is an index developed by the UNDP for global use, and is base d on three indicators: 1) length of life, as measured by life expectancy at birth, 2) educa tional attainment, measured by the adult literacy rate and the combined gross primary, secondary a nd tertiary enrollment ratio and 3) standard of living, as measured by GDP per capita. Poverty incidence is the per cent of the provincial population that is living below th e poverty line, which is set by NESDB and varies spatially based on varying cost-of-living and s ubsistence needs (Deolalikar, 2002). Due to the 1997 financial crisis, as well as the time period representation of the human indices IHD and HAI, which represent short-term pre-1997 and post-1997 conditions, the socioeconomic variables were processe d to represent thr ee time periods: 1981-2003 (referred to as Period A), 1981-1995 (Period B) and 1998-2003 (P eriod C). The omission of years 1996 and 1997 in Periods B and C is due to the occurrence of 3 provinces splitting into 6 new provinces, and associated data discrepancies which alloca ted newly split population and income to the new

PAGE 53

53 provinces in different years. Gross provincial product (GPP), G PP per capita and GPP per area were averaged for each time period. In addi tion, compound annual growth rate of GPP (CAGR) was calculated by taking the nth root of the total percentage gr owth rate for each period, where n is the number of years in the period being considered. Transformations were selected for the data as described for the envi ronmental variables. Table 2-2 lists the final set of socio-economic variables and th e associated transformations. Composite Variables via Prin cipal Components Analysis Due to the large amount of interrelated vari ables investigated, a multivariate technique, principal components analysis (PCA), was us ed to reduce the dimensionality within the environmental and socio-economic datasets (Minita b module used with standardized variables). Principal components analysis re duces the complexity of the dataset into re presentative components that capture the majority of the variability across samples by transforming the original variables into new composite variables called prin cipal components (PCs), which are orthogonal, uncorrelated, linear co mbinations of the original va riables (Jackson, 1991). Since the orthogonal PCs are constructed to maximize the desc ription of the data va riance, the first PC describes the largest portion of the data variability, the seco nd PC explains the maximal remaining variance uncorrelate d to the first, and so on. Principal components analysis serves two main purposes in this study. First, it facilitates a better understanding of the interrelationships am ong variables within the environmental and socio-economic categories. In addition, it condenses the environmental and socio-economic data into fewer, more representative variables for insertion into regression models. Components were chosen for continued analysis based on the amou nt of original variation captured by each component.

PAGE 54

54 Principal components analysis is vulnerable to outliers in the data due to its least squares nature. Outliers strongly affect the data varian ce and the true correlation structure of the data (Stanimirova et al., 2007). To account for outliers, variables were first transformed as described earlier, then standardized automatically by Minitab as part of the PCA process. Distance to Bangkok In order to account for the e ffect of proximity to the ma rket potential and ports of the Bangkok area, a simple Euclidian distance measure was included in the analysis. Using Idrisi GIS software, the distance of all cells to Bangkok center was calculated. Then, the average of these distances was extracted for each province and used as a variable in the statistical analysis. Correlation Analysis Correlations were performed be twee n the selected environment and wealth variables, including principal components. In addition, relationships with the distance variable were investigated. After generating a correlation matri x, scatter plots matrices were investigated in order to discover regional patter ns of correlation that would no t be apparent looking only at correlation coefficients for the full dataset. Regression Analysis Regression analysis was used to investigate th e type and strength of relationships between the environmental and socio-economi c variables. Multiple regression in particular was used because it can establish that a se t of independent variables explai ns a proportion of the variance in a dependent variable at a significant level (through a si gnificance test of R2), and can establish the relative predictive importance of the independent variables ( by comparing beta weights). The least squares method was used to derive the equa tions, with the environmen tal variables or their principal components as the independent variab les and the socio-economic variables or their principal components as dependent variables. Variab les were standardized to aid in interpretation

PAGE 55

55 of beta coefficients (the unstanda rdized coefficients are a functi on of the variable variance) and to enable direct comparison of the root m ean square error (RMSE) of various models. After reduction of variables following investig ation of correlation a nd scatter plots, the following socioeconomic variables from Table 2-2 were specified as dependent variables for model exploration: GPP per cap ita for periods B and C, GPP per area for periods B and C, agricultural GPP per area for periods B and C, HAI, income component of HAI, poverty incidence, first principal com ponents of wealth variables for periods B and C, and population density for periods B and C. For each dependent variable, four best fit models were sought for each of 4 categories of independent environmental variables: water flows in mm/yr, PCs of water flows (PCw), emergy flows in sej/m2/yr and PCs of emergy flows (PCem). Regional dummy variables for Central, East and Northeast, as well as distance to Bangkok were also included for each category during model investigations. Best subsets regression in Minitab was used as an exploratory method to begin identifying potential models for each dependent variable and the four sets of independent variables. Best subsets identifies the best-fitting regression models that can be constructed with the predictor variables specified by the user. With this met hod, Minitab examines all possible subsets of the predictors, beginning with all models containing one predictor, and then all models containing two predictors, and so on. By default, Minitab displays the two best models for each number of predictors, based on the maximum R2 criterion. Also reported for each model is Mallows Cp, a statistic used as an aid in choosing between al ternative multiple regression models. Mallows Cp compares the precision and bias of the full model to models with the best subsets of predictors. A Mallows Cp value that is close to the number of pred ictors plus the constant indicates that the model is relatively precise and unbiased in es timating the true regression coefficients and

PAGE 56

56 predicting future responses (Mallows, 1973). Only models with Cp less than 1 plus the number of predictors were considered. After the best subsets method was used to inde ntify several possible m odels for each of the dependent variables with each of the four indepe ndent variable categories, the reduced set of models were run using SPSS softwa re, and t-statistics for the coeffi cients were evaluated. If they were less than 2 in magnitude (corresponding to pvalues greater than 0.05), the model was refit with the least significant variable excluded. Variables were permanently excluded if p-values for the F-change test were greater than 0.05, and root-mean-square error (RMSE) of the model did not decrease by more than 5%. Ot her diagnostic procedures were also performed to check for violation of regression assumpti ons: variance inflation factors (V IF) were checked to identify potential problems with multi-collinearity and residual plots (residual histograms and residuals versus fits) were created to check for residua l normality, residual variance and presence of outliers. For each dependent variable and each i ndependent variable category, if several models had goodness of fit diagnostics (adjusted-R2 and RMSE) within 5% of each other, the model that better fit regression assumptions of normality of residual distribution and homoscedasticity of residuals was deemed best model. Because regional dummy variables and distance to Bangkok were included during model development, partial and part correlations of each independent variable were included in the regression output in order to assess the separa te contributions of lo cation and environment. Partial and part correlations are statistical me thods for determining whether a true correlation exists between a dependent and an independent variable while cont rolling for one or more other variables (Waliczek, 1996). The Venn diagram in Figure 2-9, illustrating relationships between a dependent variable Y and tw o independent variables, X1 and X2, is useful for understanding the

PAGE 57

57 concepts of partial and part correlations (Stevens, 2003). The pa rtial correlation is correlation between the dependent variable and an independent variable when the linear effects of the other independent variables in the model have been removed from both. More specifically, the partialR2 is the portion of dependent va riable variance estimated by a variable, which is not estimated by the other independent va riables in the equation. In Figure 2-9, the partial-R2 of X1 is represented by a/(a + e). Partial co rrelation is widely used to expl ore or to test the outcome of partialing out one or more variables from the a ssociation between an explanatory and a predicted variable in both cross-sectional and longitudinal studies (Cramer, 2003). Th e part correlation is the correlation between the dependent variable and an independent variable when the linear effects of the other independent variables in the model have been removed only from the independent variable. It is related to the change in R-squared when a variable is added to an equation, and represents the unique contribution of an independent va riable that is not shared by any other independent variable in th e equation. In Figure 2-9, the part-R2 of X1 is represented by a/(a+b+c+e).

PAGE 58

58 Table 2-1. Environmental variab les for statistical analysis. Variable Original Units Transformation Description Location: reg4 none Region code (Bangkok Area,Central,East,Northeast) dist x^(0.5) Distance to Bangkok Province Water flows: strmdens km/km2 -x^(-1) Perennial stream length per area rain.mm mm -x^(-1) Average annual rainfall aet.mm mm none Average annual evapotranspiration aet.frac mm x^(0.5) AET fraction of rainfall ro.mm mm log x Runoff from rainfall irg.mm mm log x Average annual irrigation depth def.mo mm none Average monthly soil water deficit sm.mo mm log x Average monthly soil moisture Emergy signature flows: rain sej/m2 log x Rainfall, chemical potential aet sej/m2 none AET, chemical potential ROgeo sej/m2 x^(0.5) Runoff, geopotential ROchem sej/m2 log x Runoff, chemical potential ( = 0 if not coastal) rad sej/m2 log x Solar radiation wind sej/m2 log x Wind heat sej/m2 log x Heat flow irg sej/m2 log x Irrigation evapotranspired tide sej/m2 none Tide ( = 0 if not coastal or tidally influenced) Emergy aggregations: aetirg sej/m2 -x^(-1) AET from both rain and irrigation aet+ROgeo sej/m2 -x^(-1) AET chemical potential plus RO geopotential renew.W sej/m2 -x^(-1) Total renew, non-manipulated water flows renew.WI sej/m2 -x^(-1) Total renew, all freshwater flows, including irrigation renew.WTI sej/m2 -x^(-1) Total renew, including water, irrigation & tide

PAGE 59

59 Table 2-2. Socio-economic wealth va riables for statistical analysis. Variable Original Units Tr ansformation Description Time period 1981-2003(period A): pdens.A #/km2 -x^(-0.5) Population density, avg. over period A G.cagr.A % -x^(-1) Annualized GPP fraction increase Gcap.A baht/pers. -x^(-0.5) G PP per capita, average over period A Time period 1981-1993(period B): pdens.B #/km2 -x^(-0.5) Population density, avg. over period B G.cagr.B % none Annualized GPP fraction increase Gcap.B baht/pers. -x^(-0.5) G PP per capita, average over period B Gdens.B baht/km2 -x^(-0.5) GPP per area, average over period B agGdens.B baht/km2 -x^(-0.5) Agricultural GPP pe r area, average over period B IHD.comp none Index of Human Deprivation, composite, 1990-1997 IHD.inc none Income component, 1990-1997 IHD.health none Health component, 1990-1997 IHD.edu none Education component, 1990-1997 IHD.emp none Employment component, 1990-1997 IHD.hous none Housing and living conditions comp., 1990-1997 IHD.tracom none Transportation and communication comp.,1990-1997 IHD.congood none Consumer goods component, 1990-1997 IHD.wom none Womens rights component, 1990-1997 Time period 1996-2003(period C): pdens.C #/km2 -x^(-0.5) Population density, avg. over period C G.cagr.C % none Annualized GPP fraction increase, time period C Gcap.C baht/pers. -x^(-0.5) G PP per capita, average over period C Gdens.C baht/km2 -x^(-0.5) GPP per area, average over period C agGdens.C baht/km2 -x^(-0.5) Agricultural GPP pe r area, average over period C HDI none Human Development Index, 1997-2002 HAI.comp none Human Achievement Index, composite, 1990-1997 HAI.health none Health component, 1997-2002 HAI.edu none Education component, 1997-2002 HAI.empl none Employment, 1997-2002 HAI.inc none Income component, 1997-2002 HAI.hous none Housing and living conditions comp., 1997-2002 HAI.famcom none Family and community life comp., 1997-2002 HAI.tracom none Transportation and communication comp.,1997-2002 HAI.part none Participation component, 1997-2002 percapinc.00 baht/pers./mo log x Per capita personal income, 2000 percapexp.00 baht/pers./mo log x Pe r capita consumption expenditure, 2000 pov.incid.00 % x^(0.5) Poverty incidence in 2000

PAGE 60

60 Runoff Grid Rain Grid River Discharge Points AET Grid Elevation Grid Coastal nation? Total Water Emergy Determination ( START ) Net Runoff (rain RO + river import river export to other countries) chemical & geopotential Y es Use AET chemical potential + Net runoff geopotential Che m potential > Geopotential N o N o Use Rainfall Chem. Potential + Net river chemical potential Y es GIS data layers Geopotential Emergy Grid Net River Chemical Potential (river import export to other countries) Figure 2-1. Logic flowchart for determination of the renewable line item Total W ater (AET = actual evapotranspiration, RO = runoff). Figure 2-2. Global Map of Irriga ted Areas (GMIA) raster co verage, providing percent area equipped for irrigation, and the Basins polygon coverage contai ning basin boundaries which have assigned irrigation volumes.

PAGE 61

61 Figure 2-3. Flowchart of the Geographic Inform ation Systems (GIS) irrigation m odel. (OSD = on screen digitize, NN = nearest neighbor). Figure 2-4. Stream order, stream gauges and gauge watersheds used for the stream order transform ity model. EXPAND WINDOW Irrigated Area, % ( GMIA ) I rrigated % ( basin extent ) I rrigated % ( 1 km2cells ) Irrigation (m 3/ y r ) Basins resam p le Basins raste r POLYRAS Basins ( m3/ha/ y r ) Basin level irrigation data (THAICID) ASSIGN MAP ALGEBRA Basins vector Basins tiff ( DEQP ) OSD Basins control pts. OSD NN RESAMPLE Provcode (TEI, 1997) Provcode control pts.

PAGE 62

62 Figure 2-5. Schematic of transformity calcula tion, showing stream network, watersh ed outline and river discharge gauge. Figure 2-6. Flowchart depicting the GIS procedures used to estim ate the source stream of irrigation water and produce a coverage of the emergy of irrigation water evapotranspired from the land surface. ASSIGN (transformities from Stream Order Model) Perennial streams (Streamorder) Distance of all cells from streams DISTANCE Stream order allocation ALLOCATE (nearest stream order assigned to cell) Irrigation ( volume ) Irrigation (emergy) MAP ALGEBRA Irrigation (chemical potential) Stream transformity MAP ALGEBRA Rain Emergy TRF = Rain emergy (sej) / discharge chemical potential energy (J) Rain (sej) = Rain (J) Rain chemical potential TRF Rain energy(J) = volume*density*Grain Volume (m3) = Average rain watershed area Density = 1000 kg/m3 Grain = Gibbs free energy = 4740 J/kg Rain chemical poten tial TRF = 30,500 sej/J Discharge energy (J) = volume*density*Griver Volume (m3) = Average annual discharge(m3/sec) sec/yr Density = 1000 kg/m3 Grain = Gibbs free energy = 4720 J/kg Discharge at gauge

PAGE 63

63 0. 4 0 4 1. 0 0.7 0 .4 1 .5 0.6 1 11 .0 Figure 2-7. Map of Gulf of Tha iland with locatio ns and tidal ranges of point data obtained from the tidal prediction program WXTide32. Identify COSTGROW map values corresponding to points 4 major streams (Line raster) Tidal intrusion limit (Points) COSTGROW 1000 COSTGROW (friction = square root(elev), limit 1000) Stream width River channel area (m2) ASSIGN elev Stream classRerun COSTGROW (limit = avg. values, friction = sqrt(elev)*strmclass) Average COSTGROW values at 4 intrusion points Full stream network COSTGROW 200 and 500 Interpolate 2m to 0m in 200 and 500 steps River Tidal Amplitude ASSIGN (1=300, 2=25, 3=10m) SCALAR *500m (stream length per cell) Tidal emergyMAP ALGEBRA (emergy calculation) Figure 2-8. Flowchart depicting the GIS procedures used to es tim ate the tidal energy dissipation in stream channels connected to the Gulf of Thailand. Tidal gauges

PAGE 64

64 Figure 2-9. Venn diagram illustrating part and partial correlations between a dependent variable Y and two independent variables, X1 and X2 (Stevens, 2003). Where R2 indicates the squared correlation coefficient, the partial-R2 of X1 with Y in the presence of X2 is represented by a/(a + e), and the part-R2 of X1 with Y in the presence of X2 is represented by a/(a+b+c+e).

PAGE 65

65 CHAPTER 3 RESULTS National Emergy Analysis First, results are presented for the emergy ev aluations which were conducted at the national scale in order to provide an ini tial view of the scale above the pr ovincial level, and to provide an overview of the major flows and character istics of the country of Thailand. Systems Diagram An energy system s diagram of the Thailand environment-economic system is presented in Figure 3-1. This diagram uses energy circuit diagramming language (Appendix B) to show the flows of energy and materials from external ecosystem sources (sunlight, rainfall, tide); depletion of internal stocks (mineral deposits); transformations (agricultu re, industry); import of fuels, goods, and services; export of goods and services; and the flow of money (dashed lines) through the economy, always paired with goods and services, but only exchanged among people. The main internal primary production systems directly transforming biospheric inputs such as water and tidal energy include forests, estuar ies, marine fisheries a nd croplands, particularly rice. Direct harvest occurs from these production systems, with subsequent processing and export for certain fractions. Mineral resources are also ta pped and used directly or transformed further. Industry utilizes some of the natural resources while supporting and being supported by the human population. Fuels and other good are importe d and interact with indigenous flows in transformation processes. Associat ed with these goods is a high emergy flow in the form of foreign services. Goods, along with their associated services are exported in exchange for money from the international market. The chemical potential of evapotranspired ra in is the most important renewable emergy flow. Water flows, shown in blue, highlight the role of rainfall in rivers, terrestrial ecosystems,

PAGE 66

66 and the estuarine areas which support aquaculture. Dams are constructed for both water storage and electricity generation, and also a lter the historic sediment flows. National Emergy Table Table 3-1 presents annual emergy flows for Th ailand, circa 2000. The table is divided into renewable flows, internal production processes, indigenous non-renewa ble extraction, imports and exports. For each lin e item, energy, material or dollar flows (J, g or $) are indicated, followed by the unit emergy value (sej/J, sej/g or sej/$). The product of these two values is the emergy flow (sej/yr), and the ratio of the emergy flow to the em ergy:dollar ratio is the estimate of the macroeconomic value of that flow in units of Em$. Units of Em$ are used to distinguish these values from the actual $ valu es of the flows. Footnotes to the table, including raw data, unit emergy values (UEVs), data sources, and energy conversion equations, ar e provided in Appendix C. The major renewable emergy inputs are the eva potranspiration of water, the kinetic energy of tides, and deep heat flow in the crust. To avoid double-counting the external biospheric inputs used to calculate the renewable co-product UEVs, only the total water emergy and tidal emergy are considered for the aggregate renewable flow (R), as explained in Chapter 2. The emergy of the chemical potential of evapotranspiration of ra infall is 8.6 E22 sej/yr and is the largest of the renewable flows. As the dominant renewable flow, evapotranspiration is verified as an appropriate area of focus for renewable fl ow analysis at the provincial scale. The major internal sectors of transformation and production are the agricultural, livestock, fisheries, irrigation, and electr icity sectors. The flows of fo restry activities are minor in comparison, reflecting the small fraction of remain ing forest lands. Electricity flows are on par with the livestock sector, rela ted to the achievement of 97% of households having access to electricity in the year 20 00 (Shrestha et al., 2004).

PAGE 67

67 Indigenous non-renewable extraction, defined as extraction or use at rates that exceed replacement rates, is dominated by the mineral s ector. Minerals mined at high rates to fuel the cement and construction industry include clays, g ypsum, dolomite, limestone and shale. Next in magnitude is natural gas production, followed by so il organic matter losses due to erosion, coal mining, oil extraction, fish catch, metals extraction and forestry. The imports are dominated by fuel in energy units and fuel, metal and human service flows in emergy units. Machinery and chemicals are the next largest inflows, followed by meat products, finished products (upgraded paper, text ile, and mineral products), agricultural products, plastics, other refined goods, minerals, and electr icity. The largest export is the human services embodied in each material flow, represented by pr ice of the goods exported, as all of the money exchanged is paid to humans, not to the environm ental energies and materials that flow into the production process. Machinery and transport equipment and other refined goods have the next highest emergy values. Following that is tourism. Other exports in decreasing order of magnitude are metals, meat and fish products, plastics, finished products, fu els, agriculture products and electricity. Flow aggregation and summary diagram National line item em ergy flows are aggregated into summary flows in Table 3-2. Flows include renewable; indigenous non-renewable production; use and export without use; imported materials and services; and expor ted materials and services. Also included are macroeconomic data for the nation: gross domestic product (G DP), payment for imports, and revenue from exports. Figure 3-2 presents a diagram of these aggregated flows of emergy in the Thai economy. Material and energy flow along so lid lines and money flows along do tted lines counter-current to material and energy. Imported services are computed by multiplying the money (US$) paid for imports by the world emergy to dollar ratio (P2). Similarly, the service embodied in exports is

PAGE 68

68 the money paid for those exports multiplied by the emergy to dollar ratio computed for Thailand with the emergy account data (P1). The three-arm diagram in th e lower right shows a further aggregation illustrating the indigenous flows and the trade balance in common emergy units. This shows that the majority of inflows to th e economy are coming from indigenous sources in approximately a 1.8 to 1 ratio. Most of the summ ary emergy indices can be computed from the numeric flows presented in these diagrams. Summary emergy indices Table 3-3 presents a suite of indices for Th ailand that are commonly computed in emergy analysis. Included are total emergy use, em ergy balance of trade, environmental loading ratio (ELR), environmental yield ratio (EYR), investment ratio (IR), emergy sustainability index (ESI), and fraction of total em ergy use that is from various source flows including indigenous, renewable, purchased, fuels, et cetera. The fr action of use from all indigenous sources is 0.62, the fraction from dispersed local sources is 0.13, and the fraction from only locally renewable sources is 0.10. The relatively low fraction of dispersed sources (renewable and slowly replenished non-renewables such as forests and fisheries) results in a value of 6.7 for the ratio of concentrated to rural emergy flows. The value of 1.8 for ratio of expor ts to imports reveals Thailand as a net exporter of emergy. Large fract ions of use include fuel (0.23) and the overall fraction of use that is purchased (0.38). The investment ratio (IR) reflects the larger fraction of indigenous use to purchased use, and may be thought of as one unit of indigenous flow attracts and outside investment of 0.6 purchased use from the larger economy. The environmental loading ratio (ELR) of 8.7 re sults from the nonrenewable use per unit renewable use. The summary indices are generally more useful when comparing across countries or over time within a country. Figure 3-3 shows bar graphs for a set of 3 emergy indices computed for 142 countries using the National Environmental Accounting Databa se (NEAD) (Sweeney et al.,

PAGE 69

69 2006). Thailand is labeled, along with the countr ies surrounding it, based on parameter value. Countries at either end of the spectrum for each index are also identified for additional perspective. Thailands score for renewable fraction of total emergy use places it near the beginning of the lower third tier of countries. This indica tes that regardless of the magnitude of renewable flows, non-renewable flows are more dominant in overall emergy use. Total emergy use per area, or empower density, in Thailand is relatively high, within the first 30% of all country values. Thailands per capita emergy use has a mid-level ranking relative to other countries of the world. Provincial Emergy Analysis Renewable flow raster coverages were used for provincial analys es. Additionally, the provincial analyses involved an irrigation model, tabulation of discharge and eleva tion values for river entry and exit points along provincial borders, allocation of tidal energy amongst coastal provinces and allocation of majo r river discharge into the inne r Gulf of Thailand amongst the coastal provinces in that area. Systems Diagram Many of the study provinces have similar flows and storages of energy, though the ma gnitudes of these components may vary. Major differences include leve l of industrialization, amount of irrigation infrastructure, agricultura l productivity, rainfall and evapotranspiration patterns, and if the province is located on the Gulf of Thailand. Three general representative types of provincial systems are shown in Figure 3-4: A) a poor, agricultural Northeast province, B) a Central agricultural province with favorable water flow s and C) a coastal industrial province, as may occur in either the East of Bangkok region. Many other t ypes could be depicted depending on the focus of study, but these three repr esent the majority of provinces in the threeregion area of Thailand under cons ideration. Differences among provinces regarding rainfall

PAGE 70

70 patterns within the annual cycle are represente d in the diagrams by the red line chart figure within the rainfall source compone nt. Internal processes that vary greatly among provincial types are shown in yellow and include the magnitude of the agricultural a nd industrial sectors, as well as the amount of irrigation. Flows that vary amongs t these three system types, such as water or imported materials flows, are depicted via line thickness. In general, Northeast provinces are economical ly poor and have little industry (Figure 34A). Several of the East and Central provinces also fall within this general system type. Irrigation flows are small re lative to many provinces in the Central and Bangkok regions. Rainfall is higher in some areas of this region, bu t often it is concentrated in a smaller window of time and may not be as available in the soil for evapotranspiration. Central provinces are generally more agricultural than industrial, but do have highe r levels of industry than the Northeast (Figure 3-4B). They have highe r much higher irrigation application, higher agricultural productivity a nd are also wealthier and able to am ass more societal assets. Coastal industrial provinces occur in both the Central and Bangkok regions (Figure 3-4C). Agricultural still plays a role, albeit a smaller one compared to the large scale of i ndustrial production. Some of these provinces include the refining areas that process the natural gas pumped from the Gulf of Thailand. In the agricultural areas of these provinces, irriga tion levels are high. In addition, these systems have the coastal inputs of tidal and wave energies. Marine fisheries and aquaculture play a role in the economy. Asset and mone tary wealth accumulation is highest in these areas. Renewable Emergy Flows Renewable emergy flow values for provinces we re obtained from the renewable water flow coverages generated in the water balance m odel, as well as the additiona l models for provincial tidal flows, irrigation flows, a nd river inputs to the coastal prov inces. Results are presented first

PAGE 71

71 for the models providing the input data, and seco nd for the resulting emergy values aggregated by province. See Appendix A for the pr imary water balance model results. Spatial surfaces of renewable emergy flows Renewable energy flow surfaces were inter polated from existing point measurem ent databases for solar radiation, heat flow, wind, and rain fall. Evapotranspiration and runoff were estimated in the water balance model as descri bed in Appendix A. Standard unit emergy values were applied ( Table C-2 ) and the emergy raster coverage s generated are displayed in Figure 35. Solar, wind and heat flow raster coverages retain the imprint of the original point data, which seem to lack a strong directional spatial pattern (Figures 3-5A, 3-5B, 3-5C). Values for wind emergy are generally higher in the coastal area s and Central region. The runoff geopotential raster, generated with an equati on using runoff and elevation head as the variable terms, looks very similar to the base elevation map, with areas of higher runoff having relatively higher values (Figure 3-5D). The large linear discontinuities seen are artifacts from the original ETOPO5 global elevation coverage. Figure 3-5E is rain chemical potential emergy. It is a linear transformation of the rainfall coverage in mm (Fi gure A-5A), due to the use of a constant rain transformity in the emergy conversion calcula tion. The AET emergy (Figure 3-5F) coverage results from application of the water balance m odel on a monthly, per pixel basis, and is a linear transformation of the AET coverage in mm (Figure A-5C) due to the use of a constant AET transformity in the emergy conversion calculation. Spatial variation of em ergy flows is discussed further in the section addressi ng provincial aggregations. Stream order transformity model The 19 evaluated watersheds in the stream order transformity m odel demonstrate the relationships typically found in stream networks among log(discharge), log(area) and order

PAGE 72

72 (Horton, 1945; Leopold, 1968). Discharge and ar ea have a squared corre lation coefficient (R2) of 0.98, discharge and order have an R2 of 0.97, and area and order have an R2 of 0.98 (Figure 3-6). Rain to a watershed provides water chem ical energy, which is partly used in evapotranspiration and partly transported downstr eam. With the assumption that it takes all of the rainfall to generate the obser ved discharge, transformity fo r stream water chemical energy was calculated by dividing rainfall chemical potential emergy (solar emjoules) by discharge chemical potential energy (joules). Table 3-4 contains the watersheds data and the results of the transformity calculation. Figure 3-7A displays a plot of the resulting transformities for the sample watersheds coded by stream order, and Figure 3-7B shows the av erage transformity per order and the fitted equation used to generate the final stream order transformities. In general, transformities appear to increase linearly wi th stream order, repres enting the convergence and transport of emergy and energy through stream ne tworks. Other than the natural variation that may occur within a given stream order, such as slope, rainfall and la nd cover differences, the extensive human modification of water flow via dams and paddy field dikes contributes to the variation in transformities within stream orders. The stream transformities applied to river wa ter withdrawn for irrigation are calculated using the fitted equation from Figure 3-7B. The da ta for stream order 8 is from a gauge on the Mekong River shortly after it reaches the Thai border. The Mekong is assumed to be stream order 8 due to the fit with the other data as it is plotted Figure 3-7. B ecause the watershed lies beyond the spatial extent of the GIS work done fo r this study, average rainfall was not obtained and a direct transformity was not calculated. The transformity used for water from the Mekong in Thailand is calculated from the fitted equation in Figure 3-7B, solved for a stream order 8.

PAGE 73

73 Irrigation model The GIS irrigation model generated a raster coverage with estim ated volume per pixel. Estimated values for irrigation volume applied to agricultural lands ar e shown in Figure 3-8A. Irrigation volume is far greater in the Central region than ot her areas of the country. The Northeast region has higher irrigation volumes in areas along major rivers. After reducing applied volume by 50% to account for transmission and runoff lo sses (as explained in Chapter 2), the remaining total volume per year per provi nce, normalized by area, gives the average depth of irrigation evapotranspired pe r province in mm (Figure 3-8B). Again, the Central region has the highest irrigation AET flows, ranging from 130 to 500 mm/yr. The southernmost of the Northeast provinces and some of the East provin ces have the lowest ra tes, ranging from roughly 10-30 mm/yr. Middle values are seen in the west ern and eastern portions of the Central Region (further from the Chao Phraya River) as well as the northern portion of the Northeast region. In order to estimate the emergy of the irri gation water used, the order of the nearest perennial stream was assigned to each pixel with irrigation, enabling the us e of the stream order transformities in the em ergy calculation. Figure 3-9 displays the final emergy map (Figure 3-9D) as well as the intermediary maps generated in th e estimation of the chemi cal potential emergy of river-derived irrigation water. The stream order coverage sh ows a high density of stream channels in the lower Chao Phraya River valley which loses the adherence to a typical dendritic stream network pattern as seen in the rest of the country (Figure 3-9A). Figure 3-9B shows the results of a distance operation which produced a coverage of each pixels distance from the nearest perennial stream. The allo cate operation used the distance coverage to assign the stream order of the nearest perennial stream to each cell (Figure 3-9C). Performing the emergy calculation for chemical potential of evapotranspired irrigation water resulted in the coverage in Figure 3-9D.

