<%BANNER%>

Coupled Simulation Modeling of Flatwoods Hydrology, and Nutrient and Vegetation Dynamids


PAGE 1

COUPLED SIMULATION MODELING OF FLATWOODS HYDROLOGY, NUTRIENT AND VEGETATION DYNAMICS By LEI YANG 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 2006

PAGE 2

Copyright 2006 by Lei Yang

PAGE 3

This dissertation is dedicated to my parents, my high school teacher Zhiting Wang, and my friend Joseph S. Smith. I recognize and appreciate the life-long influence they brought to me at different stages of my life.

PAGE 4

iv ACKNOWLEDGMENTS I am greatly indebted to my supervisor Dr. Wendy D. Graham, for her constant guidance, insight, encouragement, and conti nuous support as well as confidence in my research over the past five years. Her thorough and thoughtful coaching with all aspects of my research was unselfishly tireless, a nd her enthusiasm for research and quest for excellence have left me an everlasting im pression. I would like to express sincere appreciation to Dr. Kenneth L. Campbell for his invaluable advice on addressing each of my technical problems and concerns. Wit hout his constant supervision and guidance throughout the model development, the comp letion of this model would have been impossible. I am grateful to Dr. James W. Jones for his insight and invaluable advice on the methodology of the vegetation dynamic m odel and support in offering important crop growth model related literature; to Dr. Mark W. Clark for introducing me to the niche theory and the interactive relationshi p between wetland hydrology and vegetation dynamics and his help in identifying pastur e vegetation species for simulation; to Dr. Gregory A. Kiker for introducing me to the Java programming language with his great enthusiasm and his technical support rega rding the ACRU2000 modeling system during the model development. I deeply benef ited from many hours of precious discussions with each of these committee members on a mu ltitude of perspectives regarding my research. Without their combined supervisi on of each step throughout my research, this study would not have been possible.

PAGE 5

v I would like to acknowledge my debt to Ch ris Martinez for his great cooperation as a teammate throughout the development of hydro logic and nutrient models. I wish also to thank Dr. Michael D. Annable for hi s generous comments on specific technical problems during many brown-bag group meetin gs; Dr. Patrick Bohlen in MacArthur Agro-ecology Research Center, Florida, for in troducing me to the pasture sites at Buck Island Ranch and offering useful documents; Mr Gregory S. Hendricks for offerings of data for Buck Island Ranch and nutrient relate d documents; Dr. Stuart J. Rymph from the University of Wisconsin for providing important bahiagrass related documents; Ms. Cheryl H. Porter for offering crop modeling documents; and Dr. William Wise and his graduate student Min Joong Hyuk fo r helping with DHI software. Also, special thanks go to a few friends in cluding Joseph S. Smith, Tricia G. Smith, Donna L. Miller, Paul Miller, and David R. Murphy for their friendship and support throughout the past few years, especially during some tough times. I would like to acknowledge all the good friends for their friendship. I am also grateful to several labmates for their friendship and encouragem ent, and faculty, staff and students in the Agricultural and Biological Engineering Department for the quality academic environment. Finally I particularly appreciate my pare nts and siblings fo r their unconditional love, understanding, patience and encouragement. Without their affection, it would have been even more difficult to complete this research.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...............................................................................................................x LIST OF FIGURES.........................................................................................................xiii ABSTRACT....................................................................................................................... xx CHAPTER 1 INTRODUCTION........................................................................................................1 Study Background........................................................................................................1 Overview of the Coupled Modeling System................................................................8 Study Objectives...........................................................................................................9 2 LITERATURE REVIEW...........................................................................................12 Overview of Previous Modeling Efforts.....................................................................12 ACRU2000 Modeling System....................................................................................20 Model Testing Procedures..........................................................................................24 Model Calibration................................................................................................25 Model Validation.................................................................................................25 Sensitivity Analysis.............................................................................................26 Model Evaluation................................................................................................28 Statistics.......................................................................................................28 Graphic representation.................................................................................31 3 HYDROLOGIC SIMULATION MODEL.................................................................34 Introduction.................................................................................................................34 Vertical Hydrologic Components...............................................................................38 Rainfall................................................................................................................38 Canopy Interception............................................................................................38 Evapotranspiration...............................................................................................39 Infiltration............................................................................................................42 Soil Water Redistribution....................................................................................42 Upward Flux........................................................................................................43 Deep Seepage......................................................................................................44

PAGE 7

vii Horizontal Hydrol ogic Components...........................................................................44 Overland Flow.....................................................................................................45 Overland flow calculation............................................................................46 Surface water storage apportionment...........................................................49 Groundwater Flow...............................................................................................51 Lateral groundwater flow calculation..........................................................54 Groundwater storage apportionment............................................................56 Canal Flow...........................................................................................................56 Initial and Boundary Conditions.................................................................................57 Model Testing and Validation....................................................................................58 Simulation Sequence and Model Performance Accuracy Analysis....................58 Case 1: Overland flow along a flat rectangular plane..................................60 Case 2: Overland and groundwater flow over an axisymmetric domain.....65 Application at Dry Lake Dairy #1, Kissimmee River Basin, Florida.................76 Site description.............................................................................................76 Results and discussion..................................................................................79 Concluding Remarks..................................................................................................82 4 NUTRIENT SIMULATION MODEL.......................................................................87 Introduction.................................................................................................................87 Nutrient Components..................................................................................................89 Nitrogen Cycle Components...............................................................................89 Mineralization..............................................................................................91 Immobilization.............................................................................................92 Denitrification..............................................................................................92 Runoff, sediment transport and percolation.................................................93 Uptake, evaporation, and fixation................................................................94 Rainfall and fertilizer...................................................................................95 Ammonia volatilization................................................................................95 Surface and subsurface lateral ni trate nitrogen transport.............................95 Surface and subsurface lateral ammonium nitrogen transport.....................97 Phosphorus Cycle Components...........................................................................98 Mineralization..............................................................................................99 Immobilization...........................................................................................100 Runoff, sediment, percolation....................................................................100 Uptake and evaporation..............................................................................101 Rainfall and fertilizer.................................................................................101 Surface and subsurface lateral labile phosphorus transport.......................102 Conservative Solute Transport Components.....................................................104 Initial and Boundary Conditions...............................................................................105 Model Testing and Validation..................................................................................106 Conservative Solute Test...................................................................................106 Scenario description...................................................................................107 Results and discussion................................................................................107 Application at Buck Island Ranc h, Lake Okeechobee Basin, Florida..............111 Project description......................................................................................111

PAGE 8

viii Sensitivity analysis.....................................................................................117 Results and discussion................................................................................122 Concluding Remarks................................................................................................133 5 VEGETATION DYNAMICS SIMULATION MODEL.........................................189 Introduction...............................................................................................................189 Methodology.............................................................................................................192 Model Structure.................................................................................................192 Plant Growth......................................................................................................193 Potential growth.........................................................................................193 Reduced growth..........................................................................................196 Dry matter partitioning...............................................................................197 Leaf Area Index.................................................................................................197 Plant Senescence...............................................................................................198 Evapotranspiration.............................................................................................200 Nitrogen Uptake................................................................................................201 Phosphorus Uptake............................................................................................204 Growth Reduction Factor..................................................................................205 Water stress and logging factors................................................................206 Nitrogen stress factor.................................................................................208 Phosphorus stress factor.............................................................................209 Hypothetical Scenario Model Testing......................................................................209 Scenario Description.........................................................................................209 Results and Discussion......................................................................................213 Water and nutrient responses to water retention BMP...............................213 Influence of differences in temperature sensitivities among species.........214 Species composition dynamics due to water retention BMP.....................216 Concluding Remarks................................................................................................219 6 SUMMARY AND CONCLUSIONS.......................................................................234 Hydrologic Simulation Model..................................................................................235 Nutrient Simulation Model.......................................................................................236 Vegetation Dynamics Simulation Model..................................................................237 Implications of the Research....................................................................................238 Future Research Recommendations.........................................................................238 Model Preand Post-processing Capacity.........................................................238 Use Consistent Units for Pa rameters and Variables..........................................239 Documentation..................................................................................................239 Potential Changes to Existing Objects..............................................................239 Sub-Daily Time Step.........................................................................................240 Herbivore Movement Module...........................................................................240 Hydrologic Model.............................................................................................240 Nutrient Model..................................................................................................241 Vegetation Model..............................................................................................241

PAGE 9

ix APPENDIX A FLOW CHART FOR THE COUPLED MODELING SYSTEM AND MODEL PROCESSES............................................................................................................242 B FLOW CHART OF LATERAL GR OUNDWATER FLOW SIMULATION.........247 C LISTS OF NEW AND MODIFIED OBJECTS.......................................................248 D LISTS OF NEW INPUT AND OUTPUT VARIABLES.........................................262 E LISTS OF MODEL INPUT FILES..........................................................................265 F ALGORITHMS OF WATER STORAGE APPORTIONMENT.............................266 LIST OF REFERENCES.................................................................................................282 BIOGRAPHICAL SKETCH...........................................................................................293

PAGE 10

x LIST OF TABLES Table page 3-1 List of strategies for water tran smission and water storage updating for simulation sequence experiments..........................................................................59 3-2 Statistics for the surface water dept hs predicted by the modified ACRU2000 model and MIKE SHE for the three experiments with the same simulation sequence.................................................................................................................62 3-3 Statistics for the surface water dept hs predicted by the modified ACRU2000 model and MIKE SHE for the three experiments..................................................69 3-4 Statistics for the simulated groundw ater table depths by the modified ACRU2000 model, MIKE SHE, and M ODFLOW for the three experiments with the same simulation sequence........................................................................75 3-5 Summary of annual water budge t on Dry Lake Dairy #1 site...............................78 3-6 Model input parameters for the Dry Lake Dairy #1 site........................................85 3-7 Statistics for the simulated surface ru noff and groundwater table depths by the modified ACRU2000 and FHANTM....................................................................85 4-1 Statistics for the solute mass predicted by the modified ACRU2000..................108 4-2 List of stocking activities fo r pastures S1, S4, W6 and W7................................137 4-3 List of fertilization activi ties for pastures S1 and S4...........................................137 4-4 List of burn activities for pastures S1, S4, W6 and W7.......................................137 4-5 Percent of area occupied by different soil series and we tlands in selected summer and winter pastures.................................................................................137 4-6 Model input parameters for th e winter pastures W6 and W7..............................138 4-7 Model input parameters for th e summer pastures S4 and S1...............................139 4-8 Selected hydrologic parameters for sensitivity analysis for W6..........................140 4-9 Selected nutrient parameters for sensitivity analysis for W6...............................141

PAGE 11

xi 4-10 Selected hydrologic parameters for sensitivity analysis for S4...........................142 4-11 Selected nutrient parameters for sensitivity analysis for S4................................143 4-12 Sensitivities of total surface runoff of the selected input parameters throughout the simulation period for the calibration site W6.................................................144 4-13 Sensitivities of maximum water table to the selected input parameters depth throughout the simulation period fo r the calibration site W6..............................144 4-14 Sensitivities of total P loads to the selected input parameters throughout the simulation period for the calibration site W6......................................................144 4-15 Sensitivities of total N loads to the selected input parameters throughout the simulation period for the calibration site W6......................................................144 4-16 Sensitivities of total surface runoff to the selected input parameters throughout the simulation period for the calibration site S4..................................................145 4-17 Sensitivities of maximum water table to the selected input parameters depth throughout the simulation period fo r the calibration site S4................................145 4-18 Sensitivities of total P loads to the selected input parameters throughout the simulation period for the calibration site S4........................................................145 4-19 Sensitivities of total N loads to the selected input parameters throughout the simulation period for the calibration site S4........................................................145 4-20 Annual water budget for the calibration site W6.................................................146 4-21 Annual water budget for the verification site W7................................................146 4-22 Annual water budget for the calibration site S4...................................................146 4-23 Annual water budget for the verification site S1.................................................146 4-24 Monthly statistics fo r the calibration site W6......................................................147 4-25 Annual statistics for the calibration site W6........................................................147 4-26 Monthly statistics for the verification site W7.....................................................147 4-27 Annual statistics for th e verification site W7.......................................................147 4-28 Monthly statistics fo r the calibration site S4........................................................148 4-29 Annual statistics for the calibration site S4..........................................................148 4-30 Monthly statistics for the verification site S1......................................................148

PAGE 12

xii 4-31 Annual statistics for th e verification site S1........................................................148 5-1 Percent cover of vegetati on on summer pastures S4...........................................222 5-2 Parameter values for the different plant species for the vegetation model..........222 5-3 Initial composition percentage and am ount for species in each land segment....223 5-4 Variables calculated from other models..............................................................223 5-5 Output variables from the vegetation model........................................................223 5-6 Testing scenarios for temperature functions........................................................223 E-1 New input and output variables require d towards the multi-directional spatial simulation beyond the existing variables in ACRU2000.....................................262

PAGE 13

xiii LIST OF FIGURES Figure page 1-1 Schematic of the feedback rela tionships among hydrology, nutrient, and vegetation dynamics in the coupled model system..................................................8 2-1 General structure of the lumped ACRU (v3.00) model.........................................21 2-2 Land segment configuration of the ACRU2000 model.........................................23 3-1 Schematic of an example catchment and its spatial discretization........................35 3-2 Hydrological processes in the m odified ACRU2000 hydrologic model...............36 3-3 Configuration of overlan d flows between source land segment and adjacent land segments.................................................................................................................46 3-4 The relationship between Manning's r oughness-coefficient (n) to water depth and to plant height, bo th of which change dynamically in the model...................48 3-5 Structured boundary...............................................................................................49 3-6 Schematic representation of the late ral groundwater flow from the higher head hs to the lower head hd. (A) and (B) represent the situations without and with overland flow, respectively....................................................................................52 3-7 Diagram of lateral groundwater flow between two land segments. (A) and (B) represent the generalized situations without a nd with overland flow, respectively............................................................................................................53 3-8 A rectangular plane with 20 land segm ents (arrow indicates water movement direction and digits assigned in each gr id cell indicate the number for each land segment).................................................................................................................60 3-9 Comparisons of the simulated surf ace water depth from the modified ACRU2000 model on the thre e experiments between two opposite simulation sequences...............................................................................................................63 3-10 Comparisons of the simulated surfa ce water depths between the MIKE SHE’s overland flow model and the modifi ed ACRU2000 hydrologic model on the three experiments with the simula tion sequence from LS1 to LS20.....................64

PAGE 14

xiv 3-11 Schematics of two-dimensional axisym metric domain (left) and its threedimensional discretization (right). The digits assigned in the left diagram indicate the number for each land segment............................................................66 3-12 Comparisons of simulated water depths of the selected land segments between the modified ACRU2000 model and the MIKE SHE’s overland flow model for the three experiments.............................................................................................68 3-13 Comparisons of the simulated water depths of the sele cted land segments between the modified ACRU2000 and MI KE SHE’s overland flow model for Experiment 3..........................................................................................................70 3-14 Comparisons of the groundwater table depths at the selected land segments between and the modified ACRU200 0 hydrologic model and MIKE SHE’s groundwater flow model for the three experiments...............................................74 3-15 Location of Dry Lake Dairy #1 site.......................................................................76 3-16 Dry Lake Dairy # 1 site map, topographi c survey and locatio n of well stations and tracer application compound. Distance scales are in feet and elevation contours are in feet above mean sea level..............................................................77 3-17 Comparisons of the continuous simu lations of runoff from the modified ACRU2000 and FHANTM against the observed data..........................................84 3-18 Comparisons of the cumulative simu lated runoff from the modified ACRU2000 and FHANTM with the cumulative observed data................................................84 3-19 Comparisons of the continuous simulati on of groundwater table depths from the modified ACRU2000 and FHANTM against the observed data...........................85 3-20 Scatterplots of observed vs. simulate d surface runoff depths predicted by the modified ACRU2000 and FHANTM....................................................................86 3-21 Scatterplots of observed vs. simulate d groundwater table depths predicted by the modified ACRU2000 and FHANTM....................................................................86 4-1 Schematic representation of the GLEAMS nitrogen cycle....................................90 4-2 Schematic representation of the GLEAMS phosphorus cycle...............................98 4-3 Comparisons of the solute mass in the 3rd and 4th soil layers of selected land segments between the modified ACRU2000 and PMPATH...............................110 4-4 Location of MacArthur Agro -ecology Research Center......................................111 4-5 General layout of the project field ......................................................................113

PAGE 15

xv 4-6 Relative sensitivities of the total surface runoff of 6-year simulation period over the selected input parameters fo r the pasture sites W6 and S4............................149 4-7 Relative sensitivities of the maximum water table depth of 6-year simulation period over the selected input paramete rs for the pasture sites W6 and S4.........150 4-8 Relative sensitivities of the total P load of 6-year simulation period over the selected input parameters and variab les for the pasture sites W6 and S4............151 4-9 Relative sensitivities of the total N load of 6-year simulation period over the selected input parameters for the pasture sites W6 and S4..................................152 4-10 Continuous simulation of groundwat er table depth from September 2000 through December 2003 for the calibration site W6............................................153 4-11 Continuous simulation of groundwat er table depth from September 2000 through December 2003 for the verification site W7..........................................153 4-12 Continuous simulation of groundwat er table depth from September 2000 through December 2003 for the calibration site S4.............................................154 4-13 Continuous simulation of groundwat er table depth from September 2000 through December 2003 for the verification site S1............................................154 4-14 Continuous simulation of surface runoff from July 1998 through December 2003 for the calibration site W6...........................................................................155 4-15 Continuous simulation of surface runoff from July 1998 through December 2003 for the verification site W7.........................................................................155 4-16 Continuous simulation of surface runoff from July 1998 through December 2003 for the calibration site S4............................................................................156 4-17 Continuous simulation of surface runoff from July 1998 through December 2003 for the verification site S1...........................................................................156 4-18 Continuous simulation of P loads from July 1998 through December 2003 for the calibration site W6.........................................................................................157 4-19 Continuous simulation of P loads from July 1998 through December 2003 for the verification site W7........................................................................................157 4-20 Continuous simulation of P loads from July 1998 through December 2003 for the calibration site S4...........................................................................................158 4-21 Continuous simulation of P loads from July 1998 through December 2003 for the verification site S1.........................................................................................158

PAGE 16

xvi 4-22 Continuous simulation of N loads from July 1998 through December 2003 for the calibration site W6.........................................................................................159 4-23 Continuous simulation of N loads from July 1998 through December 2003 for the verification site W7........................................................................................159 4-24 Continuous simulation of N loads from July 1998 through December 2003 for the calibration site S4...........................................................................................160 4-25 Continuous simulation of N loads from July 1998 through December 2003 for the verification site S1.........................................................................................160 4-26 Continuous simulation of surface r unoff P concentrations from July 1998 through December 2003 for the calibration site W6............................................161 4-27 Continuous simulation of surface r unoff P concentrations from July 1998 through December 2003 for the verification site W7..........................................161 4-28 Continuous simulation of surface r unoff N concentrations from July 1998 through December 2003 for the calibration site W6............................................162 4-29 Continuous simulation of surface r unoff N concentrations from July 1998 through December 2003 for the verification site W7..........................................162 4-30 Continuous simulation of surface r unoff P concentrations from July 1998 through December 2003 for the calibration site S4.............................................163 4-31 Continuous simulation of surface r unoff P concentrations from July 1998 through December 2003 for the calibration site S1.............................................163 4-32 Continuous simulation of surface r unoff N concentrations from July 1998 through December 2003 for the verification site S4............................................164 4-33 Continuous simulation of surface r unoff N concentrations from July 1998 through December 2003 for the verification site S1............................................164 4-34 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the calibration site W6...............................................................................................165 4-35 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the verification site W7..............................................................................................166 4-36 Linear plots of monthly and annual wa ter table depth from 1998 to 2003 for the calibration site W6...............................................................................................167 4-37 Linear plots of monthly and annual wa ter table depth from 1998 to 2003 for the verification site W7..............................................................................................168

PAGE 17

xvii 4-38 Linear plots of monthly and annual P load from 1998 to 2003 for the calibration site W6.................................................................................................................169 4-39 Linear plots of monthly and a nnual P load from 1998 to 2003 for the verification site W7..............................................................................................170 4-40 Linear plots of monthly and annual N load from 1998 to 2003 for the calibration site W6.................................................................................................................171 4-41 Linear plots of monthly and a nnual N load from 1998 to 2003 for the verification site W7..............................................................................................172 4-42 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the calibration site S4.................................................................................................173 4-43 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the verification site S1...............................................................................................174 4-44 Linear plots of monthly and annual wa ter table depth from 1998 to 2003 for the calibration site S4.................................................................................................175 4-45 Linear plots of monthly and annual wa ter table depth from 1998 to 2003 for the verification site S1...............................................................................................176 4-46 Linear plots of monthly and annual P load from 1998 to 2003 for the calibration site S4...................................................................................................................177 4-47 Linear plots of monthly and a nnual P load from 1998 to 2003 for the verification site S1...............................................................................................178 4-48 Linear plots of monthly and annual N load from 1998 to 2003 for the calibration site S4...................................................................................................................179 4-49 Linear plots of monthly and a nnual N load from 1998 to 2003 for the verification site S1...............................................................................................180 4-50 Duration curves of daily surface runoff and water table depth for the calibration site W6.................................................................................................................181 4-51 Duration curves of daily P and N loads for the calibration site W6....................182 4-52 Duration curves of daily surface runoff and water table depth for the verification site W7.................................................................................................................183 4-53 Duration curves of daily P and N loads for the verification site W7...................184 4-54 Duration curves of daily surface runo ff and water table depth from 1998 to 2003 for the calibration site S4.....................................................................................185