PAGE 74

74 Provincial renewable emergy flows Values for emergy variables (s ej/yr) were ex tracted from the raster coverages using a vector coverage of provincial boundaries. Emergy variables were normalized by province area for final units of sej/m2/yr. Figure 3-10 contains maps of selected emergy flows by province on a per area basis. Maps of provincial rain and AET emergy will be identical to the maps in mm units, due to the use of constant transformitie s in the conversion of energy to emergy units. Average annual emergy of solar radiation a nd wind emergy are higher in the coastal areas (Figure 3-10A, Figure 3-10B). Heat flow is generally higher in the Korat Plateau of the northeast, and lowest in the Chao Phraya Rive r valley in the Central region (Figure 3-10C). Runoff geopotential emergy from rainfall is higher in the provinces with more elevation change, and highest in those with both high rainfall values and larger elevation gradients (Figure 3-10D). Tidal emergy is highest for those provinces with relatively longer coastlines that reside in the Inner gulf (Figure 3-10E). Emergy of the chemical pot ential of freshwater runoff into the Gulf is highest in the Inner Gulf where 4 major rivers di scharge, and for those Inner gulf provinces with longer coastlines (Figure 3-10F). Maps of irrigation emergy and several renewabl e emergy aggregates at the provincial scale are shown in Figure 3-11. The more agricultural provinces of the Chao Phraya River valley have the highest values for emergy of irrigation water ev apotranspired (Figure 3-11A), as they are in a zone of streamflow convergence where runoff from large areas to the north is concentrated into the Chao Phraya River. Lower values are found in most of the Northeast region and the easternmost portion of the East region which ha s no rivers above order 5, other than the Mekong, and relatively lower irrigation volu mes. The map of the aggregate of total AET from both rainfall and irrigation (Figure 3-11B) is si milar to that of irrigation only. The final two maps of this figure depict two variations on Total Renewa ble Emergy summary flow: Figure 3-11C excludes

PAGE 75

75 irrigation, while Figure 3-11D in cludes the irrigation flow. The dramatically higher values for irrigation in the Central region account for the differences seen between these two maps. Table 3-5 contains summary statistics for th e provincial renewable emergy flows on a per area basis. Variable descriptions are located in Table 2-1. Many distributions are far from normal as indicated by the skewness values. The relativ e magnitudes and shapes of the untransformed distributions are shown in box plots in Figure 3-12, and Figure 3-13 illustrates the shift to normal distributions for each variable via the data tran sformations described in Chapter 2 and noted in Table 2-1. Details of scale units are not shown as the purpose of this figure is to simply illustrate the change in the distributions due to transformation. After tran sformation, the distributions are relatively normal and many outliers are elimin ated. Outliers still exist for rain, runoff geopotential, and total renewable water (as calculated without irrigation flows). Evapotranspiration did not require transformation. Regional variation In order to aid in visualizing regional vari ation, provincial values for renewable emergy flows have been aggregated by region and displa yed in box plot diagrams. Figure 3-14 displays a suite of transfor med emergy flow variables by region. Relative values for emergy in rain and AET are the same as with the flows measured in mm, due to the application of a constant transformity. Irrigation emergy is highest in the Bangkok and Central regions, lower in the Eastern region and lowest overall in the Northe ast region (Figure 3-14C ). Runoff geopotential emergy is lowest in the Bangkok area and highest in the Central and Eastern regions (Figure 314D). Runoff chemical potential emergy is zero in the Northeast because there are no coastal provinces. The other regions have similar ranges of values overall, and included inland provinces with zero runoff chemical potential emergy (Fi gure 3-14E). Similarly, tidal emergy does not exist in the Northeast and varies in the othe r regions depending on the average tidal amplitude

PAGE 76

76 along provincial coastlin es (Figure 3-14F). Total renewabl e water flow calculated without considering irrigation (renew.W) averages th e highest in the East, though the Bangkok area has some provinces with higher values (Figure 3-14G). The Central and Northeast regions have lower values overall, with some Central provinces having higher values. Total renewable water flow calculated with irrigation included (re new.WI) is highest in the Bangkok area and significantly lower in the Northeast (Figure 3-14 H). The Central and East regions are similar overall and values lie between the Bangkok and Northeast regions. Correlation Among Environmental Variables A Pearson correlation matrix of the transfor med environm ental variables is located in Table E-1 in the Appendix (see Table 2-1 for variable definitions and transformations). Correlation coefficients greater than 0.60 are highlighted with asterisks to highlight stronger correlations. This table includes variables which are directly calculated from other variables as well as aggregations of variables within the table. Strong correlations among those variables directly derived from other variab les within the table are not explored due to the obvious nature of their relationships. As might be expected, rainfall in mm/yr is positively correlated with runoff in mm/yr ( r = 0.86, p<0.01). Landscapes have an upper limit for absorbing and evapo-transpiring water such that larger volumes of rainfall would result in larger runoff volumes and a reduced AET to rainfall ratio. This limitation of the landscape to absorb large volumes is particularly apparent when looking at the relationship of rainfall a nd AET in a monsoon climate, with rain and AET calculated on a monthly basis, then aggregated, as is done in this study. If AET were calculated as a constant percentage of rainfall, the relatio nship would have a correlation coefficient of 1.0. Because this study employed a monthly water bala nce, the relationship is quite different ( r = 0.06, p=0.71). Upon looking at a bivariate plot of these two variables, it is apparent that there is a

PAGE 77

77 positive relationship which breaks down as rainfall increases (Figure 3-15). Figure 3-15A plots all provinces and Figure 3-15B plots those with less than 1450 mm of rainfall and also excludes the outlier Kanchanaburi, a large mountainous provi nce in the westernmost portion of the central region. Rainfall is especially concentrated into the monsoonal rainy season for the Northeast and East provinces that have higher overall rainfall values. The monthly water balance discerns this excess water on the landscape duri ng the rainy season, when the so il is not capable of holding more water so that it may evapo-transpire. For provinces with less than 1450 mm of rainfall, AET fraction of rainfall averages 0.85, while th e fraction for province with greater than 1450 mm rainfall averages 0.56. Stream density shows significant positive co rrelation with average irrigation depth (r = 0.80, p=0.00), irrigation emergy (r = 0.79, p=0.00), and the total renewable aggregate that includes irrigation, renew.WI (r = 0.81, p<0.01). This is not surpri sing given that almost all of Thailands irrigation is ob tained from stream flow. Figure 3-16 contains a matrix of bivariate plots for relationships among aet.mm irg.mm, aetirg.mm, irg and aetirg, color-coded by region. Th e regional coding reveals that the nature and strength of most of the correla tions vary by region. The total AET aggregate (aetirg) which sums rainfall AET (aet) and irrigation AET (irg), has a strong positive relations hip with irrigation AET in all provinces other than th e Northeast (Figure 3-16). Irriga tion is generally lower in the Northeast and therefore aetirg has a stronger positive correlatio n with aet in this region. Of note is the correlation be tween irrigation depth in mm (irg.mm) and irrigation emergy (irg). Due to the use of stream order transformities, the relationship does not exhibit a perfect positive correlation. Provinces with higher order str eams deviate toward relatively higher values. However, the deviation is small and the pos itive relationship is still very strong ( r = 0.99,

PAGE 78

78 p<0.01). Accounting for the energy quality of the irrigation water does not radically alter the variability of irrigation volume in mm. Regarding the renewable aggregates, renew.W (total renewable wit hout irrigation) shows strong correlations with only tide and chemical potential of runoff ( r = 0.87and r = 0.69, respectively, with p<0.01). This indicates that ROchem is the do minant component of this total renewable aggregate. Tide is corr elated with renew.W because it is positively related to chemical potential of runoff based on coastal location of a province. Only coastal provinces have non-zero values for these two variables. The total renewable aggregate renew.WI includes irrigation, and because the variables aet and irg are part of the sum used to calculate renew.WI, a positive relationship is expected with th ese two flows. However, of thes e two flows, only irrigation has a strong correlation with renew.WI ( r = 0.78, p<0.01). Additional variables in this aggregate account for most of the variation in the corr elation, overshadowing the role of aet in the summation of total renewable flow: runoff geopot ential in the case of inland provinces, and runoff chemical potential in the case of coastal provinces. Among the emergy measures in the correlati on matrix (Table E-1), runoff chemical emergy and tidal emergy are highly correlated ( r = 0.77, p<0.01). Only coastal provinces have non-zero values for the chemical emergy in runoff and river flow reaching the coast, therefore this result is not surprising. The positive rela tionship is also caused in part by the use of provincial shoreline length to allocate both the river inputs to the estuarine area and the tidal energy on the continental shelf. Ti dal emergy is also strongly corre lated with renew.W, the total renewable water flow that does not include irrigation or tide. This is because the chemical emergy in runoff is a significant portion of th e RENEW.W aggregation for coastal provinces.

PAGE 79

79 Principal Components Analysis (P CA) of Environmental Variables Principal components analysis (PCA) was performed to reduce the num erous, often intercorrelated, variables under consider ation into fewer, uncorrelated principal component variables. Variables are automatically standard ized as part of the PCA proce ss in most statistical software packages, including Minitab software used in this study. A box plot was constructed of the standardized variables to check for any remaini ng outliers (Figure 3-17). Despite transformation and standardization, outlier s are still present for se veral of the variables, though they are within three standard deviations. PCA of standard unit water variables The PCA of water variables, henceforth called PCAw, includes strm dens, rain.mm, aet.mm, irg.mm, ro.mm, sm.mo a nd def.mo (see Table 2-1 for defi nitions). Table 3-6 lists the eigenvalues and loading values fo r the first three principal com ponent of PCAw. Variables with loadings around 0.5 or greater for a principal component (PC) are considered to be associated strongly with that PC. Principal component 1 (PC1w) has variance (eigenvalue) 3.12 and accounts for 45% of the total variance in the suit e of water flow variables. PC1w is positively associated with rainfall and runo ff and negatively associated with soil deficit, suggesting that PC1w is reflecting an abundance of water ava ilability, though this is most likely occurring during the limited rainy season and might be c onsidered water excess. Principal component 2 (PC2w) has variance (eigenvalue) 1.77 and accounts for an additional 25% of the total variance. PC2w is negatively associated with both soil moisture and irrigati on, suggesting this PC reflects water shortage, both from lower throughput via A ET and lower storage in the soil. Principal component 3 (PC3w) is dominated by a str ong association with aet .mm, suggesting this component represents water use, or transformation.

PAGE 80

80 Figure 3-18 displays the loadi ng plot and score plot for PC1w and PC2w. When colorcoded by region, the score plot reveals that Bangkok and the Central region score lower on PC1w (water abundance/excess), wh ile the East and Northeast sc ore higher. Additionally, the Bangkok and Eastern regions score lower on PC2w (water shortage), while the Central region scores higher, but with a wide range, a nd the Northeast scor e generally higher. PCA of renewable emergy measures In the set of standardized emergy flows, not including the aggregated summary flows, 3 variables have outliers: rainfall chemical potent ial (rain), runoff geopoten tial (ROgeo), and tidal energy (tide) (Figure 3-17). Trad and Chantha buri, the two easternmos t coastal provinces, are upper outliers for rainfall. Chanthaburi, Kanchanaburi and Singburi are uppe r outliers for runoff. All coastal provinces are the outlie rs for the tidal measure because all interior provinces have a value of zero for tide. Due to the extreme nature of the highest rain outlier, and because rainfall as a whole does not represent a flow that is entire ly absorbed or used by the system (it is not part of the RENEW calculation), rain is omitted from the emergy PCA. The ROgeo outliers are not as extreme and this variable is left in with cau tion. The tidal emergy, though a unique component of the emergy signature, is also omitted as it is prob lematic both in terms of the outliers, and the assumption when calculating the energy values that the fraction of the total tidal energy in the Gulf which is received by a province is proportional to the provincial coastline length. The PCA of the emergy flow variables, henceforth called PCAem, includes chemical potential of rain-derived AET (aet), irrigation derived AET (irg), geopotential of runoff (ROgeo), solar radiation (rad), wind, deep heat (heat), a nd chemical potential of runoff reaching coastal waters (ROchem). Table 3-7 lists the eigenvalues and loading values for the first three principal component of PCAem. Because the emergy of water flows is of sp ecial interest as a renewable subsidy to the system, the signs on the scores and loadings for the first PC will be reversed to

PAGE 81

81 reflect positive values for these subsidies, thus making interpretation easi er. Principal component 1 (PC1em) has variance (eigenva lue) 2.72 and accounts for 39% of the total variance. PC1em is positively associated with irri gation and wind and runoff chemi cal potential, and negatively associated with solar radiation and heat. A higher score on PC1em reflects relatively higher inflows of the types of emergy experienced in th e coastal areas (wind, runoff chemical, generally higher irrigation values). Prin cipal component 2 (PC2em) has variance (eigenvalue) 1.75 and accounts for an additional 25% of the total variance. PC2em has a strong positive association with aet and a negative association with RO geo, suggesting that PC2em represents general terrestrial use of water flow s via evapo-transpiration. Principal component 3 (PC3em) has variance (eigenvalue) 1.23 and accounts for an additional 18% of the total variance. Together, all 3 PCs account for 82% of the total varian ce in this emergy signature dataset. Figure 3-19 displays the loadi ng plot and score plot for PC 1em and PC2em. Note that these figures use the original signage for PC1em. The loading plot reveals the relationships among the variables, showing a general positive association among irg, wind and ROchem, and their negative loading on PC1em with the orig inal sign. When color-coded by region, the score plot reveals that the Bangkok re gion scores lower on the original PC1em, while the Northeast scores higher. This is mainly due to the inla nd location of the Northeas t provinces, resulting in zero values for runoff chemical emergy. The provinces of the Bangkok region and most of the Northeast region score higher on PC2w (terrestr ial water use), while the East and Central regions score relatively lower. The PC scores for PC1w, PC2w, PC3w, PC1e m, PC2em and PC3em will be included in further exploratory analysis of relationships among environmental and socio-economic wealth measures.

PAGE 82

82 Socioeconomic Wealth Variables Socio-economic wealth variables include m oneta ry flows, as well as indices of welfare. Values for variables as attained from sources detailed earlier are aggr egated by province and categorized by several time periods as noted in the Methods section: Period A from 1981-2003, Period B from 1981-1993 and Period C from 1996-2003. Figure 3-20 contains maps produced from some of these provincial scale variables for period C. The population density map (Figure 320A) shows the high values in the Bangkok are as well as a few provinces in the Northeast. Lowe r values tend to be on the periphery of the 3 region area of study. Gross Provincial Product (GPP) per capita (Figure 3-20B) is much higher in the Bangkok area and Central and East regions re lative to the Northeast. Gross Provincial Product normalized by land area (Figure 3-20C) show s a similar distribution to GPP normalized by population, with differences mainly occurring along the western side of the Central region. Compound annual growth rate (CAGR) is highly vari able over space as shown in Figure 3-20D. Figure 3-20E shows the amount of GPP due to agriculture, normalized by area. The highest values are found in the Central valley around the Chao Phraya River, and in a few of the coastal provinces. Figure 3-21 contains four a dditional coverages of provinc ial wealth measures. The composite values for the Human Achievement Index (HAI) are generally higher in the Central and East regions, with the lowest values occurr ing exclusively in the Northeast (Figure 3-21A). This map is very similar to the GPP per capita map in terms of relative order of values. The income component of the HAI has a similar distribution among provinces (Figure 3-21B) as both the HAI composite and GPP per capita. Per capita e xpenditure is again similar to the formerly mentioned maps (Figure 3-21C). Poverty incidence appears to be correlated to the earlier maps

PAGE 83

83 as well, though in the reverse, with the highest values occurrin g exclusively in the Northeast Region (Figure 3-21D). Table 3-8 contains the summary statistics fo r all of the provincial wealth variables. Variables that have outliers a nd high skewness values were transformed as described in the methods section and are detailed in Table 2-2. Variation Preand Post-Financial Crisis Comparing data for periods B and C provide s some insight regarding differences in variables before and after the 1997 financial crisis. Figure 3-22 is a series of box plots showing the distributions over the three time-periods for both raw data and transformed data. Transformations reduced skew and reduced the occurrence of outliers. Population density (not shown) varies little over this relatively short time period, but other variab les show distinctions preand post-crisis. Gross Provincial Product per area (Gdens) a nd agricultural fraction of GPP per area (agGdens) increase from time period B to time period C (Figure 3-22B). Per capita GPP is also higher post-1997 (Figure 3-22D). Compound annual growth rate (CAGR) was lower in the post-crisis time period, droppi ng from an average of 7.3% to 1.4% for time periods B and C respectively (Figure 3-22E). However, GPP pe r capita (in constant 1988 baht) is higher postcrisis, with the average valu e increasing from 30,389 to 54,488 ba ht/capita/yr (Figure 3-22C). Regional Variation in Socioeconomic Measures In order to visualize regional variation, provincial socio-econo mic data were aggregated by region and organized into box plot diagram s (see Table 2-2 for deta ils on transformations applied). Figure 3-23 contains 4 sets of box plots which were c onstructed for both time periods B and C (preand post-financial crisis) for furthe r insight into regional differences in changes across time periods.

PAGE 84

84 Per area GPP (Gdens) rises moderately and re latively evenly among the provinces between the two time periods (Figures 3-23A, 3-23B), though it appears th at the other regions close the gap somewhat with the Bangkok region. Agricultu ral fraction of GPP per area (agGdens) decreases for the Bangkok region, but rises for th e other regions (Figures 3-23B, 3-23C). Per capita GPP (Gcap) rises across the board, but le ss so for the Bangkok region (Figures 3-23C, 323D). Compound annual growth rate of GPP (G.c agr) is much lower post-financial crisis (Figures 3-23G, 3-23H). Though all four regi ons have lower CAGR in period C, the Bangkok and East regions show larger de creases overall. In addition, average growth rates for the Central, East and Northeast regions are higher than th at of the Bangkok area for Period C, as opposed to the relative positions in Period B. The Index of Human Deprivation (IHD) is a composite index representing roughly 1990 through 1996 (UNDP, 1999). Figure 3-24 displays regional box plots for IHD and its eight components. The Northeast region stands out with higher deprivation values for the composite index and all of the components, particularly income, education, employment and ownership of consumer goods. The Human Achievement Index (HAI) is a composite index representing roughly 1998 through 2002 (UNDP, 2003). Figure 3-25 displays regional box plots for HAI and its components. Trends are generally similar to th e respective components in the IHD, with some more noticeable differences. The East scores re latively higher in the HAI income component versus the IHD income component (Figure 3-25B). The Bangkok area scores higher in the education component in the HAI (Figure 325D). Two components of the HAI are not represented in the IHD: family and communication (Figure 3-25H) and participation (Figure 3-

PAGE 85

85 25I). Regional score positions for these components are generally opposite the positions for the other components, with the Northeast scoring highest and Bangkok the lowest. Keeping in mind that IHD and HAI differ w ith respect to component categories, time periods, and calculation method, pr ovincial values for these two indices are compared in the bivariate plot in Figure 3-26A. The correla tion is negative because the IHD measures deprivation rather than achievement, and deviations from the relationship occur mainly in the Northeast and Bangkok regions. This may be due to the fact that th e IHD methodology treats all the scores above the median equally and fo cuses on the lowest and second quartiles because IHD is meant to highlight deprivation, not ex cellence. Thus, IHD does not capture differences among the above-the-median provinces. It pena lizes poor performance but does not recognize excellence. Provinces that perform well on IHD ar e those that manage to stay above average on all indicators, but may not be exceptional on a ny one. In contrast, the HAI methodology uses the calculation method of the HDI for its indicators: (act ual value minimum value) / (maximum value minimum value). Data for individual Human Development Index (HDI) components were not published in the UNDP report (2003). The composite index data are shown in Figure 3-26B plotted against HAI, which is calculated for the same time period. Though there is a positive relationship overall, it does not hold for the Central and Bang kok regions in isolation. This is most likely due to the difference in components included for each index: HDI includes edu cation, life expectancy and income, while HAI includes health, education, employment, income, housing and living environment, family and community life, transport and communication, and community participation.

PAGE 86

86 Three additional economic measures, repres enting the year 2000, ar e per capita income, per capita expenditure an d poverty incidence. Box plots of th e transformed variables are shown in Figure 3-27. Income and expenditure show similar regional variation, with the Bangkok region highest, the Central and East regions w ith mid-range values, a nd the Northeast region with significantly lower values (Figures 3-27A, 3-27B). The Northeast also stands out as having dramatically higher values fo r poverty incidence (Figure 3-27 C). The other regions have overlapping ranges of values, with th e lowest values in the Bangkok region. Correlation among Socioeconomic Measures Table E-2 in the Appendix contains a Pearson correlation matrix of the socio-econom ic variables (see Table 2-2 for variable definitio ns and any transformations applied). Larger correlation coefficients ( r > 0.75) are highlighted with as terisks to iden tify the higher correlations. Because period A is an aggregation of periods B and C, with period B having 13 years compared to 8 years in period C, correlations between period A and period B for a given variable were extremely strong (not shown). To simplify further analysis, only periods B and C will be analyzed in the regressions and PCA. Correlation coefficients are large and signi ficant among GPP per area, GPP per capita, and poverty incidence. In addition, these economic m easures are also highly correlated to the HDI and HAI. Distance to Bangkok is st rongly correlated to all ec onomic measures other than compound annual growth rate. Regarding the agricultural GPP per area, co rrelations are less str ong with the broader wealth measures (HAI, IHD), but they are still significant. Period B agGdens has a coefficient of 0.74 (p<0.01) with HAI and Period C agdens has a coefficient of 0.72 (p<0.01) with HAI. Unlike the majority of the wealth measures compound annual growth rate does not exhibit many strong relationships. Within time periods B and C, compound annual growth rate and GPP

PAGE 87

87 per capita are weakly positively related, though the degree of correlation is noticeably weaker in post-crisis period C (0.58, p<0.01; 0.42, p<0.01, respectively). Growth rate s became negative for some provinces in period C due to the financial crisis, which may explain the differences seen in CAGR correlations for period C (G.cagr.C) as co mpared to the respective relationships for the pre-1997 time period. Among the indices, HAI is strongl y correlated with both IHD ( r = -0.84, p<0.01) and HDI ( r = 0.81, p<0.01). The patterns of the relationships can be seen in the bivariate plots in Figure 326. In addition, per capita measures of GPP, income and expenditure show strong positive relationships with both HAI and HDI ( r > 0.8, p<0.01, Table E-2). Pove rty incidence also has a strong negative relationship with HAI ( r = -0.91, p<0.01) and a sligh tly less strong negative relationship with HDI ( r = -0.78, p<0.01), indicating that the HAI index is capturing more information about poverty at the provincial level. Visualizing the data in bivariate plots reveals interesting differences in the relationship between GPP per cap ita and the indices when looking at regions (Figure 3-28). Though positive across all regions for HAI, the Bangkok region actually has a negative correlation between GPP per capita and HAI and the Central region shows no strong relationship. Plots of both HDI and HAI versus GPP per capita for periods A and B look very similar to those in Figure 3-28. Principal Components Analysis of Socioeconomic Variables Principal components analysis was performe d to reduce the nume rous variables under consideration into fewer, unc orrelated principal component variables. Variables (some transformed as listed in Table 22) were standardized as part of the PCA process in Minitab. A box plot was constructed of the standardized variables for periods B and C to check for any remaining outliers (Figure 3-29). Outliers are present for some of the variables: CAGR for both periods, the education compone nt of IHD, the employment component of HAI, and the

PAGE 88

88 participation component of HAI. Th e variables with outliers are incl uded, with atte ntion given to the potential affect on the creati on of the principal components. Time period B (1981-1993) The PCA of the wealth variab les for Period B, henceforth called PCAb, includes GPP per area, GPP per capita, compound annual growth rate (CAGR) of GPP, each of the IHD components, and the IHD composite variable. Th e IHD composite was in cluded to account for the weighting scheme used by th e developers of this index. Table 3-9 lists the eigenvalues and loading values for the firs t three principal component of PCAb. Principal component 1 (PC1b) has vari ance (eigenvalue) 5.65 and accounts for 51% of the total variance. PC1b has strong positive association with the consumer goods IHD component, as well as some positive association w ith all other IHD components, suggesting that PC1b represents the IHD in general. PC1b is also negatively associated with GPP per capita, as expected given that GPP per capita and the IHD are negatively correlate d. Because this wealth component will be easier to interpret with the signs reversed, PC1b will henceforth have opposite sign from original scores and reflect higher we alth. Principal component 2 (PC2b) has variance (eigenvalue) 1.67 and accounts for an additional 15% of the total variance. PC2b is dominated by a strong negative association with CAGR. Pr incipal component 3 (PC3b) has variance (eigenvalue) 1.01 and accounts for an additional 9% of the total variance. PC3b is dominated by a negative association with the womens issues component of the IHD. Together, all 3 PCs account for 77% of the total variance in this particular dataset. Figure 3-30 displays the loadi ng plot (A) and score plot (B ) for PC1b and PC2b. Note that these graphs show the original sc ores for PC1b, with sign opposite that used in further analysis. When color-coded by region, the score plot reveal s the Northeast region squarely on the positive side of PC1b, reflecting the high de privation scores of these provin ces in the IHD, as well as the

PAGE 89

89 lower GPP per capita in the Northeast. The PC2b axis, related more to CAGR than any other variable in this set, has all regions other th an Bangkok spread about the range of values. The Bangkok region scores lower on this PC, i ndicating higher annual growth rates. Time period C (1996-2003) The PCA of the wealth variab les for Period C, henceforth called PCAc, includes GPP per area, GPP per capita, compound annual growth rate (CAGR) of GPP, poverty incidence, each of the HAI components, and HDI. Table 3-10 lists the eigenvalues and loading values for the first three principal component of PCAc. Principal component 1 (PC1c) has va riance (eigenvalue) 9.43 a nd accounts for a large portion of the total variance (67%). PC1c has pos itive association with all income variables and HAI components except for the family and community component. Poverty incidence is negatively associated with PC1c. Obviously, PC 1c represents monetary welfare. Health, education and employment are posi tively associated with PC1c, but with less strength. Principal component 2 (PC2c) has variance (eigenvalue) 1.10 and accounts for an additional 8% of the total variance. PC2b is dominated by a str ong negative association with CAGR. Principal component 3 (PC3c) has variance (eigenvalue) 0.90 and accounts for an additional 6% of the total variance. PC3c is positively correlated with CAGR and the employment component, and negatively correlated with the HAI community participation component. Together, all 3 PCs account for 82% of the total variance in this particular dataset. Figure 3-31 displays the loadi ng plot (A) and score plot (B) for PC1c and PC2c. The loading plot demonstrates the association of the income variables and the majority of the HAI components. The family and communication component is on the opposite with poverty incidence. In Thailand, areas with more poverty tend to score high on the family and community components, perhaps because there is more re liance on relatives and the local community for

PAGE 90

90 assistance and resources. When color-coded by region, the score plot reveals the Northeast region squarely on the negative side of PC1c, refl ecting the low scores of these provinces in the economic welfare variables. The PC2b axis, related more to CAGR than any other variable in this set, has all regions represented across the component. The Bangkok region generally scores lower on this PC, indicating lower annual growth rates, the opposite of the results in PCAb. The PC scores for PC1b, PC2b, PC1c, and PC2c will be included in further exploratory analysis of relationships among environmen tal and socio-economic wealth measures. Bivariate Analysis of Environmen tal and Socioeconomic Variables Correlation matrices and bivariat e scatter plo t matrices were used to explore patterns of correlation among environmental and socioeconomic variables. Results are presented for the set of matrices created that are relevant to the dist illation of a variable set for further investigation using regression analysis. Standard Unit Water Flows and Wealth Table E-3 in the Appendix contains a correlation matrix of selected water flow variables, water flow principle com ponents (PCs), and wealth measures. Water flow measures (defined in Table 2-1) include rain.mm, aet .mm, ro.mm, irg.mm, aetirg.mm, def.mo, sm.mo, PC1w (water excess) and PC2w (water shortage ) and PC3w (water us e). Socioeconomic meas ures (defined in Table 2-2) include popd.c, Gdens.C, agGdens.C, Gcap.C, G.cagr.B, G.cagr.C, pov.incid, IHD, HAI.inc, HAI, PC1b and PC1c. Variables PC1b and PC1c represent income and welfare for periods B and C. The second wealth components for both time periods (PC2b, PC2c) represent compound annual growth rate and did not exhibit any strong correlations with water measures, thus they are not present in th e correlation matrix. Correlation co efficients greater than 0.60 are highlighted with asterisks and i ndicate candidate variables for incl usion in regres sion models in combination with other variables. The strongest correlations betw een water measures and wealth

PAGE 91

91 measures are found with irrigation AET, and to a lesser extent, PC1w. Correlation of wealth with other water flow variables tends to be weak or non-existent when including all 45 provinces. Bivariate plots coded by region were create d to look for any regional differences in correlations that might be causing lower correlation coefficients for the full dataset. The bivariate plots in Figure 3-32 reveal the regional patt erns of the relationships found between socioeconomic measures and rainfall rainfall AET (aet.mm), irriga tion AET (irg.mm), and AET of both rainfall and irrigation (aetie rg.mm). The plots of PC1w and PC2w are not shown due to the similarity with other plots: the PC1w plots look almost identical to the rainfall plots, and the PC2w plots are very similar to th e inverse of the irrigation plots. Rainfall and wealth When all regions are included, rainfall does not have a strong overall correlation with any socioeconomic variable. However, the bivariate pl ots (F igure 3-32) indica te that correlations may be stronger if region is accounted for, particularly the Northeast. The Northeast provinces generally clumped on the low end of the Y-axes for wealth variables (o pposite for poverty), effectively spreading the scat ter plot along the wealth axes and significantly reducing the correlation coefficients. In addition, it appears that rain is negativel y correlated with wealth only in the Northeast and East regions. Evapotranspired rainfall and wealth Though the correlation matrix does not reveal any strong po sitive relationships between aet.mm and any of the wealth variables, the bivariate plots coded by region suggest some positive relationships if the Northeast region is omitted or considered separately (Figure 3-32). For example, when looking at AET in relation to the poverty variable, the Northeast exhibits no clear relationship, while the other regions exhibi t a weak negative correlation (r = -0.42, p<0.05). A similar phenomenon is seen in the plots of AET and both GPP per capita and PC1c, where

PAGE 92

92 positive relationships are present when the Northeast region is omitted from the correlation (r = 0.64 and 0.57, p<0.001, respectively). AET does not seem to be a driving f actor in per capita wealth in the Northeast, but it may play a role in wealth per area in the Northeast as well as the determination of per capita and per area wealth in the Central, East and Bangkok regions. When plotting AET versus population density, a fa irly linear positive relationship is seen. Irrigation and wealth The moderately strong correlation of the wealth variables with AET de rived from irrigation does not apply within the Northeast or East regi ons (Figure 3-32). In addition there are 1 to 2 prom inent outliers in the East region, Chonbur i Province, the larges t outlier, and Rayong province. These are two of the w ealthiest provinces, as measured by GPP, due to their industrial sectors which include the processi ng of natural gas piped in from the Gulf of Thailand. If the correlations are performed again with all four regions, but with Chonburi only omitted, correlation coefficients with irg.mm are as follows (p<0.01): r = 0.77 for HAI, 0.82 for HAI.inc, -0.74 for pov.incid, 0.77 for Gcap.C and 0.81 for PC1c. When aet.mm and irg.mm are combined to form one total AET variable (aetirg.mm), and all provinces are included, the correlations decrease from the coefficients for irrigation A ET alone (irg.mm) for most wealth variables, perhaps due to the poor correla tion between irrigation AET and rainfall AET which may cause values to become more homogenized (Table E-3) However, coefficients do increase for GPP per area and agricultural GPP per area. Soil moisture and wealth The bivariate plots f or soil moisture reveal the familiar trend of a potential separate relationship for the Northeast region for some of the socio-economic va riables (Figure 3-32). Gcap.C tends to decrease with increasing soil moistu re in the Northeast, but increases slightly in the other regions. PC1c and HAI show similar trends more distinc tly. Poverty incidence

PAGE 93

93 increases with soil moisture in the Northeast a nd decreases with increasing soil moisture in the other provinces. Emergy Flows and Wealth Table E-4 in the Appendix contains a correlation matrix of selected em ergy flow variables, emergy flow principle components (PCs), and wealth measures. Correlation coefficients greater than 0.60 are highlighted with asterisks and c onsidered strong candidates for inclusion in regression models in combination with other va riables. The emergy aggregate of rainfall AET plus runoff geopotential (aetROgeo) is not shown in the table due to lack of correlation with any wealth variables. However, this variable is important because, for inland provinces, it is equivalent to the total renewable flow aggreg ate (renew.W). The bivari ate plots in Figure 3-33 reveal the pattern behind the lack of correlation with wealth, usi ng GPP per area as an example. If the relationship for the Northeast is consider ed separately, AET is positively correlated (Figure 3-33A) and runoff negatively corre lated (Figure 3-33B), with Gdens.C. Because AET and runoff are also negatively correlated with one another, summing them results in a loss of correlation with Gdens.C and other wealth variables (Figure 3-33C). Figure 3-34 contains bivariat e plots revealing the regional patterns of some of the relationships in Table E-4. As with the non-emer gy water flow variables, the Northeast appears as a separate block of values do to its uniform ly lower scores for socioeconomic measures. In addition, correlations that exist for other regions generally are not present for the Northeast. These plots suggest that a dummy variable for the Northeast should be used in any multiple regression analysis. Terrestrial water emergy flows and wealth Emergy of the chemical potential of both rain fall and AET are directly proportional to the water flow values in mm, thus these two vari ables are not shown in Table E-4 or Figure 3-34.

PAGE 94

94 Emergy of irrigation AET (irg) e xhibits positive correlation with GPP per capita in periods B and C (r = 0.55, 0.62), HAI (r = 0.66), income component of HAI (r = 0.73), and per capita expenditure (r = 0.70), all at p< 0.01. Correlations with the wea lth principle components for periods B and C are also positive (r = 0.66 a nd 0.71 respectively, p<0.01). Irrigation is also negatively correlated with poverty incidence (r = -0.66, p<0.01). Emergy of AET from both sources (aetirg) displays stronger correlations th an the non-emergy counterpart of this variable, aetirg.mm, due to the use of the stream order transformities in deri ving emergy. Aetirg is positively correlated at p<0.01 with population de nsity (r = 0.77), GPP per area (r = 0.80), agricultural GPP per area (r = 0.77), GPP per capita in pe riod C (r = 0.58), HAI (r = 0.64), income component of HAI (r = 0.73), and negative ly correlated with pov erty incidence (r = 0.64). Correlations with the wealth principle com ponents for periods B and C are also positive (r = 0.62 and 0.68 respectively, p<0.01), and slightly weaker than the same correlations for only irrigation AET. Coastal emergy flows and wealth Emergy in tide and runoff chem ical energy at the coast are special cases, as the inland provinces have values of zero for these variab les. These variables do have some weak positive correlations with the wealth variables. Bi-variate relationships will not be explored further with these two variables, but they will be consider ed in regression analysis, and runoff chemical emergy is a significant part of the emergy RENEW aggregations for coastal provinces. Total renewable emergy flow and wealth Correlations f or the two interpretations of to tal renewable flow are shown in Table E-4. Renew.W, which includes water flows other than irrigation, shows weak positive correlation with GPP per area and GPP per capita (r = 0.43, 0.49; p<0.01). Renew.WI which adds irrigation AET to renew.W, exhibits stronger correlations with GPP per area and GPP per capita ( r = 0.81,

PAGE 95

95 0.78), HAI (0.81), income component of HAI ( r = 0.79), and both PC1s from periods B and C ( r = 0.75, 0.84), all at p<0.01. Renew.WI is also negatively correlated with poverty incidence (r = 0.80, p<0.01). The bivariate plots in Figure 3-34 show the regiona l patterns of correlation and indicate that the correlations do not hold at the regional scale, but must be viewed at the larger sub-national scale to see a correlation. In addi tion, the clumping of the Northeast region on one end of the plot tends to strengthe n the correlation coefficient in th e absence of a consistent linear trend in the relationship for all provinces. Emergy principal components and wealth PC1em, PC2em and PC3em in Table E-4 are the first, second and third principle components from the renewable emergy signature PCA. PC1em (generally correlated to coastal flows) is correlated to all wealth variables with higher strength than any of the other emergy variables. The highest correlations for PC1em are with PC1c (r = 0.88), GPP per capita period C (r = 0.87), GPP per capita period B (r = 0.85) and HAI (r = 0.85). PC2em (water use) appears to have no notable correlation wi th any of the wealth variables when considering the full set of provinces, but is correlated str ongly with population density in period C. However, if the Northeast region is omitted PC2em does show a weak positive correlation with HAi.inc (r = 0.64, p<0.01). The bivariate plots in Figure 3-34 indicate that th e PC1em relationships with wealth occur only outside the Northeast, but remain strong statistically when viewed at the larger scale because of the clumping of the Northeast provinces on one end of the plot. Despite the lower correlation coefficients found with PC2em, the PC2em relationships appear strong when excluding the Northeast, or when looking at the Northeast inde pendently for GPP per area and agricultural GPP per area.

PAGE 96

96 Regression Analysis of Environment and Wealth Based on the correlation analysis and extensive study of a host of bivariate plots, a subset of the enviro nmental and socioeconomic variab les was selected for consideration in the regression analyses. For each dependent variable a best fit model was found for each of 4 categories of independent environmental variable s: water flows in mm/yr, PCs of water flows (PCw), emergy flows in sej/m2/yr and PCs of emergy flows (PCem). Regional dummy variables for Central, East and Northeast, as well as distance to Bangkok were also included for each category during model investigations. Also, because distance to Bangkok has such strong correlations with all of the soci o-economic variables (other than population density), a univariate distance regression was run for each wealth vari able for comparison to models composed of environmental variables. For some of the socioeconomic variables investigated, the environmental variables uniquely accounted for only a small fraction of th e variance relative to regional location (part-R2 < 0.05), though partial-R2s were higher. Details of the regression models will only be shown for the socioeconomic dependent variables for which e nvironment appears to play a larger role (partR2 > 0.05). Gross Provincial Product Per Capita Table 3-11 contains information for the best fit models developed for GPP per cap ita for each of the five dependent variables categorie s (distance to Bangkok as the only independent variable, and the other f our categories of environmental variables), for both time periods B and C (preand post-financial crisis). Sa Kaeo provi nce is an extreme outlier (standardized residual > 3.0) in all models and was omitted. This province was created at the beginning of period C (split off from another province) and it is possible that economic reports may have been in error in the beginning of period C. Using dist2bkk as the only predictor results in an adjusted-R2 of 0.75 for

PAGE 97

97 period C and 0.74 for period B. When including envi ronmental variables in the model for all of the 4 environmental cate gories, dist2bkk is not a significant variable. Instead, the NE regional dummy enters the equation as the significant lo cation-related variable. In addition, because the slope of the relationship between AET and GPP/capita differs for the Northeast region (Figure 332), an interaction term streng thens the model for the water, water PCs and emergy categories (NE*aet.mm, NE*PC3w and NE*aet, respectively), with the exception of the water category for period B. Water flow models for Gross Pr ovincial Product per capita In the water flow category, the stronges t m odel for period C has an adjusted-R2 of 0.83 and includes NE, aet.mm and NE*aet.mm. Model strength as defined by adjusted-R2 and RMSE is almost identical for period B, although the independent variables differ (def.mo and rain.mm are present, rather than NE*aet.mm). In period C, using water PCs in place of individual water flows results in a slightly strong er model with an adjusted-R2 of 0.85, though models are very similar because aet.mm is the dominant variable of PC3w The additional variation in PC3w due to the other water flow variables loading on this PC is presumably responsible for the decrease in model error of about 7% relative to the root mean square error (RMSE) of the water flow model. The water PC model for period B is identical to period C, though slightly weaker in strength. The partial-R2 and part-R2 columns in Table 3-11 provide additional information about the role of each independent variable in explaining variance in the dependent variable. The partial correlation is the correlation betw een the dependent variable and an independent variable when the linear effects of the other independent vari ables in the model have been removed from both. The squared partial correlation (partial-R2) is the portion of Y varian ce estimated by a variable, which is not estimated by the other independe nt variables in the equation. The partial-R2 of aet.mm in describing GPP per capita for peri od C is 0.32, indicating that if the NE and

PAGE 98

98 NE*aet.mm variables were held constant, aet.mm would explain 32% of the variation in GPP per capita. In the water PC model, PC3w estimates 44% of variation in GPP per capita if the other independent variables are held constant. Values are similar for period B. Additionally for period B, def.mo and rain.mm occur in the water flow model and ae t.mm, def.mo and rain.mm each respectively estimate 38%, 25% a nd 23% of GPP per capita variati on if all other variables are hypothetically held constant. The part correlation is the co rrelation between the dependent variable and an independent variable when the linear effect s of the other independent vari ables in the model have been removed only from the independent variable The squared part correlation (part-R2) is related to the change in R-squared when a variable is a dded to an equation, and represents the unique contribution of an independent variable to total variation in the dependent variable that is not shared by any other independent va riable in the equation. For Gca p.C and water flow variables, the part-R2 of aet.mm is 0.08. The interaction term NE*aet.mm adjusts the model further for aet.mm in the Northeast region and has a part-R2 of 0.04. Effectively aet.mm uniquely explains 12% of the variance in Gcap.C, wh ile location in the Northeast re gion uniquely explains 74% of the variance. Results are similar for the water PC category, with part-R2 of 0.74 for the NE and 0.17 for the PC3w terms. For the water flow mode l in period B, NE uniqu ely explains 67% of the variance in GPP per capita, while collectively the water flow variable s uniquely explain 21% of the variance. Results are similar to pe riod C for the water PC category, with part-R2 of 0.76 for the NE and 0.15 for the PC3w terms. Emergy flow models for Gross Provincial Product per capita When emergy flow variables are used for m ode l development, the best fit model for period C has independent variables NE, aet and the in teraction term NE*aet. Because AET in sej/m2/yr is directly proportional to aet.mm, the model is id entical to the water flow category model (Table

PAGE 99

99 3-11). For period B, the best fit model has the same terms, but differs from the water flow model for period B which included def.mo and rain.mm. Pa rtial and part correlations are very similar to the same model for period C. Emergy PC best f it models are the same for periods B and C, both including NE, PC2em and NE*PCem2 as independent variables. For both periods PC2em represents roughly 25-30% of the variation in GPP per capita if NE where held constant, and only about 10% of unique variation in GPP pe r capita. Because AET emergy loads heavily on PC2em, these models are virtually identical to the water flow and emergy flow models in period C, and the emergy flow model in period B. Gross Provincial Product Per Area Table 3-12 contains diag nostics for the best f it models developed for GPP per area for each of the five dependent variable s categories (distance to Bangk ok only, water flows, water PCs, emergy flows and emergy PCs), and time periods B and C (preand post-financial crisis). Sa Kaeo province is an extreme outlier (standardized residual > 3.0) in all models for period C and was omitted due to reasons stated earlier. As w ith GPP per capita, the distance only models are both robust and simple, but adjusted-R2 values are lower relative to models including environmental variables, and re sidual diagnostics are not as favorable, showing less normality and in some cases less randomness in error variability. Water flow models for Gross Pr ovincial Product per area For the water category, the strongest model for period C has an adjusted-R2 of 0.88 and includes NE, aet.mm and rain.mm. Mode l strength as defined by adjusted-R2 and RMSE is slightly higher for period B, possibly due to the addition of sm.mo, wh ich was not a significant term for period C. In period C, the use of water PCs as an alternative to individual water flows results in a slightly strong er model with an adjusted-R2 of 0.89, and all water PCs are included as

PAGE 100

100 terms in the model. The water PC model for peri od B is very similar to period C, and slightly stronger with an adjusted-R2 of 0.91. The partial-R2 of aet.mm in describing GPP per area fo r period C is 0.72, indicating that if the NE and rain.mm variables were held constant aet.mm would e xplain 72% of the variation in GPP per area. If NE and aet.mm were constant, rain.mm would explain 45% of the variation. For period B, soil moisture is an additional term in the model and partial-R2s for NE, aet.mm, rain.mm and sm.mo are 0.86, 0.62, 0.59, and 0.25, respec tively. In the water PC model for period C, PCw2 (shortage), PCw3 (use) and PCw1 (ex cess) respectively estim ate 52%, 44% and 40% of variation in GPP per area if each of the other independent variables is held constant. For Gdens.C and water flow variables, the part-R2s of aet.mm and rain.mm are 0.29 and 0.09. Combined, the water terms in this model uniquely explains 38% of the variance in Gdens.C, while location in the Northeast region uniquely explains 61% of the variance. For period B, water flow terms uniquely explain 29%, while NE explains 50% of the variance in Gdens.B. In the model for the water PC cate gory, collectively the wa ter PCs have a part-R2 of 0.26, and the NE part-R2 is 0.20. Values are similar for peri od B. The low value of the part-R2 for the NE term and water PC terms relative to the other models indica tes a higher portion of shared variance among the NE term and the water PC terms, thus the portions of unique variance explained are lower. Emergy flow models for Gross Provincial Product per area When utiliz ing emergy flow variables for model development, the best fit models for GPP per area in both periods contain independent variables NE, aet and rain. In the case of period C, the model coefficients are identical to the water flow model. For period B, the water flow model additionally includes soil moisture, which allows for an interesting co mparison of the part correlations between the water flow and emergy flow models for this period. The combined part-

PAGE 101

101 R2 for emergy flows is 0.41, and that of the water flows is 0.29. The additional variable in the water flow model causes the combined part-R2 to be lower due to the additional shared variance with the other independent variab les which gets partialled out. Emergy PC best fit models include the same terms for periods B and C, both including NE and PC2em as independent variables. For both periods PC2em represen ts roughly 75% of the variation in GPP per area if NE where held constant, and about 33% of unique variation in GPP per area. Because AET emergy loads heavily on PC em2, these models are very similar to the water flow and emergy flow models. Agricultural Gross Provincial Product Per Area Table 3-13 contains information for the best fit models developed for agricultural GPP per area for each of the five dependent variables ca tegories, and tim e periods B and C (preand postfinancial crisis). Two provinces are extreme outliers (standardized residual > 3.0) in all models for period C. One is Bangkok, which may be an outli er due to its extreme industrial nature, and the other is Trad, a poor coastal province in the East region whic h derives virtually all of its agricultural GPP from fishing-related endeavors. These two provinces are omitted from models in period C in order to obtain more robust equati ons, and because of the apparent reasons for the absence of the variable relationships which ar e occurring within the re st of the provinces. Distance only models have lower adjusted-R2 values than the two prev ious dependent variables, and models for agGdens which include environmen tal variables exhibit better fit than those for GPP per capita, and somewhat weak er fit than those fo r GPP per area. In addition, environmental variable contributions to agGdens variance rela tive to the NE dummy variable contribution are higher than in the cases where GPP per capita and GPP per area are the dependent variables.

PAGE 102

102 Water flow models for agricultural Gross Provincial Product per area For the water category, the strongest model for period C has an adjusted-R2 of 0.83 and includes NE, rain.mm, aet.mm and sm.mo. For pe riod B, independent variables are the same. Model strength as defined by adjusted-R2 and RMSE is slightly higher for period B. In period C, the use of water PCs as an alte rnative to individual water flow s results in a slightly stronger model with an adjusted-R2 of 0.85, and all water PCs are incl uded as terms in the model. The water PC model for period B is nearly identical to period C, with a different order of importance and slightly higher adjusted-R2 of 0.88. The partial-R2s of rain.mm, aet.mm and sm.mo in describing agGdens for period C are 0.56, 0.46, and 0.21 respectiv ely, while part-R2s are 0.20, 0.14 and 0.04. These values are of similar magnitude to the part and partial correlati ons of the water flow variables in the total GPP per area model, but the partial and part-R2s of NE (0.71 and 0.39) are the lowest seen so far, indicating that location may play less of a role in determining agGdens relative to water flows as compared to GPP per capita and GPP per area. The NE dummy variable uniquely accounts for 39% of the total variance in agGdens.C as comp ared to 74% and 61% for Gcap.C and Gdens.C. For period B, NE apparently plays a gr eater role in determining agGdens. The partial-R2s of PCw1 (excess), PCw2 (shortage) and PCw3 (use) in describing agGdens for period C are 0.53, 0.48, and 0.31 respectively, while part-R2s are 0.15, 0.12 and 0.06. Values for period B are similar, and again NE apparently plays a greater role than waterPC flows in determining agGdens as compared to period C. The NE dummy variable uniquely accounts for 13% of the total variance in agGdens.C as compared to 17% for agGdens.B. Emergy flow models for agricultural Gross Provincial Product per area When emergy flow variables are considered in model development, the best fit models for agricultural GPP per area in both time periods contain independent variables NE, aet and rain.

PAGE 103

103 Period C additionally includes irrigation, though th is variable uniquely explains only an additional 4% of the total variation in agGdens.C. Model adjusted-R2 values for emergy flow models are less than those of water flow and water PC models. Of note is the part-R2 value of aet in the emergy flow models for both periods. Aet uniquely accounts fo r 25-28% of the total variance in agGdens as compared to 11-14% for aet.mm in the water flow models, due to the additional variance explained by so il moisture. Including soil moistu re strengthens those models relative to the emergy models, but reduces the apparent unique contribution of aet.mm to the dependent variable variance. The combined part-R2 for emergy flows is 0.36 for both time periods, and that of the water flows is 0.38 for period C and 0.30 for period B. Emergy PC best fit models include the same terms for periods B and C, both including NE and PC2em as independent variables. For bot h periods PC2em uniquely accounts for about 30% of total variation in agricultural GPP per ar ea. Because AET emergy loads heavily on PCem2, these models are very similar to the water flow and emergy flow m odels, though without the extra variance explained by the additional variables included in the other models. Other Wealth Indices Other wealth indices for which regression m odels were investigat ed include the HAI composite index, the income component of HAI (HAI.inc), poverty incidence, and the first principle component of wealth variables for pe riods B and C (PC1.c and PC1.b). In general, individual environmental variables uniquely acco unt for less than 10% of the variance in these wealth variables if the NE dummy variable is included in the model. The region-coded scatter plots in Figure 3-32 suggest that for these wealth variables, th e NE provinces as a group have values that are even further removed from those of the other regions th an is the case for the wealth variables for which regressions have be en shown. Therefore, th e NE dummy variable

PAGE 104

104 dominates the model and the contribution of th e environmental variables is statistically diminished. Rather than trying to account fo r the NE effect within the regression model, regressions were investigated for all provinces outside of the Northeast region for wealth Principal Component 1. Table 3-14 shows three models pred icting period C Principal Component 1 of the wealth variables for the 25 provinces of the Bangkok, Central and East regions. The distance only model for PC1.c excluding the NE has an adjusted-R2 of 0.63. When including water flow variables, aet.mm and rain.mm are significan t (p < 0.001 and p < 0.06) and the adjusted-R2 increases to 0.74. Because of mutual variance sh ared with PC1.c, the independent variables aet.mm, dist2bkk and rain.mm have partial-R2 values of 0.32, 0.15 and 0.27, respectively. In other words, if the NE region is excluded and if distance and rain were held constant, aet.mm would account for 32% of the variance in PC1.c, wh ile if the two water flow variables were held constant, distance to Bangkok would account for 15%. Because of the nature of the Index of Huma n Deprivation (IHD), which is a component of wealth PC1 for period B, no suitable regression models were found for the group of provinces which excludes the Northeast. The IHD scores have very little variance fo r provinces outside the NE, as this index targets level of deprivation rather than level of wealth. When including the NE provinces, the NE dummy variab le and dominates regression equations, with environmental variables accounting for less than 5% of unique variance in PC1.b. When looking only at the Northeast provinces, distance to Bangkok is not a si gnificant predictor, and water flow variables were borderline significant, with p-values for co efficients of about 6-7%, and p-values for the models around 3-5%.

PAGE 105

105 Population Density Relationships between environment and populati on density we re also investigated. Table 3-15 contains the best fit models for period C for the same five categories of independent variables used with the wealth models. Period B re sults were very similar and are not presented. All categories have robust models, except for th e distance-only model. In fact, distance to Bangkok was not a significant predictor for popul ation density. In addition, the NE dummy variable was not significant in any of the cate gories. The water category best-fit model includes aet.mm, rain.mm, irg.mm and sm.mo, and has an adjusted-R2 of 0.86. As with most of the wealth models, aet.mm is the water flow which explai ns the highest fraction of population density variability with a partial-R2 of 0.62 and a part-R2 of 0.21. The water PC model has similar strength and includes all PCs. The emergy flow model contains aet, irg and rain and has an adjusted-R2 of 0.80. Model strength is w eaker than for the water fl ow and water PC models because of the absence of the additional vari ables accounted for in those models. For emergy PCs, the best-fit model includes PC1em (coastal flows) and PC2em (water use) and has an adjusted-R2 of 0.79, similar to the emergy flow model. The province of Singburi in the Central region is a major outlier in the emergy PC model (standardized residual of 3.77). If the emergy PC model is rerun with Si ngburi omitted, the adjusted-R2 increases to 0.86, on par with the water flow and water PC models. Wealth Model Comparisons The socio-economic variables with the strongest overall models, and which also exhibited stronger relationships with e nvironmental variables, include GPP per capita, GPP per area, agricultural GPP per area and popul ation density. In order to co mpare models to the simple distance model, and to evaluate the best mode l for each dependent variable, RMSE differences were calculated. Table 3-16 recaps the model variab les and contains values for the decrease in

PAGE 106

106 RMSE for each model when compared to the simp le distance model, as well as the increase in RMSE when compared to the best-fit model. The best-fit model is defined as the model with the lowest RMSE. Best-fit models a nd models with less th an 10% RMSE increase over the best-fit models are highlighted with asterisks. A RMSE difference of less than 10% roughly equates to adjusted-R2 values that are within a few percent of each other. Though the distance-only models are robust and simple, if environmental terms are included in wealth equations, a nd presence in the NE region is accounted for, model error is reduced 15%-39% for period C and 16-42% for peri od B (Table 3-16). In all cases, the best-fit models were the water flow models and the wa ter PC models. Water PC models had the best absolute values of adjusted-R2 and RMSE, but the water flow models were within 0.01-0.02 of the adjusted-R2 values and within 1-8% of the error values. Differences in the models between period B and period C are slight, with few change s in independent variables included, and only minor differences in adjusted-R2 values. Table 3-16 also shows the sum of the individual part-R2 values for the environmental terms in each model. With the caveat that the part-R2 sum generally decreases with increased number of terms in the model, the part-R2 sum gives a general sense of th e portion of dependent variable variance that is uniquely attributed to the environmental components of the model. Using the highest part-R2 sums for each dependent variable, envir onmental flows unique contribution to variance in GPP per capita, GPP pe r area, agricultural GPP per ar ea and population density is around 17-21%, 29-38%, 30-48% and 75%, respectively.

PAGE 107

107 Table 3-1. Emergy account for Thailand (c.200 0). All flows are on an annual basis. Flow UEV Emergy Em$ # Line item Flow units (sej/unit) (E20 sej) (E6 US$) RENEWABLE FLOWS: 1 Sunlight 3.1E+21 J 1.0E+00 30.8 207 2 Deep heat 8.8E+17 J 5.8E+04 512.2 3430 3 Tide 9.6E+17 J 7.4E+04 709.7 4754 4 Wind 2.8E+17 J 2.5E+03 6.8 45 5 Total water 1179.3 7900 ET, chemical potential 2.8E+18 J 3.1E+04 862.0 5775 Runoff, geopotential 6.7E+17 J 4.7E+04 317.2 2124 6 Waves 5.6E+17 J 5.1E+04 285.6 1913 INTERNAL TRANSFORMATIONS: 7 Agriculture Production 1.5E+18 J 8.6E+04 1242.0 8320 8 Livestock Production 2.3E+16 J 3.7E+06 839.9 5626 9 Fisheries Production 7.8E+15 J 8.4E+06 655.7 4392 10 Fuelwood Production 1.3E+17 J 3.7E+04 48.7 326 11 Indust. Roundwood Prod. 4.0E+16 J 9.2E+04 36.9 247 12 Water extraction 4.3E+17 J 8.1E+04 1051.9 7046 13 Hydroelectricity 2.1E+16 J 2.8E+05 59.5 398 14 Total Electricity 3.0E+17 J 2.9E+05 867.9 5814 INDIGENOUS NONRENEWABLE EXTRACTION: 19 Forestry, net loss 5.8E+16 J 3.8E+04 22.4 150 20 Fisheries, net loss 1.6E+15 J 8.4E+06 133.7 896 21 Water, net loss 0.0E+00 J 8.1E+04 0.0 0.0 22 Topsoil losses (OM) 1.1E+17 J 2.9E+05 330.9 2216 23 Coal 4.4E+17 J 6.6E+04 285.5 1912 24 Natural Gas 7.1E+17 J 6.8E+04 483.6 3239 25 Oil 2.4E+17 J 9.4E+04 230.7 1545 26 Minerals 8.9E+13 g 8.9E+09 8291.2 55545 27 Metals 3.9E+10 g 2.6E+11 99.7 668 IMPORTS: 28 Fuels 1.7E+18 J 9.2E+04 1586.9 10631 29 Metals 9.3E+12 g 1.7E+10 1728.5 11579 30 Minerals 1.0E+12 g 3.5E+09 54.7 366 31 Food & agriculture products 7.1E+16 J 2.7E+05 196.5 1316 32 Livestock, meat, fish 7.3E+15 J 3.7E+06 266.0 1782 33 Plastics & synthetic rubber 1.0E+12 g 1.2E+10 129.9 870 34 Chemicals see notes g, J see notes 390.6 2616 35 Finished products see notes g, J see notes 207.0 1386 36 Machine, transp. equipment 2.7E+10 $ 2.6E+12 713.8 4781 37 Other refined goods 4.3E+09 $ 2.6E+12 113.2 758 38 Electricity 1.1E+15 J 2.9E+05 3.1 21 39 Service in imports 6.1E+10 $ 2.6E+12 1606.5 10762 EXPORTS: 40 Fuels 2.2E+17 J 1.3E+05 291.0 1949 41 Metals 2.9E+12 g 3.0E+10 677.7 4540 42 Minerals 2.1E+13 g 6.2E+09 1057.6 7084 43 Food & agriculture products 3.1E+17 J 8.1E+04 261.2 1750 44 Livestock, meat, fish 7.9E+15 J 8.1E+06 477.3 3197

PAGE 108

108 Table 3-1 continued. Flow Emergy Em$ # Line item Flow units UEV E20 sej E6 $ 45 Plastics & synthetic rubber 2.9E+12 g 1.2E+10 354.4 2374 46 Chemicals see notes g, J varies 130.7 875 47 Finished products see notes g, J varies 329.5 2207 48 Machine, transp. equipment 2.9E+10 $ 1.5E+13 4312.1 28888 49 Other refined goods 1.3E+10 $ 1.5E+13 1893.3 12683 50 Electricity 5.7E+13 J 2.9E+05 2.1 14 51 Service in exports 6.9E+10 $ 1.5E+13 10261.6 68745 52 Tourism 7.5E+09 $ 1.5E+13 1117.9 7489 Note: Table footnotes and raw data and UEV s ources provided in Table C-2 in Appendix C. Table 3-2. Summary flow s for Thailand (c. 2000). Code Summary f low Calculation Units Value R Renewable use items 5 + 3 sej 1.9E+23 N Total non-renewable extraction sum of items 19 to 27 sej 9.9E+23 N0 Dispersed NR usesum of items 19 to 22 sej 4.9E+22 N1+N2 Concentrated NR prodsum of items 23 to 27 sej 9.4E+23 N1 Concentrated NR useN N0 N2 sej 8.9E+23 N2 Total export w/out usesee notes sej 4.7E+22 N2(m) Mineral, metal N2see notes sej 0.0E+00 N2(f) Fuels N2see notes sej 4.7E+22 F(i) Fuel, metal, mineral imports items 28 + 29 + 30 sej 3.4E+23 G(i) Rest of imports + electricity sum of items 31 to 38 sej 2.0E+23 I $ spent for imports item 39, $ value $ 6.1E+10 P2I Services imported item 39, sej value sej 1.6E+23 F(e) Fuel, metal, mineral exports items 40 + 41 + 42 sej 2.0E+23 G(e) Rest of exported goods + elec sum of items 43 to 50 sej 7.8E+23 E $ received for exports item 51, $ value $ 6.9E+10 P1E Services exported item 51, sej value sej 1.0E+24 X GDP Gross domestic product $ 1.2E+11 P2 World emergy:$ (EMR) World U/World GDP sej/$ 2.6E+12 P1 Thailand emergy:$ (EMR) U/GDP sej/$ 1.5E+13

PAGE 109

109 Table 3-3. Summary emergy indices for Thailand (c.2000). Item Index name Calculation Value 1 Renewable emergy flow R 1.9E+23 2 Flow from indigenous nonrenew reserves N 9.9E+23 3 Flow of imported emergy F(i)+G(i)+P2I 7.0E+23 4 Total emergy inflows R+N+F(i)+G(i)+P2I 1.9E+24 5 Total emergy use, U R+N0+ N1+F(i)+G(i)+P2I 1.8E+24 6 Total exported emergy F(e)+G(e)+P1E 2.1E+24 7 Export to Imports, without services [F(e)+G(e)] / [F(i)+G(i)] 1.8 8 Fuel, metal, mineral import/other imported goodsF(i) / G(i) 1.7 9 Fuel, metal, mineral export/other exported goodsF(e) / G(e) 0.26 10 % raw material export ("w/out use") N2 / total export 0.05 11 Fraction of use, indigenous sources (N0+N1+R) / U 0.62 12 Fraction of use, locally renewable R / U 0.10 13 Fraction of use, dispersed (R+N0) / U 0.13 14 Fraction of use, electricity (electricity emergy) / U 0.05 15 Fraction of use, fuel (Fuel production + import export)/U 0.23 16 Fraction of use, purchased [F(i)+G(i)+P2I] / U 0.38 17 Fraction imported service P2I / U 0.09 18 Fraction of use, soil loss Soil loss / U 0.02 19 Fraction of use, fishery depletion Fishery, net loss / U 0.01 20 Fraction of use, forestry net loss Forestry, net loss / U 0.00 21 Fraction capital stock depletion (Soil, Fish, Forestry net losses) / U 0.03 22 Ratio of concentrated to rural [F(i)+G(i)+P2I+N1] / (R+N0) 6.7 23 Use per unit area, Empower Density U / Area (square meter) 3.6E+12 24 Use per capita U / population 3.0E+16 25 Renewables per capita R / population 3.1E+15 26 Non-renewables per capita NR / population 1.6E+16 27 Fuel use per capita Fuel / population 4.2E+15 28 Renewable carrying capacity (R/U)*(population) 6.3E+06 29 Ratio of use to GDP, emergy/dollar ratio P1=U / GDP 1.5E+13 30 Investment Ratio, imports/indigenous use [F(i)+G(i)+P2I] / (R+N0+N1) 0.6 31 Environmental Loading Ratio, (NR use)/R [F(i)+G(i)+P2I+N0+N1] / R 8.7 32 Yield ratio, total use / imports U / [F(i)+G(i)+P2I(i)] 2.6 33 ESI, Emergy sustainability index EYR / ELR 0.3 See Table 3-2 for definitions of letter codes.

PAGE 110

110 Table 3-4. Data for sample basins in the stream order transformity m odel. Basin Order Discharge Area Avg. rain Runoff Rain energy Rain emergy TRF ID m3/sec km2 mm/yr fraction Joules/yr sej/yr sej/J 1 2 0.77 67 1054 0.34 3.3E+14 1.0E+19 8.9E+04 2 2 1.02 77 1104 0.38 4.0E+14 1.2E+19 8.1E+04 3 3 3.91 471 1118 0.23 2.5E+15 7.6E+19 1.3E+05 4 3 2.5 359 1046 0.21 1.8E+15 5.4E+19 1.5E+05 5 3 2.27 323 1056 0.21 1.6E+15 4.9E+19 1.5E+05 6 3 4.18 326 929 0.44 1.4E+15 4.4E+19 7.0E+04 7 3 3.66 263 1383 0.32 1.7E+15 5.3E+19 9.7E+04 8 3 2.62 219 1162 0.32 1.2E+15 3.7E+19 9.4E+04 9 4 6.3 1281 1095 0.14 6.6E+15 2.0E+20 2.2E+05 10 4 15.77 1500 1039 0.32 7.4E+15 2.3E+20 9.6E+04 11 4 8.13 945 1135 0.24 5.1E+15 1.6E+20 1.3E+05 12 4 17.17 1920 1119 0.25 1.0E+16 3.1E+20 1.2E+05 13 4 8.6 703 1118 0.35 3.7E+15 1.1E+20 8.9E+04 14 4 35.59 2900 1377 0.28 1.9E+16 5.8E+20 1.1E+05 15 5 40.48 3341 1540 0.25 2.4E+16 7.4E+20 1.2E+05 16 6 88.29 23200 1058 0.11 1.2E+17 3.5E+21 2.7E+05 17 6 279.23 25879 1286 0.26 1.6E+17 4.8E+21 1.2E+05 18 7 597.86 106740 1152 0.15 5.8E+17 1.8E+22 2.0E+05 C.P. 7 637.38 121834 1096 0.15 6.3E+17 1.9E+22 2.0E+05 Mekong 8 3625.36 268000 C.P. = Upper Chao Phraya Basin and Mekong = Upper Mekong Basin. Delineation of the Mekong Basin was not performed in this study, thus some parameters were not calculated for this basin.

PAGE 111

111 Table 3-5. Summary statistics for provincial emergy flows, mean in units of sej/m 2/yr. Variable N Mean SE Mean StDev CofV SumofSq Skew Kurtosis rain 45 1.9E+11 8.2E+09 5.5E+10 2.8E+01 1.8E+24 2.32 7.12 aet 45 1.4E+11 2.1E+09 1.4E+10 9.7E+00 9.5E+23 -0.68 0.38 ROgeo 45 2.4E+10 4.0E+09 2.7E+10 1.1E+02 5.9E+22 1.76 2.48 ROchem (all) 45 5.3E+10 2.0E+10 1.3E+11 2.5E+02 9.1E+23 2.48 4.52 ROchem (coastal) 11 2.2E+11 6.0E+10 2.0E+11 9.1E+01 9.1E+23 0.18 -2.35 rad 45 5.1E+09 1.4E+07 9.4E+07 1.8E+00 1.2E+21 0.24 -0.32 wind 45 8.5E+09 9.3E+08 6.2E+09 7.4E+01 4.9E+21 2.14 6.27 heat 45 1.8E+10 2.7E+08 1.8E+09 9.7E+00 1.6E+22 0.33 0.00 irg 45 1.1E+11 2.1E+10 1.4E+11 1.3E+02 1.4E+24 1.35 0.46 tide (all) 45 9.0E+10 2.4E+10 1.6E+11 1.8E+02 1.5E+24 1.67 1.45 tide (tidal only) 14 2.9E+11 4.5E+10 1.7E+11 5.8E+01 1.5E+24 0.00 -1.27 aetirg 45 2.5E+11 2.1E+10 1.4E+11 5.5E+01 3.7E+24 1.35 0.41 renew.W 45 3.8E+11 1.1E+11 7.1E+11 1.8E+02 2.9E+25 3.57 11.55 renew.WI 45 4.9E+11 1.1E+11 7.5E+11 1.5E+02 3.5E+25 3.4 10.78 SE Mean = standard error of the mean, StDev = standard deviation, CofV = coefficient of variance, SumofSq = sum of squares. See Table 2-1 for variable descriptions. Table 3-6. Eigenvalues and loadings for pr incipal components for water flows (PCw). PC1w PC2w PC3w Eigenvalue 3.12 1.77 1.30 Proportion 0.45 0.25 0.19 Cumulative 0.45 0.70 0.88 Variable loadings: strmdens -0.21 -0.59 -0.35 rain.mm 0.52 -0.12 -0.02 aet.mm 0.18 -0.35 0.71 irg.mm -0.36 -0.49 -0.25 ro.mm 0.48 0.04 -0.36 sm.mo 0.34 -0.52 0.17 def.mo -0.42 0.04 0.39 Table 3-7. Eigenvalues and loadings for pr incipal components for emergy flows (PCe m). PC1em (original)PC1em PC2em PC3em Eigenvalue 2.722.72 1.75 1.23 Proportion 0.390.39 0.25 0.18 Cumulative 0.390.39 0.64 0.82 Variable loadings: aet 0.09-0.09 0.63 0.24 irg -0.430.43 0.17 -0.54 ROgeo 0.07-0.07 -0.67 -0.10 rad 0.44-0.44 0.34 -0.25 wind -0.490.49 0.05 0.19 heat 0.47-0.47 -0.13 0.41 ROchem -0.390.39 0.01 0.61 Note that PC1em will have the opposite sign of the original PC in further analysis.

PAGE 112

112 Table 3-8. Summary statistics for socio-economic wealth variables. Variable N Mean SE MeanStDevVarianceCofVSumof Sq. SkewKurtosis pdens.B 42 262 895773.3E+05219.91.7E+07 5.7 34.7 Gden.B 42 1.6E+07 9.6E+066.2E+073.9E+15381.21.7E+17 6.0 37.2 agGden.B 42 9.6E+05 1.9E+051.2E+061.5E+12126.39.9E+13 3.3 11.7 G.cagr.B 42 7.2 0.42.87.638.12518.1 1.0 0.6 Gcap.B 42 28839 4261276127.6E+0895.76.6E+10 1.8 2.5 IHD.comp 45 0.33 0.020.160.0247.85.93 0.4 -1.1 IHD.inc 45 0.34 0.040.280.0881.58.81 0.4 -0.8 IHD.health 45 0.36 0.020.150.0241.56.73 0.0 -0.3 IHD.edu 45 0.37 0.030.200.0454.47.99 0.8 0.9 IHD.emp 45 0.33 0.040.290.0886.48.54 0.5 -0.5 IHD.hous 45 0.26 0.030.200.0477.04.74 0.2 -1.1 IHD.tracom 45 0.34 0.030.170.0349.56.41 1.0 1.4 IHD.congood 45 0.35 0.050.360.13102.611.19 0.6 -1.3 IHD.wom 45 0.33 0.020.140.0242.05.85 0.6 -0.4 pdens.C 45 300 956344.0E+05211.42.2E+07 5.5 32.7 Gden.C 45 2.9E+07 1.3E+078.7E+077.7E+15305.33.7E+17 5.2 30.0 agGden.C 45 8.9E+05 1.3E+058.7E+057.6E+1198.66.9E+13 3.3 14.3 G.cagr.C 45 1.4 0.42.35.5171.3327.4 0.9 2.3 Gcap.C 45 52744 9129612423.8E+09116.12.9E+11 2.1 3.7 HDI 45 0.71 0.010.060.008.322.57 0.6 -0.9 HAI.comp 45 0.60 0.010.060.0010.516.11 -0.2 -1.2 HAI.health 45 0.67 0.010.070.0111.020.39 -0.2 -0.6 HAI.edu 45 0.51 0.010.060.0011.811.85 0.2 -0.3 HAI.empl 45 0.59 0.020.120.0120.416.46 -0.2 0.8 HAI.inc 45 0.50 0.020.160.0332.512.26 -0.2 -0.8 HAI.hous 45 0.68 0.020.150.0222.621.71 -0.4 -1.3 HAI.famcom 45 0.66 0.010.080.0111.719.60 -0.3 -0.7 HAI.tracom 45 0.54 0.020.130.0223.613.89 0.2 -0.2 HAI.part 45 0.62 0.010.080.0113.617.43 -2.2 5.5 percapinc00 45 3138 22815282.3E+0648.75.5E+08 1.4 2.4 percapexp00 45 2310 15310241.0E+0644.32.9E+08 1.3 1.5 pov.incid.00 45 16 21523997.121902 0.9 -0.4 dist2bk 45 217 261712937278.93416479 0.5 -1.3SE Mean = standard error of the mean, StDev = standard deviation, CofV = coeffici ent of variance, SumofSq = sum of squares. See Table 2-2 for variable descriptions.