PAGE 18

xviii 4-55 Duration curves of daily N and P lo ads from 1998 to 2003 for the calibration site S4...................................................................................................................186 4-56 Duration curves of daily surface runo ff and water table depth from 1998 to 2003 for the verification site S1....................................................................................187 4-57 Duration curves of daily N and P lo ads from 1998 to 2003 for the verification site S1...................................................................................................................188 5-1 Diagram for daily plant growth in re lation to weather and water and nutrient availabilities (DM = dry matter; N = nitrogen; P = phosphorus; and SLA = specific leaf area).................................................................................................193 5-2 Diagram of temperature function for species i....................................................195 5-3 An example relationship between plan t biomass nitrogen concentration and growth ratio..........................................................................................................202 5-4 An example relationship between pl ant biomass phosphous concentration and growth ratio..........................................................................................................204 5-5 Aerial photo showing the layout of the improved summer pasture site S4 and location of associated instrumentation. S1 to S6 indicate the individual summer pasture sites and LS1 to LS12 indicate th e land segments divided for the site S4. The dotted lines were made to show the boundary between land segments........210 5-6 Continuous simulation of surface runo ff throughout the simulation period from 1998 to 2003........................................................................................................224 5-7 Continuous simulation of water table depths at land segment 9 throughout the simulation period from 1998 to 2003...................................................................224 5-8 Continuous simulation of P loads thr oughout the simulation period from 1998 to 2003......................................................................................................................225 5-9 Continuous simulation of N loads th roughout the simulation period from 1998 to 2003.................................................................................................................225 5-10 Continuous simulation of P concentr ations throughout the simulation period from 1998 to 2003................................................................................................226 5-11 Continuous simulation of N concentr ations throughout the simulation period from 1998 to 2003................................................................................................226 5-12 Comparison of predicted potential aboveground biomass for species in land segment 11 with different temperature functions................................................227

PAGE 19

xix 5-13 Comparison of temperature factors am ong the three selected species in land segment 11...........................................................................................................228 5-14 Species distribution (percent of total aboveground biomass at the end of each year) over land segments for bahia, floralta and panicum throughout the simulation period from 1998 to 2003...................................................................230 5-15 Comparisons of continuous simulati on of aboveground biomass for all species in the selcted land segments 8, 11, and 12 before water retention BMP.............231 5-16 Comparisons of continuous simulati on of aboveground biomass for all species in the selcted land segments 8, 11, and 12 after water retention BMP................232 5-17 Growth limiting factors for bahiagrass in land segment 8 before water retention BMP was applied.................................................................................................233 G-1 Single land segment type ne ighbor receiving overland flow...............................266 G-2 Single river type neighb or receiving overland flow............................................267 G-3 Configuration of multiple directional overland flows from source land segment to adjacent land segments....................................................................................269

PAGE 20

xx 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 COUPLED SIMULATION MODELING OF FLATWOODS HYDROLOGY, NUTRIENT AND VEGETATION DYNAMICS By Lei Yang May 2006 Chair: Wendy D. Graham Cochair: Kenneth L. Campbell Major Department: Agricultur al and Biological Engineering Lake Okeechobee, located at the center of the Kissimmee-Okeechobee-Everglades aquatic ecosystem in south Florida, is experi encing water quality degradation. Non-point agricultural runoff from dairies and cow-calf operations in the northern watershed of the lake is considered to be the primary sour ce of excess phosphorus (P ) loading discharged into the lake. In orde r to evaluate alternative land ma nagement practices that result in reduced P loading from the watershed to the lake, a coupled modeli ng system integrating hydrology, nutrient and vegetation dyna mics simulation was developed. The coupled modeling system was deve loped within the Java-based, objectoriented framework of the ACRU2000 m odeling system by adding new hydrologic and nutrient components and a vegetation model to enable multi-directional spatial simulation of hydrological, chemical, and biological proce sses simultaneously in a daily time step. The coupled model was tested for accur acy by comparing performance with wellaccepted models including MIKE SHE and MO DFLOW. Results indicate that the

PAGE 21

xxi coupled model is capable of simulating, with reasonable accuracy, hydrological and solute transport processes for the hypotheti cal scenarios. Additionally, the model was tested in the Kissimmee River Basin and La ke Okeechobee Basin by comparing with the FHANTM model and against measured data. These applications demonstrated that the coupled model is statistically close to th e performance of FHANT M with respect to hydrologic response in the Kissimmee River Ba sin, but much better than FHANTM with regard to hydrologic and nutrient responses in the Lake Okeechobee Basin. From the testing, it was concluded that the model is able to continuously simulate the surface runoff and groundwater tables wi th adequate accuracy. However the model’s capacity to simulate nutrient loading needs further testing after sufficient reliable nutrient data becomes available. The vegetation model, coupled with the hydrologic and nutrient models, was tested for a hypothetical scenar io based on the conditions in the Lake Okeechobee Basin. The test results show th at the temporal and spatial vegetation composition pattern can be an indicator of the ecohydrological impact s of alternative land management practices. However, for actual application of this model, further testing is required when more vegetation data are available. Recommended future research includes fu rther development of the coupled model to enable a user-friendly preand post-proc essing graphical interface, an option for subdaily time steps, beef-cattle roaming simulati on, and plant competition. Further testing of the coupled model should be conducted at larger watershed scales a nd for the nutrient and vegetation simulations when addi tional data are available.

PAGE 22

1 CHAPTER 1 INTRODUCTION Study Background Lake Okeechobee is a large, multi-functional lake located at the center of the Kissimmee-Okeechobee-Everglades aquatic ecosy stem in south Florida. The lake provides regional flood protection, water supply for agricultural, urba n and natural areas, and is a critical habitat for fish, birds and ot her wildlife. Water quality in the lake has declined over recent decades due to urban development, channeliz ation of the Kissimmee River and agricultural operations. The 1997 Lake Okeechobee Surface Water Improvement and Management Plan (Sout h Florida Water Management District [SFWMD], 1997a) found that excess phosphorus (P ) loading is one of the most serious problems the lake is facing. The Lake Okeechobee watershe d, with an area of 12,000 km2 extending from Orlando to the Everglades, lies predominat ely in the southern Florida flatwoods physiographic region characteri zed by flat, poorly drained, hi gh-water table, and fine sandy soils, which consist of Spodosols, Entisols, and Histos ols (United States Department of Agriculture [USDA], 1990). Th e majority of the soils in the northern watersheds of the lake are Sposodols, w ith 8-20 cm thick surface horizons underlain by spodic horizons at depth of 0.5 m to greater than 2 m (USDA, 1990). These soils, with greater than 90% sand, are characterized by high infiltration rates and poor internal drainage due to low permeability of the spodic horizon. When rainfall occurs, the soils often become saturated in a short time. Th e water table is commonly within 1 m of the

PAGE 23

2 ground surface during the wet season and may recede to 2 m depth during the dry season (Knisel et al., 1985). With highly permeable surface soils there is little surface runoff until the soil pore space is filled with wa ter and the water table reaches the surface (Heatwole et al., 1987); slow downward or lateral movement of water and solutes occurs with water table recession. The land use in the Okeechobee watershed is pr imarily beef cattle, dairy, and citrus. Since 1930, ranching has intensified from native pastures to improved pastures with high quality grasses and legumes, drainage and fertilization. During the 1950s, the dairy industry first moved to Okeechobee County, and by the mid-1980s, there were 49 milking barns and 50,000 milk cows in the lower Kissimmee River (LKR) and Taylor Creek/Nubbin Slough (TCNS) regions (Flaig and Reddy, 1995). The primary feed for dairy cows, including high P containing mate rials, was imported into the watershed. Animal waste management was almost non-ex istent until the 1970s (Flaig and Reddy, 1995). Historically, the majority of P load to the lake was derived from the LKR and the TCNS basins with 13% contri buted by the LKR basin and 22% by the TCNS basin. With the implementation of improved management pr actices during recent decades, the P load from the TCNS basin has decreased by 17% (Gunsalus et al., 1992), and the LKR basin now provides the greatest P load to the lake. Previous studies have indicated that nonpoint agricultural runoff in the northern waters hed of the lake is considered to be the primary source of excess P being discharged into Lake Okeechob ee, which typically exceeds the recommended total maximum daily loading (TMDL = 140 metric ton/year). P in agricultural runoff mainly originates from one or more of four sources: fertilizers,

PAGE 24

3 animal manures, mineralization of organic ma terials, and/or atmos pheric deposition. The first three sources can be controlled by agricultural best management practices. In order to protect the water quality of Lake Okeechobee and reach environmental restoration goals, a variety of best manageme nt practices (BMPs) have been implemented in the Lake Okeechobee watershed. BMPs are on-farm activities designed to reduce nutrient losses in drainage waters to an environmentally acceptable level, while simultaneously maintaining an economically viable farming operati on (Bottcher et al., 1995). To reduce P loading one must either reduce P concentration or water volume. According to Bottcher et al. (1995), there are th ree ways to reduce P concentrations in the runoff water from agriculture: 1) reduce th e amount of P on the farm by minimizing P inputs to the farm and maximizing non-runo ff P output from the farm; 2) reduce the hydrologic mobility of the P that is on the farm by limiting water contact and/or reducing the solubility or erodibility of phosphatic mate rials; 3) edge of fi eld/farm pre-discharge treatment using uptake, adsorption, deposition, or precipitation tec hnologies, such as wetland and/or chemical additives. There ar e two methods used to reduce water runoff volume: 1) increase the evapotranspiration fr om the farm; 2) decrease off-farm or groundwater irrigation water input s to the farm by improved i rrigation efficiency or by using storage runoff as a substitute irriga tion supply. From numerous studies in the Okeechobee basin as well as generally accepted pr actices from other parts of the country, Bottcher et al. (1995) summarized the BMPs appropriate for the Lake Okeechobee basin as follows: 1) fertility BMPs including calibra ted soil testing (CST), banding of fertilizer, prevention of misplaced fertilizer, and sp lit application; 2) animal manure BMPs including dairy high intensity area (HIA) drainage control, collection and distribution of

PAGE 25

4 barn manure, watering, feed and shade faciliti es placement, fencing animals from ditches and streams, grazing management, selecting hi gh P uptake crops from manure application areas, and composting; 3) ge neral BMPs including crop management, irrigation and drainage management, maximum flow distan ce for P control, flow-way buffer strips, limit drainage of organic and/or wetland so ils, and alternative land use; 4) edge-offield/farm treatment including runoff reten tion/detention system, use of wetlands, and chemical treatment. Only a few of the above listed BMPs have been field tested, and even those were tested for only a limited set of conditions (B ottcher et al., 1995). Flaig and Reddy (1995) indicated that implementation of BMPs has not been sufficient to meet P load reduction goals, and additional P control practices to further reduce P are needed. Recent BMPs efforts to reduce nutrient loading from th e Lake Okeechobee watershed have focused on restoration of wetlands for their particular capacities to reduce nutri ent loadings, thereby reducing eutrophication in adjacent water bodies. Wetlands are an important component of the Lake Okeechobee watershed. However, many of these wetlands have disa ppeared or have been degraded due to hydrologic alteration, urbanizati on and agricultural practices. Seasonal and year-round isolated and connected wetlands used to occ upy 25% of the watershed area (McCaffery et al., 1976). Now many isolated wetlands are connected by shallow drainage ditches and have been converted into pastures. Current ly, wetlands represent about 15% of the land area in the Lake Okeechobee watershed (Flaig and Havens, 1995). Wetland loss inevitably leads to loss of biological, envi ronmental quality and so cio-functions such as

PAGE 26

5 flood storage, groundwater recharge, sedime nt trapping, retenti on and removal of nutrients and pollutants, and wildlife and recreational habitat (Davis and Froend, 1999). Among all wetland functions, transport a nd transformation processes including sedimentation, sediment adsorption, nutrien t uptake, microbial assimilation and transformations, and denitrifi cation may be the most important mechanisms as they are responsible for nutrient removal or retention. Fisher and Acreman (2004) investigated 57 wetlands from around the world and indicated that the majority of wetlands reduced nutrient loading and there was little difference in the proporti on of wetlands that reduced nitrogen (N) to those that reduced P loading. However, they also pointed out that some wetlands increased nutrient loading by increasi ng the loading of soluble N and P species, thus potentially driving aqua tic eutrophication. Busnardo et al. (1992) researched the effect of hydroperiod on nutrien t removal efficiency in replicate wetland mesocosms and concluded that alternate drai ning and flooding of sediments ( pulsed discharge) increased nutrient removal efficiency compared to th e continuous-flow “control”. Uusi-Kamppa et al. (1997) indicated that in ma ny wetlands the retention of soluble P is much less efficient than that of particulate P. Also, wate rlogged sediments are known to release P into overlying waters (Mortimer, 1941), where it is more likely to be exported from the watershed. Denitrification, which occurs under anaerob ic conditions to release N into the atmosphere, is believed by many to be the major mechanism of N loss in wetlands. Denitrification rates can be limited by carbon ava ilability and, in this way, vegetation can influence denitrification rate s indirectly (Broa dbent and Clark, 1965). Vegetation may also influence nitrification and denitrifica tion by influencing the oxygen concentration of

PAGE 27

6 the wetland substrate within the rhizosphere (Armstrong, 1964) or by providing bacteria which can fix N in root nodules. There is evidence that N removal efficiency is not affected by the length of time the wetland has re ceived N pollution while, in contrast, the ability of a wetland to remove P is know n to decline with time (Nichols, 1983; Richardson, 1985). Wetlands are not stand-alone ecosystems. They impact hydrology, water quality and vegetation dynamics throughout the waters hed. Wetlands are ch aracterized by the periodic excess of water inflow over outflow that provides a saturated substrate. The effect of this characteristic is substantial water storage within wetlands and the development of a readily identifiable wetla nds flora and fauna which are adapted to periodic anoxic conditions (Bradley and Gilvear, 2000). P load ing can alter plant communities through increased plant productivit y, tissue P storage, soil P enrichment, and shifts in plant species composition (Dav is, 1991; Urban et al., 1993; Chiang et al., 2000). Altered hydrologic regime, caused by water management infrastructure and operations, can also affect the pattern of vegetation communities. Newman et al. (1996) showed that P concentration and water depth are two important driving forces for cattail invasion into the Everglades. Urban et al. (1993) also indicated that a combination of prolonged hydroperiod and P loading stimulat es cattail reproduction and encroachment and thus results in the degradation of vegetative habitats a nd other ecological characteristics in wetlands. Conversely, the physical and physiologica l characteristics of vegetation influence hydrological response a nd nutrient cycling. The flora and fauna impact the hydrology of many wetlands in that their incomplete decomposition leads to the progressive development of an organic subs trate that itself influences the pattern and

PAGE 28

7 direction of water flow thr ough wetlands and the quantity of water storage (Bradley and Gilvear, 2000). Gurnell et al (2000) indicated that interc eption, evapotranspiration, and infiltration processes are particularly heav ily influenced by the characteristics and dynamics of the vegetation cover. Many ecologists have recognized that changing vegetation pattern/structure causes feedback that can alter rates of hydrologic processes and nutrient cycling, which, in turn, can cause additional changes in vegetation structure (Lauenroth et al., 1993). From the above discussion, it is clear that it is critical to unders tand the role of both BMPs and wetland functions in nutrient removal, retention and storag e in order to reduce P loading into Lake Okeechobee. However, it is impractical to test all BMPs for their effectiveness through field experiments. Th e use of computer models is therefore beneficial because simulation results can not only predict how well a proposed BMP or combined BMPs will reduce P loads, but can also quantitatively evaluate specific hydrologic and biogeochemical processes asso ciated with management activities for BMP design and optimization. Ecohydrologic modeling that simulates hydr ological, biochemical and ecological processes and their interrelat ions in soil and water bodies has captured the attention of hydrologists and other scientists in recent ye ars. Modeling provide s a useful tool for gaining insight into ecohydrol ogical processes and evaluatin g management practices if model predictions are accurate. Many models have been developed but they are typically linked to the regions where they were devel oped and tested for a specific purpose and are often limited to specific sp atial scales. The unique fl atwoods hydrology in the Lake Okeechobee watershed requires the design of a model that can simulate integrated

PAGE 29

8 multidimensional surface and subsurface wate r, nutrient, and vegetation dynamics at multiple temporal and spatial scales. Overview of the Coupled Modeling System For this study, a coupled modeling system th at enables the distributed simulation of hydrology, water quality and vegetation dynamics was developed for Okeechobee flatwoods watersheds that incorporate uplands wetlands, and transition zones located in between these landscapes. The proposed coupled modeling system and the feedback relationships among its submodels are depict ed in Figure 1-1. This model system contains a hydrologic submodel as a critical component wh ich simulates precipitation, interception, evapotranspiration, infiltrati on, water movement within unsaturated soil zones, upward flux, deep seepage, overland flow and groundwater flow. It also includes a submodel for nutrient cycling to simulate th e transport and transf ormation processes for nitrogen and phosphorus. Another necessary pa rt of the model is a vegetation submodel that simulates plant growth dynamics under the combined influence of hydrology, nutrient availability, and land management prac tices including beef-cat tle stocking rates, fire, fertilization, etc. Roughness Nutrient stress (N and P) Decomposition Uptake Water stress/logging Interception Transpiration Transport Hydrologic Model Nutrient Model V egetation Model Figure 1-1. Schematic of the feedback relationships among hydrology, nutrient, and vegetation dynamics in th e coupled model system.

PAGE 30

9 The three submodels are coupled in that the interactions a nd feedbacks between processes of these submodels are simulated together. As outlined in Figure 1-1, plant communities respond to available nutrients and water, which are drivers for plant growth; dynamics of live and standing dead vegetati on alter surface water runoff through changes in canopy structure and thus surface roughne ss. Hydrology in the model responds directly to the vegetation via linkages such as Manni ng’s roughness coefficient and transpiration losses. Water losses by plants vary with changes in biomass (leaf area index) and physical canopy structure. Availa bility of water in surface, unsaturated and saturated zone storage is one control on pl ant growth and mortality. Both surface and subsurface water transport dissolved nutrients and the soil water conditions affect the biogeochemical processes, while nutrient av ailability and uptake kinetics can control plant growth. Dead organic ma tter in different forms is a major source of nutrients. The coupled model was developed with in the Java-base d, object-oriented framework of the ACRU2000 model (Campbell et al., 2001; Clark et al., 2001; Kiker and Clark, 2001), an agrohydrological modeling sy stem originally cr eated by Schulze (1989, 1995) in South Africa. The coupled mode l was developed by adding hydrological components capable of multi-directional spatia l simulation of overland flow and lateral groundwater flow, nutrient components capable of multi-directional sp atial simulation of conservative solute, nitrogen and phosphorus transport, as well as a new vegetation dynamics simulation model. Study Objectives The overall purpose of this study was to develop a coupled model system capable of simulating hydrology, nutrient and vege tation dynamics simultaneously for south

PAGE 31

10 Florida flatwoods watersheds that incorporat e wetlands, uplands and transition zones. Specific objectives of this research can be summarized as follows: Modify the existing ACRU2000 modeling syst em to enable the multi-directional spatial simulation of flatwoods hydrol ogy, nutrient and vegetation dynamics. Test the accuracy of the modified hydrol ogic and nutrient models’ performance by comparing with the existing mode ls MIKE SHE and MODFLOW. Validate the modified hydrological m odel by comparing with FHANTM and against the measured data in Dry Lake Dairy #1, Kissimmee Rive r Basin, Florida. Calibrate, validate and evaluate the co upled hydrologic and nutrient model using the measured data from Buck Island Ranch, Lake Okeechobee Basin, Florida. Test the coupled hydrologic, nutrient, a nd vegetation model using scenarios based on conditions at Buck Island Ranch, Lake Okeechobee Basin, Florida. Investigate the interac tions among wetlands hydrology and nutrient dynamics imposed by alternative land management pr actices and anthropogenic activities in south Florida flatwoods watersheds. This dissertation is organized into 6 ch apters. Chapter 1 briefly introduces the background and objectives of th is study, as well as giving an overview of the coupled modeling system; Chapter 2 reviews previ ous modeling efforts in hydrology, nutrient and vegetation dynamic simulation, and discus ses model testing procedures including model calibration, verification, evaluation and sensitivity analysis; Chapter 3 focuses on the development of the multi-directional spa tial simulation of the hydrologic model, including a description of m odel components, algorithms, assumptions, and model testing and validation; Chapter 4 describes the development of the multi-directional spatial nutrient (nitrogen, phosphorus, and conservati ve solute) transpor t and transformation model, including model components, algorithms, assumptions and modeling testing and validation; Chapter 5 introduces the vegeta tion dynamics simulation model, including

PAGE 32

11 model algorithms and testing. Finally, Chapte r 6 gives a summary of the results found in this study, conclusions, and recommendations for future work.

PAGE 33

12 CHAPTER 2 LITERATURE REVIEW Overview of Previous Modeling Efforts Surface and subsurface water movement se rves as a major nutrient transport mechanism, both delivering essential nutrients to the biota and moving excess nutrients to receiving water bodies. The importance of hydrologic transport has been long recognized and considerable effort has been put into cr eating adequate models for various landscapes (Beven and Kirkby, 1979; Beasley and Huggins 1980). The complexity of a specific watershed simulation model depends on the te mporal and spatial resolution, and on the extent to which important hydrological a nd biochemical processes are considered (Krysanova et al., 1998). Over the past several decades many models, ranging from lumped conceptual models to semi-distribut ed models to fully distributed physicallybased models, have been developed. Early examples of lumped hydrologic mode ls are the Stanford Watershed Model (Crawford and Linsley, 1966), the SSARR (Streamflow Synthesis and Reservoir Regulation) model (Rockwood et al., 1972), the Sacramento model (Burnash et al., 1973), the tank model (Sugawara et al., 1976) HEC-1 (Hydrologic Engineering Center, 1981) and the HYMO (Williams and Hann, 1983). In these models, both differential equations based on simplified hydraulic laws and empirical algebraic equations were used for different processes. More recent conceptual models have incorporated soil moisture replenishment, depletion and redist ribution for the dynamic variation in areas contributing to direct runoff (Arnold and Fohrer, 2005).

PAGE 34

13 Progress in developing coupled hydrological/w ater quality models is more evident at the field scale or in small homogeneous watersheds than at large watershed and regional scales. Starting from the early 1970s, non-point source models have been developed in the USA in response to th e Clean Water Act. CREAMS (Chemicals, Runoff, and Erosion from Agricultural Ma nagement Systems) (Knisel, 1980) was developed to simulate the impact of land ma nagement on water, sediment, nutrients and pesticides leaving the edge of a field. Subsequently, severa l field-scale models evolved from the original CREAMS including G LEAMS (Groundwater Lo ading Effects of Agricultural Management Systems) (Leona rd et al., 1987) to si mulate groundwater pesticide loadings, EPIC (Erosion-Productivity Impact Calculator) (Williams et al., 1984) to simulate the impact of erosion on cr op productivity, and OPUS (Smith, 1992) to estimate the effects of management pr actices on non-point source pollution. Spatially-distributed models in larger watersheds represent a more complicated problem (Krysanova et al., 1998). Semi-distr ibuted physically-based hydrologic models have the advantage of a simple model stru cture and fewer model parameters together with a realistic representation of the wate rshed hydrologic proce ss. SWAT (Soil and Water Assessment Tool) (Arnold et al., 1993) is a continuous-time di stributed simulation watershed model to predict the effects of alternative management decisions on water, sediment and agricultural chemical yields with reasonable accuracy in watersheds and large river basins. SWAT was originally developed from CREAMS to a basin-scale model SWRRB (Arnold et al., 1990), and then combined with the ROTO model (Arnold, 1990) to form the more comprehensive mode l SWAT. The latest version, SWAT2000, has several significant enhancements that in clude: bacteria transport routines; urban

PAGE 35

14 routines; Green and Ampt infi ltration equation; improved weat her generator; ability to read in daily solar radiati on, relative humidity, wind speed and potential ET; Muskingum channel routing; and modified dormancy cal culations for tropical areas (Arnold and Fohrer, 2005). SWIM (Soil and Water Inte grated Model) (Krysa nova et al., 1998, 2005), a hybrid of SWAT and MATSALU (Krysanova et al., 1989), is a continuous-time spatially semi-distributed model, which in tegrates hydrological processes, vegetation growth (agricultural crops and natural ve getation), nutrient cy cling (nitrogen and phosphorus) and sediment transport at the ri ver basin scale. AGNPS (Agricultural NonPoint Source pollution model) (Young et al ., 1989) was developed to examine water quality as it is affected by soil erosion fr om agriculture and urban areas during single precipitation events at watershed scale. TOPMODEL (a TOPogra phy based hydrological MODEL) (Beven and Kirkby, 1979) is a variable contributing area conceptual model, in which the major factors affecting runoff ge neration are the catchment topography and the soil transmissivity that diminishes with depth. ANSWERS (Areal Nonpoint Source Watershed Environment Response Simula tion) (Beasley and Huggins, 1980) was developed to evaluate the effects of BM Ps on surface runoff and sediment loss from agricultural watersheds. Th e current version of the model, ANSWERS-2000, is a continuous simulation model that was devel oped in the mid 1990s (Bouraoui and Dillaha, 1996) with the revised nutrient submodels, im proved infiltration (Green and Ampt), and new soil moisture and plant growth components to permit long-term continuous simulation. Another class of models ar e based on differential equations for conservation of mass, energy and momentum, and are called fu lly physically-based distributed models.

PAGE 36

15 The spatial distribution of catchment para meters in such models is achieved by representing the basin on a grid network. Examples of physically-based distributed models include MIKE SHE (Refsgaard and Storm, 1995) and MODFLOW (McDonald and Harbaugh, 1988). MIKE SHE is a physi cally-based, distributed, integrated hydrological and water quality modeling system It simulates the hydrological cycle including evapotranspiration, overland flow, channel flow, soil water and ground water movement. Related water quality modules incl ude: 1) advection-dispersion, 2) particle tracking, 3) sorption and degradation, 4) ge ochemistry, 5) biodegradation, and 6) crop yield and nitrogen consumption. This modeling system can be used to predict pollutant loading and transport, pesticide leaching, a nd outcomes of alternate BMPs on watersheds and their underlying aquifers. MODFLOW is a three-dimensional finite difference groundwater model of the U. S. Geological Su rvey, which can simulate various stresses to the system such as wells, rivers, drai ns, head-dependent boundaries, recharge and evapotranspiration. MODFLOW can simula te homogeneous/heterogeneous systems, isotropic/anisotropic media, a nd steady-state/transient flow. Model development is often linked to wher e the model was developed and tested. To simulate the unique flatwoods ecohydrologi cal issues in south Florida watersheds, several models have been developed. These models can be approximately sorted into 3 types according to their f unctionalities, includ ing (1) hydrologic models such as DRAINMOD (Skaggs, 1980), the weighted implic it finite volume model (Lal et al., 1998), SFWMM (South Florida Water Manage ment Model) (SFWMD, 2005) and NSM (Natural System Model) (SFWMD, 1998); (2 ) water quality models such as CREAMSWT (Chemicals, Runoff, and Erosion from Agricultural Management Systems-Water

PAGE 37

16 Table) (Heatwole, 1986), EAAMOD-FIELD (Everglades Agricultural Area MODelField) (Bottcher et al., 1 998) and FHANTM (Field Hydrologic And Nutrient Transport Model) (Tremwel, 1992; Fr aisse and Campbell, 1996, 1997) ; and (3) ecohydrological models such as ELM (Everglades La ndscape Model) (Fitz et al., 2002) and FLATWOODS (Sun et al., 1998). Some of these models are field-scale, lumped conceptual models, incapable of spatially distributed simulation of ecohydrol ogic variations. For example, DRAINMOD (Skaggs, 1980) was designed to evaluate the e ffects of drainage on the water table depth below agricultural fields in the coastal plain of North Carolina where it performed well for the silty-clay soils, but required modifica tions to improve its ability to predict the outflows from a field with sandy soils (R ogers, 1985). CREAMS-WT (Heatwole, 1986) was modified from CREAMS (Knisel, 1980) to better represent the low phosphorus buffering capacity of sandy soils and the hydrology of flat, sandy, high-water-table flatwoods watersheds. FHANTM (Tremwel, 1992) is based on the DRAINMOD model but was modified to include simulation of phosphorus movement and routing of overland flow. The model was further modified and released as FHANTM 2.0 (Fraisse and Campbell, 1996, 1997) for generalized use in modeling cow/calf operations. This new version incorporates the nut rient component of GLEAMS (Leonard et al., 1987). EAAMOD-FIELD (Bottcher et al., 1998) was deve loped to assess the e ffects of different agricultural practices on phosphorus losses from fields in th e Everglades Agricultural Area. Zhang et al. (1995) compared CREAMS-WT and FHANTM and concluded that there is little difference in performance of these models in estimating runoff, phosphorus concentration and loads. Zhang et al. (1999) further compared FHANTM, FHANTM

PAGE 38

17 2.0, and EAAMOD-FIELD and concluded EAAM OD-FIELD was the best model to use for the regulatory program because it has the capability of simulating land use changes in the middle of a simulation period and the most pot ential to be enhanced so that it can be used in the Lake Okeechobee WOD regulator y program. Hendricks (2003) compared FHANTM and EAAMOD, and concluded that EAAMOD predicted water table levels more accurately than FHANTM, and EAAMOD was able to remove the depressional storage water more quickly than FHANTM, allowing EAAMOD’s water table level to respond more quickly when rainfall events subside. Some of the other models, although they are physically distributed, generally are designed for regional, long-term applications to watersheds at a sc ale of thousands of square kilometers. Although scalable, perf ormance constraints may impose practical limits on the time and space scales. These models are not intended for local-scale decision-making support because the details of local-scale watersheds may not be sufficiently simulated. For instance, SF WMM (SFWMD, 2005) is a regional-scale computer model that simulates the hydrology a nd the management of the water resources system from Lake Okeechobee to Flor ida Bay using a mesh of 2 mile 2 mile cells. NSM (SFWMD, 2005) uses the same clima tic inputs, time ste p, calibrated model parameters and algorithm as the SFWMM, but it differs from the SFWMM in that it does not simulate the influence of any man-made f eatures and uses estimates of pre-subsidence topography and historical vegeta tion cover. ELM (Fitz et al., 2002) is a regional-scale, integrated ecological assessment tool design ed to understand and predict the landscape response to different water management scenar ios in south Florida, USA. In simulating changes to habitat distributions, the EL M dynamically integrates hydrology, water

PAGE 39

18 quality, soils, periphyton, and vegetation in the Everglades region. Due to the computational complexity in the hydrologic modules, the model gene rally constrains the spatial resolution to 1.0 km2 grid cells. Other models either require relatively small time steps to maintain stability and accuracy, such as the weighted implicit fini te-volume model (Lal et al., 1998), or are specially designed for certain applicati on such as FLATWOODS (Sun et al., 1998), which was developed for the cypress wetlandpine upland landscape using a distributed, physically based approach especially us eful in the study of forest hydrology. Vegetation models with different comple xities range from simple empirical formulae to complex physiologically-based mode ls. These models differ as a result of the objectives of model development, and hen ce the required scale a nd degree of detail and comprehensiveness (Van Ittersum et al., 2003). Over the past decades, a considerable number of vegeta tion models have been develo ped to target many different aspects of the terrestrial carbon cycle. At th e core of most of th ese models is a net primary productivity submodel, although many of the mechan isms behind terrestrial productivity are still not prope rly understood. As a consequence of this, as well as the availability of data and computational cap acity at the time of development, models developed to calculate net primary productiv ity are quite diverse in their approaches (Adams et al. 2003). At one end of the sp ectrum is the simple, empirically derived, correlation of net primary productivity with air temperature and pr ecipitation used, for example, in the Miami model (Leith, 1975a). At the other end is the detailed simulation of process-oriented, biochemistry used in the Hurley Pasture Model (Thornley and

PAGE 40

19 Cannell, 1997), DSSAT (Jones et al., 2003) and the Wageningen crop models (Van Ittersum et al., 2003) Several process-based grassland models ha ve been reported such as ELM (Innis, 1978), PAPRAN (Seligman and Van Keule n, 1981), CENTURY (Parton et al., 1987, 1996), ERHYM-II (White, 1987), GEM (Hunt et al., 1991), CCGRASS (Verberne, 1992) and GEMT (Chen and Cougheour, 1994). Comp ared with these models, the Hurley Pasture Model (Thornley and Cannell, 1997) is more comprehensive in simulating ecosystem processes. A limited number of previous efforts at modeling vegetation dynamics in south Florida flatwoods watersheds have been report ed in the literature. Fitz et al. (1996) developed a processed-oriented General Ecosystem Model (GEM) to capture the response of macrophyte and algae communities to simulated levels of nutrients, water and environment inputs such as light intens ity, temperature and fi re. The vegetation submodel in SFWMD wetlands model (SFW MD, 1997b) was developed for sawgrass (Cladium) and cattail (Typha) to simula te aboveground (leave s and shoots) and belowground (roots and rhizomes) biomass and nutrient content. This model includes the interactions between light, temperature a nd nutrients on plant growth and decomposition but it does not consider the effects of hydrology, grazers or other influence on plant growth or survival. Wu et al. (1997) developed SAWCAT using Markov Chain probabilities to simulate vegetation dynamics in wetlands in response to levees, water depth, and phosphorus. The tr ansitional probability model only simulates the number, mean size, and largest size of patches of each vegetation type, but does not simulate biomass production.

PAGE 41

20 Process-oriented vegetation models genera lly require comprehensive physiological and phenological parameters so th at application and testing of these models is difficult in watersheds where little data are available. In this research a simple and generic vegetation dynamics model for simulating th e growth dynamics of pasture and wetland grasses found in south Florida watersheds was developed. Details of the vegetation model development are presented in Chapter 5. ACRU2000 Modeling System The ACRU (Agricultural Catchment Res earch Unit, v3.00) model was originally developed in FORTRAN by Schulze (1995) in the department of Agricultural Engineering, University of Natal, Piet ermaritzburg, South Africa for simulating catchment, forest and wetland hydrology, fl ood routing, and soil erosion, dam design, irrigation modeling and crop yiel d modeling within South Afri ca. Schulze (1995) gives a comprehensive description of ACRU (v3.00) and particular aspects of the model, such as hydrologic sensitivity to climate change, global climate-change and agricultural productivity are described in a range of publicatio ns (Schulze et al ., 1993; Smithers and Caldecott, 1993; New and Schulze, 1996). ACRU (v3.00) has been enhanced by many additions in response to the need for answers to a range of water-re lated issues in South Africa and elsewhere (C lark et al., 2001). ACRU (v3.00) is described as a physical, con ceptual model. It is physical in that the hydrological processes are represented as ex plicitly as possible and conceptual in that the components, relationships and processes in the system are idealized. The hydrologic model in ACRU (v3.00) can operate either as a lumped small catchment model with relatively homogeneous soil and la nd cover attributes, or as a distributed cell-type model where complex catchments are separated into subcatchments or land segments.

PAGE 42

21 Figure 2-1. General structure of the lu mped ACRU (v3.00) model (Schulze, 1995). In the lumped ACRU (v3.00) model, the core of the model comprises the water budget routines for multiple soil layers on a catchment, the general structure of which is illustrated in Figure 2-1. Water enters the subcat chment as precipitation and/or irrigation. The vegetated or impervious land surface may intercept all or part of the precipitation, and this intercepted water in turn is ev aporated back into the atmosphere. For precipitation reaching the soil surface, runo ff is calculated using the modified SCS equation, and the balance infilt rates into the topsoil horiz on (Clark et al. 2001). Soil water evaporation takes place in the topsoil horizon, and transp iration takes place in those soil horizons that contain roots. Soil water movement takes place between soil horizons. Soil water can percolate from the bottom soil horizon to the groundwater storage. Baseflow is generated and released from th e intermediate and groundwater stores as a part of daily surface runoff in this subcatchment. Surface runoff is routed to the subcatchment outlet as a combination of quick fl ow (storm released into the stream on the

PAGE 43

22 same day as the rainfall events) and delaye d storm flow, a surroga te for post storm interflow. In the distributed ACRU mode l, a catchment is divided in to a set of subcatchments where the lumped model is applied indivi dually throughout the entire simulation time period, then the generated daily runoff from each subcatchment can be convoluted into a runoff hydrograph using the triangular unit hydro graph approach for that subcatchment. The runoff hydrograph from each subcatchment is then routed along the pre-specified pathways to the outlet of the catchment us ing the Muskingum-Cunge method. No spatial exchange of water flows is considered among subcatchments in ACRU (v3.00). This design may be sufficient to present hydr ology in watersheds where topographical gradients dominate water movement a nd subcatchment hydrology is relatively independent, but is not desirable for low-relief catchments. To enable code expansion and functionality to meet different needs, ACRU (v3.00) was entirely restructured into ACRU2000 (C ampbell et al., 2001; Clark et al., 2001; Kiker and Clark, 2001) within a Java-based, object-oriented fram ework. In ACRU2000, the internal hydrological mode ling features of ACRU (v3.00) were retained but the underlying foundation was changed to allow new features to be developed. The restructured version revised the surface runoff routing algorithm in ACRU (v3.00) by executing each subcatchment every time step and then routing runoff from each subcatchment to the outlet of watershed using the pre-specified pathway before continuing to the next time interval. Additionally, a nutri ent module, ACRU-NP (Campbell et al., 2001), patterned after transformation and tran sport concepts used in the GLEAMS model (Knisel et al. 1993), was added into ACRU2000. Campbell et al.

PAGE 44

23 (2001), Clark et al. (2001), and Kiker and Clark (2001) de scribe the restructured ACRU2000. Figure 2-2. Land segment configuration of the ACRU2000 model (Kiker et al., 2001). Figure 2-2 shows an example of the land segment configuration of a simulation domain when ACRU2000 runs in a distributed mode. Hydrologic processes, such as rainfall, canopy interception, evapotranspiration, infiltration, runoff, and percolation, etc., simulated in the lumped ACRU model are c onsistently applied in each land segment when operating the model in a distributed m ode. Land segments are connected to one another in a pre-specified pathway, which is used to deliver the runoff from upstream land segments down to outlet of the simulation domain without interactions in land segments along the way. Groundwater is treated as part of surface runoff from each land segment. Physical water exchange is i gnored between land segments because water flows bypass downstream land segments. This design may be sufficient to present hydrology in watersheds where topographica l gradients dominate water movement and subcatchment hydrology is relatively indepe ndent. However, it is not adequate in

PAGE 45

24 simulating south Florida flatwoods hydrol ogy where water movement is primarily dominated by hydraulic gradient s and ecohydrologic variation is spatially inte ractive in a significant manner. Furthermore, the nut rient model in ACRU 2000, patterned after transformation and transport concepts used in GLEAMS (Leonard et al. 1987), does not consider the spatial transport of nutrients through lateral subsurface water movement. In order to make the ACRU2000 model a pplicable in south Florida flatwoods watersheds, modifications were needed to en able multi-directional spatial simulation of water and nutrients. With the object-orien ted design of ACRU2000, it is feasible to add new modules while retaining the existing capaciti es of the model. The following Chapters 3, 4 and 5 describe the details of modifications for hydrology and nutrient transport simulation, and the a ddition of the vegetation dynamics model, respectively. Model Testing Procedures Model testing is a very important part in model development in that it can detect inappropriate design in model algorithms, increase the model robustness and accuracy through model calibration and validation, and de termine the limitations and constraints of model application through model evaluation. However, general methodologies related to model calibration and validation have been subj ect to considerable discussion and dispute during the past decades (Refsgaard, 1997). Distributed models have the capacity to simulate spatial variations. Therefore the number of parameters and variables are ofte n several orders of magnitude higher than those required for lumped models of the sa me area. Lumped and distributed models should have different requirements with regard to model calib ration and validation procedures. Unfortunately, much attention has been given to sp ecific calibration and validation procedures for lumped models, wher eas very limited attention has so far been

PAGE 46

25 devoted to the more complicated tasks in c onnection with distribute d models (Refsgaard, 1997). The procedure used for model calib ration and validati on in this study are summarized in the following sections. Model Calibration Calibration is the process by which model para meters are adjusted to give the best fit between simulated results and observed data at a particulate site. In other words, calibration involves adjusting certain model parameters by systematically comparing simulated results with observations while model structure remains the same. Model calibration is built on the assumption that a well calibrated model enhances its predictive capability by incorporating the best available data and adjusting calibration parameters to obtain a close agreement between mo del output and historical data. To calibrate a model adequately, the calib ration period should include events that significantly stress the simulation system. Calib ration of distributed integrated models is more challenging, as discussed above, due to complex model structures and large parameter sets, and requires calibration over both time and space. Model Validation Validation is the process by which the cal ibrated model parameters are used to predict state variables during periods when comparisons ca n be made to an independent set of historical data, which were not previously used in the calibration process. The purpose of validation is to determine if the model is sufficiently accurate for its application as defined by objectives of the simulation study. The model is said to be validated if its accuracy and predictive capac ities in the validation period have been proven to lie within well defined limits that may depend on intended model use.