PAGE 113

113 Table 3-9. Eigenvalues and loadin gs for principal components of wealth measures in period B. PC1b (original)PC1b* PC2b PC3b Eigenvalue 5.655.651.67 1.01 Proportion 0.510.510.15 0.09 Cumulative 0.510.510.67 0.76 Variable loadings: Gdens.B -0.34 0.34 -0.23 0.28 Gcap.B -0.37 0.37 -0.18 0.15 G.cagr.B -0.21 0.21 -0.58 -0.20 IHD.inc 0.29 -0.29 0.30 0.25 IHD.health 0.29 -0.29 -0.27 -0.39 IHD.edu 0.30 -0.30 -0.38 0.05 IHD.emp 0.31 -0.31 -0.11 0.46 IHD.hous 0.29 -0.29 0.13 -0.09 IHD.tracom 0.27 -0.27 -0.39 -0.29 IHD.congood 0.38 -0.38 0.07 -0.10 IHD.wom 0.24 -0.24 -0.29 -0.64 *PC1b has the inverse sign of the originally co mputed PC and is used in further analysis. Table 3-10. Eigenvalues and loadin gs for principal components of wealth measures in period C. PC1c PC2c PC3c Eigenvalue 9.43 1.10 0.90 Proportion 0.67 0.08 0.06 Cumulative 0.670.75 0.82 Variable loadings: Gden.C 0.30 -0.03 0.10 Gcap.C. 0.31 -0.07 0.11 G.cagr.C 0.11 -0.69 0.53 percapexp 0.31 0.13 -0.04 pov.incid -0.30 0.03 0.12 HAI.health 0.23 0.22 -0.18 HAI.edu 0.25 -0.25 -0.12 HAI.empl 0.25 0.05 0.36 HAI.inc 0.30 -0.04 -0.13 HAI.hous 0.27 -0.12 -0.34 HAI.famcom -0.26 0.15 0.25 HAI.tracom 0.30 0.05 -0.19 HAI.part -0.19 -0.57 -0.47 HDI 0.29 0.13 0.23

PAGE 114

114 Table 3-11. Best fit models fo r Gross Provincial Product (GPP) per capita for five independent variable categories and 2 tim e periods, using standardized values for all variables. N = 44 for period C and N = 42 for period B. Response IV category Predictor Coef.t (p-value) Adj-R2RMSE Partial-R2Part-R2 Period C: Gcap.C Distance dist2bkk -0.87-11.6 (0.00)0.75 0.50 na na Gcap.C Water NE -1.78-13.44 (0.00)0.83 0.42 0.83 0.74 aet.mm 0.334.41 (0.00) 0.32 0.08 NE*aet.mm -0.46-3.06 (0.00) 0.19 0.04 Gcap.C Water PCs NE -2.05-12.36 (0.00)0.85 0.39 0.85 0.74 PC3w -0.60-5.55 (0.00) 0.44 0.11 PC3w*NE 0.584.08 (0.00) 0.29 0.06 Gcap.C Emergy NE Model is identical to the Water category aet NE*aet Gcap.C Emergy PCs NE -1.77-13.17 (0.00)0.82 0.43 0.81 0.74 PC2em 0.314.10 (0.00) 0.30 0.07 NE*PC2em -0.41-2.56 (0.01) 0.14 0.03 Period B: Gcap.B Distance dist2bkk -0.86-10.76 (0.00)0.74 0.51 na na Gcap.B Water NE -1.72-12.24 (0.00)0.82 0.43 0.81 0.67 aet.mm 0.364.76 (0.00) 0.38 0.05 def.mo -0.31-3.52 (0.00) 0.25 0.10 rain.mm -0.32-3.35 (0.00) 0.23 0.06 Gcap.B Water PCs NE -2.03-12.82 (0.00)0.81 0.43 0.81 0.76 PC3w -0.58-4.87 (0.00) 0.38 0.11 PC3w*NE 0.493.10 (0.00) 0.20 0.04 Gcap.B Emergy NE -1.75-11.43 (0.00)0.78 0.47 0.77 0.71 aet 0.323.77 (0.00) 0.27 0.08 NE*aet -0.36-2.12 (0.04) 0.11 0.03 Gcap.B Emergy PCs NE -1.73-11.30 (0.00)0.77 0.48 0.77 0.71 PC2em 0.303.62 (0.00) 0.26 0.07 NE*PC2em -0.36-2.02 (0.04) 0.10 0.02 IV = independent variable, Coef. = beta coefficient, t = t-statistic, Adj-R2 = adjusted R2. See Tables 2-4 and 2-5 for variable descriptions.

PAGE 115

115 Table 3-12. Best fit models for GPP per area for five independent variable categories and 2 time periods. N = 44 for period C, and N = 42 for period B. Response IV category Predictor Coef.t (p-value) Adj-R2RMSE Partial-R2Part-R2 Period C: Gdens.C Distance dist2bkk -0.87-12.62 (0.00)0.770.48 nana Gdens.C Water NE -1.61-14.55 (0.00)0.880.35 0.840.61 aet.mm 0.5810.13 (0.00) 0.720.29 rain.mm -0.33-5.75 (0.00) 0.450.09 Gdens.C Water PCs NE -1.43-9.47 (0.00)0.890.33 0.710.22 PC2w -0.34-6.39 (0.00) 0.520.10 PC3w -0.35-5.41 (0.00) 0.440.07 PC1w 0.315.09 (0.00) 0.400.06 Gdens.C Emergy NE -1.61-14.55 (0.00)0.880.35 0.840.61 aet 0.5810.13 (0.00) 0.720.29 rain -0.33-5.75 (0.00) 0.450.09 Gdens.C Emergy PCs NE -1.68-15.61 (0.00)0.880.35 0.860.69 EMpc2 0.5610.51 (0.00) 0.720.31 Period B: Gdens.B Distance dist2bkk -0.88-12.11 (0.00)0.780.46 na na Gdens.B Water NE -1.51-14.81 (0.00)0.910.30 0.860.50 aet.mm 0.477.83 (0.00) 0.620.14 rain.mm -0.44-7.23 (0.00) 0.590.12 sm.mo 0.243.44 (0.00) 0.250.03 Gdens.B Water PCs NE -1.36-9.53 (0.00)0.910.30 0.720.20 PC2w -0.36-7.43 (0.00) 0.610.12 PC3w -0.33-5.63 (0.00) 0.460.07 PC1w 0.335.54 (0.00) 0.460.07 Gdens.B Emergy NE -1.57-13.76 (0.00)0.880.35 0.840.57 aet 0.5910.37 (0.00) 0.740.32 rain -0.33-5.61 (0.00) 0.460.09 Gdens.B Emergy PCs NE -1.64-15.45 (0.00)0.890.33 0.860.66 EMpc2 0.5711.23 (0.00) 0.770.35 IV = independent variable, Coef. = beta coefficient, t = t-statistic, Adj-R2 = adjusted R2. See Tables 2-4 and 2-5 for variable descriptions.

PAGE 116

116 Table 3-13. Best fit models for agricultural GPP per area for five independent variable categories and 2 time periods. N = 43 for period C and N = 40 for period B. Response IV category Predictor Coef.t (p-value) Adj-R2RMSE Partial-R2Part-R2 Period C: agGden.C Distance dist2bkk -0.77-7.86 (0.00)0.580.65 nana agGden.C Water NE -1.34-9.65 (0.00)0.830.42 0.710.39 rain.mm -0.62-6.93 (0.00) 0.560.20 aet.mm 0.485.73 (0.00) 0.460.14 sm.mo 0.303.17 (0.03) 0.210.04 agGden.C Water PCs NE -1.06-6.04 (0.00)0.850.40 0.490.13 PC1w 0.506.52 (0.00) 0.530.15 PC2w -0.38-5.86 (0.00) 0.480.12 PC3w -0.31-4.12 (0.00) 0.310.06 agGden.C Emergy NE -1.06-5.47 (0.00)0.820.45 0.800.14 aet 0.557.19 (0.00) 0.530.25 rain -0.37-3.87 (0.00) 0.310.07 irg 0.292.76 (0.01) 0.310.04 agGden.C Emergy PCs NE -1.55-9.62 (0.00)0.740.52 0.690.58 EMpc2 0.607.01 (0.00) 0.550.30 Period B: agGden.B Distance dist2bkk -0.81-8.87 (0.00)0.660.59 nana agGden.B Water NE -1.46-12.60 (0.00)0.880.35 0.820.47 rain.mm -0.47-6.72 (0.00) 0.560.14 aet.mm 0.415.94 (0.00) 0.500.11 sm.mo 0.324.05 (0.00) 0.310.05 agGden.B Water PCs NE -1.26-6.96 (0.00)0.880.34 0.620.17 PC2w -0.39-5.27 (0.00) 0.580.14 PC1w 0.35-2.41 (0.02) 0.420.08 PC3w -0.294.73 (0.00) 0.330.05 agGden.B Emergy NE -1.58-11.31 (0.00)0.820.43 0.780.58 aet 0.537.78 (0.00) 0.620.28 rain -0.29-4.12 (0.00) 0.310.08 agGden.B Emergy PCs NE -1.61-12.26 (0.00) 0.830.41 0.790.62 EMpc2 0.548.60 (0.00) 0.660.31 IV = independent variable, Coef. = beta coefficient, t = t-statistic, Adj-R2 = adjusted R2. See Tables 2-4 and 2-5 for variable descriptions.

PAGE 117

117 Table 3-14. Distance and water flow models for wealth PC1 for time period C, omitting provinces in the NE region. N = 25. Response IV category Predictor Coef.t (p-value) Adj-R2RMSE Partial-R2Part-R2 Period C: PC1.c Distance dist2bkk -0.76-6.65 (0.00)0.630.33 nana PC1.c Water aet.mm 0.253.17 (0.00)0.740.28 0.320.11 dist2bkk -0.41-2.79 (0.01) 0.150.04 rain.mm -0.12-1.94 (0.06) 0.270.09 Table 3-15. Best fit models for population density for five independent variable categories and 1 time period (period C). Response IV category Predictor Coef.t (p-value) Adj-R2RMSE Partial-R2Part-R2 popdens distance dist2bkk (relationship not significant) popdens Water aet.mm 0.608.15 (0.00)0.860.37 0.620.21 rain.mm -0.51-6.23 (0.00) 0.490.12 irg.mm 0.415.96 (0.00) 0.470.11 sm.mo 0.374.12 (0.00) 0.300.05 popdens Water PCs PC1w 0.6111.25 (0.00)0.870.36 0.760.37 PC2w -0.65-11.97 (0.00) 0.780.42 PC3w -0.30-5.62 (0.00) 0.440.09 popdens Emergy aet 0.7110.07 (0.00)0.800.45 0.710.46 irg 0.567.34 (0.00) 0.570.25 rain -0.24-3.02 (0.00) 0.180.04 popdens Emergy PCs EMpc2 0.8011.67 (0.00)0.790.46 0.760.64 EMpc1 0.405.76 (0.00) 0.440.16 popdens Emergy PCs EMpc2 0.8515.04 (0.00)0.860.37 0.850.71 (singburi omitted) EMpc1 0.396.93 (0.00) 0.540.15 IV = independent variable, Coef. = beta coefficient, t = t-statistic, Adj-R2 = adjusted R2. See Tables 2-4 and 2-5 for variable descriptions.

PAGE 118

118 Table 3-16. Comparison of regressi on models using root mean s quare error (RMSE) differences and part-R2sums. Y IV Variable category Independent variables (IV) adjR2RMSE RMSE decrease compared to dist model RMSE increase compared to best model Part-R2 sum for env. IVs Gcap.C location dist2bk 0.75 0.50 dist 0.28 na water NE, aet.mm, NE*AET 0.83*0.42* -0.15* 0.08* 0.12* PCw NE, PC3w, NE*PC3w 0.85*0.39* -0.22* Best* 0.17* PCem NE, PC2em, NE*PC2em 0.82 0.43 -0.13 0.11 0.10 Gcap.B location dist2bk 0.74 0.51 dist 0.19 na water NE, aet.mm, def.mo, rain.mm 0.82*0.43* -0.16* best* 0.21* PCw NE, PC3w, NE*PC3w 0.81*0.43* -0.16* 0.01* 0.15* emergy NE, aet, NE*aet 0.78 0.47 -0.08 0.10 0.11 PCem NE, PC2em, NE*PC2em 0.77 0.48 -0.06 0.12 0.09 Gden.C location dist2bk 0.77 0.48 dist 0.46 na water NE, aet.mm, rain.mm 0.88*0.35* -0.26* 0.08* 0.38* PCw NE, PC2w, PC3w, PC1w 0.89*0.33* -0.32* best* 0.23* Pcem NE, PC2em 0.88 0.35 -0.26 0.08 0.31 Gden.B location dist2bk 0.78 0.46 dist 0.55 na water NE, aet.mm, rain.mm, so.mo 0.91*0.30* -0.34* 0.01* 0.29* PCw NE, PC2w, PC3w, PC1w 0.91*0.30* -0.35* best* 0.26* emergy NE, aet, rain 0.88 0.35 -0.25 0.17 0.41 PCem NE, PC2em 0.89 0.33 -0.29 0.10 0.35 agGden.C location dist2bk 0.58 0.65 dist 0.64 na water NE, rain.mm, aet.mm, sm.mo 0.83*0.42* -0.35* 0.08* 0.48* PCw NE, PC1w, PC2w, PC3w 0.85*0.39* -0.39* best* 0.33* emergy NE, aet, rain, irg 0.82 0.45 -0.30 0.15 0.36 PCem NE, PC2em 0.74 0.52 -0.20 0.31 0.30 agGden.B location dist2bk 0.66 0.59 dist 0.72 na water NE, rain.mm, aet.mm, sm.mo 0.88*0.35* -0.41* 0.01* 0.30* PCw NE, PC2w, PC1w, PC3w 0.88*0.34* -0.42* best* 0.27* emergy NE, aet, rain 0.82 0.43 -0.27 0.25 0.36 PCem NE, PC2em 0.83 0.41 -0.30 0.20 0.31 popd.c (distance not significant) water aet.mm,rain.mm, irg.mm, sm.mo 0.86*0.37* na 0.04* 0.49* PCw PCw1, PCw2, PCw3 0.87*0.36* na best* 0.75* emergy aet, irg, rain 0.80 0.45 na 0.25 0.80 PCem PC2em, PC1em 0.79 0.46 na 0.27 0.86 Best model is the model with the lowest RMSE for each dependent variable. See Tables 2-4 and 2-5 for variable descriptions. Asterisks indicate best fit models. Part-R2 sum is for environmental variables in the respective model.

PAGE 119

119 Figure 3-1. Systems diagram of material and energy flows in Thailand, with water flows shown in blue.

PAGE 120

120 Figure 3-2. Aggregated systems diagram for the Thailand economy with the standard national three-arm diagram (circa 2000). Source: Sweeney et al., 2006.

PAGE 121

121 Figure 3-3. Bar graphs of sel ected emergy indices for 142 count ries, highlighting the relative placement of Thailand. A) Renewable fraction o f total use, B) log of total use per area and C) log of total use per cap ita. Source: Sweeney et al., 2006.

PAGE 122

122 C B A Figure 3-4. Systems diagrams representing three general provincial types in Thailand. A) a Northeast province, B) a Central agricultu ral p rovince and C) a coastal industrial province. Components in yellow and water fl ows in blue highlight some items that vary among provinces (presence of tide, irrigation infrastructure, magnitudes of water flows and relative magnitudes of ag ricultural and indus trial systems).

PAGE 123

123 Figure 3-5. Raster coverages of average annual renewable emergy flows. A) solar radiation, B) wind, C) heat flow, D) geopotential of rainfall runoff, E) chem ical potential of rain and F) chemical potential of AET.

PAGE 124

124 Figure 3-6. Typical Horton an alysis (Horton, 1945) relati onships for the 19 evaluated watersheds. A) discharge as a f unction of watershed area with R2 = 0.98, B) discharge as a function of stream order with R2 = 0.97 and C) watershed area as a function of stream order with R2 = 0.98. y = 23895x + 32695 R = 0.9418 0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 012345678 Stream order y = 23735x + 36446 R = 0.52 0.0E+00 5.0E+04 1.0E+05 1.5E+05 2.0E+05 2.5E+05 3.0E+05 3.5E+05 012345678Stream water transflormity, sej/JStream orderB A Figure 3-7. Stream transformities versus stream order for sa mple basins. A), calculated stream transformities versus stream order and B) average stream transformities by stream order. Note, the point at stream order ze ro is the transformity for rainfall.

PAGE 125

125 Figure 3-8. GIS coverages of estimated irrigation volume per year. A) raster image of applied volum e per area in cubic meters/km2 with basins overlay and B) and average depth of evapotranspired irrigation water per province in mm.

PAGE 126

126 Figure 3-9. Intermediary coverages generated by the emergy irrigation model which assigns stream order transform ity to irrigated cells, and final coverage of the estimated emergy value of irrigation water used. A) st ream order, B) distance from perennial streams, C) allocated stream order, D) em ergy of irrigation evapotranspired per year.

PAGE 127

127 Figure 3-10. Provincial maps of renewable emer gy flows. A) solar radi ation, B) wind, C) heat flow, D) runoff geopotential, E) tide and F) chemical pote ntial of rain runoff to the coast.

PAGE 128

128 Figure 3-11. Provincial maps of irrigation emergy flow and selected emergy flow aggregates. A) irrigation evapotranspired, B) rain and irrigation evapotra nspired, C) total renewable flow, calculated without incl uding irrigation (ren ew.W) and D) total renewable flow, calculated w ith irrigation AET (renew.WI).

PAGE 129

129 Figure 3-12. Box plots of provincial rene wable emergy flows be fore transfor mation. Figure 3-13. Box plot of provincial renewable emergy flows afte r transforma tions detailed in Table 2-1. Note that for display purposes only, RO.chem and tide are distributions for only those provinces with non-zero values. R O c h e m l o g i r g l o g h e a t l o g w i n d l o g r a d l o g r a i n l o g 16 15 14 13 12 11 10 9 Renew flows, transformed data t i d e s q r t R O g e o s q r t 2E6 1.5E6 1E6 5E5 0 r e n e w W I n i r e n e w W n i a e t i r g n i 0 -1E-12 -2E-12 -3E-12 -4E-12 -5E-12 -6E-12 -7E-12 -8E-12 r e n e w W I r e n e w W t i d e R O c h e m 4000 3000 2000 1000 0 a e t i r g i r g h e a t w i n d r a d R O g e o a e t r a i n 600 500 400 300 200 100 0 Renew flows, E9 sej/m^2/yr

PAGE 130

130 Figure 3-14. Box plots of emergy variables. A) rain, B) aet, C) irg, D) ROgeo, E) ROchem F) tide, G) renew.W and H) renew.WI. Plot s are aggregated by region: Bangkok (bkk), Central (c), East (e) and Northeast (n e). Suffixes on variable names indicate transformation performed on the em ergy density measured in sej/m2/yr.

PAGE 131

131 Figure 3-15. Bivariate plots of AET and rain fall. A) including all provinces and B) only provinces with rainfall less than 1450m m and excluding Kanchanaburi.province. Figure 3-16. Matrix bivariate plot of aet.mm, irg.mm, aetirg.mm, irg (emergy), and aetirg (emergy). The legend denotes regions Ba ngkok (bkk), Central (c), E ast (e) and Northeast (ne). All variable s are transformed as noted in Table 2.4, thus units and scales vary.

PAGE 132

132 Figure 3-17. Box plot of standardized environm ental variable s (z-scores) used in principal components analysis (PCA). Figure 3-18. Loading and score plots for prin cipal components PC1w a nd PC2w of the water variable PCA. A) Loading plot and B) score plot.

PAGE 133

133 Figure 3-19. Loading and score plots for principal components PC1em and PC2em of the emergy variable PCA. A) Load ing plot and B) score plot.

PAGE 134

134 Figure 3-20. Provincial maps of socioeconomic measures for period C. A) population density, B) gross provincial product (GPP) per capita, C) GPP density, D) compound annual growth rate (CAGR) and E) agricultural GPP density.

PAGE 135

135 Figure 3-21. Provincial maps of wealth in dicators. A) HAI composite, B) HAI income component, C) per capita expenditu re and D) poverty incidence.

PAGE 136

136 Figure 3-22. Box plots of socio-economic vari ables across time periods A) Gross provincial product (GP P) per area, including agricultural fraction, B) GPP per area transformed, C) GPP per capita, D) GPP per capita transformed, E) GPP average growth rate (no transformation applied). Time periods: B (1981-1993) and C (1996-2003). GPP is in 1988 baht.

PAGE 137

137 Figure 3-23. Box plots of transformed socio-ec onomic variables by region and time period. A) Gross provincial product (GPP) per area (trans form ed), B) GPP per area, agricultural fraction (transformed), C) GPP per capit a (transformed), and D) compound annual growth rate.

PAGE 138

138 Figure 3-24. Box plot diagrams of the Index of Human Deprivation and its com ponents. A) IHD composite index, B) income, C) health, D) education, E) employment, F) housing and living conditions, G) transportation and communication, H) consumer goods and I) womens rights. Regions along the x-axis ar e Bangkok (bkk), Central (c), East (e) and Northeast (ne).

PAGE 139

139 Figure 3-25. Box plot diagrams of the Human Achievement Index and its com ponents. A) HAI composite index, B) income, C) health, D) education, E) employment, F) housing and living conditions, G) transportation and communication, H) family and community and I) participation. Regions along the x-axis are Bangkok (bkk) Central (c), East (e) and Northeast (ne).

PAGE 140

140 Figure 3-26. Bivariate plots for pr ovincial values of th ree alternative wealth indices, color coded by region. A) IHD versus HAI and B) HDI versus HAI. Figure 3-27. Regional box plots for the variables income, expenditure and poverty. A ) per capita income, B) per capita expenditu re and C) poverty incidence.

PAGE 141

141 Figure 3-28. Bivariate plots of provincial values for GPP per cap ita versus the wealth indices HDI and HAI for time period C. A) Gcap.C versus Hum an Development Index and B) Gcap.C versus Human Achievement Index. Regression lines are shown by region and provinces are color-coded by region. Figure 3-29. Box plot of standardized socio-economic variab les (z-score s) used in principal components analysis (PCA).

PAGE 142

142 Figure 3-30. Loading and score plots for principal components PC1b and PC2b of the socioeconomi c variable PCA for time period B. A) Loading plot and B) score plot. Figure 3-31. Loading and score plots for principal components PC1c and PC2c of the socioeconomi c variable PCA for time period C. A) Loading plot and B) score plot.

PAGE 143

143 Figure 3-32. Bivariate plots of selected water flow variables (rain.mm, aet.mm, irg.mm, sm .mo) and socio-economic variables (popd.C, Gden.C, Gcap.C, pov.incid, HAI, PC1c), coded by region.

PAGE 144

144 Figure 3-33. Bivariate plots of GPP per km2 ve rsus the emergy flows aet, ROgeo and (aet + ROgeo). A) Gdens.C versus aet, B) Gd ens.C versus runoff geopotentail and C) Gdens.C versus the sum of aet and runoffg eopotential. The sum of aet and ROgeo is equivalent to the renewable aggregate, re new.W for inland provinces and is the major component of renew.WI for inland provinces.

PAGE 145

145 Figure 3-34. Bivariate plots of selected emer gy flow variables (renew.W, r enew.WI, PC1em, PC2em, PC3em) and socio-economic vari ables (popd.C, Gden.C, Gcap.C, pov.incid, HAI, PC1c), coded by region.

PAGE 146

146 CHAPTER 4 DISCUSSION Summary Disparity in income levels is a global c oncern, with some areas exhibiting double-digit percent growth rates annually, while other areas experience pe rpetually high poverty incidence rates. Many approaches have been used in a ttempts to understand why trajectories of economic growth differ among countries and regions. Many of these approaches come from the economics discipline and employ complex, mathematically intense economic models. These modeling efforts often include relevant industrial and in stitutional parameters (transportation networks, government interventions, access to credit, et cet era), as well as social parameters (education level, age class distribution, et cetera). However, free environmental flows from nature, other than concentrated non-renewable resources, are rarely included. This di ssertation explored the relationship of free, renewable e nvironmental flows, to a variety of wealth measures across four regions of Thailand. Various measures of environmental flow s were calculated within a Geographic Information Systems (GIS) framework from wh ich continuous spatial coverages of each parameter were produced. Subseque ntly these data were extracted at the provincial scale for analysis. A monthly water balance was used to capture temporal differences in intra-annual rainfall distribution among re gions, and produced values for evapotranspiration, runoff, soil moisture and soil moisture defi cit in mm units. In order to account for a wider variety of environmental flows, as well as energy quali ty differences, environmental accounting using emergy was also employed. The emergy analysis evaluated a suite of environmental flows in emergy density units (sej/m2/yr). Principal compone nts analysis (PCA) was then used to create a

PAGE 147

147 set of principal components (PCs) for the suite of water flow measures, and a set of PCs for the emergy measures. These four categories of environmental flow measurement (water flows in mm, water flow PCs, emergy flows in solar emjoules, emergy flow PCs), along with publically available socioeconomic data at the provincial scale, provided the foundation for using regression techniques to assess the relati onship of environmental flows a nd wealth in Thailand. Regional location and distance to Bangkok were also included in models a nd partial correlation was used to assess the separate contributi ons of location and environment. The over-arching research question of this dissertation was: Is wealth based on renewable, indigenous energy flows? Additi onally, if significant relationships were found, a secondary goal was to determine if the strength of those relati onships varied with other system characteristics, such as geographic location or economic growth phase (rapid growth versus recession and recovery). Principal Conclusions There is a strong, positive univari ate relationship between averag e distance of a province to Bangkok and all wealth measures. However, incl uding water flow variables in models of monetary income and accounting for possibl e location in the Northeast (NE) region resulted in a stronger model than the univariate distance model. Data analysis indicates a m easureable, unique influence of renewable water flows on some measures of economic wealth, after accounting for location inside or outside of the NE region. Of the environmental flows investigat ed, evapotranspiration (AET) exhibited the strongest link with income and rainfall generally had the next str ongest link. The income variable with the most variance explained by models was gross provincial product (GPP) normalized by area. Adjusting energy flows for energy quality did not improve the strength of regression models of income. When using the pool of emergy flows as independent variables, the evapotransipration (AET) and rain water flow s were often the significant independent variables, and provincial values for these terms in emergy units have identical variance to their counterparts measur ed in units of mm.

PAGE 148

148 The total renewable emergy flow index, one of the major indices used in emergy analysis, did not have a strong, positive relationship with income. Using principal components as variables increased model strength only slightly for water flows in mm units, and not at all when used as the dependent variable in place of other economic measures. Relationships between income and environment a ppear to differ little in the preand postcrisis periods. A strong relationship was found between populati on density and water flows, regardless of region or distance to Bangkok. Discussion of Principal Conclusions Water Flow and Location Univariate models for income measures wh ich included distance as the sole predictor showed a strong positive relationship between income and distance to Bangkok. Distance to Bangkok accounted for 74% to 78% of the variation in gross provincial product (GPP) per capita and GPP per area. When looking only at agricultural GPP, distance accounted for 58-66% of the variation. This is not surprising given the dominating market influence of Bangkok and its position as the gateway for trade of many of the goods produced in the Thailand. At a smaller spatial scale, Felkner and Townsend (1997) also found that wealth and enterprise decrease significantly with increasing geographic distance from city an d town markets in Thailand. Even though these distance-only models met th e criteria of good fit, simplicity and strong theoretical basis, introducing meas ures of water flow resulted in stronger mode ls with more explanatory power and bette r fit according to a suite of criteria: adjusted-R2 values are higher, root mean square error (RMSE) values are lower and residuals have more normal distributions with variance that is more random. This supports the idea that water flows do indeed affect average income generation at the provincial scal e. The water flow model for GPP per capita in period C has 15% less error than the distance only model, for GPP per area, 26% less error and

PAGE 149

149 for agricultural GPP per area, 35 % less error (Table 3.16). Percenta ge increases in the amount of wealth variance explained by water models in period C over that of distance-only models are 8% for GPP per capita, 10% for GPP per area and 25% for agricultural GPP per area. In addition to increasing model strength, wh en entering water flow variables into the models, distance to Bangkok is no longer a significant predictor. Instead, the northeast region dummy variable (NE) dominates th e water flow models. This variab le is related to distance, but perhaps better captures the regi onal nature of econo mic disparity than distance to Bangkok alone. Repeatedly, the NE region stands out in box plots and scatter plot s as having significantly lower values for wealth measures and sometimes differe nt relationships with environmental variables as compared to the other provinces outside of the Northeast. Distinguishing the Unique Contribution of Water Flow to Wealth Variance Based on the information in the region-coded sc atter plots of wealth versus water flows, the inclusion of a NE dummy variable was necessary to capture much of the variation in wealth among provinces. Once a regional dummy variable was included with environmental flow variables, it then became necessary to find a me thod for assessing the separate contributions of location and environment in order to assess the degree to which environment is contributing to model strength. To this end, part ial and part correlations were investigated as part of the regression output. The partial-R2 is the portion of dependent va riable variance estimated by an independent variable, which is not estimated by the other independent va riables in the equation. Partial correlation is, hypothetically, the correl ation between any two va riables if all other variables are held constant. The part correlation is the correlati on between the dependent variable and an independent variable when the linear eff ects of the other indepe ndent variables in the model have been removed only from the independe nt variable. Part corr elation represents the unique contribution of an independent variable to total variation in the dependent variable that is

PAGE 150

150 not shared by any other independe nt variable in the equation. Discussion of partial and part correlation as it relates to this study will be limite d to time period C for the sake of brevity and because models for the two time periods do not differ dramatically. Because of the additional wealth disparity shown in the Northeast provinces as compared to the aggregate of provinces outside the Nort heast, the NE dummy variable dominates the regression models as evidenced by th e relatively higher beta, partial-R2 and part-R2 values. PartR2 values for the water flow models for each peri od C independent wealth variable indicate that presence in the NE uniquely accounts for 74% of the total variation in GPP per capita, 61% of the total variation in GPP per area and 39% of the total variation in agricultural GPP per area, after controlling for any water flow variables that are in the model. For the GPP per capita models in period C, the part-R2 values for water flows and PCs are fairly low, with the larg est value of 0.11 occurring for PCw3 (Tab le 3-11). This indicates that the unique contribution of water flows to total vari ance in GPP per capita is 11% or less, after controlling for presence in the NE. If the vari ance explained by the NE dummy variable is removed from GPP per capita, partial-R2 values indicate that 32% of the remaining variance can be uniquely explained by aet.mm in the water fl ow model, and 44% of the remaining variance can be uniquely explained by PCw3 in the water PC model. Though water fl ows are statistically significant in the GPP per capita models, because the part-R2 values for individual water flows across independent variable cate gories are around 10% or less, a strong claim cannot be made that water flows have a definite, unique contribu tion to GPP per capita variance when taking into account possible errors introduced by the water balance modeling that produced the water flow estimates. Water flow variables in models for GPP per capita may have a lesser role than for other wealth variables due to the areal basis of the water flow units versus the per capita basis of

PAGE 151

151 this wealth variable. In addition, even when water flows are found to account for much of the variance in population density (as is the case with this study), di fferential production per capita might be responsible for the lack of corre lation between water flows and GPP per capita. The role of water flows is more pronounced in models for GPP normalized by area. A partR2 of 0.29 for evapotranspiration (A ET) suggests that 29% of the to tal variance in GPP per area in period C can be uniquely attributed to AET, af ter statistically controlling for rainfall and the NE region (Table 3-12). If pres ence in the NE is partialled ou t of GPP per area, AET accounts for 72% of the remaining varian ce, according to the partial-R2. Rainfall is also a significant independent variable in the GPP per area model, though it has a negative effect on wealth as compared to the positive effect of AET. This is likely due to the widesp read flooding that occurs in areas receiving high rainfall, and especially those areas receiv ing a large percentage of the annual flow in a short period of time during the monsoon. As one might expect given the contribution of water to agricultural processes, the significance of water flows is al so more pronounced in models for agricultural GPP normalized by area, than in models for GPP per capita. Additi onally, the unique ro le of the NE region is less. A part-R2 of 0.20 for rain suggests that 20% of the to tal variance in agricultural GPP per area in period C can be uniquely attributed to rainfall, after statistically controlling for NE region, AET and soil moisture (Table 3-13). Part-R2 values for AET and soil moisture are 0.14 and 0.04. Summing the water flow part-R2 values indicates that 38% of the total variance in agricultural GPP per area can be uniquely attributed to wate r flows, in comparison to location in the NE uniquely explaining 39% of GPP per area total variance. This is in contrast to role of the NE variable in the corresponding G PP per capita and GPP per area water flow models, with part-R2 values of 0.74 and 0.61, respectively.

PAGE 152

152 The best water flow model for agGdens.C (i ndependent variables = NE, rain.mm, aet.mm, sm.mo) exhibits some interesting differences co mpared to the corresponding model for total GPP per area. First, the strongest water flow variable in the model is rain, with a negative influence on income, rather than AET. This suggests that th e negative effects of flooding account for more of the variability in agriculturally-produced wealth, than do the positive effects of increased evapotranspiration in the agricultu ral landscape. In addition, soil mo isture is significant in this model, with a positive, t hough minimal influence (part-R2 is 0.04). The positive effect of AET on both total and agricultural income per area may occur through several avenues. A proximate effect is th e translocation of nutrien ts through plant tissues via transpiration. Higher AET levels suggest higher crop prod uction, leading to increased agricultural revenues. Areas with increased agricultural surplu ses have produced additional resources that can be traded for higher qual ity inputs from outside the system (machinery, fertilizer, et cetera) to further capture renewable resources and amplify productive processes. In addition, resources produced within provinces ma y be further transformed inside the system before exchange. In addition to the current link between AET a nd agricultural revenue, there may also be an effect of histor ically higher AET levels on current wealth. Areas that historically transformed more water and nutrients into exces s crop production via relati vely higher renewable water flow endowments, have had the opportunity to build and develop more non-agricultural assets over many centuries, and especially sin ce the industrial revolution. This would lead to more varied economic production pathways and ultimately to higher economic prosperity. This may explain the significant rela tionship between AET and total G PP per area, regardless of the portion of GPP due to agricultural activities.

PAGE 153

153 Use of Emergy Analysis to Characterize Environmental Flows Adjusting energy flows for energy quality, by using transformities to convert flows in joules to emergy units of solar emjoules, did no t improve the strength of regression models of wealth in this study. This was due solely to the pa rticular variables that were significant in the models, namely water flows only. When drawi ng independent variables from the pool of provincial emergy flows, the water flows AET and rainfall are generally the significant independent variables, and non-water emergy flows were never included in any of the stronger, best-fit regression models. Emergy values for ra in and AET have identical variance to their counterparts measured in units of mm due to the use of a constant transformity for both rainfall and AET, thus emergy flow models and water flow models were identical if rain and AET were the independent variables. The irrigation emergy flow, which does vary from irrigation in mm, is a significant variable in the agricultural G PP per area model; however, it has minimal unique contribution to overall variation in agricu ltural GPP, with a part-R2 value of 0.04. The lack of model improvement when using emergy variables in place of standard unit variables in mm does not imply that emergy analysis is not a usef ul tool in research at the environment-economy interface. One of the strengths of emergy analysis lies in the ability it gives the user to analyze and compare com ponents from the whole system, whereas this particular study was only investigating the rene wable flow inputs to the system. Renewable flows lie at the lower end of the energy c oncentration hierarchy, with relatively higher abundance and more dispersal. Variation of renewable system components based on emergy measures is more similar to that of energy or ot her standard unit measures than is the case in higher regions of the energy concentration hierarchy (soils, minerals, human labor, goods and services, et cetera). One of the reasons for this is the use of global average transformities for rainfall, wind, tide, runoff geopotential and deep heat, which result in direct proportionality

PAGE 154

154 between energy and emergy measures for these flows. Emergy analysis research has not focused extensively on the area of spatially variable tran sformities for the renewable flows, thus most investigations rely on the global average transf ormities for the renewable flow portion of the analysis. This is an area ripe for research. For materials and processes at higher levels of the energy transformation hierarchy, variable tran sformities within component types are more common. For instance, transformitie s vary for fuel, metal and mineral types (Brown and Ulgiati, 2004); electricity based on source fuel and pl ant processes (Odum, 1996); soil organic matter based on turnover times (Cohen et al., 2006,); an d human labor based on education (Odum, 1996). Therefore, emergy analysis used in co mparative efforts (e.g., comparing regional economic systems), is perhaps more useful when analyzing a suite of vari ables that extend across the energy transformation hierarchy, or sets of variables at higher leve ls in the hierarchy. Total Renewable Emergy Flow Total renewable emergy flow (RENEW) is a major summary index in emergy accounting, and is calculated by taking the larg est renewable line item flow. It is of interest in this study due to its ubiquity in emergy studies and its designation as the main descriptor of renewable flows in a system. For every study province, the larges t renewable flow was Total Water. For inland provinces, Total Water is the chemical potentia l emergy of AET plus geopotential emergy of runoff. For coastal provinces, where freshwater fr om streams enters provincial estuaries, Total Water is the larger of: 1) chemical potential of AET plus chemical potential of runoff entering estuaries, or 2) AET chemical potential plus runoff geopotential, as with inland provinces. For this study, 2 versions of RENE W were investigated: 1) Tota l Water as described above (renew.W), and 2) Total Water plus the chemical potential of irrigati on water evapotranspired (renew.WI). The second version is still referre d to as renewable because when inputs to 2

PAGE 155

155 irrigation dams in Thailand were analyzed in an emergy synthesis study in 1996 (Brown and McClanahan), the emergy of irrigatio n water was 85 to 90% renewable. The RENEW flow was a major item of interest in this study as a potential pred ictor of wealth, however, renew.W did not have a strong relationship with any measure of income, and did not show up as a significant independent vari able in any of the multi-variable regression models. The bivariate plots of GPP per area versus aet, ROgeo and aet+ROgeo in Figure 3.33 reveal the pattern behind the la ck of correlation between renew.W and wealth for the inland provinces where RENEW equals emergy of AET plus that of runoff geopotential. AET is positively correlated with Gdens.C and runoff geopotential is negatively correlated with Gdens.C. Because emergy of AET and runoff are al so negatively correlated with one another, summing them results in a loss of correlation with GPP per area, as well as other wealth variables. While renew.W may be a useful indicator of ove rall system configuration and dynamics, it is not useful as an indicator of th e role of dispersed renewable flows acting as a subsidy to economic processes in a cross-s ectional investigation such as this study. The variation of RENEW that included irri gation (renew.WI) does ha ve strong, positive, statistically significant bivariate correlations with wealth variables (Table E-4). However, this variable is also was strongly correlated to distance to Bangkok, with a Pearson correlation coefficient of -0.83, which makes teasing out th e unique contribution of RENEW to wealth difficult. The correlation of renew.WI to distance to Bangkok is pa rtly due to the larger emergy flows occurring in coastal provinces (coastal prov inces have the chemical potential of runoff in their RENEW aggregates), and partly due to the correlation between irrigation emergy and distance to Bangkok, which has a Pearson correlati on coefficient of -0.75. When a regression model is created predicting G PP per capita with both distance to Bangkok and renew.WI as

PAGE 156

156 independent variables, the part-R2s of the respective independ ent variables are 0.12 and 0.03, suggesting that distance can uniquely explain only 12% of to tal variance and renew.WI can uniquely explain only 3%, due to the large amount of variance that these two variables share. One might argue that the spatial variability of historic renewable environmental flows was present before Bangkoks rapid rise as a metropolis, and set the stage for development to favor areas near Bangkok and the Chao Phraya delta. In other words, renewable flows (and the early stages of irrigation) preceded the existence of distance to Bangkok, and therefore distance to Bangkok is related to wealth because of the correla tion of distance to renewable flows. However, this study cannot claim strong evidence for this line of thought. Additionally, the obvious economic effects of international trade on the growth of Bangkok, a port city, and the surrounding areas, play a role in the relationship of distance to Bangkok and wealth, and are not directly related to renewable flow variability. Use of Principal Components to Represent Environmental and Economic Flows Using principal components obtained through principal components analysis (PCA) as variables increased model strength only sligh tly for water flows in mm units, and did not increase model strength for emergy measures. Using principal components (PCs) of economic measures as the dependent variable in regression models resulted in much weaker models than those which used original measures of wealth. Water PC models for both GPP per area and ag ricultural GPP per area included all 3 water PCs. Even though this essentially means that all water flows in this study were represented, as opposed to only 2 to 3 individual water flows fo r the water flow models, model strength only increased by 1% and 2% for Gdens.C and agGdens.C respectively, according to adjusted-R2 values. Root mean square error va lues indicate an increase in error of 8% in the water flow models versus the water PC models. Even though water PC models had slightly more

PAGE 157

157 explanatory power, the 2 to 3 water flows presen t in the water flow models (AET, rain, soil moisture) account for the vast majority of the vari ability in the dependent va riable that is related to the water flows. In addition, the water flow mo dels are easier to interp ret than the water PC models. The emergy flow PCs were conceived as a methodological enhancement enabling inclusion of the variability in all emergy flows of a province as opposed to considering only the largest flow, as is done with the RENEW su mmary index. According to standard emergy analysis, one cannot add together all of the renewable emergy flows due to potential doublecounting of the contributions of the biospheric in puts to each of the renewable flows line items. However, each province has a unique suite of individual emergy flow magnitudes, not to mention the complete absence of some flows fr om the inland provinces (e.g., tide and chemical potential of runoff). The deriva tion of PCs for the provincial em ergy flows was an approach to finding a proxy for the unique emergy signature of each province. In retrospect, the lack of significance of non-water emergy flows in models of wealth indicates that the emergy PCs would not add additional useful information to wealth models. Indeed, only PC2em is significant in the wealth models, and this PC is strongly correlated to AET and runoff geopotential emergy (Pearson correlation coefficien ts are 0.83 and -0.89, respectively). Though emergy PCs were not useful in models of wealth, they do contain more information about system flows than the RENEW aggregate alone, and may be relevant fo r investigating relationships of emergy with system properties that are not restricted to human wealth generation. Principal components of wealth measures used as the dependent vari able in regression models produced much weaker mo dels than those which used orig inal measures of wealth. This is in part due to the inclusi on of compound annual growth rate (CAGR) in the PCA. CAGR as an

PAGE 158

158 individual variable has no correlation with any of the inde pendent variables used in the regression models. It was included in the PCA in order to have a measure that included both the variance of CAGR and the varian ce of the other wealth measures Although CAGR loaded most heavily on the second PC, which had no correlati on to other study variables, some of the variance in the first principal com ponent is also related to CAGR. In addition, the wealth PCs also included non-monetary indices of wealth: the Human Achievement Index (HAI) for period C and the I ndex of Human Deprivation (IHD) for period B. The inclusion of these indices and their indivi dual components in the PCA does add additional information to the wealth PCs, and almost all index components load heaviest on the first PC. However, the non-income components of th ese indices do not co rrelate strongly to environmental flows, and thus the first PCs for both periods B and C have less correlation to the environment, and result in weaker regressi on models, than the economic wealth measures. Additionally, the nature of the Index of Human Deprivati on (IHD), which is a component of wealth PC1 for period B, contributes to the rela tive weakness of mode ls predicting PC1.b. IHD scores have very little variance for provinces outside the NE, as this index targets level of deprivation rather than level of wealth. Preand Post-crisis Time Periods In general, relationships between income and environment appear to differ little in the preand post-crisis periods. Model strength according to adjusted-R2 values are generally within a few percentage units between corresponding models in period B and C. Even though income levels differ between time periods, it does not ap pear that the financial shock changed overall system dynamics in such a way as to change the main drivers of wealth generation investigated in this study: location an d long-term average annual renewable water flows.

PAGE 159

159 Environmental Water Subsid y and Population Density In addition to looking at rela tionships between environment and socio-economic flows, regression analysis was also used to investigate links be tween environment and population density. A strong relationship was found between population density and wate r flows (water flow model adjusted-R2 = 0.86), and unlike the models for w ealth measures, region and distance to Bangkok were not significant. Similar to wealth models, AET was the dominant independent variable according to values for the beta coefficient, partial-R2 and part-R2 (Table 3-15). These results provide some evidence for the general systems principle that systems or regions with relatively larger flow s of energy will develop networks and storages to utilize these flows. A possible narrative involving Thailand might start with the idea that areas richer in certain renewable flows, namely the ability to transform rainfall through evapotranspiration in the landscape, supported more dense populations of humans through increased crop production. Subsequently, populations grew a nd utilized these water flows to create additional flows and storages of food. Extrapolating further, these ex cess flows and storages of food, relative to areas with poorer water flows, may have allowed for the build-up of additional non-food system assets, as well as increased trad e with other systems. Suggestions for Further Research This study offers initial steps toward estab lishing a link between renewable environmental flows and wealth in Thailand, but refinement is necessary to claim robust evidence for a causal role. Improvements could be made in both details of the models used, as well as the nature of the data selected for study. Temporal Window of Investigation One of the difficulties in this study was teasi ng out the links between water and wealth in the presence of the strong link between dist ance to Bangkok and wealth. One might argue that

PAGE 160

160 Bangkok developed into a thriving metropolis due to the renewable flows available to it via the convergence of materials and water energies tr ansported from the Chao Phraya watershed; however, its location also makes it an international port, which certainly plays a role as well. It is likely that the potent combination of port city a nd site of convergence of environmental energies propelled Bangkok into the position of dominant center of wealth, tr ade and population in Thailand. Though it would be difficult, and perhaps impossible, to obtain wealth data for a preBangkok Metropolis/pre-int ernational market/pre-industrial study, such data would allow investigation of environment-w ealth relationships in the abse nce of the effects of a huge, dominant economic market like Bangkok. Prana khon Si Ayudhya, a seat of power and central market with access to trade by sea did exist be fore the rise of Bangkok; however, the links between Ayudhya as core and the rest of the coun try as periphery were probably less than in modern times with Bangkok as the core. Addition ally, a pre-industrial Th ai society likely had relatively more localized, self-sufficient economie s. Odum and Odum (2001) noted that before the era of fossil fuels, human settlements we re organized around the hydrological cycle and adapted to the zones with higher stream transformity. Perhaps the role of renewable flows from nature would be more apparent and more important in an earlier era in Thailand. Other Ecosystem Subsidies In looking for links between environmental and wealth, this study focused solely on renewable environmental flows. Of interest was the wealth generating human processes that developed within the matrix of environmental flows that existed historic ally, and that benefit from these flows in the present. Another importa nt and likely more domin ant contribution of the environment to human economies is in the fo rm of non-renewable flows from nature. Nonrenewable is defined here as resources that are extracted and us ed faster than they are being replaced. In Thailand, this could include soil orga nic matter lost to erosio n, fisheries extraction

PAGE 161

161 beyond replacement rates, natural gas, minerals and metals. Including these spatially variable environmental flows in the analysis may reveal further links between environment and wealth. An aspect of extractive non-renewable flows that would need to be taken into consideration is their transportability relative to the dispersed renewable flows. Most of the wealth generation from these flows accrues to the portion of society that transforms them further, which often takes place at a site other than the site of extraction. Water Balance Issues A limitation of large-scale, GISbased empirical modeling is that the modeling is subject to error. The results may be of questionable general value in specific areas without some validation process that would require inte nsive field sampling. Although field sampling at the scale of this study was not feasible, a well-designed spat ial sampling regime would improve model development and allow for some level of veri fication. In the absence of field sampling the comparisons of this studys modeled AET with another published dataset (Ahn and Tateishi, 1994), and with AET predicted by rainfall minus di scharge values from GRDC (2004) provides a moderate level of assurance that the data represent reality when aggregated to the level of province. In the comparison of this studys AET with that of Ahn and Tateishi (1994), average absolute error was 11%. It is not assumed that Ahn and Tateishis data are accurate values which this studys estimated AET should match. Because the input data are different between the two models, tight agreement is not expected. However, because the methodology is very similar, the level of agreement between the two datasets provides some assurance in using the data aggregated into provincial aver ages in subsequent analysis. The second model comparison used watersheds ra ther than provinces as the spatial unit, and again, this studys AET estim ation is not expected to completely agree with the rainfall

PAGE 162

162 minus published discharge values. Average absolute error between these two datasets was 13%. Primary reasons for differences include the temporal resolution of the two datasets: average discharge values are for varying numbers of y ears, from any number of decades back to the 1950s, while this study estimated AET from rainfall averages for 1981 through 1990, and other climatic variables with averages from 1961 through 1990. In addition, the AET estimation represents a theoretical value based on rainfa ll as the only hydrologic input, with no human interventions such as damming and irrigation, wh ile the discharge values will reflect these human interventions. Nonetheless, checking th e magnitude of the estimation against the discharge values does provide anothe r check on the water balance model. Other improvements to the water balance m odel could include using a Priestly-Taylor alpha coefficient that varies s easonally, or with rainfall amount. V ourlitis et al. (2002) found that the seasonal variation in estimated monthly was highly correlated with total monthly rainfall (R2 = 0.84; p < 0.01). The close correspondence between and precipitation is expected given that relates evapotranspiration to surface features such as th e amount of available water (Priestley and Taylor, 1972). Conclusion The link between renewable environmental flow s and wealth was investigated in Thailand, with an emphasis on water. Evidence from this ca se study suggests that renewable water flows in the environment uniquely account for part of the vari ability in income measur es of wealth at the provincial level in Tha iland, based on a cross-sectional anal ysis utilizing long-term average climatic and hydrologic conditions paired with curre nt socio-economic data. In addition, the data show that the Northeast region of Thailand should be given sp ecial consideration in both research efforts to understand economic processes and policy efforts to reduce income disparity. For instance, in the case of ev apotranspiration (AET) as part of the basis of average provincial

PAGE 163

163 income, regression analysis indicated that income increases with increasing evapotranspiration rates in all regions; however, in the Northeast, this relationship occurs at a much lower overall level of income. Interventions in the Northeast to increase AET values (e.g., dams for irrigation, increasing water holding capacity of soil) may increase income along the income continuum present for that region, however, AET increases s hould not be expected to cause rises in income to the levels seen in the Central or Bangkok regi ons for the same AET values. Another area of interest for policy-makers involves the negative relationship seen between rainfall and income in the Northeast (and the East to some extent). This negative relationship is likely occurring due to the negative impacts of flooding. If so, interventions aimed at flood attenuation may have a positive impact on income. Moreover, locally targ eted flood attenuation efforts involving dams may have the dual benefit of reducing floods a nd increasing AET via a longer temporal window for application of water to agri cultural fields. This type of intervention would tap two possible channels though which in come may be increased. In a larger general context, understanding th e environmental contributions to economic wealth and human well-being is a global impera tive as world resources are being depleted, landscapes are being altered and climate is changing. Research efforts explicitly addressing the inter-connected web of wealth and free renewable resources from nature may give policy-makers useful results for decision making regarding deve lopment plans and investment schemes in areas of varying renewable resource endowment. Moreove r, quantification of th e direct contributions of climate-based renewable water flows to sy stem productivity and human income would be useful in contingency planning for predicted climat e change scenarios. It is hoped that the results of this particular study contribu te to a growing body of research investigating th e environmental

PAGE 164

164 basis of wealth. Additionally, it is hoped that the results will prove to be intriguing to economists who often leave out environmental contri butions to models of human economies.

PAGE 165

165 APPENDIX A WATER BALANCE MODEL Introduction In order to investigate spatially variable environmental energy flow contributions to Thailands human systems, it was critical to fo rmulate a water balance model to capture the major flows of water that enter and exit the region s of interest, particularly evapotranspiration. In addition, water flow data were required for i nput into the provincial emergy analysis. This Appendix serves to provide introductory and me thodological information for the water balance implemented in this study. It also presents the primary data output from the model. Water Balance Models in the Literature A literature review of water balance models was performed in order to make an appropriate methodological choice for this study. Included are various options for estimating the major parameters used in a water balance, and the reasoning behind the choice of methods used in this study. Research on the development and application of water balance models has been carried out since the 1940s (Thornthwaite, 194 8). Water balance models are ge nerally classified into three types: empirical, conceptual, and theoreti cal (Xu and Singh, 1998). This study follows the approach of Thornthwaite and Ma ther (1955), a conceptual appro ach in which the form of the model equations is suggested by consideration of the physical processes acting upon inputs and outputs in a highly simplified form Monthly conceptual water bala nce models simulate selected hydrological processes by conceptual izing the spatial area as an assemblage of interconnected storages over time, through which water passes from input as rainfall to ou tput as runoff and ET; the controlling equations satisfy the water bala nce requirement.

PAGE 166

166 Water balance calculations are important in both climate research and biosphere studies because they provide essential information on th e amount of water circulating in the hydrologic cycle and thus, the amount of renewable water available for ecosystems and human society (Fekete et al., 2004). The water budget over a unit land surface area may be expressed as dW/dt = P E S, where W is soil moisture, P is precipitation, E is evapotranspiration, and S is surplus (Thornthwaite, 1948; Thornthwaite and Mather, 1955). This model does not consider horizontal motion of water on the land surface. Utilizing this equation within a given area or spatial grid is often called the bucket model appro ach. Despite uncertainties associated with these simple soilwater budgets, many researchers have applied this type of macro-scale model to problems ranging from catchment scale studies to global water balance and climate change scenarios (Thornthwaite, 1948; Wilmott et al., 1985; Voro smarty, 1989, Ahn and Tateishi, 1994). In the simple water budget model, precipitation (P) is th e only hydrologic variable measured directly on a regular basis. The evapotranspiration term of the water balance equation is normally not measured directly over broad domains, therefore it has to be estimated based on the potential evapotranspiration (PET), water-holding capacity of the soil (AWC), and a moisture extraction function. Potential evapotranspiration (PET) component The difference between runoff and rainfall is largely explained by ev apotranspiration. This link in the hydrologic cycle ha s a tremendous impact on ecology, economic activity, and human welfare. Transpiration is the loss of water from th e cuticle or the stomatal openings in the leaves of plants. This transpired water is replaced by wa ter taken into the roots of the plant from the soil. Through transpiration, plan ts control their temperature a nd perform other vital functions such as photosynthesis and translocation of nut rients. When computing water loss from a

PAGE 167

167 vegetated land surfaces, it is usually impossible to separate tran spiration and evaporation from the soil surface, so the two processes are usua lly considered together under the title of evapotranspiration (ET). Evapotranspiration is the component of the water balance method that involves the most uncertainty and a plethora of estim ation choices, as measurement of ET in the field is not nearly as widespread as measurements of rainfall and runoff. Estimating PET is usually the first step toward an estimate of actual evapotranspira tion (AET). When a vegetated surface is losing water to the atmosphere at a rate unlimited by deficiencies of water s upply, the process is known as potential evapotranspiration (P ET). Thornthwaite (1948) first used the concept of PET as a meaningful measure of moisture demand to replace two common surrogates for moisture demand, temperature and pan evaporation. Actua l simulated evapotranspiration (AET) is distinguished from PET through limitations imposed by rainfall amount and soil water deficit. The number of variables required for the calculation of PET va ries by the amount of detail in the estimation method. The simpler methods re quire air temperature an d/or solar radiation, while the more complex methods may additionally require vapor pressure, wind speed, land cover, and diurnal variation of air temperat ure (Federer et al., 1996). Penman (1948, 1961, 1963) and Thornthwaite and Mather ( 1955) have stressed that PET is largely controlled by weather, with vegetation and soil factors playing only a mi nor role. However, vegetation and soil factors become increasingly important when water supp ly to the plant is limited (Dunne & Leopold, 1978). Computation procedures are the subject of lively debate. A brief review of the Penman equation serves as a starting point for discussing both the estimati on of PET in general, and the eventual choice of the Priestly-T aylor method used in this study.

PAGE 168

168 Penman method. Two requirements for evaporation to occur are an energy input and a mechanism for the transport of water vapor away from the saturated surface. Subsequently, two traditional approaches to modeling evaporat ion are an energy budget approach and an aerodynamic approach (Dunne and Leopold, 1978). With the energy budget approach, the net radiation available at the surface must be partitio ned between latent heat flux and sensible heat, typically achieved using the Bowen ratio of sensib le heat flux to latent heat flux. Approximating the Bowen ratio requires measurements of temp erature and humidity at two heights. The aerodynamic approach involves a vapor transport co efficient multiplied by the vapor pressure gradient between the saturated surface and an arbitrary measurement height (Dunne and Leopold, 1978). Determination of the vapor trans port coefficient requires measurement of wind speed, humidity, and temperature. Brutsaert (1 982) presents equations for calculating Bowen ratio and vapor transport coefficients. In 1948, Penman combined the energy budget and aerodynamic approaches into an equation (1-1) for PET. The Penman equation is a weighted averag e of the rates of evaporation due to net radiation ( Er, Equation 1-2) and turbulent mass transfer ( Ea, Equation 1-3). It also includes the term which is the rate of change of va por pressure with temperature, and which is the psychometric constant. E = {[ /( + )] Er} + {[ /( + )] Ea} (1-1) where, Er = Rn/ lvw (1-2) and, Ea = K (u)(es-e) (1-3) In Equation 1-2, Rn is net radiation [W m-2], lv is latent heat of vaporization [J kg-1] and w is density of water [kg m-3]. In Equation 1-3, K(u) is a mass transfer coefficient, es is saturated vapor pressure at air temperature and e is the actual vapor pressure.

PAGE 169

169 Provided that model assumptions are met and ad equate input data are available, various forms of the Penman equation can yield the mo st accurate estimates of evaporation from saturated surfaces (Maidment et al., 1997). Howeve r, across landscapes of varying surface cover and wetness conditions, the Penman method has been found to generally overestimate evaporation (Vorosmarty et al., 1998). This may be due to the concept pu t forth by Thornthwaite and Mather (1955) that any formula for determ ining potential evapotra nspiration containing a humidity term will give excessive values in dry areas or dry periods. They submit that There is no relationship between potential evapotranspiration and expression s relating to the evaporating power of the air such as rela tive humidity or saturation deficit; they are not synonymous. The moisture content of the air is strongly influenced by the evapor ation regime, thus it is not a conservative property. Consequently it is not possi ble to determine potential evapotranspiration by considering either relative humidity or saturation deficit. Simple temperature-driven models. Despite the comprehensive nature of the Penman formulation, some studies have found that simp ler methods may be more accurate under certain conditions. One of the more basic formulati ons is the empirical, temperature-driven Thornthwaite (1948) model. Based on lysimeter and watershed observations of water loss in the central and eastern United States, Thornthwaite refined the formulation containing humidity terms to utilize only climatic data that are gene rally available, eliminating all factors except for mean temperature and average length of day. Th e equation (1-4) estimates PET in mm per month for a reference crop (typically a short, comple te, green plant cover such as grass), and is summarized by Wilmott et al. (1985). PETi = 16 (10Ti / I )a (1-4)

PAGE 170

170 T is mean surface air temperature in month i ( C) and I is the heat index defined in Equation 1-5. The exponent a in equation 1-4 is a fourth-ord er function of the heat index ( I ). I = ( Ti / 5)1.514 (1-5) Monthly estimates of PET calculated with equati on 1-4 need to be adjusted for day length because 30-day months and 12-hour days were assumed when Thornthwaite developed this relationship. The adjusted PET (APET) accountin g for both month length and daylight duration is given in Equation 1-6. APETi = PETi [( d / 30)( h / 12)] (1-6) APET is in mm/month, d is the length of the month in days, and h is the duration of daylight in hours on the fifteenth day of the month. Thornthwaite and Mather (1955) attribute the satisfactory results obtained without the use of wind, humidity, or solar radiation, to the f act that all of these important influences on evaporation including temperature vary together. However, this is not always the case in monsoon climates, like Thailand, where warmer temperatures may be paired with the extensive cloudiness of the wet season, resulting in lower solar radiation. Dunne and Leopold (1978) state that The methods based on air temperature work be st in the regions for which they were developed, namely, mid-latitude continental c limates, where air temperature is a fairly good index of net radiation. In the tropics, how ever, these methods often give erroneous results, and may seriously underestimate the am plitude of seasonal fluctuations of water demand. In such areas, it is preferable to us e the energy balance appro ach even if radiation must be estimated. For this reason, the Thornthwaite model, and other temperature-driven models, were not used for this study. Priestly-Taylor method. Priestly and Taylor (1972) s how that under certain conditions, knowledge of net radiation and ground dryness may be sufficient to determine vapor and sensible

PAGE 171

171 heat fluxes at the Earths surface. When large land areas (hundreds of kilometers) become saturated, Priestly and Taylor reasoned that net radiation is the dominant constraint on evaporation and analyzed numerous datasets over land and ocean to show that the advection or mass-transfer term in the Penman combination e quation tends toward a constant fraction of the radiation term under equilibrium conditions. Th e Priestly-Taylor definition considers the radiation term in the Penman equation to be a lower limit for the evaporation from a moist surface. The form of the evaporation equation developed by Priestly-Taylor is a constant () times Penmans radiation term (Equation 1-7, see Equation 1-1 for explanation of terms). E = [ /( + )] Er (1-7) Using micrometeorological observations ove r ocean surfaces and over saturated landsurfaces following rainfall, Priestly and Taylor came up with a best estimate of 1.26 for the parameter Since 1972, several other researchers have confirmed that values in the range of 1.26-1.28 are consistent with observations under si milar conditions. However, there have been indications that the coefficient may exhibit significant seasonal variation, as well as spatial variation dependent on vegetation t ype (Brutsaert, 1982). Often, lower values are estimated for forested landscapes (Viswanadham et al., 1991, Vourlitis et al., 2002). When looking at humid climates, Priestly-Taylor estimates have shown good agreement with lysimeter measurements for both peak a nd seasonal evapotranspira tion (Kumagai et al., 2005). Shuttleworth (1993) states that the Prie stly-Taylor method is the preferred radiationbased method for estimating reference crop evapotra nspiration, and notes that errors using the Priestly-Taylor method are on the order of 15% or 0.75 mm/day, whichever is greater, and that estimates should only be made for periods of ten days or longer.

PAGE 172

172 Vorosmarty et al. (1998) compared 11 PET models by estimating ET in a global water balance at 0.5-degree resolution (latitude x longitude), and comp aring the results to measured runoff values at stream gauges. They found that the most commonly used measures, Penman and Thornthwaite, produced the poorest agreement of all methods when compared with measured runoff. They found that Priestly-Taylor gave ge nerally unbiased results for cultivated land over the entire range of evapotranspiration values. Ahn and Tateishi (1994) produced one of th e few global PET and AET coverages finer than one degree resolution. They estimated PET using the Priestly-Taylor equation ( = 1.27), and AET using the Thornthwaite and Mather mont hly water budget, producing global surfaces at 0.5 degree resolution. Because their product is at a fairly coarse re solution relative to some of the Thailand provinces, it was not used as th e source for AET values for this study, though it is used for comparison purposes given the similarity of the methodology. Due to results in the literature comparing P ET formulations, the gene ral climatic type of Thailand (monsoonal, semi-humid), and dominant la nd cover of Thailand (cultivated land), the Priestly-Taylor equation was chosen to model PET. Available water capacity (AWC) Once PET is estimated, a water balance model can be used to estimate AET as a function of PET and the water available for evapotranspi ration. In most water balance equations, the available water includes some por tion of the rainfall and/or previous soil water present at the selected spatio-temporal interval. The amount of so il water available for extr action is particularly important in months with little or no rainfall, when ET may consist entirely of former soil water. The soil water available for extraction via ET depends on rooting depth and available water capacity (AWC) of the soil. As with estimating PET, there are a variety of methods used to

PAGE 173

173 model AWC over landscapes, and most of them involve estimating soil hydraulic properties such as field capacity (FC) and wilting point (PWP). The AWC of a soil is generally defined as the difference between FC and PWP (Dunne and Leopold, 1989). For a small area, soil parameters may be obtained by sampling, but measuring soil conditions such as texture and AWC for regiona l landscapes at fine spatial resolution is practically impossible (Gijsman, et al, 2003). Often, soil characteristics that are not readily available are expressed in terms of basic soil da ta that are more widely available through soil surveys. Bouma (1989) introduced the term pe dotransfer function (PTF ) for this purpose. Estimating soil hydraulic properties through multip le regression analysis has dominated PTF research, and several reviews on PTF developmen t and research have been published over the years, including Rawls et al. (1991) and Timlin et al. (1996). For the water balance in this study, a global, spatial AWC dataset was used that employs Saxtons (1986) PTF regressions estimating AWC from soil texture data and is described fully in a technical report by Reynolds et al. (1999). The AWC raster image has the same 5-minute spatial resolution of the original SMW to preserve the integrity of the original data. Grid-based water balance models With the advent and improvement of geogra phic information systems (GIS) technologies in the 1980s and 90s, it became possible to perform water balance calculations over much larger areas, limited only by the extent a nd accuracy of the input data. Many implementations of GIS water balance models are described in the literature and are app lied from regional to the global scale (Vorosmarty et al., 1989; Vo rosmarty et al., 1998; Ahn and Tateishi, 1994). Because of the availability of climatic point data across Thailand and the intent to capture spatial variability in

PAGE 174

174 environmental flows, GIS was used in this study to apply the water balance calculations on a cell by cell basis. Methods The monthly water balance used in this study is based on the methodology of Thornthwaite and Mather (1955). Initial data inputs required for the water balance model included available water capacity (AWC), pr ecipitation, and potential evapotrans piration (PET). In turn, the PET calculation required primary data inputs of sun dur ation, temperature and de wpoint temperature. Table A-1 contains the equations for the PET derivation, while Table A-2 details all equation parameters, with data sources, methodology notes, and values if applicable. Because of the availability of climatic point data across Thailand GIS is used in this study to apply the water balance calculations on a cell by cell basis. A 1 km cell size was chosen based on the resolution of the land use polygon coverage, wh ere a conversion to raster re sulted in a representation of four 1 km cells for the smallest land use polygon. Concerning temporal scale, all datasets do not cover the same time-frame due to lack of appropriate data. The rainfall gauge data were decadal averages covering 1981-1990, while the other clima tic variables are long-term averages from 1950-1990. Land cover is based on 1990 land use maps. PET Calculation Due to results in the literature comparing P ET formulations, the gene ral climatic type of Thailand (monsoonal, semi-humid), and dominant la nd cover of Thailand (cultivated land), the Priestly-Taylor equation was chos en to model the PET component of the water balance (Priestly and Taylor, 1972). The primary inputs of sun dur ation, temperature and dewpoint temperature were required to derive the net radiation term. The data were obtained as geographically located point values obtained by anonymous FTP from Kyoto University (Kyoto University, accessed

PAGE 175

175 2002). Values are long-term averages of a pproximately 1961-1990, and were inverse-distance interpolated using Idrisi32 soft ware. Fifty-seven points were avai lable for temperature, 57 points for dewpoint temperature, a nd 18 points for sun duration. A land-use shapefile obtained from Thailand on a Disc (TEI, 1996) was also employed in the PET calculation to account for land-c over variation in th e Priestly-Taylor coefficient that may be due to varying land cover. According to Kumagai et al. (2005), should be less than its typical 1.26 value for forested systems because of additional boundary layer, leaf, xylem, and root resistances. Kumagai et al. obtained values of 0.6 to 0.8 for in dipterocarp forest in Borneo, and they comment that this range a ppears reasonable when compared to other dry canopy studies (De Bruin, 1983). Visw anadham et al. (1991) found that varied from 0.67 to 1.16 in the Amazon forest. Vourlitis et al. (2002) estimated seasonal variations of of a transitional tropical forest of Mato Grosso, Brazil, and reported a maximum value of 1.07 and an average monthly value of 0.72. Based on these studies, a value of 0.7 for is used in this study for naturally forested areas, and a value of 1.0 is used in orchards, which are assumed to have resistance characteristics similar to forests, but of a lower magnitude due to the assumed homogeneity, sparser planting distances, and lesser coverage of shorter herbaceous and shrubtype layers. Water Balance Model The water balance method developed for this study is based on the monthly Thornthwaite and Mather (1955) accounting method, as de scribed in Dunne and Leopold (1978). The calculations were performed in a GIS environm ent using Idrisi32 softwa re. Raster coverages were made for each of the input variables of the water balance and map algebra was used to