PAGE 47

26 Sensitivity Analysis Sensitivity analysis (SA) is the process of varying model input parameters and evaluating how model output changes with su ch variation (SFWMD, 2005). If a small change in a parameter results in relatively large changes in the model output, the model is said to be sensitive to that parameter. This may mean that the parameter has to be determined very accurately. Prior to accepting the final set of calibrated model parameters, a sensitivity analysis should be performed to determine the relative magnitude of model response to ch anges in selected parameters. The most common form of SA is indepe ndent parameter perturbation in which parameters are varied individually by a fixed percentage around the base value (Ferreira et al., 1995). An example of this approach is first-order analysis (Hann and Zhang, 1996), in which the first order derivatives in the Taylor series approximation of the output variables are estimated: i i j i i j i j jix ) x ( y ) x x ( y x y s (i = 1, 2, n; j = 1, 2, m) (2-1) in which, sji is the sensitivity of output yj to the change of parameters xi; yj (j = 1, 2, m) are the m predictions; xi (i = 1, 2, n) are the n parameters These sensitivities sji have units and thus are difficult to compare over para meters of interest directly. Thus one can employ normalized or dimensionless sensitivit ies, Jones and Luyten (1998) used the relative sensitivity, rsji: j i ji i i j j jiy x s x x y y rs (2-2) where xi is the calibrated value for input parameter xi, and yj is the prediction by the calibrated model parameters. However, inte ractions between two or more parameters

PAGE 48

27 (non-linear effects) may affect the model out puts, even if the out put is linear in each parameter. To check for two parameter inte ractions, second or high order derivatives in the Taylor series approximati on must be evaluated. In prac tice, the computation load to find all possible significant deriva tives may be too excessive. Distributed integrated models are struct ured to enable simulation of spatial variations in catchment charac teristics and thus require more parameters, in principle, than the lumped models. Thus it is necessa ry to investigate the spatial and temporal sensitivities of input parameters. The first order met hod as shown in Equation (2-1) and (2-2) was used to conduct the SA on majo r model outputs including surface runoff, groundwater table depth, and P and N loads. If the output is a time se ries, the sensitivity is a function of time, which may help to de tect the influences of input parameters on extreme values and trends of outputs. If the output is an individual value, an individual sensitivity is calculated, which may help to de tect the influences of input parameters on the overall model response. The spatial sensit ivities of outputs can be calculated using the same equations when the spatial observa tion of output variables are available. The significance of model sensitivity anal ysis is two-fold in that it provides information on the behavior of model output to input parameters which, in turn, can be used in model calibration; also, it gives insight in establishing priorities related to future data collection efforts. Sensitivity analysis is distinguished from uncertainty analysis in that it is a measure of the re lative importance that each input parameter has on the range of simulated outputs. Uncertainty analysis qu antifies the confidence in particular output variables given the probability distributions for input parameters. While sensitivity analysis is often limited to parameter sens itivity, uncertainty may be generated by a

PAGE 49

28 number of factors including parameter uncertainty, boundary and initial condition uncertainty, model spatial and temporal resolu tion, availability and quality of data, and model errors. Model Evaluation For sake of the model calibration and va lidation, and the comparison of models, one needs quantitative information to meas ure model performance compared to observed data or other model predictions Statistical approaches to quantify the accu racy of model predictions provide standardized measures of model performance although even these methods do not provide completely clear-cut conclusions about the accuracy of model predictions (Ramph, 2004). Give n these caveats, the use of se veral different measures of performance to evaluate a model may pres ent a more complete picture of model performance than any single measure and may allow the user to weight individual results according to their priorities (Ramph, 2004). Statistics A number of statistics were used to ev aluate model results in this study, including model bias, average relative e rror (RE), root mean square error (RMSE), coefficient of variation (CV), Pearson product-mo ment correlation coefficient (R2), and Nash-Sutcliffe (NS) coefficient (Nash and Sutcliffe, 1970). Br ief overviews of these statistical measures are provided below. Bias, also called the averag e error, determines the average deviation of the predicted values from the measured values gi ven N simulated-measured value pairs. As can be seen from Equation (2-3), where Pi is the observed value, Oi is the modelsimulated value, and N is the number of obser vations, bias is calculated as the mean

PAGE 50

29 differences between paired observed and simula ted values. Bias valu es closer to zero indicate better overall model performance. N 1 i i i) O P ( N 1 bias bias (2-3) The relative error is a unitless, normalized parameter that also quantifies the bias of the predicted values. The arithmetic average relative error (RE) as shown in Equation (24) determines if the model overestimates (pos itive deviation) or unde restimates (negative deviation) the measured values, which is most useful for comparing models and data sets (Hession et al., 1994). Howeve r, Hession et al. (1994) indi cated that the arithmetic average relative error can be significantly influenced by one or two outliers. N ) O P ( O 1 REN 1 i i i RE (2-4) The root mean square error (RMSE), or st andard deviation/erro r, provides a direct measure of the error between the model and the observed data (Thomann, 1982) as seen in Equation (2-5). This statistic is used to measure the discrepancy between modeled and observed values, and indicates the overall pr edictive accuracy of a model. Due to the quadratic term, greater weight is given to la rger discrepancies. With this measure, smaller values indicate better mode l performance (Evans et al., 2003). N 1 i 2 i i) O P ( N 1 RMSE RMSE 0 (2-5) The RMSE is essentially the overall sum of squares errors normalized to the number of observations (Hession et al., 1994).

PAGE 51

30 The coefficient of variation (Young and Alward, 1983) defines a normalized error as shown in Equation (2-6), which is th e RMSE normalized to the overall mean. N 1 i 2 i i) O P ( N 1 O 1 CV CV (2-6) Using the RMSE, along with the CV, Hedden (1986) suggests that for screening applications a model should be able to repl icate observed data within an order of magnitude, and for site-specific applications th e predictions should be within a factor of two. In the criterion of within “a factor of two” will be satisfied when the CV value is less than one and the criterion of within “an order of magnitude” will be met when it is less than nine. The Pearson product-moment correlation coefficient (R2) value, also called the goodness of fit, is a measure of the degree of linear association between two variables. Depending on the strength of the linear relationship, the R2 can vary from 0 to 1. The closer R2 is to 1, the better the regression explai ns the relationship between simulated and observed. However, it does not explain how cl ose the relationship is to the perfect linear fit (R2 = 1) between observed and predicted values. 2 2 i 2 i N 1 i i i 2) P P ( ) O O ( ) P P )( O O ( R 1 R 02 (2-7) Like the R2 measure described above, Nash-Sut cliffe (NS) coefficient (Nash and Sutcliffe, 1970) is another indicat or of goodness-of-fit, and is one of the statistics that have been recommended by the American Soci ety of Civil Engineers (ASCE, 1993) for evaluation of the performance of hydrological models. The NS can vary from 0 to 1,

PAGE 52

31 with 1 indicating a perfect fit. Computati onally, the NS could be negative but this becomes rather meaningless as far as interpreta tion or results are concerned. For NS = 0, the interpretation can be made that the model is predicting no better than using the average of the observed data (ASCE, 1993). The statistics works best when the coefficient of variation for the obser ved data set is large (ASCE, 1993). N 1 i 2 i N 1 i 2 i i) O O ( ) P O ( 1 NS 1 NS (2-8) Graphic representation Graphic representation enables visualizi ng the relationship between predicted (Pi) and measured (Oi) values and provides a straightforward way to compare the fit between predicted and measured values Scatterplots and duration curves are two commonly used graphic representation techniques in the hydrological literature. Scatterplot is a simple method to visualize the relationship between Pi and Oi by fitting a regression line between Pi (Y-axis) and Oi (X-axis), relative to a line designating a 1:1 relationship. The resulting pattern indi cates the type and asso ciation/strength of the relationship between predicted and observed values. A positive association would be indicated by upward trend (pos itive slope) and a negative asso ciation would be indicated by a downward trend (negative slope). Or, ther e might not be any notable association, in which case a scatterplot would not indicate a ny trends whatsoever. The slope of the regression line and its Y-intercept may provide evidence of systematic error in the model, providing quantities that can be compared across models. A slope of 1.0 with an intercept equal to 0.0 in dicates perfect fit of the model predictions.

PAGE 53

32 A duration curve represents the relationshi p between magnitude and frequency of a variable for a particular watershed providi ng an estimate of the percentage of time a given variable was equaled or exceeded ove r a period of time. There are different approaches to generate duration curves fo r satisfying different purposes. Among these, the flow duration curve (FDC) has widespread application in hydrol ogic studies such as hydropower, water-supply and irriga tion planning. In recent y ears, duration curves have also been applied in other areas such as lo ad assessment and TMDL development. For this study, the flow and load duration curves were compared for simulated and observed variables. The basic procedure for generating a dur ation curve includes: 1) ranking the daily/weekly/annually simulated and measured data separately from highest to lowest for each dataset of interest; 2) calculating percen t of days/weeks/years these variables were exceeded (= rank/number of tota l data points); 3) plotting both simulated and measured data vs. percent of time interval exceeded on the same plot. Comparison of duration curves between model predictions and observed data provides a direct evaluation of the accuracy of particular flow/load ranges. However, there are some shortcomings of the durati on curve method. For example, there is no commonly accepted method to deal with repeat ed values. Therefore, using duration curves altogether with other statisti cs is advisable in model evaluation. No one of the abovementioned model eval uation approaches will be best in all situations. Reviewing several of these meas ures together will provide a more complete description of model performa nce (Ramph, 2004). The results should also be viewed in the context of the intended us e of the model. Users must decide for themselves what

PAGE 54

33 level of performance is acceptable and, likewise, which approach is most appropriate to their interests.

PAGE 55

34 CHAPTER 3 HYDROLOGIC SIMULATION MODEL Introduction In the coupled modeling system, the hydrol ogic model plays an important role in that it not only simulates hydrologic processe s but also provides a framework for the nutrient and vegetation models di scussed in Chapters 4 and 5. Therefore, the design and development of the hydrologic model has to take these other models into account. An example catchment, shown in Figure 3-1A is discretized into a network (Figure 3-1B) of multiple land segments, each of which is a unique hydrologic unit with varying shape and size depending on land cover, t opography, vegetation type s and soil types. Each land segment is bordered by either the internal boundaries, which are artificially defined separations between two land segmen ts, and/or the external boundaries, which are the separation between th e simulated catchment and its neighboring catchments. A continuous unconfined groundwater system is a ssumed to underlie the discretized model domain. For each land segment, it is assu med that the soil beneath is laterally homogenous with heterogeneous vertical soil la yers that are divided according to natural soil horizons and soil properties. Additionall y, soil layers in neighboring land segments have the same number of soil layers of the same thickness, and adjacent soil layers are laterally connected. A fixed time step of one day is used in the coupled model. The selection of the time step is a tradeoff between time scales required for different model processes and general temporal scales of input data, such as rainfall and evapotranspiration.

PAGE 56

35 Figure 3-1. Schematic of an example cat chment and its spatial discretization. An overall flow chart of the coupled m odeling system and major model processes included in the model are illustrated in Appendix A. In the natural condition, hydrological processes, toge ther with nutrient and ve getation processes, occur continuously in a simultaneous manner, but in the model these processes are discretized and simulated in a sequential manner within the one day time step. In the model these processes are grouped into vertical and hor izontal processes for each land segment according to the flow direction each process describes and whether a process interacts with others in a neighboring land segment. Vertical processes deal with the water movement, nutrient cycling, and vegetation dynamics through the soil profile whereas horizontal processes simulate lateral wate r movement, nutrient cycling, and vegetation dynamics over the domain and require the si mulation of water and nutrient exchange between adjacent land segments. The vertical process group is the basic building block of the model, simulating the temporal dynami cs of important hydrological, chemical, and growth processes within one land segment. The horizontal process group is the core of the distributed modeling system in that it spatially connects all land segments in a watershed together through processes such as lateral water movement, nutrient transport, A. Example Catchment B. Catchment Discretization External boundary Internal boundary Land segment

PAGE 57

36 and biomass expansion. In the coupled m odel, the vertical processes are executed sequentially throughout the simulation doma in for each land segment followed by the horizontal processes across all land segments. Nutrient and vegetation processes are also separately grouped into vertical and horizontal processes. Di scussions of these processes are presented in detail in Chapters 4 and 5. Figure 3-2. Hydrological processes in th e modified ACRU2000 hydrologic model. Hydrologic processes simulated in the m odified ACRU2000 model are depicted in Figure 3-2, including rainfall, interception, evapotranspiration (ET), infiltration, soil water redistribution, upward flux, deep s eepage, overland flow, canal flow, and groundwater flow. Although some of these pr ocesses including rainfall, interception, ET, infiltration, and soil water re distribution existed in the original ACRU2000 model, Martinez (2006) added a lternative processes to simulate ET, infiltration, and soil water redistribution and created new processes for deep seepage and upward flux in order to better represent the soil wate r dynamics for flat, poorly-d rained sandy, and high-watertable flatwoods soils in south Florida. Multi-directional spatial simulation of overland Canal Flow Rainfall Eva p otrans p iration Overland Flow Infiltration Groundwater Flow U p ward Flu x Dee p See p a g e Soil Water Redistribution Interce p tion

PAGE 58

37 and lateral groundwater flow, and canal flow did not exis t in the original ACRU2000 model. Developing algorithms for these proc esses was the primary purpose of this study. In the following hydrologic components sec tions, the methodology used to simulate each process is described, with more detail given to the new overland flow (including canal flow) and lateral gr oundwater flow processes. Schulze (1995) gave a comprehensive description of the hydrologic processes incorporated into ACRU (v3.00). The hydrologic model was developed base d on the existing hydrologic model in ACRU2000 (Campbell et al., 2001; Clark et al., 2001; Kiker and Clark, 2001) by adding new components and modifying existing com ponents needed for multi-directional spatial simulation, with an emphasis of considering lateral surface and subsurface flow exchange throughout all soil layers betw een adjacent land segments. Changes were made so that the original model features were retained for other potential applic ations while enabling new functionality. The objectives of this Chapter are To modify the ACRU2000 hydrologic m odel to add new hydrologic components that enable multi-directional spatial simulation of overland flow and lateral groundwater flow; To analyze the influence of the simulati on sequence on simulation results from the modified model; To test the ability of the modified m odel to accurately si mulate hydrological processes by comparing with well-ac cepted models including MIKE SHE (DHI, 2004) and MODFLOW (McDonald and Harbaugh, 1988); To test the reliability of the modifi ed model in predicting major hydrologic processes by comparing with the F HANTM model simulations of a field experiment in the Kissimmee River Basin, Florida. In the following sections, the methodology for each hydrologic component simulated in the modified version of the hydrologic model is described, followed by the

PAGE 59

38 description of boundary and ini tial conditions required for th e model. Then, simulation sequence analyses, hypothetical scenario tests and an appli cation in Dry Lake Dairy #1, Kissimmee River Basin, Florida are presented and discussed and conclusions are drawn. Appendix D lists and describes the added or mo dified objects. Appe ndix E lists all newly added variables associated with the coupled modeling system. Vertical Hydrologic Components Rainfall Rainfall is the fundamental driving force behind most hydrological processes. The success of hydrological simulati on studies depends to a large extent on the precision with which the rainfall data are observed temporally and spatially and how they are processed in the model. Several techniques in ACRU (V3.00) are available for estimating aerial rainfall for input files, including Driver station method ACRU-300 trend surface approach Thiessen polygon method Inverse distance interpolation technique and Spline interpolation technique The design of input files in the model al lows each land segment to have its own daily time series of rainfall depth input. Th erefore, the temporal and spatial distribution of rainfall can be differentiated when such spat ial rainfall data are available. But within one land segment, or when the model is used in the lumped mode, the rainfall is assumed to have a uniform spatial distribution. Canopy Interception Canopy interception is calculated as the mi nimum of the available rainfall and the amount of interception storag e capacity availabl e on the intercepting vegetation leaf canopy. The available rainfall first fills av ailable canopy storage, then the remaining

PAGE 60

39 rainfall is assumed to reach the land surface and is available for infiltration or/and surface runoff. If the rainfall is less than or equal to the available canopy storage, all rainfall is intercepted. Evapotranspiration The calculation of actual evapotranspiration requires three steps including reference potential evapotranspirati on (RET), potential evapotra nspiration (PET), and actual evapotranspiration (AET). The latter is furthe r separated into soil evaporation and plant transpiration for later use in nutrient uptake discussed in Chapter 4. The reference evapotranspiration (RET) refe rs to the maximum evaporation from a short grass surface, or the maximum evaporati on from an actively growing alfalfa crop, at least 0.3 m tall, standing erect and cove ring an extensive area. There are many methods of estimating RET in ACRU (v3.00) ranging from complex physically based equations to simple measurements and even simpler surrogates based on single variables such as temperature: The daily United States Weather Bureau Class A pan evaporation. The 1948 Penman equation Temperature based equations in ACRU (v3.00) The Linacre suite of equations (1977, 1984, 1991) The Hargreaves and Samani methods (1982 & 1985) The Blaney and Cri ddle (1950) equation The Thornthwaite (1948) equation In addition, Martinez (2006) added the Penm an-Monteith grass reference potential evapotranspiration approach (Allen et al. 1998 ), referred to as FAO56PM in the model. The RET can be calculated separately and input as a time series or calculated directly by the model with required input parameters and data.

PAGE 61

40 For all methods, evapotranspiration is calcu lated in a top-down approach. First the evaporation demand is applied to intercepte d water, next to ponded water on the ground surface, and then to the soil as evaporation and transpira tion. The PET for a crop is obtained by multiplying RET with a month-bymonth crop coefficient which integrates numerous dynamic processes relating to cr op transpiration and soil evaporation. Two existing AET calculation methods of ACRU (v3.00) have been adapted for use, which calculate AET either as a lumped quantity or by determining soil evaporation and transpiration separately using Ritchie’ s method (Ritchie 1972). In addition, a third simple actual evapotranspiration me thod was added by Martinez (2006). When calculating evapotranspiration as a lumped quantity, the RET is multiplied by the crop coefficient (V1CAY) to get the PET, then the layer AET within the root zone is calculated as the minimum between the PET and the available water storage (beyond the wilting point) for a specific soil layer. Using an approximate method (Childs and Hanks, 1975), the layer AET is split be tween soil water evaporation (AEs) and plant transpiration (AEt) for use for plant uptake of solutes and nutrients: t sAE AET AE (3-1) 0.2 V1CAY 0 0.2 V1CAY ) 0.8 0.2 V1CAY ( AET 0.95 AEt (3-2) When using Ritchie’s method to determin e evapotranspiration, the potential soil evaporation and plant transpiration are sepa rated as a function of the leaf area index (V1LAI). t sPE PET PE (3-3)

PAGE 62

41 7 2 LAI V1 PET 95 0 7 2 LAI V1 PET ) 21 0 LAI 1 V 7 0 ( PE5 0 t (3-4) The potential transpiration (PEt) is further adjusted by being multiplied by a crop coefficient. The actual plant transpiration ca n be either equal to maximum transpiration for a given soil layer or be less than maximu m transpiration because of a deficiency of available soil water or an excess of soil wa ter. According to Martinez (2006), if the FAO56PM RET is used, the potential evaporat ion is multiplied by a factor of 1.15 to account for the difference in albedo between bare soil and a vegetated surface as recommended by Allen et al. (1998). The pot ential soil evaporat ion (PEs) is further adjusted for the percent surface cover by mulc h. The soil water evaporation takes place down to a user defined depth of the soil w ith recommended values ranging from 0.1 to 0.15 m (Allen et al. 1998). A ccording to Ritchie’s method actual evaporation from the soil surface continues at a maxi mum rate equal to the potentia l rate (stage 1) until the accumulated soil water evaporation exceeds the stage 1 upper limit. After that, soil water evaporation proceeds at a reduced (stage 2) rate. The new actual evapotranspiration method th at has been added to the model is a simplification to the lumped evapotrans piration method described above. The simplifications include ne glecting water stress an d the apportionment of evapotranspiration to different layers. Po tential evapotranspiration is applied to individual soil layers in the original model according to the fraction of plant roots within that layer. Similarly, the new method app lies potential evapotranspiration from the topmost soil layer until the wilting point is reached and then ap plies the remainder to the layer below until the bottom of the root zone (the last layer containing roots) is reached.

PAGE 63

42 Infiltration Infiltration is a process by which water on the soil surface enters the soil. Considering the high infiltrating capacity of fl atwoods soils, the rainfa ll rate is assumed to be less than the maximum infiltration ra te and the water on the ground surface is allowed to infiltrate until the soil profile becomes completely saturated. Thus in the modified ACRU2000, saturated-excess ponding a nd runoff is simulated. The volume of infiltration is taken as the minimum of the rainfall rate multiplied by grid cell area and time step and the available void space be tween the water table and land surface. Infiltrating water is moved into the top-most soil layer and is further moved into deeper soil layers through a process discussed below. Water that infiltrates into the soil profile will cause a rise in the water table after any depleted root zone soil moisture is replenished. Soil Water Redistribution In ACRU (v3.00), unsaturated soil water redistribution downwards can take place from the top to a subsoil horizon and from the subsoil horizon to groundwater when the soil moisture content in th e upper soil horizon is above the field capacity. Upward redistribution, which is similar to capillary movement, takes place from the bottom soil horizon and works up through the soil horizons. If a soil horizon is above porosity then the excess water is moved to the soil horizon ab ove, or in the case of the top soil horizon the surface of the land segment, wher e it is added to ponded water. Martinez (2006) created an alternative pro cess specifically for the flatwoods soils. This process redistributes in filtrating water based on the drained-to-equilibrium soil moisture content of each soil la yer. According to Martinez ( 2006), this process has to be turned on after infiltration, net infiltrati on, new water table de pth (if any), and new

PAGE 64

43 drained-to-equilibrium soil moisture contents (i f any) have been determined. The process proceeds from the top down, and each soil layer is drained to equilibrium. Then, starting from the bottom-most layer, excess water is moved upwards from the lower to upper layers, or to the ground surface if necessary, to ensure that no layer is above its drainedto-equilibrium water content. Upward Flux Evapotranspiration depletes th e soil moisture content in the root zone and causes an upward gradient between the soil below and the root zone, which induces an upward flux of water within the soil. This depleted r oot zone has to be replenished first before infiltration occurs to cause a rise in the water table. The upward flux algorithm was added by Martinez (2006) using an approximate method (Anat et al., 1965) to calculate the maximum steady-state upward flux: b 2 s ufh d 1 886 1 1 K q (3-5) where quf is the amount of upward flux on a given day [L/T]; Ks is the saturated hydraulic conductivity of the soil [L/T]; d is the distance between the water table and the depleted root zone [L]; hb is the bubbling pressure head [L]; and is a function of the pore size distribution index, of the Brooks and Corey (1964) model, i.e., 3 2. This relationship for upward flux assumes that the soil profile is homogenous. For each soil layer a set of parameters, including Ks, hb, and are input. These parameters can be best obtained by fitting the above equati on to a steady-state solution of Richard’s equation for all water table depths below the layer in question. Alternatively, SOILPAR (v2.00) (Acutis and Donatelli, 2003) can be used to estimate these parameters as well.