PAGE 176

176 execute each of the water balance steps. Table A-3 shows the calculations used in the monthly water balance, and Figure A-1 shows a flowchart of the major steps in the GIS implementation. Precipitation was derived by inter polating point source data to a 1 km resolution grid using the inverse-distance weighted in terpolation algorithm in Idrisi Rain gauge point data were obtained from Thailand on a Disc (TEI, 1996), and consisted of monthly rainfall amounts for individual years from 1960 through 1990. Only data for the time period 1981 through 1990 were used, as the vast majority of points had no data prior to 1981. Quality control was performed on the raw data by deleting records with blanks, deleting records with three c ontinuous months of no rain during the monsoonal rainy season, and dele ting obvious data entry errors. An inverse distance weighted interpolation was performed (no exponent, search radius of 6) using each points monthly rainfall average, creating a contin uous grid of spatially varying rainfall values for each month. Available water capacity (AWC) was obtained from a global, spatial AWC dataset that employs Saxtons (1986) PTF regressions estimati ng AWC from soil texture data. The dataset is described fully in a technical report by Reynolds et al. (1999). The water balance as implemented in GIS pr oduced output surfaces of the AET, runoff, soil moisture and soil moisture deficit estimations for the 3 regions of interest. A coverage of administrative boundaries at the prov incial level could then be used to extract averages and sums for each province. Water Balance Model Evaluation Direct measurement of AET over large areas is difficult, resulting in a lack of data for rigorous validation of AET estimations. Two comp arative evaluations were performed in an attempt to reduce uncertainty in input parameters and therefor e, model output: 1) a comparison

PAGE 177

177 of estimated AET to the AET of a global land surface dataset (Ahn and Tateishi, 1994) at the provincial scale, and 2) a comparison of es timated AET to the difference between the interpolated rainfall grid and reported river discharge values (G RDC, 2004), at the water basin scale. For both comparisons, a suite of descriptive, pa ir-wise difference statistics were calculated: mean error, mean percent error, mean absolute er ror, mean absolute percent error and root mean square error. These statistics are commonly used in climatic and geophysical model evaluation (Fox, 1981; Wilmott, 1981; Tateishi a nd Ahn, 1996). In the following formulas, O is used to indicate observed data, and P is used to indicate this studys model-predicted values. Note, that Ahn and Tateishis AET values are also modele d estimates and not directly observed values. Mean error (ME) describes average bias, retain ing sign to indicate direction of overall bias (Equation 2-1). To express ME in percentage te rms, mean percent error was also calculated (Equation 2-2). Both ME and MPE are very sens itive to the presence of negative and positive error terms. If large positive error terms cancel out large nega tive error terms, both measures might be very close to zero. Therefore, mean absolute error (MAE) and mean absolute percent error (MAPE) should also be reported as well, and thus were calculated in this study (Equations 2-3, 2-4). ME = (1/ n) (Pi Oi) (2-1) ME = (1/ n) [(Pi O i)/Oi] (2-2) MAE = (1/ n ) (|Pi Oi |) (2-3) MAPE = (1/ n) [(|Pi Oi |)/Oi] (2-4) An alternative way of handling the problem of positive and negative error canceling is to square the difference between the predicted a nd observed values, producing the mean square

PAGE 178

178 error (MSE). Taking the root of MSE results in th e root mean square error (RMSE), which is one of the more popular measures in model evaluation. The RMSE is more sensitive to extreme values, due to the physically arti ficial exponentiation that is an artifact of the statisticalmathematical reasoning from which RMSE co mes (Willmott, 1982). MSE and RMSE were also calculated as part of the suite of error measures (Equation 2-5). RMSE = MSE = (1/ n) (Pi Oi )2 (2-5) For the first comparison, the pr ovincial boundary coverage wa s used to extract average AET values from both the estimated AET covera ge generated in this study, and the Ahn and Tateishi (1996) global estimated AET coverage. Th e error measures were then calculated with the pairs of AET values for each province. In the second comparison, AET values estimate d in this study were compared to rainfall minus the discharge values reported at river gauge stations (GRDC, 2004). Rainfall was determined by creating polygons for the gauge wa tersheds and using the polygons to extract rainfall values. Watersheds were digitized initia lly as part of the stream order transformity model, described later. Watersheds for the gauge station points were digitized on-screen with ArcGIS 8.2, using the stream network provided in Thailand on a Disc (TEI, 1996), with the ETOPO5 digital elevation model (NOAA, 1988) in the background as a guide. Area in square kilometers was calculated for the digitized wate rsheds to ensure agreement with reported watershed areas for the gauge stations. Area values for the digitized watersheds were within 10% of the GRDC reported areas. For this second AET comparison, watersheds wi th areas less than the smallest province were eliminated from the set of watersheds digi tized for the stream order model. It is assumed that measurement error may be larger when look ing at very small wate rsheds, and it was not

PAGE 179

179 desirable to include error at a scale smaller than the scale of this study, namely the provincial level. Because the smallest provincial area is 392 km2, watersheds digitized for the stream order model that had areas of less than 350 km2 were eliminated for the r unoff validation, leaving 14 of the 19 watersheds used in that model. In order to have additional waters heds, points were chosen from the remaining gauge station dataset fo r which reported area exceeded the smallest provincial area. This resulted in an additional 12 samples, bringing the total sample size to 26. The GRDC points available and c hosen are shown in Figure A-2, w ith the stream network in the background. Unfortunately, gauge data were not available for the East region. Figure A-3 shows a closer view of several digitized watersheds, with the river networ k and elevation in the background that were used as a guide. Digiti zation was performed at a higher level of magnification than what is shown in Figure A-3. Figure A-4 displays the set of watersheds used for extracting estimated AET from the water ba lance model for comparison to the difference between extracted average rainfall and the point discharge values obtai ned from GRDC. Because of slight area differences, the average AET and rainfall in mm were extracted from the grids and multiplied by the GRDC reported area, then compared to the reported volume discharge. Results Major outputs from the monthly water bala nce model include rainfall, PET, AET and runoff, with AET and runoff becoming line items in subsequent emergy tables. Figure A-5 presents grids of the long-term average yearly flows for the following variables: the rainfall surface interpolated from the rain gauge data (A), the PET surface generated using map algebra with grids of the input variable s required for the Priestly-Taylor PET equation (B), and the AET, runoff, soil moisture, and soil moisture deficit su rfaces resulting from the implementation of the monthly water balance model (C, D, E, F).

PAGE 180

180 Comparison of Estimated AET and Runo ff to Previously Published Datasets The AET values produced in this study were compared to those of the global land surface grid of AET produced by Ahn and Tateishi (199 4). Methods for each of the estimations were similar, with the same PET and water balance calc ulations, but different source data for rainfall, solar radiation, and available water capacity (AWC). Pair-wise comparison for average provincial AET for the 45 study provinces shows an average absolute error of 11%, mean bias of -118 mm, mean absolute error of 118 mm, and RMSE of 192 mm (Table A-4). Error values and absolute error values are virtually the same in this comparison due to the systematically lower values of this studys AET estimation. This is due at least in part, to the difference in rainfall values used in the respective AET models. This study uses average rainfall for the years 19811990, which is lower than average for the time period 1950-1980, used in Ahn and Tateishis AET estimation. When comparing this studys estimated AET for 26 watersheds, with AET derived from the rainfall grid minus GRDC gauge runoff values, average absolute error is 13%, mean bias is 34 mm, mean absolute error is 114 mm, and RMSE is 138 mm (Table A-4). Regional Comparisons Annual averages for the Bangkok, East, Central, and Northeast regions, as well as some component ratios, are shown in Table A-5. To pr ovide a more detailed view of regional water budgets, average monthly values for the major water balance parameters (rain, PET, AET, soil moisture deficit and runoff) were extracted fr om the GIS water balance model using regional boundaries, and plotted in series (Figure A-6). The graphs illustrate the seasonal monsoonal nature of the climate affecting all four regions, and highlight periods of soil moisture deficit, when AET exceeds precipitation. The four regions have similar water budget values, although

PAGE 181

181 the East region has significantly more rainfall and runoff, utilizing only 58% of the rainfall in evapotranspiration processes. Evapotranspiration in the Central and Northeast regions averages around 75% of rainfall, and Bangkok has the highest percentage of rain evapotranspired at 86%. Soil moisture deficits total around 400 mm for most areas, though in the East, the value is much less, estimated at 341 mm. Data Summary of Water Flows at the Provincial Level Provincial averages for water flow variable s were extracted from the national raster coverages using a vector coverage of provincial boundaries. Also calculated for each province is a measure of perennial stream density in units of km per square km. Water flow values for each province (depth in mm) are visually di splayed in map form in Figure A-7. Orographic influences cause much of the discontinuous rainfall patte rn (Figure A-7A). Rainfall is highest along the eastern coast of the Gulf of Thailand, the eas ternmost part of the Northeast region and the two Central provinces along the eastern side of the Bangkok Plain where winds are forced to ascend the Khorat escarpment that rises abruptly from sea level (Figure 1-2). The western slopes of the mount ain chain on the coast of Burma run through Kanchanaburi, the westernmost province in Figur e A-7, which causes higher rainfall than in neighboring provinces (up to 1500 mm of precipitation annually). These mountains remove the moisture from the air on the easte rn side of the mountain system, reducing the rainfall in a northsouth string of Thai provinces to around 900 mm/yr. A drier area is also found in western Khorat (Northeast Region), where the landscape is in th e rain shadow of the escarpment and hills between Khorat and the Central Valley. Fraction of rain evapotranspired varies over space, thus provincial AET values are not directly proportional to rainfall values (Figure A-7B). Provinces to the nort h and east of the inner

PAGE 182

182 Gulf of Thailand, as well as most of the provinc es in the southern section of the Northeast Region, have the highest AET values. Provinces with higher AET rate s consequently have relatively lower runoff values, and vice versa (Fi gure A-7C). Also, the provinces with extremely high rainfall values also tend to have relativel y higher runoff rates. Ir rigation is also shown (calculations given in Chapter 3) and is much more extensive in the Central region, which is reflected in the provincial map of irrigation evapotranspire d (Figure A-7D). The final two maps of Figure A-7, soil moisture (E) and soil moisture deficit (F) are given in units of average monthly mm for 12 months of the water budget. Soil moisture values are higher in the Chao Phraya delta area of the Central Region (Figure 1-2) and the easternmost provinces of the Northeast for which rainfall is higher. Average monthly soil moisture deficit values are variable across pr ovinces, and are generally highest for the provinces with higher runoff values. The soil moisture deficit map is shown with a reversed legend so that interpretation is similar to th e other mapped variables: areas with lower deficit are darker. Summary statistics for the stream density and water flow variables at the provincial scale are shown in Table A-6. Variable definitions and units are summa rized in Table 2-1. Other than AET and monthly soil deficit, variables ex hibit varying degrees of skew from a normal distribution. Box plots are shown in Figure A-8 to further explor e the data distributions. These figures indicate skewed distribut ions and the presence of outliers in all variables other than AET and monthly soil moisture. Figure A-9 contains box plots for the transformed variables (see Table 2-1 for information on transformations). Other than 1 remaining outliers for the rain variable, outliers have been eliminated and distributions are closer to normal. Using the untransformed provincial data valu es, box plots were also constructed with values grouped by region (Figure A-10). Perennial stream density is much higher in the Bangkok

PAGE 183

183 and Central regions as compared to the East and Northeast. Annual ra infall is highest in the East and Northeast regions. When modeled on a monthly ba sis first, then aggregated to annual values, AET is highest in the Bangkok region, lower in the Central region and in the middle of the spectrum for the East and Northeast regions, th ough there is overlap am ong all regions. The fraction of rainfall evapotranspi red is generally higher in the Bangkok and Central regions, with the Northeast and East having lower values. Acco rding to the monthly water balance, average monthly soil moisture is highest is the Bangkok region and lowest in the Central region, with the East and Northeast roughly in the middle. Averag e monthly soil deficit is generally higher in the Central and Northeast regions, and lowest in the East, though all re gions show overlapping values to some extent. Irrigati on evapotranspired is highest in the Bangkok and Central regions. Runoff is much higher in the East, which also has significantly higher rainfa ll, and is lowest in the Bangkok and Central regions, which exhibit a higher fraction of rainfall evapotranspired.

PAGE 184

184 Table A-1. Equations used in the monthly Priestly-T aylor potential ev apotranspiration (PET) model. Variable name Equation PET, mm PET = 8.64E7* *[(Rn)/L*w]*[( / )/( / +1)] Net radiation, W/m2 Rn = 0.484*[(1)*Qs]-Qlw SW radiation, cal/cm2/d Qs = Io[a + b(n/N)] LW radiation, cal/cm2/d Qlw = 1.17E-07*T4*(0.56-(0.08 ea))*(0.01+(0.09*n/N)) Actual vapor pressure, kPa ea = 0.1 0.6108 2.7183[17.27T /( T+273 ) ] Note: PET equation is from Priestly and Taylor (1 972); net, shortwave (SW) and longwave (LW) radiation equations are from Dunne and Leopold, 1978. Table A-2. Parameters used in the P ET model equations, with data sources. N ame Symbol Value Source Constants: Albedo 0.18 assumed Freshwater density, kg/m3 w 1000 Dunne and Leopold, 1978 Latent heat of vaporization, J/kg L 2.448E6 Dunne and Leopold, 1978 SW radiation equation constant a a 0.23 Dunne and Leopold, 1978 SW radiation equation constant b b 0.48 Dunne and Leopold, 1978 Values vary by month: aRadiation at atmospheric surface, cal/cm2/d Io varies Dunne and Leopold, 1978 aMax potential sun duration, hr/month N varies Dunne and Leopold, 1978 Values vary across space (GIS map generated): Preistly-Taylor alpha: varies TEI, 1996 (land use classes) forest 1.00 Kumagai et al, 2005 non-forest 1.26 Priestly and Taylor, 1972 orchards 1.13 Avg. of forest and non-forest b Average monthly temperature, K T4 varies Kyoto University, 2002 b Average monthly dewpoint temperature, C T varies Kyoto University data, 2002 b Actual sun duration, hr/month n varies Kyoto University data, 2002 cPenman's dimensionless parameter: / varies Dunne and Leopold, 1978 rate of ea change with temp., Pa/K varies Dunne and Leopold, 1978 psychometric constant, Pa/K 66.5 Dunne and Leopold, 1978 a Values for each month obtained by averaging the 10 and 20 N values in Table 4-2 and Table 4-4 of Dunne and Leopold (study area is from 12 to 18N), for Io and N, respectively. b Point values with coordinates obtained by anon ymous FTP (Kyoto University, 2002). Values are longterm averages of approximately 1961-1990, and we re inverse-distance interpolated using Idrisi32 software. 57 points were available for temperature, 57 points for dewpoint temperature, and 18 points for sun duration. c Values are temperature dependent and were obtained from Table 4-6 in Dunne and Leopold (1978) and assigned to the GIS temperature map.

PAGE 185

185 Table A-3. Thailand water balance calculations based on Thornthwaite and Mather (1955) as presented in Dunne & Leopold (1978). Parameter Equation/Def inition Rainfall P, Primary input to wate r balance (interpolated coverage, this study) Potential Evapotranspiration PET, Primary input to water bala nce (Priestly-Taylor estimate, this study) Available Water Capacity AWC, Primary i nput to water balance (Rawls et al., 1991) Accumulated Potential Water Loss APWL = IF P > PET, THEN APWL = 0, ELSE APWL = previous month APWL + (P PET) Soil Moisture1 SM = IF (P PET) > 0 AND (previous month [SM + (P PET)] < AWC), THEN SM = (previous month SM + (P PET)) ELSE SM = AWC EXP(coefficient*AWPL) Change in Soil Moisture dSM = present month SM previous month SM Actual Evapotranspiration AET = IF (P PET) <= 0, THEN AET = P + ABS(dSM), ELSE AET = PET Soil Moisture Deficit D = IF (P PET) <= 0, THEN D = PET AET, ELSE D = 0 Soil Moisture Surplus S = IF D > 0, THEN S = 0, ELSE S = (P PET) dSM Total Available For Runoff TAFR = (0.5 previous month TAFR ) + present month S Runoff RO = 0.5 TAFR 1 Coefficients for 11 AWC's were calculated from char t lines in 8-3 in Dunne and Leopold, 1975 [e.g., for AWC=150mm; SM=150*EXP(0.00678*APWL)].

PAGE 186

186 Rain (point) Rain (grid) INTERPOL PET (grid) MAP CALC (P-PET) APWL SM ReynoldsAWC Land Cover (g rid ) ASSIGN Alpha Value Net radiation Sun duration (point) Temp (point) Dewpt. Temp (point) INTERPOL INTERPOL INTERPOL Sun duration Temp Dewpt. Temp MAP CALC MAP CALC (Priestly-Taylor) AET dSM MAP CALC MAP CALC Deficit MAP CALC MAP CALC Surplus MAP CALC Runoff MAP CALC Figure A-1. Flowchart for the Thailand water ba lance model, imple mented on a monthly basis using Geographic Information Systems (GIS). Figure A-2. Discharge gauges for the four region study area show n with the stream network.

PAGE 187

187 Figure A-3. Digitized watersheds, shown over th e river network and elevation coverages that were used as a guide for the digitization process. Figure A-4. Digitized watersheds used for compar ing estimated runoff to discharge. Three of the larger watersheds (hatched lines) encompass smaller watersheds.

PAGE 188

188 Figure A-5. Average flow surfaces generated by the water balance model. A) annual rainfall, with point coverage overlay, B) annual P ET, C) annual AET, D) annual runoff, E) monthly soil moisture and F) monthly soil moisture deficit.

PAGE 189

189 Table A-4. Summary and difference statistics for modeled pr ovincial AET values versus Ahn and Tateishi AET estimation, and m odel wate rshed values versus TEI rainfall minus GRDC runoff. Provincial values: Watershed values: Statistic AET AET* AET Rain GRDC Runoff N 45 45 26 26 Mean 959 mm 1077 mm 932 mm 898 mm Std Dev 93 mm 112 mm 107 mm 147 mm Comparison, provinces: Comparison, watersheds: ME -118 mm 34 mm MPE -0.10 0.06 MAE 118 mm 114 mm MAPE 0.11 0.13 RMSE 192 mm 138 mm AET* is from values published by Ahn and Tateishi, 1994. ME = mean error, MPE = mean percent error, MAE = mean absolute error, MAPE = m ean absolute percent error, RMSE = root mean square error. Table A-5. Water balance model results as average annual values, summar ized by region. Water Balance Averages: Ratios: Region Rain PET AET Deficit Runoff AET/Rain RO/Rain AET/PET Bangkok 1148 1380 989 408 216 0.86 0.19 0.72 Central 1093 1205 811 394 294 0.74 0.27 0.67 East 1670 1309 971 341 675 0.58 0.40 0.74 Northeast 1289 1375 968 408 309 0.75 0.24 0.70

PAGE 190

190 0 50 100 150 200 250 300 123456789101112Parameter average in mmBangkok Region rain p et aet def ro 0 50 100 150 200 250 300 123456789101112Central Region 0 50 100 150 200 250 300 123456789101112Parameter average in mmMonthEast Region 0 50 100 150 200 250 300 123456789101112MonthNortheast RegionAB D C Figure A-6. Regional water budget s graphs derived from the monthly water balance model showing historical averages for rain, potential ev apotranspiration (pet), actual evapotranspiration (aet), soil moisture deficit (def) and runoff (ro) for four regions. A) Bangkok, B) Central, C) Ea st and D) Northeast.

PAGE 191

191 Figure A-7. Provincial maps of wa ter flows. A) annual rainfall, B) annual AET from rain, C) annual runof f, D) annual AET from irrigati on, E) average monthly soil moisture and F) average monthly so il moisture deficit.

PAGE 192

192 Table A-6. Summary statistics fo r provincial water balance averag es, perennial stream density, and AET as a fraction of rainfall. Variable Units Mean SE Mean StDev CofVar SumofSq Skew Kurtosis strmdens km/km2 0.24 0.03 0.23 96.2 5.0E+00 1.75 2.44 rain.mm mm 1285 55 365 28 8.0E+07 2.32 7.12 aet.mm mm 959 14 93 10 4.2E+07 -0.68 0.38 RO.mm mm 335 55 372 111 1.1E+07 2.71 9.11 irg.mm mm 232 38 254 109 5.3E+06 1.23 0.23 def.mo mm 34 1 7 22 5.5E+04 0.00 0.69 sm.mo mm 46 2 14 30 1.0E+05 0.73 0.53 SE Mean = standard error of the mean, StDev = standard deviation, CofVar = coeffici ent of variance, SumofSq = sum of squares. Figure A-8. Box plots of provincial water flows before transformation.

PAGE 193

193 Figure A-9. Box plots of transformed water fl ow variables for use in PCA and regression modeling. Units vary depending on the type of transforma tion.

PAGE 194

194 Figure A-10. Box plots of untransformed water fl ow variables. A) perenn ial stre am density in km/km2, B) rain in mm, C) AET in mm, D) AET as fraction of rain, average monthly soil moisture in mm, E) average monthly soil deficit in mm, F) irrigation in mm, and G) runoff in mm. Plots are aggregated by region: Bangkok (bkk), Cent ral (c), East (e) and Northeast (ne).

PAGE 195

195 APPENDIX B ENERGY CIRCUIT DIAGRAMMING SYMBOLS USED IN SYSTEMS DIAGRAMS Figure B-1 provides images and definitions of som e standardized symbols used in energy circuit diagramming (adapted from Odum, 1996). Figure B-1. Standardized symbols used in ener gy circuit diagramming (a dapted from Odum, 1996).

PAGE 196

196 APPENDIX C FOOTNOTES TO EMERGY EVALUATION TABLES These tables provide reference information for energy, material and unit em ergy values (UEV) reported for national and provincial emer gy evaluations. Presented first are data sources and internet addresses for the pr imary input data used in the na tional emergy analyses (Table C1). Given next are the line item notes for the national scale (Tab le C-2) evaluation of Thailand.

PAGE 197

197Table C-1. Data sources and websites for national scale primary data inputs. Variable Dataset Accessed through URL for dataset Land area The World Factbook Central Intelligen ce Agency www.cia.gov/ci a/publications/factbook/ N et solar radiation ERBE DAWWB, Version 1.0 www. ce.utexas.edu/prof/maidment/gishyd97/atlas/atlas. htm Continental shelf area GMBD UNEP, GEO-3 Data Compendium, 1.1 geocompendium.grid.unep.ch/ Tidal range Typology Data Set LOICZ www.loicz.org N umber of tides Typology Data Set LOICZ www.loicz.org Rainfall Wilmott grid V.2.01 Center for Climatic Research climate.geog.udel.edu/~climate/html_pages Evapotranspiration Ahn and Tateishi, AET grid UNEP, GEO Data Portal, GNV183 www.grid.unep.ch/data/data.php Elevation ETOPO5 N ational Geophysical Data Cente r www.ngdc.noaa.gov/mgg/global/etopo5.HTML Rain runoff volume UNH/GRDC Composite Runoff Water Systems Analysis Group, UNH www.grdc.sr.unh.edu/index.html River flow at border GRDC discharge database Global Runoff Data Center grdc.bafg.de/servlet/is/1035/?lang=en Wind speed Climate Research Unit CL 1.0 Climate Resear ch Unit www.cru.uea.ac.uk/~timm/grid/CRU_CL_1_0.html Coastline length The World Factbook Central Intellig ence Agency www.cia.gov/ cia/publications/factbook/ Wave height Typology Data Set LOICZ www.nioz.nl/loicz/welcome.html Heat flow Global Heat Flow Database International Heat Flow Commission www.heatflow.und.edu/index2.html Agriculture & livestock FAOSTAT Food and Agriculture Organization faostat.fao.org/ Fishery extraction FIGIS Fo od and Agriculture Organization faostat.fao.org/ Wood extraction FAOSTAT Food and Agri culture Organization faostat.fao.org/ Water extraction AQUASTAT database Food and Agriculture Organization www.fao.org/ag/agl/aglw/aquastat/main/index.htm Hydroelectric production International Energy Annual 2004 Energy Information Administration http://www.eia.doe.gov/iea/ Electricity use International Energy Annual 2004 Energy Information Administration http://www.eia.doe.gov/iea/ N onrenewable fisheries FAO Fisheries Technical Paper 457 Food and Agriculture Organization ftp ://ftp.fao.org/docrep/fao/007/y585 2e/y5852e00.pdf Wood biomass per area IPCC report, Table 3A.1.4 IPCC www.ipcc-nggip.iges.or.jp/public/gpglulucf/ Annual forest extent lost GFRA, 2000 UNEP, GEO-3 Data Compendium, 1.1 geocompendium.grid.unep.ch/data_sets/forests/ Gas, coal, oil production International Energy Annual 2004 Energy Information Administration http://www.eia.doe.gov/iea/ Metals minerals World Mineral Production, 99-03 British Geological Survey http://www.miner alsuk.com/free_downloads.html# Soil organic matter WISE (version 2) ISRIC http://www.isric.org Soil degradation GLASOD database ISRIC www.grid.unep.ch/data/grid/soils.html Gas, coal, oil, elec. trade Worl d Energy Database EIA, International Energy Annual 2001 www.eia.doe.gov/emeu/world/main1.html All other trade flows COMTRADE United Nations Statistics Division unstats.un.org/unsd/comtrade/default.aspx GDP UNCDB United Nations Statistics Division http://unstats.un.org/unsd/cdb Tourism expenditure UNCDB United Nations Statistics Division http://unstats.un.org/unsd/cdb DAWWB = Digital Atlas of the World Water Balance, ERBE = Earth Radiation Budget Experiment, GFRA = Global Forest Resources Asse ssment, GMBD = Global Maritime Boundaries Dbase, IPCC = Intergovernmental Panel on Climate Change, LOICZ = Land -Ocean Interaction in Coastal Z one.

PAGE 198

198 Table C-2. National emergy an alysis footnotes containing energy and emergy conversion calculations. Note Variable Value Units Source RENEWABLE FLOWS 1 Sunlight Land area 5.1E+11 m2 CIA, 2005 Continental shelf area 1.9E+11 m2 CIA, 2005 Net radiation 139.8 W/m2 ERBE grid, from Maidment, 1997 (1983-1991 average) Energy 3.1E+21 J/yr Total area radiation 3.154e7 sec/yr UEV 1 sej/J Odum, 1996 2 Deep heat Land area 5.1E+11 m2 CIA, 2005 Heat flow 54.5 mW/m2Pollack et al., 1993, Glob al Heat Flow Database Energy 8.8E+17 J/yr Area (heat flow/1000) 3.154e7 sec/yr UEV 5.8E+04 sej/J Odum, 2000, Folio2 3 Tide Cont. shelf area 1.9E+11 m2 CIA, 2005 Avg. tidal range 1.89 m LOICZ Typology Data Set, 1998 (v.3) Number of tides 1.58 #/day LOICZ Typology Data Set, 1998 (v.3) Seawater density 1.0E+03 kg/m3 Energy 9.6E+17 J/yr Area*0.5*# *range2*1025kg/m3* 9.8 m/sec2*0.5 UEV 7.4E+04 sej/J Odum et al, 2000, Folio1 4 Wind Avg. surf. Windspeed 1.4E+00 m/sec New et al., 1999, CRU CL 1.0 grid (1961-1990 average) Avg.geostrophic speed 2.4E+00 m/sec Assume surface winds are 0.6*geostrophic Air density 1.23 kg/m3 Odum, 1996, p.294 Drag coefficient 0.001 na Energy 2.8E+17 J/yr Total area 1.23 0.001 geostrophic speed^3 3.154e7 s/yr UEV 2.5E+03 sej/J Odum et al, 2000, Folio1 5 Total water Total inland area 5.1E+11 m2 CIA, 2005 Land Area 5.1E+11 m2 CIA, 2005 Continental shelf area 1.9E+11 m2 Pruett and Cimino, 2000, GMBD Avg. rain on land 1.69 m/yr Wilmott et al., 1998, Precipitation grid V.2.01 (1920-1980) Avg. rain on shelf 2.23 m/yr Wilmott et al., 1998, Precipitation grid V.2.01 (1920-1980) AET estimate 1.11 m/yr Ahn and Tateishi, 1994, grid (1920-1980) Runoff estimate 0.48 m/yr Fekete et al., 2000, UNH/GRDC CRF V.1.0 Elevation varies m ETOPO5 DEM, NOAA, 1988 River inflow 1.2E+11 m3/yr GRDC, 2005, Global Runoff Data Centre gauge data River outflow 2.1E+11 m3/yr GRDC, 2005, Global Runoff Data Centre gauge data Rain chem. potential (L) 4.3E+18 J/yr Land area rain on land *1000kg/m3 4940J/kg Rain chem.pot. UEV (L) 3.1E+04 sej/J Odum et al, 2000, Folio1 Rain chem. Potential (S) 2.0E+18 J/yr Shelf area rain on shelf *1000kg/m3 4940J/kg Rain chem.pot. UEV (S) 7.0E+03 sej/J Baseline (sej)/Global shelf rain (J) from Wilmott et al., 1998 AET chem. potential 2.8E+18 J/yr Inland area AET 1000kg/ m3 4940J/kg AET UEV 3.1E+04 sej/J Odum et al, 2000, Folio1 Rain, runoff geopotential 6.7E+17 J/yr GI S cellcalc: **1000 kg/m3* 9.8 m/sec2 River inflow geopotential 1.9E+17 J/yr SUM(river IN(m3)*border elev (m)*1000kg/ m3*9.8m/sec2) River outflow geopotential 1.9E+17 J/yr SUM(riverOUT(m3)*border elev(m)*1000kg/ m3*9.8m/sec2) Net total runoff geopotial 6.7E+17 J/yr Rain runoff geopot.+River inflow geopot.-River out geopot.

PAGE 199

199 Table C-2. Continued. Note Variable Value Units Source Water runoff geopot. UEV 4.7E+04 sej/J Odum et al, 2000, Folio1 Rain RO, chem. potential 1.2E+18 J/yr Land area(m2)*RO(m3)*1000kg/ m3 *4940J/kg Rain RO, chem. pot. UEV 3.1E+04 sej/J Odum et al, 2000, Folio1 Riverin, chem. potential 5.8E+17 J/yr River IN(m3/sec)*3.154e7sec/yr*1000kg/ m3*4940 J/kg Riverout, chem. potential 1.0E+18 J/yr River OUT(m3/sec)*3.154e7sec/yr*1000kg/ m3*4940 J/kg Net river chem. potential -4.6E+17J/yr River IN chemical potential River OUT chemical potential River chem. potential UEV 8.1E+04 sej/J Odum et al, 2000, Folio1 TOTAL WATER EMERGY DETERMINATION: Rain, chemical potential 1.5E+23 sej/yr (Land chem. J*land rain trf)+(Shelf chem. J*shelf rain trf) AET, chemical potential 8.6E+22 sej/yr AET chem. potential J*land rain chem. potential trf Rain, land, chem. potential 1.3E+23 sej/yr Rain on land chemical pot. J*land rain chem. potential trf Rain, shelf, chem. potential 1.4E+22 sej/yr Rain on shelf chemical pot. J*shelf rain chem. potential trf Water RO, geopotential 3.2E+22 sej/yr Net total water runoff geopotnetial J Runoff geopotential trf Water RO, chem. potential -3.8E+20sej/yr (Rain runo ff chem. J*rain trf)+(Net river chem. J*river trf) Total water, chem. pot. 9.4E+22 sej/yr If coastal, [Rain chem.sej)]+[Net river chem. J*river chem. trf] AET chem. pot.+ROgeopot 1.2E+23 sej/yr AET chem. potential sej + RO geopotenial sej Largest Water 1.2E+23 sej/yr Total water chem. Pot. OR AET chem. Pot. +RO geopot. 6 Waves Coastline length 3.2E+06 m CIA, 2005 Avgerage wave height 1.00 m LOICZ Typology Data Set, 1998 (version 3), average Avgerage wave speed 4.40 m/s SQRT(9.8*depth of ht. meas.) = 4.4 if at 2m Waves 5.6E+17 J/yr coastline(m)*1/8*1025kg/m3*9.8*(ht2)*speed(m/s)*3.2e7 s/yr UEV 5.1E+04 sej/J Odum et al, 2000, Folio1 INTERNAL TRANSFORMATIONS 7 Agriculture Production Agricultural production 1.2E+08 MT/yr FAOSTAT, 2005 Agricultural production 1.5E+18 J/yr Indv.items 1e6 g/MT energy conversion (J/g) UEV 8.6E+04 sej/J Wt'd avg. UEV for 71 FAO commodities 8 Livestock Production Livestock production 3.3E+06 MT/yr FAOSTAT, 2005 Livestock production 2.3E+16 J/yr Indv.items 1e6 g/MT energy conversion (J/g) UEV 3.7E+06 sej/J Wt'd avg. UEV for 17 FAO commodities 9 Fisheries Production Fisheries production 3.0E+06 MT/yr FAOSTAT, 2005 Fisheries production 7.8E+15 J/yr Indv.items 1e6 g/MT energy conversion (J/g) UEV 8.4E+06 sej/J Brown et al., 1993 10 Fuelwood Production Fuelwood production 1.3E+07 MT/yr FAOSTAT, 2005 Fuelwood production 1.3E+17 J/yr Indv.items 1e6 g/MT energy conversion (J/g) UEV 3.7E+04 sej/J Odum et al., 2000 11 Industrial Roundwood Production Roundwood production 4.0E+06 MT/yr FAOSTAT, 2005 Roundwood production 4.0E+16 J/yr Indv.items 1e6 g/MT energy conversion (J/g) UEV 9.2E+04 sej/J Odum et al., 2000 12 Water extraction Water extraction 8.7E+10 m3/yr AQUASTAT, 2005

PAGE 200

200 Table C-2. Continued. Note Variable Value Units Source Water extraction 4.3E+17 J/yr Extraction(m3/yr)* 1000 kg/m3 4940 J/kg UEV 8.1E+04 sej/J Odum et al, 2000, Folio1 13 Hydroelectricity Production Hydroelectricity prod. 6.0E+09 kwh/yr EIA, International Energy Annual 2002 Hydroelectricity prod. 2.1E+16 J/yr production(kwh/yr) 3.6e6 J/kwh UEV 2.8E+05 sej/J Odum, 1996, Brazilian hydroelectricity 14 Total electricity use Electricity use 8.4E+10 kwh/yr EIA, International Energy Annual 2002 Electricity use 3.0E+17 J/yr Use(kwh/yr) 3.6e6 J/kwh UEV 2.9E+05 sej/J Odum, 1996, average from several types of power plants INDIGENOUS NONRENEWABLE EXTRACTION 19 Forestry Forestry, NR use 3.2E+06 MT/yr Biomass density(MT/ha)*Extent change (ha), if < zero Biomass density 2.9E+01 MT/ha IPCC report, Table 3A.1.4 Avg. forest extent change -1.1E+05ha/yr GRID-GENEVA GEO-3 (get orig. ref) Forestry, NR use 5.8E+16 J/yr Forestry use (MT) 1.8e10 J/MT UEV 3.8E+04 sej/J Avg. of 4 TRFs used for wood products (see FAO subtable) 20 Fisheries Fisheries, net loss 6.1E+05 MT/yr FAO Fisheries Tech. Paper 457, amts over MSY Fisheries, net loss 1.6E+15 J/yr Loss (MT) 1e6 g/MT 2600 J/g UEV 8.4E+06 sej/J Brown et al., 1993 21 Water Water, NR extraction 0.0E+00 m3/yr AQUASTAT Water, NR extraction 0.0E+00 J/yr Extraction (m3) 1000 kg/m3 4940 J/kg UEV 2.8E+05 sej/J Buenfil, 2001 (Floridan aquifer) 22 Topsoil losses, organic matter Topsoil losses 9.4E+14 g/yr Cohen, 2006 (model based on GLASOD data) Avg. organic matter 5.3E-01 % Cohen, 2006 (model based on FAO, 2003 data) Organic matter losses 1.1E+17 J/yr Soil loss* OM content J/g (performed on cell basis) UEV 2.9E+05 sej/J Cohen, 2006 (Thai average, values vary spatially) 23 Coal Coal Production 1.8E+07 MT/yr EIA, 2004 Coal Production 4.4E+17 J/yr Poduction (MT) 2.45e10 J/MT UEV 6.6E+04 sej/J Odum, 1996, p.308 24 Natural gas Natural Gas Prod. (dry) 1.9E+10 m3/yr EIA, 2004 Natural Gas Prod. (dry) 7.1E+17 J/yr Production (m3) 3.82e7 J/m3 UEV = 6.8E+04 sej/J Bastiononi, et al., 2005 25 Oil Oil Production 4.0E+07 bbl/yr EIA, 2004 Oil Production 2.4E+17 J/yr Production (bbl) 6.12e9 J/bbl UEV 9.4E+04 sej/J Bastiononi, et al., 2005

PAGE 201

201 Table C-2. Continued. Note Variable Value Units Source 26 Minerals Mineral Production 9.4E+07 MT/yr BGS, some USGS Mineral Production 9.4E+13 g/yr Production (MT) 1e6 g/MT UEV (weighted avg.) 8.9E+09 sej/g Individua l UEVs assigned to each mineral; Cohen, 2006 g UEV UEV name and source barite 5.7E+10 9.5E+9limestone, Odum, 2000, Folio2 bentonite 2.1E+09 2.9E+9clay, Odum, 1996 kaolin 2.2E+11 2.9E+9clay, Odum, 1996 diatomite 3.9E+08 9.5E+9limestone, Odum, 2000, Folio2 feldspar 5.4E+11 1.7E+9avg. sediment, Odum, 2000, Folio2 fluorspar 4.7E+09 1.7E+9avg. sediment, Odum, 2000, Folio2 gypsum 5.8E+12 9.5E+9limestone, Odum, 2000, Folio2 talc 5.3E+10 2.9E+9clay, Odum, 1996 P rock 5.6E+09 6.5E+9P rock, Odum, 1996, FL salt 9.9E+11 6.6E+8salt, Babic, 2005 perlite 5.6E+09 1.7E+9avg. sediment, Odum, 2000, Folio2 sand 4.7E+06 1.7E+9avg. sediment, Odum, 2000, Folio2 dolomite 6.3E+11 9.5E+9limestone, Odum, 2000, Folio2 limestone 8.1E+13 9.5E+9limestone, Odum, 2000, Folio2 stone 5.5E+12 1.7E+9avg. sediment, Odum, 2000, Folio2 gemstone 1.9E+05 1E+14 assumed Minerals Summary 9.4E+13 8.9E+9(sum of mass and weighted avg. of UEV) 27 Metals Metal Production 3.9E+04 MT/yr BGS, some USGS Metal Production 3.9E+10 g/yr Production (MT) 1e6 g/MT UEV (weighted avg.) 2.6E+11 sej/g individual UEVs assigned to each metal; Cohen, 2006 g UEV UEV name and source Sb 8.4E+07 4E+12 Sb, Cohen et al., 2006 Fe ore 1.9E+08 5.8E+9Fe, Cohen et al., 2006 Pb 9.9E+09 5E+11 Pb, Cohen et al., 2006 Mn 1.1E+08 4E+11 Mn, Cohen et al., 2006 Sn 1.7E+09 2E+12 Sn, Cohen et al., 2006 W 2.8E+07 1E+12 W, Cohen et al., 2006 Zn 2.7E+10 7E+10 Zn, Cohen et al., 2006 Metals Summary 3.9E+10 2.6E11(sum of mass and weighted avg. of UEV) IMPORTS 28 Fuels Crude oil 2.4E+08 bbl/yr EIA, 2001, World Energy Database Crude oil 1.5E+18 J/yr Crude import (bbl) 5.85E9 J/bbl Crude oil UEV 9.4E+04 sej/J Bastiononi, et al., 2005 Refined oil (gasoline, etc) 1.3E+07 bbl/yr EIA, 2001, World Energy Database Refined oil 7.4E+16 J/yr Refined oil (bbl) 6.2E9 J/bbl Refined oil UEV 1.3E+05 sej/J Odum et al., 1995, Alaskan refined products Coal (hard, lignite) 4.2E+06 MT/yr EIA, 2001, World Energy Database Coal (hard, lignite) 1.2E+17 J/yr Coal (MT) 2.97e4 J/g Coal UEV 5.7E+04 sej/J Odum, 1996 Coal (coke) 8.3E+04 MT/yr EIA, 2001, World Energy Database Coal (coke) 2.4E+15 J/yr Coke (MT) 2.88e4 J/g

PAGE 202

202 Table C-2. Continued. Note Variable Value Units Source Coal UEV 5.7E+04 sej/J Odum, 1996 Natural gas 4.7E+10 ft^3/yr EIA, 2001, World Energy Database Natural gas 5.1E+16 J/yr Natural gas(ft3) 0.028317 m3/ft3 3.82e7 J/m3 Natural gas UEV 6.8E+04 sej/J Bastiononi, et al., 2005 Imp. fuels UEV(wt'd. avg.) 9.2E+04 sej/J 29 Metals Imported metals 9.3E+12 g/yr UN, COMTRADE, commodities reported in g Imported metals 1.1E+08 $/yr UN, COMTRADE, commod. reported only in $ UEV 1.7E+10 sej/g UEV wt'd avg. for 110 metals commodities 30 Minerals Imported minerals 1.0E+12 g/yr UN, COMTRADE, commodities reported in g Imported minerals 9.2E+08 $/yr UN, COMTRADE, commod. reported only in $ UEV 3.5E+09 sej/g UEV wt'd avg. for 20 mineral commodities 31 Food & agricultural products Food and ag. products 7.1E+16 J/yr UN, COMTRADE, 126 commod. (soybeans, wheat, cotton) UEV 2.7E+05 sej/J UEV wt'd avg. for 126 commodities 32 Livestock, meat, fish Livestock, meat, fish 7.3E+15 J/yr UN, COMT RADE, 59 commodities (fish prod., wool, whey) UEV 3.7E+06 varies UEV wt'd avg. for 59 commodities 33 Plastics & synthetic rubber Plastics & synthetic rubber 1.0E+12 g/yr UN, COMTRADE UEV 1.2E+10 sej/g UEV wt'd avg. for 46 commodities 34 Chemicals Chemicals 5.7E+16 J/yr UN, COMTRADE, items with UEVs in units of sej/J UEV 1.3E+05 sej/J UEV wt'd avg. for 9 commodities Chemicals 6.2E+12 g/yr UN, COMTRADE, items with UEVs in units of sej/g UEV 5.1E+09 sej/g UEV wt'd avg. for 90 commodities EMERGY 3.9E+22 sej/yr (__J/yr sej/J) + (___g/yr sej/g) 35 Finished products Wood and fiber products 4.1E+12 g/yr UN, COMTRADE,76 items w/UEVs in sej/J Wood and fiber products 1.2E+17 J/yr individual item conversion ratios UEV 1.3E+05 sej/J UEV wt'd avg. for 76 commodities Synthetic fiber, min. prod. 4.8E+11 g/yr UN, COMTRADE, 42 items with UEVs in units of sej/g UEV 1.1E+10 sej/g UEV wt'd avg. for 42 commodities EMERGY 2.1E+22 sej/yr (__J/yr sej/J) + (___g/yr sej/g) 36 Machinery & transportation equipment Machinery 2.7E+10 $/yr UN, COMTRADE, using $ value UEV 2.6E+12 sej/$ World sej/$, Sweeney et al., 2006 37 Other refined goods Other refined goods 4.3E+09 $/yr UN, COMTRADE, using $ value UEV 2.6E+12 sej/$ World sej/$, Sweeney et al., 2006 38 Electricity Electricity imports 3.0E+08 kwh/yr EIA, International Energy Annual 2002 Electricity imports 1.1E+15 J/yr Use(kwh/yr) 3.6e6 J/kwh Electricity UEV 2.9E+05 sej/J Odum, 1996, average from several types of power plants 39 Service in imports Dollar value of all imports 6.1E+10 $/yr UN, COMTRADE UEV 2.6E+12 sej/$ World sej/$, Sweeney et al., 2006

PAGE 203

203 Table C-2. Continued. Note Variable Value Units Source EXPORTS 40 Fuels Crude oil 0.0E+00 bbl/yr EIA, 2001, World Energy Database Crude oil UEV 9.4E+04 sej/J Bastiononi, et al., 2005 Refined oil (gasoline, etc) 3.8E+07 bbl/yr EIA, 2001, World Energy Database Refined oil 2.2E+17 J/yr Refined oil (bbl) 6.2E9 J/bbl Refined oil UEV 1.3E+05 sej/J Odum et al., 1995, Alaskan refined products Coal (hard, lignite) 0.0E+00 MT/yr EIA, 2001, World Energy Database Coal UEV 5.7E+04 sej/J Odum, 1996 Coal (coke) 0.0E+00 bbl/yr EIA, 2001, World Energy Database Coal UEV 5.7E+04 sej/J Odum, 1996 Natural gas 0.0E+00 ft3/yr EIA, 2001, World Energy Database Natural gasUEV 6.8E+04 sej/J Bastiononi, et al., 2005 41 Metals Metals 2.9E+12 g/yr UN, COMTRADE 103 commodities UEV 3.0E+10 sej/g UEV wt'd avg. for 103 metals commodities 42 Minerals Minerals 2.1E+13 g/yr UN, COMTRADE 22 commodities UEV 6.2E+09 sej/g UEV wt'd avg. for 22 metals commodities 43 Food & agricultural products Food and ag. products 2.9E+07 MT/yr UN, COMTRADE, 121 commodities (largest flows are rice, cassava, juice, sugar) Food and ag. products 3.1E+17 J/yr conversion to joules based on FAO reference energy contents UEV 8.1E+04 sej/J UEV wt'd avg. for 121 commodities 44 Livestock, meat, fish Livestock, meat, fish 1.9E+06 MT/yr UN, COMTRADE, 51 commodities (largest flows are fish products & shrimp) Livestock, meat, fish 7.9E+15 J/yr Conversion to joules based on FAO reference energy contents UEV 8.1E+06 sej/J UEV wt'd avg. for 51 commodities 45 Plastics & synthetic rubber Plastics & synthetic rubber 2.9E+12 g/yr UN, COMTRADE, 47 commodities UEV 1.2E+10 sej/g UEV wt'd avg. for 47 commodities 46 Chemicals Chemicals 6.6E+16 J/yr UN, COMTRADE, 10 commodities with TRFs in units of sej/J UEV 8.4E+04 sej/J UEV wt'd avg. for 9 commodities Chemicals 1.6E+12 g/yr UN, COMTRADE, 74 commodities with TRFs in units of sej/g UEV 4.8E+09 sej/g UEV wt'd avg. for 90 commodities EMERGY 1.3E+22 sej/yr (__J/yr sej/J) + (___g/yr sej/g) 47 Finished products Wood and fiber products 2.8E+16 J/yr UN, COMT RADE, 89 commodities with UEVs in sej/J units Wood and fiber products 1.1E+06 sej/J UEV wt'd avg. for 89 commodities Synth. fiber, mineral prod. 8.0E+11 g/yr UN, CO MTRADE, 42 commodities with UEVs in sej/g units UEV 4.5E+09 sej/g UEV wt'd avg. for 42 commodities EMERGY 3.3E+22 sej/yr (__J/yr sej/J) + (___g/yr sej/g) 48 Machinery & transportation equipment Mach. and trans. equip. 2.9E+10 $/yr UN, COMTRADE, using $ value UEV 1.5E+13 sej/$ Thai se j/$, Sweeney et al., 2006

PAGE 204

204 Table C-2. Continued. Note Variable Value Units Source 49 Other refined goods Other refined goods 1.3E+10 $/yr UN, COMTRADE, using $ value UEV 1.5E+13 sej/$ Thai se j/$, Sweeney et al., 2006 50 Electricity Electricity 2.0E+08 kwh/yr EIA, International Energy Annual 2002 Electricity 7.2E+14 J/yr Use(kwh/yr) 3.6e6 J/kwh UEV 2.9E+05 sej/J Odum, 1996, avg. from several types of power plants 51 Service in exports Dollar value of all imports 7.5E+09 $/yr UN, COMTRADE UEV 1.5E+13 sej/$ Thai se j/$, Sweeney et al., 2006 52 Tourism Tourist expenditures 7.5E+09 $/yr UNCDB, 2005 UEV 2.6E+12 sej/$ World sej/$, Sweeney et al., 2006

PAGE 205

205 APPENDIX D PROVINCIAL DATA TABLES These tables provide the data generated for th e provincial scale analysis. Pr esented first are the data for the water flow variables in units of mm per year. (Table D-1). Next, the renewable emergy flows are presented (Table D-2). Next, the economic data are presented (Table D-3).

PAGE 206

206 Table D-1. Provincial water flow values in units of mm/yr. Code Province Name rain.mm aet.mm aet .frac irg.mm ro.mm def.mo sm.mo 2 KANCHANABURI 1265 726 0.57 49 536 24 32 3 KALASIN 1280 983 0.77 67 299 35 45 5 KHON KAEN 1043 1004 0.96 34 37 36 34 6 CHANTHABURI 2243 907 0. 40 22 1371 20 51 7 CHACHOENGSAO 1273 1014 0.80 128 265 34 46 8 CHONBURI 1146 1032 0.90 20 135 35 37 9 CHAINAT 995 914 0.92 272 80 48 28 10 CHAIYAPHUM 968 830 0.86 29 135 34 30 15 TRAD 2839 958 0.34 30 1995 15 62 17 YASOTHON 1348 1025 0.76 19 324 40 57 18 NAKHON NAYOK 1590 967 0.61 202 626 28 58 19 NAKHON PATHOM 1009 904 0.90 477 110 37 49 20 NAKHON PHANOM 1821 977 0.54 52 863 30 57 21 NAKHON RATCHASIMA 1050 960 0.91 29 101 37 33 24 NONTHABURI 1168 1041 0. 89 390 129 36 60 27 BURIRAM 1165 1043 0.90 22 126 35 41 28 PATHUM THANI 1164 1042 0.90 443 127 42 48 29 PRACHUAP KHILIKHAN 920 798 0.87 89 186 32 26 30 PRACHINBURI 1657 961 0.58 111 693 30 45 32 BANGKOK 1281 1117 0.87 283 164 34 70 33 PRANAKHON SI AYUDHYA1047 977 0.93 461 73 45 41 38 PHETCHABURI 897 724 0.81 123 176 33 28 42 MAHA SARAKHAM 1117 1068 0.96 41 53 38 41 45 ROI ET 1212 1041 0.86 41 172 38 53 47 RAYONG 1411 1043 0.74 34 392 27 50 48 RATCHABURI 1000 807 0.81 272 196 31 41 49 LOPBURI 1089 981 0.90 120 107 41 29 52 LOEI 1127 837 0.74 26 322 22 42 53 SISAKET 1324 1045 0.79 17 289 36 57 54 SAKON NAKHON 1514 951 0.63 114 563 32 54 57 SAMUT PRAKARN 1196 1141 0.95 278 81 26 86 58 SAMUT SONGKHAM 1023 905 0.88 321 131 26 64 59 SAMUT SAKHON 1123 996 0.89 157 143 26 78 60 SARABURI 1199 961 0.80 103 243 36 34 61 SINGBURI 943 892 0.95 502 52 50 32 63 SUPHANBURI 971 845 0.87 214 123 44 31 65 SURIN 1277 1072 0.84 20 213 34 53 66 NONG KHAI 1906 1028 0.54 57 888 31 59 67 ANG THONG 951 900 0. 95 467 50 51 30 68 UDON THANI 1322 940 0.71 34 381 35 48 71 UBON RATCHATHANI 1563 1006 0.64 16 571 34 53 73 MUKDAHAN 1480 867 0.59 45 616 29 42 74 SA KAEO 1311 969 0.74 18 345 34 36 75 NONG BUA LAMPHU 1123 958 0.85 25 163 34 40 76 AMNAT CHAROEN 1453 1013 0.70 10 443 40 48 See Table 2-1 for variable descriptions.

PAGE 207

207Table D-2. Renewable emer gy flows in units of sej/m2/yr. Code Province name rad wind heat rain aet ROgeo tide irg ROchem renew.W renew.WI 2 KANCHANABURI 5.1E+09 5.5E+09 1.9E+10 1.9E+11 1.1E+11 1.1E+11 0 2.9E+10 0 2.2E+11 2.5E+11 3 KALASIN 5.2E+09 3.4E+09 1.8E+10 1.9E+11 1. 5E+11 1.2E+10 0 2.8E+10 0 1.6E+11 1.9E+11 5 KHON KAEN 5.2E+09 2.8E+09 1.9E+10 1.6E+11 1.5E+11 5.7E+09 0 1.9E+10 0 1.6E+11 1.8E+11 6 CHANTHABURI 5.0E+09 5.1E+09 1.9E+10 3.4E+11 1.4E+1 1 9.4E+10 1.0E+12 7.6E+09 2.1E+11 3.4E+11 3.5E+11 7 CHACHOENGSAO 5.0E+09 1.1E+10 1.8E+10 1.9E+11 1.5E+1 1 1.1E+10 2.4E+11 9.0E+10 2.0E+11 3.5E+11 4.4E+11 8 CHONBURI 4.9E+09 2.7E+10 1.8E+10 1.7E+11 1.6E+11 8.0E+09 2.5E+12 1.0E+10 3.1E+11 4.6E+11 4.7E+11 9 CHAINAT 5.2E+09 5.7E+09 1.8E+10 1.5E+11 1. 4E+11 5.3E+10 0 2.1E+11 0 1.9E+11 4.1E+11 10 CHAIYAPHUM 5.1E+09 5.0E+09 2.0E+10 1.5E+1 1 1.3E+11 2.7E+10 0 1.2E+10 0 1.5E+11 1.6E+11 15 TRAD 5.0E+09 6.8E+09 1.9E+10 4.3E+11 1.4E+11 7.4E+10 1.2E+12 9.1E+09 3.3E+11 4.7E+11 4.8E+11 17 YASOTHON 5.2E+09 5.2E+09 2.2E+10 2.0E+11 1.5E+11 7.7E+09 0 9.6E+09 0 1.6E+11 1.7E+11 18 NAKHON NAYOK 5.1E+09 8.6E+09 1.7E+10 2.4E+1 1 1.5E+11 6.6E+10 0 1.1E+11 0 2.1E+11 3.2E+11 19 NAKHON PATHOM 5.1E+09 1.1E+10 1.7E+10 1.5E+11 1.4E+11 1.3E+09 5.6E+09 4.0E+11 0 1.4E+11 5.4E+11 20 NAKHON PHANOM 5.0E+09 4.3E+09 1.7E+10 2.8E +11 1.5E+11 1.7E+10 0 2.0E+10 0 1.6E+11 1.8E+11 21 NAKHON RATCHASIMA 5.1E+09 3.9E+09 1.9E+10 1.6E+11 1.4E+11 1.8E+10 0 1.4E+10 0 1.6E+11 1.8E+11 24 NONTHABURI 5.1E+09 1.4E+10 1.6E+10 1.8E+11 1. 6E+11 4.9E+09 1.4E+09 3.6E+11 0 1.6E+11 5.2E+11 27 BURIRAM 5.2E+09 2.9E+09 2.0E+10 1.8E+11 1.6E+11 7.2E+09 0 9.9E+09 0 1.6E+11 1.7E+11 28 PATHUM THANI 5.1E+09 1.2E+10 1.5E+10 1.8E+11 1.6E+11 2.4E+09 0 4.1E+11 0 1.6E+11 5.7E+11 29 PRACHUAP KHILIKHAN 5.0E+09 1.7E+10 1.9E+10 1.4E+11 1. 2E+11 2.1E+10 2.4E+11 4.5E+10 2.8E+10 1.5E+11 1.9E+11 30 PRACHINBURI 5.0E+09 5.6E+09 1.8E+10 2.5E+11 1.4E+11 6.0E+10 0 5.6E+10 0 2.1E+11 2.6E+11 32 BANGKOK 5.1E+09 1.3E+10 1.6E+10 1.9E+11 1.7E+11 2.5E+08 6.6E+10 2.6E+11 2.1E+11 3.7E+11 6.4E+11 33 PRANAKHON SI AYUDHYA 5.1E+09 1.1E+10 1.5E+10 1.6E+11 1.5E+11 1.8E+10 0 4.0E+11 0 1.7E+11 5.6E+11 38 PHETCHABURI 5.0E+09 1.3E+10 1.9E+10 1.4E+11 1.1E+1 1 3.8E+10 4.5E+11 7.7E+10 1.6E+11 2.7E+11 3.5E+11 42 MAHA SARAKHAM 5.3E+09 1.9E+09 2.0E+10 1.7E +11 1.6E+11 3.7E+09 0 2.2E+10 0 1.6E+11 1.9E+11 45 ROI ET 5.3E+09 3.9E+09 2.1E+10 1.8E+11 1.6E+11 4.9E+09 0 1.8E+10 0 1.6E+11 1.8E+11 47 RAYONG 4.9E+09 3.4E+10 1.9E+10 2.1E+11 1.6E+11 1.2E+10 1.2E+12 1.2E+10 5.9E+10 2.2E+11 2.3E+11 48 RATCHABURI 5.1E+09 1.0E+10 1.8E+10 1.5E+11 1.2E+11 3.1E+10 5.6E+08 1.8E+11 0 1.5E+11 3.3E+11 49 LOPBURI 5.1E+09 7.1E+09 1.9E+10 1.6E+11 1.5E+11 9.3E+09 0 8.1E+10 0 1.6E+11 2.4E+11 52 LOEI 5.0E+09 2.5E+09 1.7E+10 1.7E+11 1. 3E+11 5.0E+10 0 1.0E+10 0 1.8E+11 1.9E+11 53 SISAKET 5.3E+09 5.3E+09 2.1E+10 2.0E+11 1.6E+11 1.7E+10 0 7.4E+09 0 1.7E+11 1.8E+11 54 SAKON NAKHON 5.1E+09 4.1E+09 1.7E+10 2.3E+11 1.4E+11 2.0E+10 0 4.9E+10 0 1.6E+11 2.1E+11 57 SAMUT PRAKARN 5.1E+09 1.2E+10 1.6E+10 1.8E+11 1.7E +11 2.9E+08 2.7E+12 2.5E+11 2.6E+12 2.8E+12 3.0E+12 58 SAMUT SONGKHAM 5.0E+09 1.4E+10 1.8E+10 1.5E+11 1.4E+11 9.9E+08 3.2E+12 2.3E+11 3.2E+12 3.3E+12 3.5E+12 59 SAMUT SAKHON 5.1E+09 1.3E+10 1.6E+10 1.7E+11 1.5E+11 3.0E+08 2.7E+12 1.4E+11 2.7E+12 2.8E+12 3.0E+12 60 SARABURI 5.1E+09 8.0E+09 1.7E+10 1.8E+11 1.4E+11 2.2E+10 0 7.5E+10 0 1.7E+11 2.4E+11 61 SINGBURI 5.1E+09 8.1E+09 1.8E+10 1.4E+11 1.3E+11 9.7E+10 0 4.2E+11 0 2.3E+11 6.5E+11

PAGE 208

208Table D-2. Continued. Code Province name rad wind heat rain aet ROgeo tide irg ROchem renew.W renew.WI 63 SUPHANBURI 5.1E+09 1.1E+10 1.8E+10 1.5E+11 1.3E+11 1.1E+10 0 1.9E+11 0 1.4E+11 3.3E+11 65 SURIN 5.3E+09 3.3E+09 2.0E+10 1.9E+11 1. 6E+11 9.3E+09 0 8.7E+09 0 1.7E+11 1.8E+11 66 NONG KHAI 5.1E+09 2.9E+09 1.9E+10 2.9E+1 1 1.5E+11 2.0E+10 0 2.6E+10 0 1.7E+11 2.0E+11 67 ANG THONG 5.1E+09 1.2E+10 1.7E+10 1.4E+11 1.4E+11 8.9E+09 0 4.3E+11 0 1.4E+11 5.7E+11 68 UDON THANI 5.1E+09 2.6E+09 1.9E+10 2.0E+1 1 1.4E+11 1.4E+10 0 1.5E+10 0 1.6E+11 1.7E+11 71 UBON RATCHATHANI 5.3E+09 8.5E+09 2.2E+10 2.4E+11 1.5E+11 2.7E+10 0 6.8E+09 0 1.8E+11 1.8E+11 73 MUKDAHAN 5.1E+09 8.1E+09 1.9E+10 2.2E+11 1.3E+11 3.4E+10 0 1.7E+10 0 1.6E+11 1.8E+11 74 SA KAEO 5.0E+09 5.2E+09 1.9E+10 2.0E+11 1.5E+11 2.0E+10 0 6.1E+09 0 1.7E+11 1.7E+11 75 NONG BUA LAMPHU 5.1E+09 2.7E+09 1.9E+10 1.7E +11 1.4E+11 7.4E+09 0 1.1E+10 0 1.5E+11 1.6E+11 76 AMNAT CHAROEN 5.2E+09 6.7E+09 2.3E+10 2.2E+1 1 1.5E+11 1.1E+10 0 3.9E+09 0 1.6E+11 1.7E+11 See Table 2-1 for variable descriptions.

PAGE 209

209Table D-3. Selected provinci al socioeconomic variables. Code pdens.B pdens.C Gcap.B Gcap.C Gdens.B Gdens.C agGden.B agGden.C pov.incid cagr.B cagr. C IHD.comp HAI.comp HAI.inc (#/km2) (#/km2) (B/pers) (B/pers) (B/km2) (B/km2) (B/km2) (B/km2) (#) (%) (%) (unitless) (unitless) (unitless) 2 32 39 33,075 38,932 1.07 1.51 0.28 0.24 10 4.3 -1.7 0.32 0.59 0.48 3 121 135 9,476 14,172 1.15 1.92 0.38 0.43 40 5.1 2.9 0.47 0.54 0.30 5 149 163 14,020 25,854 2.09 4.22 0.39 0.49 17 7.9 3.4 0.39 0.58 0.41 6 60 76 22,780 28,772 1.37 2.18 0.43 0.52 7 7.2 -0.7 0.23 0.64 0.48 7 99 119 36,702 83,789 3.64 10.01 0.83 0.92 0 11.2 3.3 0.22 0.64 0.57 8 183 233 81,288 157,909 14.84 36.75 1.11 1.36 0 9.8 3.0 0.22 0.70 0.70 9 138 147 19,915 30,927 2.75 4.54 0.89 1.32 9 3.7 0.3 0.19 0.60 0.63 10 76 87 10,904 15,527 0.83 1.34 0.27 0.41 17 7.5 0.7 0.49 0.54 0.47 15 58 77 28,509 36,672 1.64 2.81 0.57 1.07 15 8.5 -2.3 0.22 0.63 0.45 17 124 137 9,209 13,164 1.14 1.81 0.34 0.38 50 5.0 1.0 0.44 0.54 0.23 18 100 113 18,679 28,651 1.86 3.23 0.50 0.70 5 6.1 1.9 0.34 0.61 0.56 19 297 389 30,769 60,209 9.14 23.41 1.88 2.14 1 10.2 0.8 0.16 0.68 0.66 20 111 122 9,687 13,479 1.07 1.64 0.35 0.36 48 6.3 -0.4 0.38 0.48 0.25 21 110 127 14,194 21,968 1.57 2.78 0.41 0.51 21 8.1 1.4 0.45 0.57 0.40 24 839 1314 44,986 50,751 37.73 66.67 1.67 1.40 0 9.7 0.3 0.16 0.71 0.76 27 130 149 9,616 13,706 1.25 2.04 0.40 0.47 26 6.1 0.0 0.58 0.50 0.33 28 269 443 94,828 135,328 25.53 59.93 1.36 1.14 2 11.6 -3.8 0.14 0.66 0.69 29 63 71 28,473 42,655 1.80 3.03 0.65 0.61 9 7.4 1.6 0.21 0.63 0.58 30 65 84 15,549 58,980 1.00 4.97 0.32 0.51 2 6.9 3.9 0.29 0.63 0.60 32 3754 4156 106,036 132,219 398.01 549.45 2.62 0.77 0 9.1 -0.9 0.32 0.67 0.81 33 261 290 26,015 177,862 6.79 51.57 0.88 1.01 5 11.6 9.3 0.14 0.65 0.70 38 64 70 24,818 41,355 1.60 2.91 0.32 0.31 4 5.6 4.1 0.18 0.65 0.55 42 152 167 9,854 14,066 1.50 2.35 0.47 0.56 13 5.1 1.2 0.50 0.55 0.44 45 147 163 9,802 13,964 1.44 2.28 0.44 0.50 22 5.3 1.9 0.44 0.54 0.43 47 116 144 68,975 286,470 8.00 41.18 1.44 1.33 1 14.8 6.8 0.20 0.67 0.67 48 134 154 25,171 45,332 3.37 6.97 0.64 0.95 2 7.1 3.9 0.19 0.62 0.64 49 113 122 18,949 42,153 2.13 5.14 0.61 0.70 8 4.5 2.7 0.15 0.61 0.53 52 49 58 13,122 16,262 0.64 0.94 0.27 0.27 37 3.8 1.6 0.49 0.55 0.24 53 139 160 8,603 11,970 1.19 1.91 0.39 0.46 27 5.9 1.0 0.58 0.49 0.34 54 95 110 9,857 13,388 0.93 1.47 0.27 0.30 41 5.8 -0.5 0.57 0.52 0.39 57 752 994 111,409 210,752 83.77 209.46 5.08 5.27 1 7.1 5.4 0.37 0.64 0.65 58 507 517 17,410 30,382 8.83 15.70 1.78 1.36 0 6.0 -1.5 0.21 0.66 0.70 59 375 540 77,442 207,614 29.06 112.11 6.34 2.94 5 14.3 3.2 0.19 0.62 0.56 60 140 168 46,367 79,524 6.51 13.35 0.65 0.67 3 10.9 2.1 0.13 0.65 0.60

PAGE 210

210Table D-3. Continued. Code pdens.B pdens.C Gcap.B Gcap.C Gdens.B Gdens.C agGden.B agGden.C pov.incid cagr.B cagr. C IHD.comp HAI.comp HAI.inc (#/km2) (#/km2) (B/pers) (B/pers) (B/km2) (B/km2) (B/km2) (B/km2) (#) (%) (%) (unitless) (unitless) (unitless) 61 261 285 19,993 35,319 5.22 10.05 1.32 1.69 6 3.3 2.4 0.11 0.64 0.64 63 146 161 19,464 27,612 2.84 4.44 0.94 1.29 8 6.1 1.3 0.15 0.63 0.59 65 136 154 9,194 12,714 1.25 1.96 0.37 0.44 41 5.0 0.5 0.68 0.53 0.30 66 108 125 11,266 13,855 1.22 1.73 0.41 0.33 36 6.1 -0.3 0.59 0.53 0.24 67 279 291 18,699 34,101 5.21 9.91 1.22 1.60 5 4.3 0.6 0.18 0.65 0.63 68 112 133 10,526 17,690 1.18 2.36 0.34 0.41 35 5.4 -0.2 0.48 0.53 0.33 71 95 109 9,537 16,104 0.90 1.76 0.23 0.30 18 5.8 0.3 0.45 0.54 0.42 73 60 76 11,257 15,266 0.67 1.16 0.20 0.25 24 8.9 -0.4 0.44 0.53 0.40 74 75 21,422 1.61 0.41 23 1.5 0.27 0.58 0.46 75 128 9,960 1.27 0.37 50 0.3 0.44 0.50 0.15 76 119 11,433 1.36 0.39 22 -0.2 0.42 0.51 0.37 See table 2-5 for variable definitions.

PAGE 211

211 APPENDIX E CORRELATION MATRICES These tables contain the correlation matrices that were investigated in the initial stages of exploratory data analysis. Table E-1 contains the matr ix for correlations among the environmental variables. Table E-2 contai ns the matrix for correlations among the socioeconomic variables. Table E-3 contains the matrix for water fl ows and socioeconomic variables. Table E-4 contains the matrix for emergy flows and socioeconomic variables.