PAGE 65

44 The amount of upward flux that actually o ccurs on a given day is the minimum of the maximum value calculated and the amount to which the root zone was depleted (and is now fully replenished by upward flux). Deep Seepage Deep seepage can occur through a restric tive layer located below the soil profile according to Darcy’s Law. It is calculated as: d H H K qd w s ds (3-6) where qds is the amount of daily deep seepage through the restrictive layer [L/T]; Ks the restrictive layer sa turated hydraulic c onductivity [L/T]; Hw is the height of water table above the restrictive layer [L]; Hd is the hydraulic head below the restrictive layer [L]; and d is the thickness of th e restrictive layer [L]. Horizontal Hydrologic Components Horizontal hydrologic co mponents including overland flow, lateral groundwater flow and canal flow were newly added into the original ACRU2000 model. The simulation of these components requires water exchange between adjacent land segments. Lateral flow is simulated sequentially, with the simulation sequence determined by the topographic elevations of land segments and the sequence of land segment input files. More details and te sting regarding the si mulation sequence are discussed in the following section. The following assumptions were made in the simulation of multi-directional lateral flow: For each land segment, only outgoing overl and flows are estimated except when inflows occur through external boundaries;

PAGE 66

45 Overland flows happen only when a positive hydraulic gradient occurs between one land segment and its neighboring land segments, and the ponded water depth in this land segment exceeds its maximum depression storage; Potential overland flows are estimated fo r all neighboring land segments based on hydraulic gradients. If there is insufficien t storage to supply all potential outflows, actual outflows are determined through storage apportionment; Soil within each land segment is unconfined thus the hydraulic head is the water level elevation in this land segment above the datum; Soils beneath land segments have the same thickness and they are split into the same number of computational soil layers; Adjacent computational soil layers in ne ighboring land segments have the same thickness and are laterally connected. Hydraulic gradients are the dr iving force to transfer gr oundwater flow laterally in saturated soils; From each land segment only outgoing lateral groundwater flows are estimated except when lateral inflows o ccur through external boundaries; Lateral groundwater flows only happen when a positive hydraulic gradient occurs between one land segment and its neighboring land segments; Potential groundwater flows are estimated for all neighboring land segments. If there is insufficient storage to supply all potential groundwater flows, actual groundwater flows are decided th rough storage apportionment. Overland Flow Appendix B shows a flow chart of overla nd flow simulation for one source land segment. The potential outgoing overland flows from the source land segment to adjacent land segments are calculated as discussed in the following section and inflows through the external do main boundary are input or extra polated depending on the actual boundary conditions. Subsequently, a storage ch eck is conducted to guarantee there is enough surface water storage in th e source land segment to supply the potential flows. If there is insufficient storage the available water is apportioned according to assumptions discussed in the following section and the detailed algorithm is documented in Appendix

PAGE 67

46 G. External boundary inflows are introduced into boundary land segments each day. The outgoing flows are then transferred to adjacen t land segments, which causes a decrease of water storage in the source land segment and an increase of water storage in the adjacent land segment. The whole process described in Figure 3-3 is applied to each land segment in the simulation domain following the simulation sequence generated by the model. Overland flow calculation The generalized spatial relationship betw een one land segment and its adjacent land segments is shown in Figure 3-3. Two t ypes of overland flow over internal boundaries are defined in this model: Figure 3-3. Configuration of overland flows between source land segment and adjacent land segments. The digit (1~4) of ad jacent land segment index indicates the priority to receive the overland flow from source land segment, which is determined by the water elevation of that land segment and smaller water elevations correspond to higher prioriti es. An adjacent land segment with a larger index has higher priority to receive the overland flow. Unstructured boundaries, where no local obstacles ex ist to the flow (no singular head losses) and a mean resistance coefficient for a given cross secti on of the flow can be used. In this case, Ma nning’s formula is used: Neighbor 4 Neighbor 3 Neighbor 2 Qs,d,3 Qs,d,2 Qs,d,1Qs,d,4 Neighbor 1 Source

PAGE 68

47 i d s m 2 1 i d s 3 2 i d s i d s i d s i d sn U S R D W Q (3-7) where subscript s and d indicate source and de stination land segments and i the specific adjacent neighbor; Qs,d,i is the volumetric discharge from source to destination land segments [L3/T]; Ws,d,i is the width of the cross-se ctional area between source and destination land segments [L]; Ds,d,i is the water depth of the cross-sectional area between source and destination land segments [L], an average of water depths in both source and destination land segments; Rs,d,i is the hydraulic radius of the flow section between source and destination land segments and it is essentially the average surface water depths [L], roughly equivalent to Ds,d,i for wide, shallow cross-sectional area or streams; Ss,d,i is the water surface slope between sour ce and destination land segments [L/L], it is estimated as L z z Sd s i d s (3-4) in which, zs and zd are the water levels of sour ce and destination land segments, respectively [L]; L is the distance between the centers of source and destination land segments [L]; Um is the unit conversion for the Manning formula [L1/3/T]; ns,d,i is the average of Manning’s roughness coefficients of source and destination land segments, which is modified to be a function of sedime nt type and the interaction of the vegetation height/density and water depth (Fitz et al 1996). This functio n creates a dynamic feedback loop between the physical process of flow and the biological process of plant growth (Sklar et al. 2001) which is calculated as:

PAGE 69

48 Figure 3-4. The relationship between Ma nning's roughness-coefficient (n) to water depth and to plant height, both of wh ich change dynamically in the model (Fitz et al., 1996). ) 1 2 )( n n ( n n) H D 1 ( min max max i d sm i d (3-5) where nmax is the maximum Manning’s roughness coeffi cient associated with the dynamic vegetation density in the land segment [L1/3/T]; nmin is the minimum Manning’s roughness coefficient for a vegetation-free land segment [L1/3/T]; Dd is the surface water depth [L] and Hm is the dynamic height of the macr ophytes in the land segment [L] (Fitz et al., 1996). Equation (3-5) returns a positive roughness coe fficient that ranges from a vegetation-free minimum to a maximum at the point of full plant im mersion (Petryk et al., 1975). As water depth increases over th at of the macrophyte height, the roughness decreases to an asymptote at the baseline sediment roughne ss (Nalluri and Judy, 1989). Figure 3-4 shows the relationship between Ma nning's roughness-coefficient, n, to water depth and to plant height. Structured boundaries, where berms or dikes form a boundary that can be represented by a singular head loss between two land segments. In this case, two kinds of

PAGE 70

49 classical discharge formulae for broad creste d weirs (Cunge 1975) ar e applied according to the following criteria (Cunge 1975): Figure 3-5. Structured boundary (Cunge, 1975) Free flow condition (Free flow weir) w d w d 1 i d sz z ) z z ( g 2 b Q ) z z ( 3 2 z zw d w s (3-6) Submerged flow condition (Drowned weir) s d w s 2 i d sz z ) z z ( g 2 b Q ) z z ( 3 2 z zw d w s (3-7) where Qs,d,i is the volumetric discharge between source and destination land segments over a weir [L3/T]; 1 and 2 are the discharge coefficients; b is the effective width of the weir [L]; g is the gr avitational acceleration [L2/T]; zs and zd are the water levels of source and destination land segments, respectively [L]; zw is the elevation of the weir crest [L]. Figure 3-5 shows the relationship between sour ce and destination land segments when a weir exists in between. Surface water storage apportionment Potential overland flows from one land segment to all its adjacent land segments are calculated separately wit hout taking into account the storage change due to the fluxes that have been already computed. Hence, it is possible that the sum of the potential flows

PAGE 71

50 could exceed surface water storage in the land segm ent. This situation is especially likely to occur if the time step is la rge. To avoid this situation, a procedure was developed to limit the actual overland flow to its maximu m available volume of storage, hereafter simply called the available volume, which is a portion of the current volumetric water storage on ground surface allocated to a specific neighbor based on the following assumptions: The storage apportionment is allocated onl y to those neighbors that receive the overland flows; The current water storage in the land se gment is the maximum available volumetric water storage; The allocation of the maximum water storage in the land segment is made according to the priority of its neighbors; The priority of the neighbors is set up according to the elevation of the water surface boundary between the land segment and its specific neighbor. The water surface elevation of the boundary between the land segment and its neighbor is determined by the maximum of the water level or the elevation of the weir crest, whichever exists; If a weir exists in between the land segm ent and its neighbor, the elevation of the weir crest has to be higher than the elevation of the land segment; The neighbor with the lowest elevation of boundary has the highest priority for allocation of water storage from the source land segment; A neighbor with lower priority is not allo cated any water storage until the neighbor with next higher priority is filled to an equavlant water surface elevation, i.e., the two neighbours have equal water surface elevation; The available volume for a specific neighbor can not be more than required to equilibrate the water levels of the land segment and this neighbor. Once the available volumes are determin ed for all those neighbors receiving potential overland flows from the land segmen t, each of them is compared with its corresponding pre-estimated pot ential overland flow. If any potential overland flow exceeds its available volume, then the actual ove rland flow is set equal to the available

PAGE 72

51 volume. Otherwise, the actual ove rland flow is equal to its poten tial value. Details of the assumptions and algorithm for surface water a ppointment are described in Appendix G. Groundwater Flow Groundwater flow here refers to latera l groundwater flow which occurs in the saturated soils. Appendix C shows a flow ch art of lateral groundwater flow simulation for one land segment. For a source land segm ent, the potential outgoing lumped lateral groundwater flows and inflows from external domain boundary are calculated, followed by a storage check to guarantee there is su fficient groundwater stor age to provide the potential flows from the source land segment. Lumped groundwater flow available for transfer is split into layer flows for the saturated zone in proportion to the layer transmissivity. The layer flows are then tr ansferred to the neighbor ing land segment. The transmission of layer groundwater flow in creases the void space in the soil of the source land segment and decreases that in the neighboring land segments. At the end of

PAGE 73

52 Figure 3-6. Schematic representation of th e lateral groundwater flow from the higher head hs to the lower head hd. (A) and (B) represent th e situations without and with overland flow, respectively. hs hd Lateral groundwater flow Water Table Ground surface (A) Impermeable strata hs hd Lateral groundwater flow Water Table Ground surface Overland flow (B) Impermeable strata

PAGE 74

53 Figure 3-7. Diagram of lateral groundwater flow between two land segments. (A) and (B) represent the generalized situations without a nd with overland flow, respectively. Ws Qs,d,i Qs,i,1 Qs,i,2 Qs,i,3 Qs,i,4 Qs,i,5 (A) bdk L Hs zs Hdzd(B) bsj Qs,i,1Qs,i,2 Qs,i,3Qs,i,4 Qs,i,5 Ws Qs,d,i bdk L HdzdHs zs bsj

PAGE 75

54 simulation, the water table depths are upda ted for all land segments. The complete process described in Appendix C is applie d to each land segment in the simulation domain following the simulation se quence generated by the model. Lateral groundwater flow calculation Lateral groundwater flow is calculated first in a lumped manner between the saturated zone of source land segment and its adjacent land segments according to Darcy’s law. The lumped flow is then partitioned into the saturated soil layers. Figure 36 shows the generalization of lateral flows between two adja cent land segments, in case A (Figure 3-6A) the water table is below the ground surface and case B it is above the ground surface (Figure 3-6B). The generalization of (A) and (B) in Figur e 3-6 is correspondi ngly illustrated in Figure 3-7(A) and (B). First of all, using Darcy’s Law, the lumped lateral groundwater flow from the source land segment to th e destination land segment is estimated i d s i d s i d s i d s i d sH W S K Q (3-8) Where subscript s and d indicate source and destination land segments and i the specific neighboring land segment; Qs,d,i is the lumped lateral groundwater flow from source to destination land segments [L3/T]; Ss,d,i is the hydraulic gradient between source and destination land segments [L /L] and is calculated as L Z Z Sd s d s (3-9) in which zs and zd are the groundwater heads of the s ource and destinatio n land segments, respectively [L]; and L is the distance between the centroids of the source and destination land segments [L]; Ks,d,i is the effective horizontal saturated hydraulic conductivity [L/T],

PAGE 76

55 an average of horizontal saturated hydraulic conductivities of the s ource and destination land segments [L/T]. It is estimated as 2 / ) b K D 1 b K D 1 ( Kk dk dk d j sj sj s i d s (3-10) where subscript j and k indicate the number of soil layers below the water table of the source and destination land segments, respectively; Ksj is the saturated horizontal hydraulic conductivity of soil layer j of source land segment [L/T] and Kdk, that of soil layer k of destination land segment [L/T]; bsj and bdk is the thickness of saturated soil of layer j in source land segment and layer k in destination land segment [L]; Dsj and Ddk is the thickness of saturated soils for sour ce and destination and segments [L]; Ws,d,i is the boundary width of the cross-sectional area be tween source and destin ation land segments, perpendicular to the direction of lateral groundwater flow [L]; Hu,d,i is the average of the thickness of saturated soil of source and destination land segments [L], 2 H H Hd s i d s (3-11) where Hs and Hd are the thickness of saturated soil s of the source and destination land segments, respectively [L]. Next, the lateral groundwater flows from s ource to destination saturated soils are distributed to individual satura ted soil layers by the formula N 1 j j i j i j i j i i d s j i sb K b K Q Q (j = 1,…., N) (3-12) where Qs,i,j is the layer lateral flow from the saturated soil layer j of source land segment to the layer j of destination land segment [L3/T]; Ki,j is the saturated horizontal hydraulic conductivity of the la yer j of source land segment [L/T]; bi,j is the thickness of saturated

PAGE 77

56 soil of the layer j of source land segment [L]; N indicates the number of saturated soil layers. In case the water ta ble is located somewhere within a soil layer, the thickness of saturated soil in this layer is the depth from the water table to the bottom of this soil layer. Groundwater storage apportionment Similar to the procedure of surface wate r storage apportionment for overland flow, a procedure to check if the sum of the lumped groundwater flows exceeds the available saturated groundwater storage in the source land segment is needed. The available groundwater storage is assumed to be the portion of volumetric water above the field capacity. If the sum of outgoing lumped flow s is greater than the available groundwater storage, each lumped groundwater flow is adju sted using a ratio, wh ich is the difference between the sum of outgoing lumped flows a nd the available volumetric storage to the available volumetric storage. By doing th is adjustment, the actual lumped flows are limited to the available storag e. Since the groundwater flow is much slower than overland flow this procedure is rarely invoked when the model runs at a daily time step. Canal Flow The interaction of canals with the si mulation domain can occur both through overland flow and lateral groundwater flow de pending on the stage of canal and the water level in the domain. When the canal stage is higher than the water level in an adjacent land segment, back flow from the canal into the domain occurs. When the canal stage is lower than the water level in an adjacent land segment, discharge from the domain is released into the canal. Flow between th e canal and domain is estimated using the following equations:

PAGE 78

57 n S WR q2 / 1 3 5 Overland flow (3-13) KWSD q Lateral groundwater flow (3-14) where q represents the amount of overland flow to/from the canal in Equation (3-13) and lateral groundwater flow to/from the canal in Equation (3-14) [L3/T]; W is the width of the land segment-canal boundary [L]; R is th e water depth above the ground surface [L]; S is the water slope that is determined by th e difference of water level between canal and the adjacent land segment divided by the dist ance between the centroi d of the canal and the centroid of the land segment [-]; n is th e Manning’s roughness coefficient, which is assumed to be the same as the one for the adjacent land segment for this study [L1/3/T]; K is the hydraulic conductivity between the canal and the satu rated soils of the adjacent land segment, which is assumed to be the same as the saturate d horizontal hydraulic conductivity for the adjacent la nd segment for this study; an d D is the thickness of the saturated soil profile [L]. Currently canal fl ow is not an individual process in the model but is a built-in routine inside of overland fl ow and lateral groundwater flow processes. Initial and Boundary Conditions To ensure the accuracy of the solution of the simulation model, both initial and boundary conditions must be prescribed for the domain being simulated. Initial conditions in the model include water table de pth in the soil, ponded water depth if any, and depressional water storage for each la nd segment. Boundary conditions include boundary inflow, outflow, or fixed head c onditions and the description of physical hydraulic structures, levees, fences, and berm s, etc. If physical structures exist, geometric information and relevant empirical formula for flows through these structures must be prescribed. Time series of atmospheric variables such as rainfall, temperature,

PAGE 79

58 and potential ET, etc. must also be specified for the desired simulation period. Appendix F shows the input files, required for applicati ons in this dissertation, which contain the initial and boundary conditions. Model Testing and Validation Simulation Sequence and Model Performance Accuracy Analysis In the modified ACRU2000 hydrologic mode l, the simulation sequence determines the execution order of land segments. The simulation sequence is generated according to the spatial water flow conf iguration and the sequence of land segment input files specified in the control menu file (Control.me n in Appendix F). The spatial water flow configuration is arranged accord ing to the topographical elev ation of each land segment, which specifies the directions of outflow from each land segment and is deterministic once the topography in a simulation domain is specified. However, the sequence of land segment input files in Control.men is arbitrar y, for instance, one can specify the sequence from the first land segment to the last one, or its opposite, or othe r orders. Different simulation sequences can produce different si mulation results but the extent to which simulated results, resulting from different si mulation sequences, differ from one another needs to be investigated. With this consid eration, influence of simulation sequence on model predictions was analyzed. Moreover, the model accuracy is the most important indicator of whether a model is successfully developed. Simulation sequences directly affect the accuracy of the model. Theref ore it is necessary to evaluate the model performance accuracy in terms of appropr iately selected simulation sequence by comparing the modified model with well-known models. To investigate the influence of simulati on sequence, it is important to understand why the simulated results can be different when different simulation sequences are

PAGE 80

59 specified. As discussed in the introduction section of this Chapter, the vertical and horizontal hydrologic processe s are executed in two separa te loops in the modified ACRU2000 model. During the ver tical processes water is dist ributed vertically in the soil profile, which results in variation of hydraulic heads between land segments, but each land segment is simulated independently. Horizontal processes transfer surface and subsurface water between nei ghboring land segments making them interdependent. The simulation of horizontal proce sses on a particular land segm ent causes a change of water levels in adjacent land segments. These influences can be propagated to downstream land segments that are simulated later in the simulation sequence. Therefore, it is essential to investigate the difference between different simulation sequence strategies before a decision is made on what would be th e most appropriate seque nce for the model. Table 3-1. List of strategies for water transmission and water storage updating for simulation sequence experiments. Experiments Strategies to transfer water and update water storage* Experiment 1 Water transferred in lateral processes is not available for transmission to downstream land segments and also does not cause head changes until the end of the day when all land segments have been simulated. Experiment 2 Water transferred in lateral processes is available for transmission to downstream land segments but does not cause head changes until the end of the day when all land segments have been simulated. Experiment 3 Water transferred in lateral processes is available for transmission to downstream land segments and also causes heads changes immediately. Water storage either on the ground su rface or in the saturated soil layers. Three experiments, corresponding to different strategies for water transmission and storage updating, are listed in Table 3-1. These experiments were conducted based on two hypothetical scenarios, a fl at rectangular plane and an axisymmetric domain. To further evaluate the model performance accura cy, the simulated results from the first scenario were compared with the two-dime nsional overland flow model in MIKE SHE

PAGE 81

60 (DHI, 2004) and the results from the second scenario were compar ed with the threedimensional groundwater flow models in both MIKE SHE (DHI 2004) and MODFLOW (McDonald and Harbaugh, 1996). Case 1: Overland flow along a flat rectangular plane A flat rectangular plane with a size of 100 m 2000 m was used to test the influence of simulation sequence on ove rland flow in the modified ACRU2000 hydrologic model. The plane wa s divided into 20 land segmen ts, each with a size of 100 m 100 m as shown in Figure 3-8. An init ial head of 0.5 m was assigned for land segment 1 (hereafter LS1), 0.4 m for LS2, 0.3 m for LS3, 0.2 m for LS4, 0.1 m for LS5, and 0.0 m for the rest of land segments so th at all with a mild wa ter slope (= 0.001 m/m) was created from LS1 through LS5 to drive wate r propagation from the le ft to the right of the plane. A closed boundary condition was assu med and a 30-day period was simulated. The modified ACRU2000 model was first run for the three experiments (Table 3-1) from LS1 to LS20, and then the model was run again for the same experiments but with an opposite sequence, from LS20 to LS1. Figure 3-8. A rectangular plan e with 20 land segments (arrow indicates water movement direction and digits assigned in each grid cell indicate the number for each land segment). The simulated surface water depths from the three experiments each using the two simulation sequences are compared in Figure 39. For Experiment 1, the model produced identical results for both simulation sequences This is because the hydraulic gradients over the domain do not change during one tim e step for this experiment and water 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

PAGE 82

61 transmission and water storage updates do not happen until the end of each time step when all land segments have been simulated. Significant difference in surface water depths between simulation sequences is observe d in Experiment 2. Water moves faster when the simulation order is consistent with the flow direction (from LS1 to LS20) and it moves slower when the simulation order is against the flow direction (from LS20 to LS1). In this experiment the water gradie nt remains constant during one time step, however the water storages are updated within the time step which causes more water to move downstream during the time step if it is available. Differences in simulated water depths between simulation sequences also o ccur in Experiment 3 but they are not as significant as those in Experiment 2. When the water slopes within the domain get smaller (i.e. on days 20 and 30), the diffe rences between both simulation sequences diminish. These comparisons show that th e simulation order does cause a difference in surface water depths but the magnitude of the difference depends on the gradient and storage update strategies. Figure 3-10 compares the simulated surf ace water depths along the domain from the three experiments with those from MIKE SHE’s overland flow model for the simulation sequence from LS1 to LS20. On Day 2, there is a good agreement between these two models for all thr ee experiments. On day 10 and 20, the modified ACRU2000 hydrologic model moves water faster than MIKE SHE’s overland flow model for all three experiments, although the models agree with each other fairly well on timing and trends. Statistics calculated by using the predictions from MIKE SHE’s model as observed values, and the predictions fr om the modified ACRU2000 model as the simulated values quantify the differences. As shown in Table 3-2, al l statistics including

PAGE 83

62 bias, RE, RMSE, CV, R2 and NS are very close to one another among the three experiments, especially those for bias and RE For all selected days, Experiment 1 is better than Experiments 2 and 3 for the statistics RMSE, CV, R2 and NS although the differences are not that significant. These graphic and statistical comparisons indicate that for all three experiments the ACRU 2000 model compares reasonably well with MIKE SHE even though where significantly different methodologies and time steps are used. Table 3-2. Statistics for the surface wate r depths predicted by the modified ACRU2000 model and MIKE SHE for the three experiments with the same simulation sequence. Statistics* bias RE RMSE CV R2 NS (units) (m) ( ) (m) ( ) ( ) ( ) Experiment 1 0.0000 0.0000 0.0057 0.0171 0.9984 0.9984 Experiment 2 0.0000 0.0001 0.0148 0.0443 0.9899 0.9893 Day 2 Experiment 3 0.0000 0.0001 0.0087 0.0261 0.9976 0.9963 Experiment 1 0.0000 0.0001 0.0133 0.0396 0.9935 0.9874 Experiment 2 0.0000 0.0001 0.0283 0.0845 0.9610 0.9425 Day 10 Experiment 3 0.0000 0.0000 0.0372 0.1108 0.9604 0.9011 Experiment 1 0.0000 -0.0001 0.0239 0.0714 0.9749 0.9490 Experiment 2 0.0000 0.0000 0.0402 0.1197 0.9399 0.8564 Day 20 Experiment 3 0.0000 -0.0001 0.0495 0.1476 0.9182 0.7818 Statistics were calculated by assuming the predictions from MIKE SHE as the observed values while the predictions from the modified ACRU2000 as the predicted values. All three experiments were conducted with the same simulation sequ ence from LS1 to LS20. However, this rectangular plane domain is quite special in that each land segment only has water exchange with either upst ream or downstream land segments and the simulation sequence either is consistent with or against the flow direction. To further investigate the influence of simulation seque nces, a more complicated domain was used in the following experiment.

PAGE 84

63 Comparison of Surface Water Depths over Land Segments for Experiment 1 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) Day2 LS1-LS20 Day10 LS1-LS20 Day20 LS1-LS20 Day30 LS1-LS20 Day2 LS20-LS1 Day10 LS20-LS1 Day20 LS20-LS1 Day30 LS20-LS1 Comparison of Surface Water Depths over Land Segments for Experiment 2 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) Day2 LS1-LS20 Day10 LS1-LS20 Day20 LS1-LS20 Day30 LS1-LS20 Day2 LS20-LS1 Day10 LS20-LS1 Day20 LS20-LS1 Day30 LS20-LS1 Comparison of Surface Water Depths over Land Segments for Experiment 3 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) Day2 LS1-LS20 Day10 LS1-LS20 Day20 LS1-LS20 Day30 LS1-LS20 Day2 LS20-LS1 Day10 LS20-LS1 Day20 LS20-LS1 Day30 LS20-LS1 Figure 3-9. Comparisons of the simulated surface water depth from the modified ACRU2000 model on the thre e experiments between two opposite simulation sequences.

PAGE 85

64 Comparison of Surface Water Depths over Land Segments on Day 2 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) MIKE SHE Experiment 1 ACRU2000 Experiment 2 ACRU2000 Experiment 3 ACRU2000 Comparison of Surface Water Depths over Land Segments on Day 10 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) MIKE SHE Experiment 1 ACRU2000 Experiment 2 ACRU2000 Experiment 3 ACRU2000 Comparison of Surface Water Depths over Land Segments on Day 20 0.00 0.09 0.18 0.27 0.36 0.45 LS1LS3LS5LS7LS9LS11LS13LS15LS17LS19 Land SegmentWater Depth (m) MIKE SHE Experiment 1 ACRU2000 Experiment 2 ACRU2000 Experiment 3 ACRU2000 Figure 3-10. Comparisons of the simulate d surface water depths between the MIKE SHE’s overland flow model and the modified ACRU2000 hydrologic model on the three experiments with the simu lation sequence from LS1 to LS20.

PAGE 86

65 Case 2: Overland and groundwater flow over an axisymmetric domain An axisymmetric domain with its 2D and 3D gridded discretization as shown in Figure 3-11 was used to test the influence of simulati on sequence on both overland and groundwater flows in the modified ACRU2000 m odel. The domain, with a size of 600 m 600 m, is composed of 100 meter square grid cells, each of which is 2 m in depth. Four equal-thickness computational soil layers with homogenous soil properties for each soil layer were assumed for each land segment. The surface elevation for each land segment was set to 10 m. Homogeneous hydrologic a nd physical characteristics were assumed for each land segment including the Manni ng’s roughness coefficient (= 0.1 m1/3/s), soil layer thickness (= 0.5 m), horizontal/ver tical hydraulic conduc tivity (= 0.0000444 m/s), specific yield (= 0.25), and st orage coefficient (= 0.00005 m-1). An initial water depth of 0.5 m was assigned to the central four la nd segments, 0.25 m to the land segments surrounding the central ones, and 0.0 m to the re st of land segments for the overland flow simulation in this scenario. For the groundwat er flow simulations, the previous setting for initial water levels in each land segment was lowered by a depth of 1.5 m so that no surface water movement would occur. A cl osed boundary condition was assumed. A 30day simulation period was used for the overla nd flow simulation and a 2-year simulation period for the groundwater flow simulation. The objectives for this case study were to investigate the difference in simulated results for the three experiments (Table 3-1) in a more complicated domain where land segments may interact spatially with one another in multiple directions, and compare the simulated surface water depths with those from MIKE SHE (DHI, 2004) and groundwater table depths with those fr om MIKE SHE (DHI 2004) and MODFLOW (McDonald and Harbaugh, 1988). A simulation sequence, which transfers water from the

PAGE 87

66 center to boundary of the domain symmetri cally following the flow direction, was specified. An hourly time step was used fo r the 2D overland flow model in MIKE SHE, a varying time step ranging from 2 to 12 hours for the 3D groundwater flow model in MIKE SHE, and a daily time step for th e groundwater flow model in MODFLOW and the modified ACRU2000 model. The model wa s run separately with different initial water level settings so that both surface overland flow and groundw ater flow can be investigated individually. Figure 3-11. Schematics of two-dimensional ax isymmetric domain (left) and its threedimensional discretization (right). The digits assigned in the left diagram indicate the number for each land segment. For the overland flow simulation, the si mulated surface water depths from LS19, LS20, LS21, LS25, LS26, and LS31 were selected for analysis because of the axisymmetric distribution of water depths over the domain. For each selected land segment, the simulated water depths from the three experiments were compared with those from MIKE SHE’s overla nd flow model as shown in Figure 3-12. Similar to the conclusions drawn from Case 1, the compar isons from all three experiments in all selected land segments indicate that wate r moves slightly faster in the modified ACRU2000 model than in MIKE SHE. The results from Experiments 1 and 2 show y z x x y

PAGE 88

67 Surface Water Depths over Time at Land Segment 19 0.00 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Surface Water Depth over Time at Land Segment 20 0.00 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Surface Water Depths over Time at Land Segment 21 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Figure 3-12. Comparisons of simulated wate r depths of the selected land segments between the modified ACRU2000 model and the MIKE SHE’s overland flow model for the three experiments.

PAGE 89

68 Surface Water Depths over Time at Land Segment 25 0.00 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Surface Water Depths over Time at Land Segment 26 0.00 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Surface Water Depths over Time at Land Segment 31 0.00 0.10 0.20 0.30 0.40 0.50 1357911131517192123252729 Time (day)Water Depth (m) MIKE SHE Experiment 1 Experiment 2 Experiment 3 Figure 3-12. Continued.

PAGE 90

69 Table 3-3. Statistics for the surface wate r depths predicted by the modified ACRU2000 model and MIKE SHE for the three experiments. Statistics* bias RE RMSE CV R2 NS (units) (m) ( ) (m) ( ) ( ) ( ) Experiment 1 -0.0015 -0.0111 0.0078 0.0589 0.9139 0.9047 Experiment 2 0.0000 0.0002 0.0121 0.0909 0.8028 0.7726 LS19 Experiment 3 0.0011 0.0082 0.0068 0.0512 0.9315 0.9278 Experiment 1 -0.0138 -0.0861 0.0175 0.1091 0.8224 0.3598 Experiment 2 -0.0164 -0.1022 0.0205 0.1281 0.7208 0.1187 LS20 Experiment 3 -0.0130 -0.0814 0.0161 0.1008 0.8827 0.4535 Experiment 1 -0.0225 -0.1207 0.0292 0.1563 0.9524 0.8059 Experiment 2 -0.0253 -0.1353 0.0322 0.1725 0.9393 0.7637 LS21 Experiment 3 -0.0265 -0.1419 0.0335 0.1794 0.9319 0.7445 Experiment 1 0.0128 0.1092 0.0214 0.1832 0.7289 0.3180 Experiment 2 0.0141 0.1205 0.0232 0.1986 0.6855 0.1983 LS25 Experiment 3 0.0141 0.1204 0.0159 0.1358 0.9322 0.6249 Experiment 1 0.0015 0.0105 0.0221 0.1546 0.4450 -0.1851 Experiment 2 0.0015 0.0102 0.0210 0.1470 0.4703 -0.0715 LS26 Experiment 3 -0.0028 -0.0193 0.0092 0.0642 0.8471 0.7955 Experiment 1 0.0256 0.2555 0.0328 0.3271 0.7377 -0.1002 Experiment 2 0.0276 0.2759 0.0355 0.3542 0.6569 -0.2900 LS31 Experiment 3 0.0290 0.2895 0.0331 0.3307 0.7684 -0.1246 Statistics were calculated by assuming the predictions from MIKE SHE as the observed values while the predictions from the modified ACRU2000 model as the simulated values. All three experiments were performed with the same simulation sequence. fluctuations at early times of simulation but eventually trend to the results from Experiment 3 in which the curve of predicte d water depths appears smooth. However, the differences among the three experiments seem to be not that significant. Both models agree with each other fairly well on timing and trends. Table 3-3 shows the statistics that were calculated by using the predictions from MIKE SHE as the “observed” values. The st atistics indicate that the model generally produced better predictions in Experiment 3 than in the other two experiments with smaller RMSE and CV values and larger R2 and NS values in four selected land segments (LS19, LS20, LS25 and LS26) and similar pred ictions in LS21 and LS31 with the other

PAGE 91

70 two Experiments. It should be noted that s lightly larger values of bias and RE are observed in Experiment 3 in some of these la nd segments. Overall, it was concluded that the modified ACRU2000 model performed better in Experiment 3 for this scenario and thus Experiment 3 is the most appropriate strategy for producing reas onable surface water depths for this axisymmetric domain. To further visualize the comparison of simulated results for Experiment 3, Figure 3-13 displa ys a time series comparison of surface water depths for all selected land segments for these two models. 0.0 0.1 0.2 0.3 0.4 0.5 14710131619222528 Time (day)Surface Water Depth (m) LS19 MIKE SHE LS20 MIKE SHE LS21 MIKE SHE LS25 MIKE SHE LS26 MIKE SHE LS31 MIKE SHE LS19 ACRU2000 LS20 ACRU2000 LS21 ACRU2000 LS25 ACRU2000 LS26 ACRU2000 LS31 ACRU2000 Figure 3-13. Comparisons of the simulated wa ter depths of the selected land segments between the modified ACRU2000 and MI KE SHE’s overland flow model for Experiment 3.

PAGE 92

71 A groundwater simulation test was made by changing initial water levels as discussed earlier. Simila r to the above overland flow simulation, the simulated groundwater table depths also display an axisymmetric distribu tion pattern over the domain. Therefore the simulated water ta ble depths from LS19, LS20, LS21, LS25, LS26, and LS31 were used to compare with the MIKE SHE’s predictions for all three experiments as shown in Figure 3-14. From th is figure, it is fairly difficult to tell the difference among the simulated results from the three experiments for they all have very similar predictions for all selected land segments. Interestingly, the ACRU results for the three experiments are consistently located in between the predictions from MIKE SHE and MODFLOW for all selected land segmen ts, although these results agree with MIKE SHE’s predictions better in LS19, LS20, LS25, LS26, and LS31 and with MODFLOW’s predictions better in LS21. The comp arisons also reveal that the ACRU2000 groundwater movement is slightly faster th an MIKE SHE and a li ttle bit slower than MODFLOW. The difference in predicted groundwater table depths between the modified ACRU200 model and th e other two models seems la rger in the central land segments (LS20, LS21, and LS26) than it appears in the land segments along the boundary (LS19, LS25, and LS31). This could result from the larger hydraulic gradient variations between the central land segments and milder gradient s between the boundary land segments. In general, three models agree with one another on timing and tendency quite well. Table 3-4 shows the statistics that were calculated by using the predictions from MIKE SHE and MODFLOW as the “observed” values for the three experiments for all selected land segments. Over all select ed land segments, there are no significant

PAGE 93

72 differences among the bias, RE, RMSE, CV, and R2 values. However, some systematic differences in the NS values are noticed between the predictions from MIKE SHE and MODFLOW as the “observed” values: excellent NS values close to 1 for LS19, LS25, and LS31, fairly good NS values for LS21 a nd LS26 and poor NS values were obtained for LS20 when comparing to MIKE SHE’s predic tions. On the other hand, the NS values vary gradually for all selected land segm ent from excellent to fairly good when comparing to MODFLOW’s predictions. Thus it is difficult to differentiate the model performance among the three experiments and thus conclude unequivocally which experiment provides the most appropriate strategy. From the previous case studies for two di fferent hypothetical scenarios, it can be concluded that the simulation sequence aff ects the simulation results, but not very significantly. Thus picking a simulation se quence that follows the flow directions determined by the topography on the ground surface is reasonable. Moreover, among the three alternative water transmission and wate r storage updating strate gies, Experiment 3 appears to be the most appropriate one for reasonable simulation of overland flow but no significant difference was observed in simulatio n of groundwater flow. By comparing to results from MIKE SHE a nd MODFLOW, the modified ACRU2000 model has been shown to simulate hydrology reasonably well.

PAGE 94

73 Groundwater Table Depths over Time at Land Segment 19 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Groundwater Table Depths over Time at Land Segment 20 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Groundwater Table Depths over Time at Land Segment 21 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Figure 3-14. Comparisons of the groundwater ta ble depths at the selected land segments between and the modified ACRU200 0 hydrologic model and MIKE SHE’s groundwater flow model for the three experiments.

PAGE 95

74 Groundwater Table Depths over Time at Land Segment 25 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Groundwater Table Depths over Time at Land Segment 26 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Groundwater Table Depths over Time at Land Segment 31 1.00 1.10 1.20 1.30 1.40 1.50 160119178237296355414473532591650709 Time (day)Water Table Depth (m) MIKE SHE MODFLOW Experiment 1 Experiment 2 Experiment 3 Figure 3-14. Continued.

PAGE 96

75Table 3-4. Statistics for the simulated groundwater table depths by the modified AC RU2000 model, MIKE SHE, and MODFLOW for the three experiments with the same simulation sequence. Bias RE RMSE CV R2 NS Statistics [1]a [2]b [1] [2] [1] [2] [1] [2] [2] [1] [1] [2] (units) (m) (m) ( ) ( ) (m) (m) ( ) ( ) ( ) ( ) ( ) ( ) Experiment 1 -0.0010 0.0118 -0.0007 0.0084 0.0019 0.0123 0.0014 0.0020 0.9982 0.9932 0.9968 0.8774 Experiment 2 -0.0012 0.0116 -0.0009 0.0082 0.0021 0.0120 0.0015 0.0019 0.9982 0.9932 0.9963 0.8825 LS19 Experiment 3 -0.0012 0.0116 -0.0008 0.0082 0.0021 0.0121 0.0014 0.0020 0.9982 0.9932 0.9964 0.8808 Experiment 1 0.0113 -0.0163 0.0090 -0.0127 0.0140 0.0173 0.0112 0.0031 0.9979 0.9774 -1.3373 0.3915 Experiment 2 0.0111 -0.0166 0.0088 -0.0129 0.0139 0.0175 0.0110 0.0031 0.9979 0.9778 -1.2802 0.3763 LS20 Experiment 3 0.0115 -0.0162 0.0091 -0.0126 0.0142 0.0171 0.0113 0.0031 0.9977 0.9782 -1.3765 0.4006 Experiment 1 0.0289 -0.0181 0.0258 -0.0155 0.0323 0.0205 0.0289 0.0040 0.9988 0.9998 0.5957 0.9247 Experiment 2 0.0288 -0.0182 0.0257 -0.0156 0.0322 0.0206 0.0287 0.0041 0.9987 0.9998 0.5990 0.9237 LS21 Experiment 3 0.0293 -0.0177 0.0261 -0.0152 0.0326 0.0203 0.0291 0.0040 0.9987 0.9998 0.5890 0.9265 Experiment 1 -0.0004 0.0101 -0.0003 0.0070 0.0010 0.0104 0.0007 0.0017 0.9992 0.9987 0.9984 0.8414 Experiment 2 -0.0007 0.0098 -0.0005 0.0068 0.0011 0.0102 0.0008 0.0016 0.9992 0.9987 0.9980 0.8482 LS25 Experiment 3 -0.0006 0.0099 -0.0004 0.0069 0.0011 0.0103 0.0007 0.0016 0.9993 0.9987 0.9981 0.8456 Experiment 1 0.0104 -0.0155 0.0080 -0.0116 0.0113 0.0160 0.0086 0.0027 0.9986 0.9808 0.7156 0.6420 Experiment 2 0.0102 -0.0157 0.0078 -0.0117 0.0111 0.0162 0.0085 0.0028 0.9986 0.9806 0.7255 0.6328 LS26 Experiment 3 0.0106 -0.0153 0.0081 -0.0115 0.0114 0.0158 0.0087 0.0027 0.9986 0.9805 0.7110 0.6473 Experiment 1 -0.0015 0.0130 -0.0010 0.0089 0.0017 0.0150 0.0012 0.0023 0.9996 0.9937 0.9840 0.5285 Experiment 2 -0.0016 0.0129 -0.0011 0.0088 0.0019 0.0149 0.0013 0.0023 0.9995 0.9940 0.9810 0.5382 LS31 Experiment 3 -0.0016 0.0129 -0.0011 0.0088 0.0019 0.0149 0.0013 0.0023 0.9995 0.9940 0.9810 0.5383 a statistics were calculated using the predictions by MIKE SHE as th e observed values and the predic tions by the modified model a s the simulated values. b statistics were calculate using the predicti ons by MODFLOW as the observed values and the predictions by the modified model as the simulated values.

PAGE 97

76 Application at Dry Lake Dairy #1, Kissimmee River Basin, Florida Site description The Dry Lake Dairy #1 is an agricultural pasture site located within the lower Kissimmee River and Taylor Cr eek-Nubbin Slough Basins (Figur e 3-15). The site is located on Immokalee soil with surface depressions and little slope, thus the pasture often becomes flooded during the rainy season. Th e spodic horizon for Immokalee fine sand typically occurs at approximately 90 cm a nd averages 50 cm in thickness. The upper zone of the spodic horizon is black and w eakly-cemented, while the lower zone is a mottled, dark reddish brown and is even mo re weakly cemented than the upper layer. During the wet season, the water table stands near or at th e surface for short periods and recedes to below 120 cm during the dry season (Capece, 1994). Figure 3-15. Location of Dry Lake Dairy #1 site (Capece, 1994).