PAGE 212

212Table E-1. Pearson correlation matix for environm ental variables, with P-values in italics. strmd rain.mm aet.mm ro.mmir g .mmdef.mosm.morainaetRO g eo windhea t ir g ROchetideaetir g renew.W rain.mm -0.06 0.71 aet.mm -0.06 -0.06 0.71 0.71 ro.mm -0.17 *0.86 -0.09 0.2 7 0.00 0.58 ir g .mm *0.80 -0.43 -0.13 -0.44 0.00 0.00 0.39 0.00 def.mo 0.10 -0.56 0.10 *-0.660.32 0.51 0.00 0.53 0.000.03 sm.mo 0.19 0.58 *0.62 0.360.01-0.46 0.21 0.00 0.00 0.020.9 7 0.00 rain -0.13 *1.00 0.28 *0.86-0.40-0.580.53 0.40 0.00 0.0 7 0.000.010.000.00 aet -0.06 0.36 *1.00 -0.08-0.130.10*0.620.28 0.70 0.02 0.00 0.590.390.530.000.0 6 RO g eo -0.06 0.28 0.28 0.48-0.13-0.24-0.340.34-0.57 0.68 0.0 7 0.0 7 0.000.400.110.020.020.00 wind 0.64 -0.22 -0.06 -0.130.480.030.05-0.2-0.06-0.15 0.00 0.14 0.70 0.410.000.8 7 0.730.180.690.31 hea t -0.69 0.20 0.00 0.25-0.780.00-0.150.20.010.16 -0.40 0.00 0.18 0.98 0.110.001.000.310.180.9 7 0.29 0.01 ir g *0.79 -0.48 -0.10 -0.50*0.990.370.00-0.46-0.10-0.18 0.49-0.77 0.00 0.00 0.51 0.000.000.010.980.000.510.24 0.000.00 ROchem 0.34 0.06 0.06 0.090.1-0.480.30.110.01-0.14 0.57-0.190.10 0.02 0.69 0.70 0.5 7 0.530.000.050.4 6 0.720.38 0.000.210.53 tide 0.33 0.06 0.11 0.030.08-0.490.380.10.11-0.19 0.54-0.190.08*0.77 0.03 0.70 0.4 7 0.840.620.000.010.520.490.21 0.000.210.600.00 aetir g *0.76 -0.41 0.15 -0.54*0.920.470.14-0.40.15-0.32 0.50-0.68*0.940.080.10 0.00 0.01 0.32 0.000.000.000.3 6 0.010.320.03 0.000.000.000. 6 0.51 renew.W 0.38 0.18 0.17 0.100.14-0.50.490.210.17-0.02 0.4-0.230.15*0.69*0.870.16 0.01 0.23 0.2 6 0.510.3 7 0.000.000.1 7 0.2 7 0.88 0.010.120.340.000.000.30 renew.WI *0.81 -0.20 0.01 -0.27*0.770.000.26-0.140.01-0.10 *0.67-0.67*0.780.550.58*0.79*0.65 0.00 0.20 0.95 0.080.001.000.090.340.9 6 0.53 0.000.000.000.000.000.000.00 *Values with r > 0.6. See Table 2-1 for variable definitions.

PAGE 213

213Table E-2. Pearson correlation matix for socio-ec onomic variables, with P-values in italics. dist2bk p o p d.B p o p d.CGden.BGden.Ca g Gden.a g Gden.CGca p .B Gca p .CGca g r.Gca g r.C p ov.incIHDHAI.in p o p dB 0.18 0.2 6 p opd.C 0.08*0.97 0.590.0 0 Gden.B *-0.86-0.10-0.01 0.000.520.94 Gden.C *-0.87-0.100.01*0.97 0.000.550.940.00 agGden.B *-0.81-0.20-0.12*0.950.91 0.000.210.4 6 0.000.00 agGden.C *-0.77-0.29-0.19*0.870.87*0.94 0.000.0 6 0.210.000.000.00 Gcap.B *-0.86-0.19-0.10*0.840.82*0.780.69 0.000.240.530.000.000.000.00 Gcap.C *-0.87-0.22-0.10*0.810.88*0.750.71*0.94 0.000.1 7 0.510.000.000.000.000.00 G.cagr.B -0.48-0.060.040.540.600.460.420.580.66 0.000.7 0 0.810.000.000.000.010.000.0 0 G.cagr.C -0.260.030.000.230.360.210.290.220.420.30 0.090.880.980.140.020.180.050.1 6 0.0 0 0.0 6 p ovincid *0.890.230.15*-0.79-0.84-0.72-0.72*-0.82*-0.87-0.48-0.28 0.000.140.330.000.000.000.000.000.0 0 0.000.0 6 IHD 0.730.500.45-0.66-0.69-0.63-0.65-0.73*-0.78-0.39-0.25*0.75 0.000.0 0 0.000.000.000.000.000.000.0 0 0.010.100.00 HAI.inc *-0.89-0.20-0.11*0.830.86*0.750.74*0.80*0.860.460.27*-0.95*-0.77 0.000.210.4 7 0.000.000.000.000.000.0 0 0.000.080.000.00 HAI *-0.88-0.34-0.22*0.800.840.740.72*0.87*0.890.470.27*-0.91*-0.84*0.90 0.000.030.140.000.000.000.000.000.0 0 0.000.0 7 0.000.000.00 *Values with r > 0.75. See Table 2-5 for variable definitions.

PAGE 214

214 Table E-3. Pearson correlation matrix for water flows and socioeconomic variables, with Pvalues in italics. Rain.mm aet.mm ro.mmirg.mmae tirg.mmdef.mosm.moPC1w PC2wPC3w PC1w *0.80 0.21 *0.79*-0.77-0.53*-0.750.48 0.00 0.17 0.000.000.000.000.00 PC2w -0.32 -0.53 -0.100.49-0.680.23*-0.830.00 0.03 0.00 0.530.000.000.130.001.00 PC3w -0.04 *0.80 -0.42-0.280.390.470.170.00 0.00 0.79 0.00 0.000.060.010.000.281.00 1.00 dist2bk 0.35 -0.05 0.39*-0.72*-0.63-0.08-0.080.53 0.500.18 0.02 0.74 0.010.000.000.630.610.00 0.000.25 popd.C -0.27 0.59 -0.550.55*0.860.370.44-0.46 *-0.670.46 0.08 0.00 0.000.000.000.010.000.00 0.000.00 Gden.C -0.31 0.29 -0.47*0.70*0.740.200.20-0.54 -0.590.07 0.04 0.05 0.000.000.000.180.190.00 0.000.64 agGden.C -0.32 0.26 -0.49*0.65*0.740.220.17-0.54 -0.540.08 0.03 0.08 0.000.000.000.140.260.00 0.000.59 Gcap.C -0.22 0.02 -0.25*0.600.42-0.040.04-0.40 -0.44-0.24 0.15 0.88 0.100.000.000.780.780.00 0.030.11 G.cagr.B 0.07 0.27 0.000.160.26-0.190.280.02 -0.410.00 0.64 0.09 0.990.320.090.240.080.90 0.010.98 G.cagr.C -0.30 0.07 -0.320.170.150.13-0.12-0.25 -0.030.01 0.05 0.66 0.030.260.310.390.450.09 0.850.52 pov.incid 0.30 0.02 0.32*-0.64-0.46-0.03-0.020.46 0.400.22 0.05 0.91 0.030.000.000.860.920.00 0.010.15 IHD 0.33 0.29 0.24*-0.64-0.33-0.180.25-0.58 0.210.42 0.03 0.05 0.110.000.030.250.100.00 0.180.01 HAI.inc -0.36 -0.02 -0.38*0.720.550.18-0.04-0.58 -0.39-0.18 0.02 0.90 0.010.000.000.230.820.00 0.010.23 HAI -0.29 -0.08 -0.29*0.650.440.04-0.05-0.50 -0.38-0.30 0.05 0.59 0.050.000.000.820.760.00 0.010.05 PC1b -0.36 -0.21 -0.30*0.650.370.17-0.20-0.57 -0.23-0.33 0.02 0.19 0.060.000.020.270.200.00 0.140.03 PC1c -0.29 -0.02 -0.31*0.690.490.030.02-0.50 -0.45-0.26 0.05 0.91 0.040.000.000.830.920.00 0.000.08 *Values with r > 0.6. See table 2-4 and 2-5 for variable descriptions.

PAGE 215

215 Table E-4. Correlation matrix of emergy flows and socioeconomi c variables, with P-values in italics. irg aetirg ROchem renew.Wrenew.WIPC1em PC2em PC3em PC1em *-0.70 *-0.62 *-0.64-0.52*-0.84 0.00 0.00 0.000.000.00 PC2em 0.22 0.43 0.010.050.140.00 0.15 0.00 0.950.750.371.00 PC3em *-0.61 -0.51 *0.680.50-0.110.00 0.00 0.00 0.00 0.000.000.481.00 1.00 dist2bk *-0.75 *-0.74 -0.44-0.45*-0.83*-0.81 -0.20 0.14 0.00 0.00 0.000.000.000.00 0.00 0.00 popd.C 0.59 *0.77 0.170.240.570.40 *0.80 -0.15 0.00 0.00 0.260.110.000.01 0.00 0.31 Gdens.C *0.74 *0.80 0.450.43*0.81*0.77 0.45 -0.06 0.00 0.00 0.000.000.000.00 0.00 0.70 agGden.C *0.67 *0.77 0.410.40*0.79*0.67 0.38 -0.03 0.00 0.00 0.010.010.000.00 0.01 0.84 Gcap.C *0.62 0.58 0.550.49*0.78*0.86 0.09 0.03 0.00 0.00 0.000.000.000.00 0.57 0.82 G.cagr.B 0.17 0.21 0.420.260.340.55 0.28 0.29 0.30 0.19 0.010.100.030.00 0.07 0.06 G.cagr.C 0.18 0.18 0.150.080.140.26 0.08 0.04 0.25 0.24 0.330.580.370.08 0.59 0.80 pov.incid *-0.66 *-0.64 -0.51-0.47*-0.80*-0.80 -0.10 -0.01 0.00 0.00 0.000.000.000.00 0.52 0.96 IHD *-0.64 *-0.60 -0.34-0.23*-0.70*-0.77 0.16 0.14 0.00 0.00 0.020.120.000.00 0.29 0.37 HAI.inc *0.73 *0.73 0.410.36*0.79*0.80 0.14 -0.13 0.00 0.00 0.010.020.000.00 0.37 0.40 HAI *0.66 *0.64 0.510.42*0.81*0.85 0.02 -0.02 0.00 0.00 0.000.010.000.00 0.92 0.90 PC1b *0.66 *0.62 0.400.28*0.75*0.79 -0.08 -0.07 0.00 0.00 0.010.080.000.00 0.61 0.65 PC1c *0.71 *0.68 0.520.46*0.84*0.88 0.09 -0.04 0.00 0.00 0.000.000.000.00 0.55 0.78 *Values with r > 0.6. See Tables 2-4 and 2-5 for variable de finitions. Rain and aet emergy flows are not shown b ecause correlation coefficients are identical to those for rain.mm and aet.mm in Table E-3.

PAGE 216

216 LIST OF REFERENCES Ahn, C.H., Tateishi, R.,1994. Devlopm ent of gl obal land surface evapotra nspiration and water balance data sets. Journal of Japan Society of Photogrammetry and Remote Sensing. 33(5), 48-61. Araujo, C., 2003. Spatial patterns of non-agricultura l rural employment growth in Mexico during the 90s. Department of Agricultural and Resource Economic s, University of California, Berkeley, http://are.berkeley.edu/~ araujo/Araujo-paper2.pdf accessed May 2009. Bhattarai, M., Sakthivadivel, R. and Hussain, I., 2002. Irrigation impacts on income inequality and poverty alleviation: Policy issues and optio ns for improved management of irrigation systems. Working Paper 39. International Wa ter Management Institute Colombo, Sri Lanka, 29pp. Bouma, J., 1989. Using soil survey data for quantitativ e land evaluation. Advances in Soil Science 9, 177. Broad, R., 1994. The poor and the environment: Fr iend or foes? World Development 22(6), 811822. Brown, M.T., McClanahan, T.R., 1996. Emergy anal ysis perspectives of Thailand and Mekong River dam proposals. Ecological Modelling 91, 105-130. Brown, M.T., Ulgiati, S., 1997. Emergy-based indi ces and ratios to evaluate sustainability: monitoring economies and technology to ward environmentally sound innovation. Ecological Engineering 9, 51-69. Brown, M.T., Ulgiati, S., 2004. Emergy analysis and environmental accounting. In: Cleveland, C.J. (Ed.), Encyclopedia of Energy, Vo lume 2. Elsevier, Oxford, UK, pp. 329-354. Brown, M.T., Tennenbaum, S. Odum, H.T., 1991. Em ergy Analysis and Policy Perspectives for the Sea of Cortez, Mexico. Report to th e Cousteau Society. Center for Wetlands, University of Florida, Gainesville, FL, USA, 58pp. Brutsaert, W., 1982. Evaporation in the atmosphere : Theory, history, and a pplications. D. Reidel, Higham, MT, USA. 299pp. Buenfil, A., 1998. Emergy Evaluation of Water. PhD Dissertation, University of Florida, Gainesville, FL, USA, 248pp. CIA, 2005. The World Factbook, http://www.odci.gov/cia/pub lications/factbook/index.html, accessed Jan uary 2006. Cohen, M.J., Sweeney, S., King, D., Brown, M.T., 2006. Soil, water, fish and forests: natural capital in the wealth of na tions. In: Brown, M.T. (Ed.), Em ergy Synthesis 5: Theory and Applications of the Emergy Methodology. Center for Environmental Policy, University of Florida, Gainesville, FL, USA, 18.1-18.22.

PAGE 217

217 Cramer, D., 2003. A cautionary tale of two statistics: Pa rtial correlation and st andardized partial regression. Journal of Psychology 137(5), 507-511. Deolalikar, A.B., 2002. Poverty, gr owth, and inequality in Thailand. ERD Working Paper No. 8. Asian Development Bank, Manilla, Philippines, 26pp. Diamond, C., 1984. Energy basis for the regional or ganization of the Mississippi River Basin. MS Thesis, University of Florida, Gainesville, FL, USA, 129pp. Diamond, J.M., 1997. Guns, Germs and Steel: The Fa tes of Human Societies. W.W. Norton and Company, New York, NY, USA, 528pp. Donner, W., 1978. The Five Faces of Thailand. Un iversity of Queensland Press, Brisbane, Australia, 930pp. Dunne, T., Leopold, L.B., 1978. Water in Enviro nmental Planning. Freeman, New York, NY, USA, 818pp. Dupra, V., Smith, S.V., Crossland, J.I.M., Cro ssland, C.J., 2000. Estuarine Systems of the South China Sea Region: Carbon, Nitrogen and Phos horous Fluxs. LOICZ Reports and Studies No. 14. LOICZ International Project Of fice, Texel, The Netherlands, 16pp., http://www.loicz.org/imperia/md/conten t/loicz/print/rsreports/14report.pdf accessed May 2005. EIA, 2001. World Energy Database. Energy Inform ation Adm inistration, Washington, D.C., http://www.eia.doe.gov/emeu/world/main1.html accessed January 2005. EIA, 2004. International Ener gy Annual 2004. Energy Information Adm inistration, Washington, D.C., http://www.eia.doe.gov/emeu/world/main1.html accessed May 2005. Falkenmark, M., 1989. Com parativ e hydrology: A new concept. In : Falkenmark, M., Chapman, T. (Eds.), Comparative Hydrol ogy: An Ecological Approach to Land and Water Resources. The Unesco Press, Paris, France, pp. 8-34. Falvey, L., 2000. Thai Agriculture. Kasetsar t University Press, Bangkok, Thailand, 459 pp. FAO, 1999. Irrigation in Asia in Figures. Wate r Reports No. 18. Land and Water Development Division, FAO, Rome, ftp://ftp.fao.org/agl/aglw/docs/wr18.pdf accessed October 2006. FAO, 2003. Digital Soil Map of the World and Derived Soil Properties (CD-ROM). FAO/UNESCO Land and Water Digital Medi a Series, L and and Water Development Division, FAO, Rome, Italy. FAO, 2000. FAOSTAT online databa se, http://faostat.fao.org/de fault.aspx, accessed May 2003.

PAGE 218

218 Federer, C.A., Vrsmarty, C.J., Fekete, B., 1996. Intercomparison of methods for potential evapotranspiration in regional or global wate r balance models. Water Resources Research, 32, 2315-21. Felkner, J.S., Townsend. R.M., 2007. Enterprise and the wealth of villages. Working paper, University of Chicago, Chicago, IL, USA, http://enterpriseinitiative.org/Arete/WorkingPapers.aspx accessed April 2009. Fekete, B.M., Vrsmarty, C.J., Grabs, W. 2000. UNH/GRDC Co mposite Runoff Fields V1.0, University of New Hampshire, Durham, NH, USA and Global Runoff Data Center, Koblenz, Germany, http://www.grdc.sr.unh.edu accessed July 2005. Fekete, B. M., Vorosmarty, C. J., Roads, J. O., Willm ot, C. J., 2004. Uncertainties in precipitation and their impacts on runoff estimates, Journal of Climate 17, 294-304. Fox, D.G., 1981. Judging air quality model perf ormance: A summary of the AMS Workshop on Dispersion Model Performance. Bulletin of the American Meteorol ogical Society. 62, 599609. Gallup, J.L., Sachs, J.D., Mellinger, A. D., 1999. Geography and economic development. International Regional Science Review 22(2), 179. Gijsman, A.J., Jagtap, S.S., Jones, J.W., 2003. Wading through a swap of complete confusion, how to choose a method for estimating soil wa ter retention parameters for crop models. European Journal of Agronomy 18, 77. Glassman, J., Sneddon, C., 2003. Chiang Mai and Khon Kaen as growth poles: Regional industrial development in Thailand and its im plications for urban sustainability. ANNALS 590, 93-115. GRDC (The Global Runoff Data Centre), 2004. Long Term Mean Monthly Discharges and Annual Characteristics of Selected GRDC Stations dataset. D-56002 Koblenz, Germany, http://www.bafg.de/GRDC/Home/homepage__node.html accessed 2004. Hibbs, D., Olsson, O., 2004. Geography, biogeography, and why some countries are rich and others are poor. Proceedings of the National Academy of Sciences 101, 3715-3720. Horton, R.R. 1945. Erosional developments of str eams and their drainage basins: Hydrophysical approach to quantitative morphology. Geologi cal Society of Amer ica Bulletin. 56: 275370. Huang, S., Kao, W., Lee, C., 2007. Energetic mechanisms and development of an urban landscape system. Ecological Modelling 201, 495-506. Intute, 2009. World Guide, http://www.gesource.ac.uk/worldguide/html/1039.html Intute Consortium, Manchester, UK, accessed January 2009.

PAGE 219

219 Jackson, J.E., 1991. A Users Guide to Principa l Components. John Wiley and Sons, New York, NY, USA, 592pp. Kangrang, A., Chaleeraktrakoon, C., 2007. A fuzzy-GAs model for determining varied irrigation efficiency. American Journal of Applied Sciences 4(6), 339-345. Kovacs, G., 1989. Measurement and estimation of hydrological processes. In: Falkenmark, M., Chapman, T. (Eds.), Comparative Hydrology: An Ecological Approach to Land and Water Resources. The Unesco Pre ss, Paris, France, pp. 56-76. Krongkaew, M., Kakwani, N., 2003. The growth -equity trade-off in modern economic development: The case of Thailand. Journal of Asian Economics 14, 735-757. Kumagai, T., Saitoh, T.M., Sato, Y., Morooka, T ., Manfroi, O.J., Kuraji, K., Suzuki, M., 2005. Transpiration, canopy conductan ce and the decoupling coefficient of a lowland mixed dipterocarp forest in Sarawak, Borneo: dr y spell effects. Journal of Hydrology 287, pp. 237-251. Kyoto University, 2002. Online GAME-T2 Data Center, Kyoto, Japan, http://hydro.iis.utokyo.ac.jp/GAME-T/GAIN-T/index.html accessed March 2002. Lederman, D., Xu., L., 2007. Com parative advant age and trade intensity: Are traditional endowments destiny? In: Lederman, D., Mlone y, W.F. (Eds.), Natural Resources: Neither Curse nor Destiny. The World Bank, Washington, D.C., pp. 289-322. Leopold, L.B., 1968. Hydrology for urban land pl anning: A guidebook on the hydrologic effects of urban land use. U.S. Geological Circular 554. U.S. Geological Survey, Reston, VA, USA, 18pp. Leopold, L.B., Wolman, M.G., Miller, J.P. 1964. Fluvial Processes in Geomorphology. W.H. Freeman, San Francisco, California, USA, 535pp. Library of Congress, 2003. Country Studies /Area Handbook Series: Th ailand. Library of Congress, Federal Research Division, Washington, D.C., http://countrystudies.us/thailand/ accessed January 2006. Lotka, A.J., 1922. Contribution to the energetic s of evolution. Proceedings of the National Acadamy of Science. 8, 147. Maidment, D.R., Reed, S.M., Akmansov, S., Mcki nney, D.C., Olivera, F., Ye, Z., 1997. Digital atlas of the world water balan ce. Center for Research in Wa ter Resources, University of Texas, Austin, TX, USA, http://www.ce.utexas.edu/prof/maidment/atlas/atlas.htm accessed July 2006. Mallows, C.L. ,1973. Some comments on Cp. Technometrics 15, 661-675. Medhi, K., Tinakorn, P., Suphachalasai, S., 1992. Rural poverty in Thaila nd: Policy issues and responses. Asian Development Review 10(1), 199-225.

PAGE 220

220 Mekong River Commission, 2007. Report on the RAP Training Workshop and Initial Assessment of Irrigation Scheme Perform ance. Mekong River Commission, Vientiane, Lao PDR, 50pp. Mendelsohn, R., Nordhaus, W., Shaw, D., 1994. Th e impact of global warming on agriculture: A Ricardian analysis. Ameri can Economic Review 84, 753-771. Mendelsohn, R., Dinar, A., Basist, A., Kurukulasur iya, P., Ajwad, M.I., Kogan, F., Williams, C., 2004. Cross-sectional analysis of climate ch ange impacts. World Bank Policy Research Working Paper 3350. World Bank, Washington, DC, USA, 99pp. Mendelsohn, R., Basist, A., Kurukulasuriya, P ., Dinar, A., 2007. Climate and rural income. Cimate Change 81, 101-118. Minot, N., Baulch, B., Epprecht, M., 2003. Poverty and inequality in Vietnam: Spatial patterns and geographic determinants. International Food Policy Rese arch Institute Research Reports, no. 148. International Food Policy Rese arch Institute, Wash ington, D.C., USA, 90pp. NASA, 2001. Visible Earth web-based catalog, http://veimages.gsfc.nasa.gov//2316/T hailand.A2001334.0350.1km.jpg, accessed January 2006. NESDB, 2006. National Economic and Social Development Board data, NESDB, Bangkok, Thailand, http://www.nesdb.go.th/econSocial/macro/macro_eng.php accessed October 2006. NOAA, 1988. Data Announcement 88-MGG-02, Digita l relief of the Su rface of the Earth. NOAA, National Geophysical Data Center, Boulder, CO, USA, http://www.ngdc.noaa.gov/m gg/global/etopo5.HTML accessed July 2006. Odum, H.T., 1994. Ecological and General Syst ems: An Introduction to System s Ecology, 2nd edition. University Press of Colorado, Niwot, Colorado, USA. 644 pp. Odum, H.T., 1996. Environmental Accounting: Emergy and Environmental Decision Making. John Wiley and Sons, Inc., New York, NY, USA, 370pp. Odum, H.T. 2000. Emergy of Global Processe s, Folio #2 of the Handbook of Emergy Evaluation. Center for Environmental Policy, Univ ersity of Florida, Gainesville, FL, USA, 28pp. Odum H.T., Arding, J.E., 1991. Emergy Analysis Shrimp Mariculture Ecuador. Report to the Cousteau Society. Center for Wetlands, Univer sity of Florida, Ga inesville, FL, USA, 114pp. Odum, H.T., Diamond, C., Brown, M.T., 1987. Ener gy Systems Overview of the Mississippi River Basin. Report to the Cous teau Society. Center for Wetla nds, University of Florida, Gainesville, FL, USA, 107pp.

PAGE 221

221 Odum, H.T., Brown, M.T., Brandt-Williams, S. 2000. Introduction and Global Budget. Folio #1 of the Handbook of Emergy Evaluation. Center fo r Environmental Polic y, University of Florida, Gainesville, FL, USA, 17pp. Olsson, O., Hibbs, D., 2005. Biogeography and l ong-run economic development. European Economic Review, 49(4), 909-938. Parnwell, M.J.G., 1988. Rural poverty, developmen t and the environment: The case of Northeast Thailand. Journal of Biogeography 15(1), 199-208. Penman, H.L., 1948. Natural evaporation from open water, bare soil and grass. Proceedings of the Royal Society of London, A. 193, 120-146. Penman, H.L., 1961. Weather, plant and so il factors in hydrology. Weather 16, 207-219. Penman, H.L., 1963. Vegetation and hydr ology. Technical Communication No. 53, Commonwealth Agricultural Bure au, Harpenden, UK, 132 pp. Population Reference Bureau, 2006. Country Profile: Thailand. P RB, Washignton, D.C., http://www.prb.org/Datafinder/Geography/Summary.aspx?region=163®ion_type=2 accessed Decem ber 2008. Priestley, C.H.B., Taylor, R.J, 1972. On the asse ssment of surface heat flux and evaporation using large scale parameters. Monthly Weather Review, 100, 81-92. Pruett, L., Cimino, J. 2000. Global Maritime Bounda ries Database (GMBD). Fiarfax, Virginia, USA: Veridian-MRJ Technology Solutions. Online resource: http://earthtrends.wri.org /text/coastal-marine/variable-61.h tml, accessed July 2006. Rawls, W.J., Gish, T.J., Brakensiek, D.L. 1991. Estimating soil water retention from soil physical properties and characterist ics. Advances in Soil Science 16, 213. Reynolds, C.A., Jackson, T.J., Rawls, W.J., 1999. Estimated Available Water Content from the FAO Soil Map of the World, Global Soil Profile Databases, and Pedotransfer Functions, online dataset. NOAA National Geophysical Data Center, Boulder, CO, USA, http://www.ngdc.noaa.gov/ecosys/cdroms /reynolds/reynolds/reynolds.htm accessed April 2006. Romitelli, S., 1997. Energy Analysis of W atersheds PhD Dissertation, University of Florida, Gainesville, FL, USA, pp. 292. Sachs, J.D., Malaney, P., 2002. The economic and social burden of malaria. Nature 415, 680-85. Sachs, J.D., Mellinger, A.D., Gallup, J.L., 2001. The geography of poverty and wealth. Scientific American 284(3), 70-75. Saxton, K.E., Rawls W.J., Romberger J.S., Papendick R.I., 1986. Estimating generalized soil water characteristics from texture. Soil Sc ience Society of America Journal 50, 1031-1036.

PAGE 222

222 Siamwalla, A., 2003. The anatomy of a Thai cris is. In: Warr, P.G. (Ed.), Thailand Beyond the Crisis. Routledge, London, UK, 368pp. Sica, E., 2005. Climatic differences and economic growth across Italian provinces: First empirical evidence. Quaderno n.20/2005, Dipartimento di Scienze Economiche, Matematiche e Statistiche, Universit degli Studi di Foggia, 32pp. Siebert, S., Dll, P., Feick, S., Hoogeveen, J ., 2006. Global map of i rrigated areas (GMIA) version 4.0. Johann Wolfgang Goethe University Frankfurt am Main, Germany/Food and Agriculture Organization of the United Nations, Rome, Italy, http://www.fao.org/ag/agl/AGLW/aquastat/irriga tionmap/index.stm accessed August 2006. Sethaputra, S., Thanopanuwat, S., Kumpa, L., Pa ttanee, S., 2001. Thailands water vision: A case study. In: T i, L.H., Falcon, T. (Eds.), From Vi sion to Action: A Synt hesis of Experiences in Southeast Asia, The FAO-ESCAP pilot project on national wa ter visions. RAP Publication 2001/6, FAO-ESCAP, Bangkok, Thailand, http://www.fao.org/docrep/004/AB776E/ab776e00.htm accessed March 2007. Shrestha, R.M., Kumar, S., Sharm a, S. and Todoc, M.J., 2004. Institutional reforms and electricity access: Lessons from Bangladesh and Thailand. In: Prasad, K.K. (Ed.), Energy for Sustainable Development 9, pp. 41-53. Siamwalla, A., 2005. Anatomy of the Thai Econ omic Crisis. In: Warr, P.C. (Ed.), Thailand Beyond the Crisis. Routledge, London, UK, pp. 66-104. Stanimirovaa, I., Daszykowskia, M., Walczak, B. 2007. Dealing with missing values and outliers in principal component anal ysis. Talanta 72(1), 172-178. Strahler, A.N., 1957. Quantitative analysis of watershed geomorphology. American Geophysics Union Transactions. 38, 913-920. Strahler, A.N., 1964. Quantitative geomorphology of drainage basins and channel networks. In: Chow, V.T. (Ed.), Handbook of Applied Hydr ology. McGraw-Hill, New York, NY, USA, pp. 40-74. Shuttlewoth, W.J., 1993 Evaporation. In: D.R. Maidment, (Ed.), Handbook of Hydrology. McGraw-Hill, New York, NY, USA, pp. 4.1.53. Stevens, J., 2003. Partial and Semipartial Correlations. Online handout, University of Oregon, Eugene, OR, USA, online handout, http://www.uoregon.edu/~stevensj/MRA/partial.pdf accessed May 2009. Sweeney, S., Cohen, M.J., King, D., Brown, M.T., 2006. Creation of a global emergy database for standardized national emergy synthesis. In: Brown, M.T. (Ed.), Em ergy Synthesis 5: Theory and Applications of the Emergy Methodology. Center for Environmental Policy, University of Florida, Gainesville, FL, USA, 23.1-23.18.

PAGE 223

223 Tateishi, R., Ahn, C.G., 1996. Mapping evapotrans piration and water balance for global land surfaces. ISPRS Journal of Photogrammetry and Remote Sensing. 51, 209-215. TEI (Thailand Environment Institute), 1996. Thailand on a Disc. CD-ROM. Thailand Environment Institute, Bangkok, Thailand, http://www.tei.or.th/eic accessed Decem ber 2003. Thornthwaite, C.W., 1948. An appr oach toward a rational classifi cation of climate. Geographical Review, 38, 55-94. Thornthwaite, C.W., Mather, J.R., 1955. The water balance. Climatology, 8, 1-104. Timlin, D.J., Pachepsky, Y.A., Acock, B., Whisle r, F., 1996. Indirect estimation of soil hydraulic properties to predict soybean yield using GLYCIM. Agricultural Systems 52, 331. UNDP (United Nations Development Progra mme). 1999. Human Development Report of Thailand 1999. United Nations Deve lopment Programme, Bangkok, Thailand, http://origin-hdr.undp.or g/reports/detail_re ports.cfm?view=170 access ed January 2005. UNDP (United Nations Devel opment Programme). 2003. Thailand Human Development Report 2003. United Nations Development Programme, Bangkok, Thailand, http://originhdr.undp.org/reports/detail_ reports.cfm?view=169 access ed January 2005. UNEP/EAPAP (UNEP Environmenta l Assessment Programme for Asia and the Pacific), 1997. Land Cover Country Report. UNEP/EAPAP, Bangkok, Thailand, http://www.rrcap.unep.org /lc/cd/html/thailand.html accessed May 2005. Viswanadham, Y., V Filho V.P.S., Andre, R.G. B., 1991. The Priestley-Taylor parame ter alpha for the Amazon forest. Forest Ecology and Management 38, 211-225. Vorosmarty, C.J., Moore, B., Grace, A.L., Gildea, M., Melillo, J.J., Peterson, B.J., Rastetter, E.B., Steudler, P.A., 1989. Continental Scal e models of water balance and fluvial transport: An application to South Amer ica. Global Biogeochemical Cycles 3, 241-265. Vorosmarty, C.J., Federer, C.A., Schloss, A.L ., 1998. Potential evaporation functions compared on US watersheds: Possible implications fo r global scale water balance and terrestrial ecosystem modeling. Journal of Hydrology 207, 147-169. Vourlitis, G.L., Filho, P., Hayashi, M.M.S., Nogueria, J.S., Caseiro, F.T., Campelo, J.H. 2002. Seasonal variations in the ev apotranspiration of a transiti onal tropical forest of Mato Grosso, Brazil. Water Resources Research 38(6), 1-11. Waliczek, T.M., 1996. A primer on partial correla tion coefficients. Paper presented at the Meeting of the Southwest Educational Resear ch Association, New Orleans, LA, June 2006. ERIC Document Reporducti on Service No. ED 395 882. Willmott, C.J., 1981. On the validation of models. Physical Geography 2(2), 184.

PAGE 224

224 Willmott, C.J., 1982. Some comments on the evaluation of model performa nce. Bulletin of the American Meteorological Society. 63, 1309-1313. Willmott, C.J., Rowe, C.M., Philpot, W.D., 1985. Small-scale climate maps: A sensitivity analysis of some common assumptions asso ciated with grid-poi nt interpolation and contouring. The American Cartographer 12(1), 5-16. Xu, C.Y., Singh, V.P., 1998 A review of monthl y water balance models for water resources investigations. Water Resource Management 12, pp. 31.

PAGE 225

225 BIOGRAPHICAL SKETCH Sharlynn Dawn Sweeney was born in Fort Wa lton Beach, Florida. Sharlynn has a Bachelor of Science degree with a dual major in biology and marine science from the University of Miami. She worked for 4 years as a biological scientist in a neuropharmacology lab at the University of Florida, working at the microsca le with proteins and DNA. When she heard about the Systems Ecology Program at the University of Fl orida, she decided she wanted to take a leap in her scale of inquiry and pursue holistic systems science at the landscape scale. She was awarded the Graduate Assistance in Areas of National Need (GAANN) Fe llowship in order to pursue a PhD at the University of Florida, with a concentration in Geographic Information Systems. During the latter years of writing the disserta tion, she also worked full-time as the program assistant for the H.T. Odum Center for Wetlands and the Center for Environmental Policy, both research centers at the University of Florida. She received her PhD from the University of Florida in the summer of 2009.