PAGE 98

77 Figure 3-16. Dry Lake Dairy # 1 site map, topographic survey and location of well stations and tracer app lication compound. Distance scales are in feet and elevation contours are in feet a bove mean sea level (Capece, 1994). A topographic map of the Dry Lake Dairy #1 site is s hown in Figure 3-16. As observed, a low, broad berm isolated the st udy area from the rest of the pasture. The berm varied in height from 45 to 60 cm above ground surface and was approximately 5 m across at its base. The total area contained within this berm was 5.9 ha (Campbell et al., 1995). The flume station was located approxi mately 200 m from the weather station. Two supplemental wells were constructed at the weather station for continuous water-

PAGE 99

78 table monitoring. Water-table depth in one well near the flume also was continuously monitored by the flume datalogger (Campbe ll et al., 1995). A sha llow collection ditch paralleled the perimeter berm for approximate ly one third of its length nearest the flume station. This V-shaped collection ditch wa s grassed and was less than 30 cm deep and approximately 3 m wide (Campbell et al., 1995). The constructed collection ditch discharged into a trapezoidal flow measurem ent flume with a maximum capacity of 7.1 cfs (200 l/s). A total of 69 we ll stations were established on this site. Each well station was composed of two or three wells of different depths. Weather, runoff and groundwat er quality measurements were obtained at the Dry Lake Dairy #1 site to provide information concerning water and P movement in flat, sandy, high-water-table soils. Summary data for annual rainfall, ET, and observed runoff are listed in Table 3-5. These data indicate that y ears 1989 and 1990 were dry years when compared to year 1991 with less rain fall and more ET. Year 1991 was relatively wet with relatively more rainfall. A simple water budget in the last column of Table 3-5 shows year 1989 is the driest, year 1990 ranks the second, a nd year 1991 is the wettest among the three years. Table 3-5. Summary of annual water budget on Dry Lake Dairy #1 sitea. Rainfall(R)Runoff(RO)ET 100 ( ET+RO-R ) /R Years (cm) (cm) (cm) (%) 1989 (April-December) 88.66 5.15 91.37 8.87 1990 117.47 25.71 100.30 7.27 1991 133.00 44.29 92.18 2.61 Total 339.13 75.14 283.85 5.86 aData from Tremwel (1992). The purpose of this application was to validate the modifi ed ACRU2000 model by comparing with the lumped FHANTM model (T remwel, 1992), which was applied in the Dry Lake Dairy #1 site in 1992. For the m odified ACRU2000 simulation, the Dry Lake

PAGE 100

79 Dairy #1 site was divided into 4 land segm ents, each with identical hydrologic properties and soil layers since no spatially distributed data were available. The discharge from the site through the flume was estimated usi ng the empirical formula (Tremwel, 1992) 5 0 5 1 5 2h 121409 0 h 508387 0 h 282156 3 Q (3-10) in which Q is the discharge in cubic feet per second and h is the surface water depth in feet. The measured rainfall, pre-calculated potential ET, and temperature from Tremwel (1992) were used as the time series input and were applied uniformly for each land segment. Some input such as soil propert ies (field capacity, soil porosity, wilting point, and hydraulic conductivity) as shown in Table 3-6 were adapted from the calibrated parameters by Tremwel (1992) for FHANTM and the rest of input were estimated. No deep seepage process was considered in the test. No calibration was performed for the modified ACRU2000 model. The simulation period for this valida tion test was from April 1989 through December 1991. The model was run thr oughout the whole simulation period and the simulated results for surface runoff and groundw ater table were compared with those from FHANTM, for both the calibration peri od (April 1, 1989 to August 31, 1990) and the verification period (September 1, 1990 to December 31, 1991). Model prediction statistics were calculated usi ng the results from both models to quantitatively evaluate the model performance. Scatterp lots of observed vs. simulated values were prepared to visualize the predicted results from both models. Results and discussion The major output variables for this test were the surface runoff through the outlet of the Dry Lake site and the groundwater table de pth, which was taken as the field average

PAGE 101

80 from all wells for the observations and as the mathematic average of predicted groundwater table depths form the four land segments for the model simulation. Continuous simulation of surface runoff is shown in Figure 3-17 for both models against the observed data for the period from August 1989 through October 1991. Both ACRU2000 and FHANTM successfully captured major rainfall ev ents in the calibration period but missed a few rainfall events during the verification period. There were some underestimations and overestimations of event hydrograph values by both models. Figure 3-18 illustrates the comparisons of cumulative surface runoff for the whole simulation period. It appears that both models’ predictions agree with each other fairly well throughout the simulation period. Furt hermore they agree well with the observed cumulative runoff during the calibration peri od. However, they overestimated the observed data between the middle of January 1991 through July 1991, and underestimated the observed data throughout the rest of the simulation period. Similar to FHANTM, the modified AC RU2000 model performed bette r during the calibration period than it did during the verification period. Figure 3-19 compares continuous simulations of groundwater table depth from the modified ACRU2000 model and FHANTM to the observed data for August 1989 to December 1991. A better agreement is obs erved during the calibration period for both models, while a slightly larg er deviation from the observed water table depth data observed for both models in the verification period. Considering different weather conditions for calibration and verification periods, it is apparent that the calibra tion of FHANTM was done in a relatively dry period whereas the verification was made in a relatively wet period. The difference in weather

PAGE 102

81 conditions for both periods could cause th e calibrated parameters by FHANTM to be suboptimal for the verification period. Since the same calibrated parameters were used in the modified ACRU2000 model, it is not surpri sing that similar results were obtained. Statistics can be used to quantitativel y address the difference in both models’ performance in simulating the observed conditions. As shown in Table 3-7, six statistics including bias, RE, RMSE, CV, R2 and NS, introduced earlier in Chapter 2, were calculated. Statistics bias, RE, RMSE, a nd CV in runoff for the modified ACRU2000 model are slightly larger th an those for FHANTM while R2 and NS are a little smaller than those for FHANTM. And all statistics ex cept NS in groundwater table are slightly larger than those for FHANTM. The compar isons of these statistics imply that FHANTM more accurately predicted in bot h runoff and water table depth than the modified ACRU2000 model. However, cons idering the modified ACRU2000 model was not calibrated specifically for the Dry Lake si te, both model’s predictions are acceptable. Furthermore, the Dry Lake Dairy #1 site is ve ry small with an area of 5.9 ha and uniform parameters were used over the simulation domain. Therefore, the advantages of a distributed model like the modified ACRU2000 model may not be demonstrated by this example. It maybe possible that bette r results could be achieved by the ACRU2000 model if there were sufficient data to calibrate spatially distributed parameters directly for this model. Scatterplots of observed versus predicted surface runoff and groundwater table depth were prepared to further visualize the overall performance of th ese models. Figure 3-20 shows the linear plot for surface runoff dept h. FHANTM’s predictions fall closer to the 1:1 line than those from ACRU2000, while ACRU2000’s te nd to underpredict surface

PAGE 103

82 runoff more often than FHANTM. However, th e linear plot for groundw ater table depths (see Figure 3-21) shows both models match the 1:1 line very well. From this analysis, it appears that the modified ACRU2000 model’ s ability to predict groundwater table is similar to FHANTM, but its abi lity to predict surface runoff is slightly poorer than FHANTM’s for the Dry Lake Dairy #1 site. Concluding Remarks In this Chapter, a hydrologic model capab le of multi-directional spatial simulation of overland flow and lateral groundwater flow was developed based on the existing ACRU2000 hydrologic model by adding new hydrologic components and modifying existing components. A simulation sequence analysis was conduc ted using two different hypothetical scenarios to analyze the influence of the simulation sequence on m odel predictions, and also to test the accuracy of the modified model in simulating hydrologic processes by comparing with well-known physically-bas ed models MIKE SHE and MODFLOW. Results indicate that different simulation sequences do lead to slightly different simulation results, but the predictions with different sequences are consistent when compared with those from MIKE SHE and MODFLOW. It was concluded that transferring water storage from the source land segment and updating water storages in the adjacent land segments immediately pr oduces the smoothest and most reasonable results. When there are multiple simulation sequences to choose from, the one following the expected flow directions based on t opography is reasonable. In addition, by comparing to results from MIKE SHE and MODFLOW, the modified model has been shown to simulate hydrology reasonably well.

PAGE 104

83 A test was conducted to validate the modified ACRU2000 model by comparing with the lumped FHANTM model in the Dry La ke Dairy #1 site, Kissimmee River Basin, FL. From the comparisons and statistical analyses, it was concluded that the modified ACRU2000 hydrologic model is cap able of adequately simu lating overland flow and groundwater table at the Dry Lake Dairy #1 site, but the modified ACRU2000 model is not better than FHANTM. However, the m odified ACRU2000 model was not calibrated but utilized input parameters calibrated by FHANTM. Thus the advantages of a distributed model like the modified AC RU2000 may not be demonstrated by this application. It maybe possible that better results could be achieved for the modified ACRU2000 if there are sufficient data to ca librate spatially distributed parameters directly from this model.

PAGE 105

84 0 1 2 3 4 5 6 8/1/8911/1/892/1/905/1/908/1/9011/1/902/1/915/1/918/1/91 Time (day)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (am) Rainfall ACRU2000 FHANTM Observed Figure 3-17. Comparisons of the continuous simulations of runoff from the modified ACRU2000 and FHANTM against the observed data. 0 10 20 30 40 50 60 70 80 8/1/8911/1/892/1/905/1/908/1/9011/1/902/1/915/1/918/1/91 Time (day)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Cumulative ACRU2000 Cumulative FHANTM Cumulative Observed Figure 3-18. Comparisons of the cumulativ e simulated runoff from the modified ACRU2000 and FHANTM with the cu mulative observed data.

PAGE 106

85 -200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0 20 8/1/8911/1/892/1/905/1/908/1/9011/1/902/1/915/1/918/1/9111/1/91 Time (day)Water Table Depth (cm)0 20 40 60 80 100 120Rainfall (cm) Rainfall ACRU2000 FHANTM Observed Figure 3-19. Comparisons of the continuous si mulation of groundwater table depths from the modified ACRU2000 and FHANTM against the observed data. Table 3-6. Model input parameters for the Dry Lake Dairy #1 site. Parameters Unit Value Soil layer deptha m 0.10 0.10 0.10 0.10 0. 15 0.15 0.15 0.15 0.50 0.50 Soil porositya m/m 0.40 0.40 0.32 0.32 0. 32 0.32 0.30 0.30 0.30 0.30 Wilting pointa m/m 0.11 0.11 0.05 0.05 0. 05 0.05 0.10 0.10 0.10 0.10 Field capacitya m/m 0.15 0.15 0.12 0.12 0. 12 0.12 0.12 0.12 0.12 0.12 Vertical saturated hydraulic conductivityb m/s 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 3.56 E-07 Lateral saturated hydraulic conductivityb m/s 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 4.44 E-05 Manning’s coefficientb m1/3/s 0.1 Brooks-Corey hb cm 40.0 40.0 45.0 45.0 45. 0 45.0 75.0 75.0 75.0 75.0 Brooks-Corey b 1.60 1.60 0.95 0.95 0.95 0.95 0.71 0.71 0.71 0.71 Brooks-Corey b m/m 0.11 0.11 0.05 0.05 0. 05 0.05 0.10 0.10 0.10 0.10 Surface maximum depressional storageb m 0.005 Crop coefficientb 0.4, 0.5, 0.6, 0.6, 0.7, 0.85, 1.0, 0.85, 0.8, 0.8, 0.85, 0.6 for 12 months aCalibrated values by Tremwel (1992); bEstimated values. Table 3-7. Statistics for the simulated surf ace runoff and groundwater table depths by the modified ACRU2000 and FHANTM. Statistics* bias RE RMSE CV R2 NS ACRU2000 -0.27 -0.37 0.74 1.03 0.34 0.10 Runoff FHANTM -0.17 -0.24 0.61 0.86 0.56 0.38 ACRU2000 12.73 -0.15 23.26 -0.27 0.81 0.68 Water Table FHANTM 9.79 0.00 22.23 -0.26 0.80 0.71 *Statistics were calculated using the pr edictions from both calibration and verifi cation periods. Units of bias and RMSE for runoff and water tablde depth are cm. the re st of statistics are untiless for both variables.

PAGE 107

86 (FHANTM) y = 0.8371x 0.0565 R2 = 0.5611 (ACRU) y = 0.5512x + 0.0553 R2 = 0.3425 0 1 2 3 4 5 6 0123456 Observed (cm)Simulated (cm) FHANTM ACRU2000 1:1 line Linear (FHANTM) Linear (ACRU2000) Figure 3-20. Scatterplots of observed vs. simu lated surface runoff depths predicted by the modified ACRU2000 and FHANTM. (FHANTM) y = 0.9494x + 5.4304 R2 = 0.7972 (ACRU) y = 0.9776x + 10.801 R2 = 0.8131 -200 -150 -100 -50 0 50 -200-150-100-50050 Observed (cm)Simulated (cm) FHANTM ACRU2000 1:1 line Linear (FHANTM) Linear (ACRU2000) Figure 3-21. Scatterplo ts of observed vs. simulated groundw ater table depths predicted by the modified ACRU2000 and FHANTM.

PAGE 108

87 CHAPTER 4 NUTRIENT SIMULATION MODEL Introduction Nutrient components were not included in the original ACRU (v3.00). After the model was entirely restructured into an object-oriented framework using the Java programming language (ACRU2000), Campbell et al. (2001) incor porated a nutrient module (ACRU-NP), which was patterned afte r transformation and transport concepts used in GLEAMS (Leonard et al., 1987; Knisel and Davis, 1999). The goals of Campbell’s work were to add capacities in ACRU2000 to 1) simulate nitrogen (N) and phosphorus (P) losses in surface runoff, sediment transport, and leaching, 2) simulate N and P cycling in the soil-water-plant-animal system, and 3) simulate N and P mass balances in the watershed system. Since G LEAMS includes most of these capacities in its current version it was used as a guide in development (Campbell et al., 2001). The module includes rainfall, irriga tion, fertilizers, plants, and animal wastes as potential nutrient sources and represents manageme nt impacts on N and P transformation and transport (Campbell et al., 2001). Nutrient simulation requires a wide range of inputs including soil properties, irri gation parameters, crop parameters, nutrients in residues, fertilizer application and also tillage operati ons. A few adjustments were made to tailor the performance of GLEAMS when it wa s recoded into ACRU2000, including the significant modification to include the capac ity to simulate grazing of pastures. However, ACRU-NP module was incapable of simulating multidirectional lateral nutrient transport between land segments th rough either surface or subsurface water

PAGE 109

88 movement. The lateral nutrient transport component in ACRU-NP was mainly designed for transporting nutrients dissolved in runo ff and adsorbed in sediments through single outflow from one land segment. Consideri ng the important mechanisms for delivering nutrients into downstream water bodies thr ough both lateral surface and subsurface water movement in south Florida, it was necessary to modify the model to simulate lateral nutrient transport. Additionall y, the original algorithm for dealing with the soil-surface nutrient exchange is a process of nutrient ex traction and adsorption, with the ratio for partitioning the nutrients between the water and soil phases being a function of clay content. This method has the potential to overestimate the nutrients adsorbed onto soils but underestimate the nutrient released into surface water, especia lly for south Florida flatwood soils that contain low clay contents To accurately simu late nutrient dynamics, it was necessary to modify this methodology. Moreover, the testing of the nutrient module within the framework of ACRU2000 has not been reported in the literature. Consequently, the major purpose of this wo rk was to modify the existing ACRU-NP module to enable multidirectional spatial nutrien t transport and test the modified model in simulating nutrient dynamics in s outh Florida flatwoods soils. The major modifications presented in th is chapter include adding new components to enable multi-directional sp atial transport of N and P th rough surface runoff and lateral groundwater flow, and new conservative solute transport components to strengthen the capacities of the module to meet future need s. In addition, se veral existing nutrient transformation processes including ammonifi cation, nitrification, P mineralization, immobilization, and denitrification were modi fied by Martinez (2006) and utilized in the model presented here.

PAGE 110

89 Similar to the hydrologic processes, th e modified nutrient components are grouped into vertical and lateral pr ocesses as shown in Appendix A. The vertical processes consist of N and P transport and transfor mation processes, and conservative solute transport, along the soil profile. These pr ocesses are executed after the vertical hydrologic processes. The lateral processes include the lateral N, P and conservative solute transport components, which are executed after the ho rizontal hydrologic processes. Appendix A lists the individual pr ocesses and the order that these processes are simulated in the context of the coupled mode ling system. A daily time step is used in the nutrient module. In the following sections, the nutrient com ponents are briefly discussed. Details regarding the components taken from GLEAMS can be found in Knisel et al. (1993) and Fraisse and Campbell (1997). Processes de veloped in this work are discussed extensively. A conservative solute test (a n extension of the hypot hetical study in Chapter 3) is introduced to investigate the accur acy of the solute tr ansport performance by comparing with the established partic le tracking model PM PATH (Chiang and Kinzelbach, 2005). Following that, an applica tion of the modified model to beef cattle pastures in the Buck Island Ranch, Lake Okeechobee Basin, FL is conducted, which includes model calibration, verification and se nsitivity analysis. Conclusions regarding the model development and performance are summ arized at the end of this chapter. Nutrient Components Nitrogen Cycle Components A schematic representation of the N com ponents in GLEAMS is shown in Figure 4-1. N in the soil is divided into six classifications including nitrate-N (NO3-N), ammonium-N (NH4-N), fresh organic N (crop residues and roots), organic N from animal

PAGE 111

90 wastes, active soil N, and stable soil N. Each soil layer, as well as the soil surface, has its own set of these pools. Some of the compar tments delineated in Figure 4-1 are for the ground surface only (grain, stover, atmospheric N, and assimilated N); some are for both surface and subsurface computational soil laye rs (fresh organic N in crop residue and roots, fertilizer, nitrate, am monia, and organic N in anim al waste); and the active and stable soil N occurs only in the subsurface. Sources of N include atmospheric N (rainfall), fertilizer N, biol ogical N fixation, and the orga nic N pool (crop residue, roots and animal waste). Sinks for N are atmos pheric N (denitrification and volatilization), plant uptake, runoff, erosion, and percol ation. Nitrogen can be transformed by ammonification, nitrification, immobilizati on, denitrification, and volatilization. Figure 4-1. Schematic representation of the GLEAMS nitrogen cycle (AM = ammonification; NI = nitrification; DN = denitrification; VL = volatilization; IM = immobilization; UP = uptake; FX = fixation) (Knisel et al., 1993).

PAGE 112

91 Nitrogen including NO3-N and NH4-N may be transported with surface runoff and lateral groundwater flow from one land segm ent to its neighboring land segments. To calculate the amount of nitrogen mass moved with water flows, the concentrations of nutrients in the surface water and subsur face are calculated separately. These concentrations are then multiplied by the volume of water moving along each pathway to obtain the mass of nutrients lost through su rface runoff and lateral groundwater flow from each land segment to its neighboring land segments, streams or water bodies. Mineralization N mineralization is considered a two-st age process in GLEAMS ammonification followed by nitrification. In the soil, organic N is divided into two pools: a stable pool from which no mineralization occurs (C: N > 25), and an active pool available for mineralization (C: N < 25). The active pool can be replenished from flux between it and the stable pool (a function of relative pool sizes), from fr esh organic N from decomposing roots, and from decomposition of organic N in app lied animal waste. Ammonification of decomposing roots, othe r crop residues, and animal wastes in the soil is considered a first-order process a nd is a function of C:N and C:P ratios of the residue, temperature, and soil moisture content. Of the total NH4-N released by animal waste in the soil, 20% is allocated to th e active soil N pool with 80% going to an NH4-N pool. Crop residues and animal waste on the surface use a similar f unction to determine ammonification rate but liberated NH4-N stays on the surface in a soluble NH4-N pool until moved into the soil by the next rainfall or irrigation event. As the C: N ratio of the soil active N pool is set between 12:1 and 25:1, organic C is already partially accounted for and ammonification of the active N pool is a function of temper ature and soil water

PAGE 113

92 content. Nitrogen flux from the active N pool can go to either the stab le soil N pool or to the free NH4-N pool. The second stage of mineralization is n itrification, which converts N in the NH4-N pool into NO3-N. This is considered a zero-order process (not affected by the amount of NO3-N or NH4-N) and is modeled as a function of temperature and soil moisture. Nitrogen from the NH4-N pool that is nitrified is transferred to the NO3-N pool. Nitrification can also occur on the soil surfac e where it is a function of temperature and soil moisture. As with the soluble NH4-N pool, the soluble nitrate pool on the soil surface remains on the soil surface until moved in to the soil by rainfall. Once in the soil, these products join the appropriate pools in the soil. Immobilization Immobilization occurs when either NO3-N or NH4-N is used by soil microbes for growth. Once incorporated into the microbe the N is unavailable for uptake by plants and remains so until the death and decompos ition of the microbe. Thus, factors that affect the growth rate of microbes are important in driving immobilization. Immobilization of both soil and surface N is a function of the amount of NO3-N and NH4N available, C:N ratio, and C:P ratio. Immobilization is limited to 95% of total available NO3-N and NH4-N. Immobilized N in the soil can e nd up either in the fresh organic N pool or in the st able soil N pool. Denitrification Denitrification is generally thought to re quire anaerobic condi tions and available organic matter to proceed at an optimum p ace. Denitrification of N from the soil NO3-N pool to the atmosphere is driven by availabl e organic carbon, soil moisture (an indicator of available oxygen), and temperature. Ma rtinez (2006) modified the denitrification

PAGE 114

93 process using a threshold water content to determine whether denitrification will occur in each layer. The threshold content was assume d to correspond to a sa turation value of 0.8 as in the WAVE model (Vanclooster et al. 1996). Runoff, sediment transport and percolation Nitrogen lost from the top 1 cm active su rface layer of soil is primarily a function of the amount of water moving (percolation or runoff) and a combination of nutrient and soil characteristics in ACRU-N P. The concentration of NH4-N in the water is dependent on how much is adsorbed onto the soil, which is a function of the clay content of the soil. When runoff begins, the surface active layer interacts with the runoff stream, imparting some of the soil chemicals to the runoff. The amount of NH4-N adsorbed to the top 1 cm of soil is also available for erosion where loss is a function of sediment loss and adsorbed NH4-N concentration. However, the original design was base d on the assumption that no ponded water was left on the surface on a given day. Thus it ma y not be true in reality. To consider the nitrogen concentration in ponded surface wa ter, Martinez (2006) developed a process which uses a mixing-type model to determin e the exchange of nutrients between ponded water and the top-most soil layer. This pr ocess allows nutrients to move upward or downward depending on the concentration gradients between ponded water and soil water. Loss of N to percolation is a function of concentration of NO3-N and NH4-N available for movement (again, functions of partitioning between soil and water phases) in each soil layer and the volume of water percolating through the soil.

PAGE 115

94 Uptake, evaporation, and fixation Nitrogen uptake was patterned after th at in the EPIC model (Sharpley and Williams, 1990) for estimation of nitrogen demand in ACRU2000. All crops differ in their affinity for nitrate or ammonium, but it is assumed that nitrate and ammonium uptake is equal to the relative mass of each species in the soil layer from which transpiration occurs. Plant demand for N is a function of the concentration of N in the crop, the growth ratio of the plant (curre nt accumulated LAI as a proportion of potential LAI at maturity) and potential yi eld of the crop. Potential uptake of N is a function of available NO3-N and NH4-N and the amount of plant tran spiration (calculated in the hydrology component as a function of available water in the root zone). If supply is less than demand, an N stress factor that represents the N deficiency in soil (further discussion in Chapter 5), is calculated and used to lower potential growth. In modeling N uptake by legumes N fixation occurs in this model only on days when concentration of NO3-N + NH4-N in soils is below a crop/soil-specific threshold value. When fixation occurs, the amount is set equal to demand. Nitrogen fixation may also be curtailed during early crop development. Soil water can move upward in the soil profile in ACRU2000 as discussed in Chapter 2. NO3-N and NH4-N in the soil solution can move upward with the soil water to successive soil layers. The rate of upward movement of NO3-N and NH4-N is a function of the concentrations of NO3-N and NH4-N, respectively, in solution in the lower layer and upward flux amount. Soil NH4-N is not lost to the atmo sphere by volatilization in this model.

PAGE 116

95 Rainfall and fertilizer Ammonium and nitrate contained in rainfall are instantaneously available nitrogen. Their concentrations vary throughout the year bu t in the model it is assumed that all of the rainfall nitrogen is in the form of nitrate and the concen tration in rainfall remains the same throughout the model simulation period (Fraisse and Campbell, 1997). Since NO3-N and NH4-N pools are separately mainta ined, and nitrification and ammonification processes are simulated se parately as well, nitrate and ammonia fertilizers are distinguished in application. Fertilizer and animal waste can be applied on soil surface, incorporated, injected, or applied as fertigation. Ammonia volatilization NH4-N in the soil solution is not lost to volatilization. Volatilization losses from surface applied animal wastes, however, are assume d to occur. The extent of the losses is a function of the NH4-N content applied in the waste a nd the air temperature. Losses are assumed to occur for a period of one week following application. Surface and subsurface lateral nitrate nitrogen transport To calculate the mass of nutrients carried by multi-directional spatial overland flow and lateral groundwater flow, the concentra tions of nutrients are multiplied by the volume of water moving in each pathway to obtain the mass of nutrients lost through surface runoff and lateral groundwater flow to neighboring land segments, streams or water bodies. The concentra tions of nitrate in surface runoff and lateral groundwater flow are calculated: i d s i d s i d sWM 3 SNO W 3 CNO (4-1) j i s j i s j i sWM 3 SNO W 3 CNO (4-2)

PAGE 117

96 where the subscript s and d indicate the sour ce and destination land segments, and i and j are the soil layers of source and destin ation land segments, respectively; CNO3Ws,d,i and CNO3Ws,i,j are the concentrations of nitrate in surface runoff and saturated soil layers [M/L3], respectively; SNO3s,d,i and SNO3s,i,j are the nitrate masses stored in surface runoff and saturated soil layers [M/L2]; WMs,d,i is the water storage on the ground surface [L] and WMs,i,j is the water storage in the saturated so il layers [L] before overland flows and lateral groundwater flows are transfe rred to the neighboring land segments. i d s s i d sQ WS WM (4-3) j i s j s j i sQ WS WM (4-4) where WSs is the current water storage [L] and Qs,d,i is the transferred overland flow from the source land segment to its neighboring land segments [L]; WSs,j is the current water storage in the jth saturated soil layer of the source land segment [L]; Qs,i,j is the transferred lateral groundwater flow from the jth layer to its counterpart soil layer within the soil of the ith neighb oring land segment [L]. Nitrate transported through surface r unoff and lateral groundwater flow is calculated: i d s i d s i d sQ W 3 CNO 3 RONO (4-5) j i s j i s j i sQ W 3 CNO 3 RONO (4-6) where RONO3s,d,i is the nitrate mass removed from sour ce to destination land segments or water bodies through surface runoff [M/L2]; RONO3s,i,j is the nitrate removed from layer j in the source land segment to layer j in th e destination land segm ent through the lateral groundwater flow [M/L2].

PAGE 118

97 The conservation of nitrate mass is checked for both nitrate mass transported through surface runoff and lateral groundwater flows. The method for checking the mass balance can be expressed as: ANM TNM TNM TNM ANM TNM ANM TNM (4-7) where TNM is the nutrient mass to be transferred [M/L2] and ANM is the available mass in storage [M/L2] on ground surface or soil layers. Th is approach is applied to the following ammonium and labile P as well. Surface and subsurface lateral ammonium nitrogen transport The concentrations of ammonium in runoff and lateral groundwater flow are calculated: i d s i d s i d sWM AMON W 4 CNH (4-8) ) WM SOILMS CNHKD ( AMON W 4 CNHj i s j i s j i s j i s j i s (4-9) where CNH4Ws,d,i and CNH4Ws,i,j represent the concentrations of ammonium in the surface runoff and saturated soil layers [M/L3]; AMONs,d,i and AMONs,i,j are the ammonium in surface runoff and the saturated soil layer j [M/L2]; CNHKDs,i,j is the ammonium partitioning coefficient for each saturated soil layer [L3/M]; SOILMSs,i,j is the soil mass of each satu rated soil layer [M/L2]. The ammonium partitioning coefficient, CNHKDs,i,j, is estimated using an empirical relation to account for the range of clay content CLs,j [%] as j s j i sCL 083 0 34 1 CNHKD (4-10) The soil mass in each soil layer, SOILMSs,i,j, is obtained by multiplying the bulk density, SOILBDi,j, of the soil layer [M/L3], and DEPTHi,j, the depth of th e soil layer [L]:

PAGE 119

98 j i j i j i sDEPTH SOILBD SOILMS (4-11) Ammonium removed through surface runoff and lateral groundwater flow is calculated: i d s i d s i d sQ W 4 CNH 4 RONH (4-12) j i s j i s j i sQ W 4 CNH 4 RONH (4-13) where RONH4s,d,i and RONH4s,i,j are the ammonium removed through surface runoff and layer lateral flow [M/L2]. Phosphorus Cycle Components Figure 4-2. Schematic representation of the GLEAMS phosphorus cycle (MN = mineralization; IM = immobilization; UP = uptake) (Knisel et al., 1993). Phosphorus is divided into six component s (Figure 4-2). Many are analogous to the N classifications including labile-P, organic humus, fr esh organic P (crop residues and roots), organic P from anim al wastes, active i norganic P, and stab le inorganic P. Each soil layer as well as the soil surface ha s its own set of these pools. Sources of P include atmospheric P (rainfall) fertilizer P, inorganic P, and the organic P pool. Sinks

PAGE 120

99 for P are plant uptake, runoff, erosion (SED on figure), and percolation. Phosphorus can be transformed by mineralization and immobilization. Labile P may be transported with surface runoff and lateral groundwater flow from one land segment to its neighboring land segmen ts. To calculate th e amount of nutrient mass moved with water flows, the concentrations of nutrients in the water are calculated. These concentrations are then multiplied by the volume of water moving in each pathway to obtain the mass of nutrients lost through surface runoff a nd lateral groundwater flow from this land segment to neighboring land segments, streams or water bodies. Mineralization P mineralization is considered a single step first-order process (J ones et al., 1984). Mineralization of the organic humus pool supp lies the labile P pool. Mineralization of organic humus is a function of soil moisture, temperature, and P content, and uses the same temperature and moisture factors as were used in the N component. The ratio of the size of the active soil N pool to the stable soil N pool is also used to partition the soil organic humus pool into th e mineralizable fraction. Mineralization of fresh organic P (roots a nd crop residue in th e soil) supplies the soil humus pool. The rate of mineralization is a function of residue composition, C:N and C:P ratios, temperature and soil moisture. The same function is used to calculate P mineralization from residue on the soil surface as well. As in the EPIC model (Sharpley and Williams, 1990), 75% of mineralized P is ad ded to the labile P pool and 25% to the organic humus pool. Mineralization of P in animal wastes is a function of temperature and soil moisture as well as a rate constant. In accordance with Bhat et al. (1981), 75% of mineralized P is allocated directly to the labile P pool and 25% to the organic humus pool.

PAGE 121

100 A further similarity to the N component is the flux between the active inorganic P pool and the stable pool. This flux is a function of the si ze of the active and labile P pools as well as temperature, soil moisture and calcium carbonate concentration of the soil (which affects sorption). It is assumed that at equilibrium, the stable pool mass will be four times the active pool mass (Sharpley and Williams, 1990). Mineralized P from the active pool can also be a llocated to the labile P pool. Immobilization The same decomposition rate used in calcu lating immobilization of N is used in calculating immobilization of P. The decom position rate is a func tion of temperature, soil moisture and the C:N and C:P ratios of th e residue. Immobilization rate is also a function of the concentration of labile P in the soil layer. It is assumed that residue is 40% C and that the microbes assimilate 40% of the C. As with N immobilization, P immobilization is also limited to 95% of labi le P. The same factor is used for both nutrients so a shortage of N can limit P imm obilization and vice versa. Immobilization of surface P is calculated in the same way. Runoff, sediment, percolation As with N, the amount of P lost in runoff and percolation depends on the partitioning of P between the soil and wate r phases. In GLEAMS, the amount of P adsorbed to the soil is a func tion of the clay content of the soil. This method, however, greatly overestimates the partitioning coe fficient for the high water table flatwoods Spodosol soils and underestimates the P concen tration in soil wate r. An alternative method (by Portier KM according to Frai sse and Campbell (1997)) obtained through a study in Lake Okeechobee watershed determined the partitioning coefficient as a function of the content of magnesium, organic car bon and/or aluminum. Martinez (2006)

PAGE 122

101 introduced a mixing-type model to calculate P transport between the ponded water and soil water, similar to that for N. As with the N component, the actual amount of P lost in runoff or percolation is a function of soluble P available and the water flux. Losses to erosion are calculated the same as for N with corresponding losses from the animal waste, active mineral P and stable mineral P pools as well as ad ditional P lost from the humus pool. Uptake and evaporation Potential P uptake is calculated indepe ndently of demand, as a function of transpiration and concentration of labile P in the soil water. Plant P demand is the function of the concentration of P in the cr op, the growth ratio of the plant and the potential yield of the crop. If supply is less than demand, uptake is equal to the supply. Otherwise, uptake is equal to the demand. Unlike the nitrogen component, growth is assumed not to be limited by P deficiency in the original ACRU 2000. However, it may not be true in reality. A P st ress factor was introduced through this work to represent the P stress on plant growth due to P deficien cy (detailed discussion in Chapter 5). P moves upwards in the soil profile with upward soil water flux. The amount lost to the next higher layer is a function of the c oncentration of labile P in the soil water in the current soil layer and the volume of upward flux. However, it is assumed that there are no volatilizati on losses of P. Rainfall and fertilizer The GLEAMS model includes an option to input the rainfall N concentration but it does not allow input of rainfall P levels. This function was added to ACRU-NP by Campbell et al. (2001) and P in rainfall and ir rigation water is accounted for in the same way as for N.

PAGE 123

102 Both rainfall P and fertilizer P are assumed to be soluble and available to the plant as soon as they are applied. If the fertilizer is surface applied and not incorporated, it is stored on the surface and moved to the surf ace layer soluble (labile) pool on the first occurrence of rain or irrigation following th e application. All P from incorporated fertilizers and infiltrated rainfall is added im mediately to the labile P pool. Availability of P from organic wastes is more gradual and was discussed in the mineralization and immobilization sections. Surface and subsurface lateral labile phosphorus transport The concentrations of labile P in runoff and lateral flow are calculated: i d s i d s i d sWM PLAB CPLABW (4-14) ) WM SOILMS CPKD ( PLAB CPLABWj i s j i s j i s j i s j i s (4-15) where CPLABWs,d,i and CPLABWs,i,j represent the concentrations of labile P in the surface runoff and in the saturated soil layers, respectively [M/L3]; PLABs,d,i and PLABs,i,j are the labile P in surface runoff [M/L2] and in the saturated soil layer j [M/L2]; CPKDs,i,j is the labile P partitioning coefficien t in the saturated soil layer [L/M]. In GLEAMS, it was assumed that P is partia lly adsorbed to the soil clay fraction. However, this assumption is not adequate fo r sandy, high-water-table flatwoods spodosol soils. Therefore, an algorithm was deve loped by Portier (personal communication, referenced in Fraisse and Campbell (1997)) to represent the P partitioning coefficient, CPKDs,i,j, as a function of the content of doubl e-acid-extractable magnesium (DAMGRD [mg/kg]), organic carbon (OCBRD [mg/kg]) and/or oxalate-extractable aluminum (OXALRD [mg/kg]) in each soil horizon: For the A horizon

PAGE 124

103 2390 2 j i se CPKD For 2 103 DAMGRD (4-16) 2233 2 j i se CPKD For 2 103 DAMGRD when 865 1 OCBRD (4-17) 4420 1 j i se CPKD For 2 103 DAMGRD when 865 1 OCBRD (4-18) For the E horizon 2410 4 j i se CPKD For 45 496 OXALRD (4-19) 0480 1 j i se CPKD For 1 30 OXALRD (4-20) 4440 1 j i se CPKD For 45 496 OXALRD 1 30 when 95 4 DAMGRD (4-21) 7670 1 j i se CPKD For 45 496 OXALRD 1 30 when 95 4 DAMGRD with25 57 OXALRD (4-22) 3500 3 j i se CPKD For 45 496 OXALRD 1 30 when 95 4 DAMGRD with25 57 OXALRD (4-23) For the Bh horizon 7510 3 j i se CPKD For 5 1327 OXALRD (4-24) 1950 2 j i se CPKD For 5 1327 OXALRD (4-25) For the Bw horizon 2120 3 j i se CPKD For 8 570 OXALRD (4-26) 6040 1 j i se CPKD For 8 570 OXALRD (4-27) The soil mass for each soil layer, SOILMSs,i,j is calculated using Equation (4-11). Labile P removed through surface runoff and lateral groundwater flow is calculated:

PAGE 125

104 i d s i d s i d sQ CPLABW ROLP (4-28) j i s j i s j i sQ CPLABW ROLP (4-29) where ROLPs,d,i and ROLPs,i,j are the labile P removed through surface runoff [M/L2] and layer lateral flow [M/L2]. Conservative Solute Transport Components A transport module capable of simulating co nservative tracers su ch as bromide and rhodamine WT (RWT) was added into the nu trient module through this work assuming only vertical and horizontal tr ansport processes occur (i.e. no transformation processes). The vertical and horizontal trans port processes were simulated as discussed for nitrate. Martinez (2006) created several vertical processes to simulate the conservative solute transport through evaporation, soil water redi stribution, and exchange between water in the top soil layer a nd ponded water. The evaporatio n solute tran sport process calculates the upward move ment of a conservative solute in the soil due to evaporation, however solute is not allowed to move upward out of the soil surface layer by evaporation. The subsurface solute transpor t calculates the leach ing of a conservative solute downward through the soil profile. A mixing-type model is used to determine the exchange of solute between ponded water and so il water in the top-most soil layer, driven by the solute concentration gradients. Surf ace and subsurface latera l transport processes added through this work are described below. The concentrations of conservative solute in the ponded water on the ground surface and in the saturated groundwater are calculated: i d s i d s i d sWM CONSOL CCONSOLW (4-30)

PAGE 126

105 j i s j i s j i sWM CONSOL CCONSOLW (4-31) where CCONSOLWs,d,i and CCONSOLWs,i,j are the concentrations of conservative solute in the ponded water and satu rated soil layers [M/L3], respectively; CONSOLs,d,i and CONSOLs,i,j are the conservative solute mass on the ground surface and saturated soil layers [M/L2], respectively; WMs,d,i is the water storage on the ground surface [L] and WMs,i,j is the water storage in the saturated soil layers [L] as defined in Equations (4-3) and (4-4), respectively. Conservative solute removed in surface runoff is calculated: i d s i d s i d sQ CCONSOLW ROCONSOL (4-32) where ROCONSOLs,d,i is the conservative solute removed in surface runoff [M/L2]; Qs,d,i is the surface runoff generated on a given da y from the current land segment to the ith neighboring land segment [L]. Conservative solute removed in late ral groundwater flow is calculated: j i s j i s j i sQ CCONSOLW ROCONSOL (4-33) where ROCONSOLs,i,j represents the conservative solu te removed in lateral groundwater flow from the layer j to the ith neighboring land segment [M/L2]; Qs,i,j is the lateral flow in layer j to the ith neighboring land segment [L]. Mass balance is checked for the conservative solute in both surface runoff and lateral groundwater flow as was described for N and P. Initial and Boundary Conditions Running the model requires good estimates for initial nutrient pools. This is especially true for validation comparisons w ith observed data. Because of the dynamics of the nitrate and ammonium processes, inaccu rate initialization may not adversely affect

PAGE 127

106 long-term simulation results. Initial values of different nutrient pools are very site specific and are generally management depende nt (Fraisse and Campbell, 1997). This is especially true for systems with animal wast e application and management activities such as stocking, fertilization, a nd tillage operations. Model user s should make every effort to obtain good estimates of initial conditions which may involve soil sampling and analyses. However, soil samples commonly only determine certain nutrient variables such as organic matter, total nitrogen, total phosp horus, nitrate, ammonium and available phosphorus at a certain de pth of soil. Stable, active mine ral N and P and initial nutrient pools in the deeper soil layers are usually unavailable. For a spatially distributed model, the spatial distribution of initial nutrient pools is also required, which further increases the level of effort. If initial values of the nutrient pools are not availa ble, the model is set up to use default values. Agricultural activities such as stocking, fe rtilization, burning, etc. are treated as time series inputs in the modifi ed model. Nutrients from ra infall and irrigation (nitrate and labile P only) and animal wastes must be specified as time series input and are added to the appropriate nutrient pools. Model Testing and Validation Conservative Solute Test To test the accuracy of the algorithm for the conservative solute transport components added to ACRU2000, model resu lts for a hypothetical scenario were compared with those from PMPATH (Chi ang and Kinzelbach, 2005), which is an advective transport model using groundw ater pore velocities from MODFLOW (McDonald and Harbaugh, 1998).

PAGE 128

107 Scenario description The axisymmetric hypothetical domain with initial wate r levels below the ground surface used for testing groundwater movement (Case 2) in Chapter 3 was extended for this test. The hydrologic inputs remained the same as for the te sts discussed in Chapter 3. In order to run the model t ogether with the added conser vative solute components, an initial solute mass has to be specified. For the modified ACRU2000, it is assumed th at 3kg of solute was initially present in the 3rd soil layer of each of central four land segments, and that this solute mass was uniformly distributed inside of that soil layer. However, PMPATH requires that an initial number of particles, rather than an initial solute mass, be specified for each element (equivalent to a soil layer in ACRU2000), with a maximum of 5000 particles per element. Furthermore, it is difficult to set up the ini tial particle distribution pattern inside an element so that it is completely uniform. After a few experiments, 4704 particles were placed as uniformly as possible in each 3r d soil layer element for each central land segment in PMPATH to repres ent the equivalent quantity of 3kg of solute. Thus, each particle is mathematically equivalent to a mass of 3kg/4704, and it is possible to convert the integer number of particles which appear inside a certain el ement into a mass of solute in PMPATH. For comparison with the simulated results from the modified ACRU2000, the PMPATH simulation was run usi ng a daily time step for two years. Results and discussion Considering the axisymmetric distribution of solute mass, the simulated solute mass in the third and fourth soil layers of LS20, LS21 and LS26 from the modified ACRU2000 model were selected to compare with the results obtained from PMPATH (see Figure 4-3). This figure shows that th ere is a good agreement, particularly during

PAGE 129

108 the first year, between the models for the predicted solute mass in layer 3 for LS21, which is one of the central land segments that received an init ial dose of 3kg solute in this layer. During the second year a larger di screpancy is observed in layer 3 of LS21. However, a good agreement was found in the pred icted solute mass for the 4th soil layer throughout the two-year simulati on. Smaller bias and RMSE values were found for layer 4 than for layer 3 in LS21 (see Table 4-1). Similar to LS21, results for LS20 seem to be better for layer 4 than for laye r 3. However, during the second year more solute mass is predicted in layer 3 by ACRU2000 than by PM PATH. A poorer agreement is observed in both layers 3 and 4 for LS26, which is a land segment receiving water flow and solute transport through LS20 instead of from LS21 di rectly. The predicted solute mass from ACRU2000 is obviously larger th an that predicted by PMPATH for both layers. Large negative values of RE and NS in Table 41 indicate the agreement between the two models is not as strong in LS26 as it appears for LS21 and LS20. Table 4-1. Statistics for the solute mass predicted by the modified ACRU2000. Statistics* Bias RERMSECVR2 N S ( units ) ( k g) ( ) ( k g) ( ) ( ) ( ) La y er 3 0.19470.08330.25820.11040.8652 0.6387 LS21 La y er 4 -0.0297-0.10950.08720.32170.6491 0.6036 La y er 3 -0.0727-0.42160.10930.63360.7324 0.0941 LS20 La y er 4 -0.0049-0.33920.00880.61010.7497 0.4997 La y er 3 -0.0130-0.90290.01731.20110.7303 -0.9389 LS26 La y er 4 -0.0014-0.85360.00181.13400.6080 -0.7285 *Statistics were calculated using the predictions from PMPATH as the observed values. After carefully examining the outputs fo r each land segment in the domain, it was found that at the end of the two-year simula tion period PMPATH did not transport solute mass to the land segments located along the domain boundary, while in ACRU2000 some solute mass were predicted to reach the bounda ry cells in both layers 3 and 4. This finding indicates that the solute movement is slightly slower in PMPATH than it is in the

PAGE 130

109 ACRU2000 model. In ACRU2000, groundwater flow always car ries a certain amount of solute due to the assumption of uniform distribution of solute within the land segment layer. However, in the PMPATH the part icle tracking algorithm does not transport particles from one element to the next until it travels the distance from where it is originally located to the f ace between the two elements. Although the water movement was a little faster in MODFLOW than in A CRU2000, the concentra tions predicted in each element of the PMPATH model are smaller than predicted by ACRU2000. However, in spite of the difference in assumptions made by these models, the results illustrated in Figure 4-3 indicate that they produce qualitatively similar results.

PAGE 131

110 Solute Mass within the Soil of Land Segment 21 0.00 0.75 1.50 2.25 3.00 1100199298397496595694 Time (day)Solute Mass (kg) 3rd layer PMPATH 4th layer PMPATH 4th layer ACRU2000 3rd layer ACRU2000 Solute Mass within the Soil of Land Segment 20 0.00 0.13 0.25 0.38 0.50 1100199298397496595694 Time (day)Solute Mass (kg) 3rd layer PMPATH 4th layer PMPATH 4th layer ACRU2000 3rd layer ACRU2000 Solute Mass within the Soil of Land Segment 26 0.000 0.006 0.012 0.018 0.024 0.030 1100199298397496595694 Time (day)Solute Mass (kg) 3rd layer PMPATH 4th layer PMPATH 4th layer ACRU2000 3rd layer ACRU2000 Figure 4-3. Comparisons of the solute mass in the 3rd and 4th soil laye rs of selected land segments between the modified A CRU2000 and PMPATH (note change in scale).

PAGE 132

111 Application at Buck Island Ranch, Lake Okeechobee Basin, Florida Project description To evaluate the ability of the modifi ed ACRU2000 model to predict non-point source nutrient pollution, especi ally P loading, from beef cattle ranches in the Lake Okeechobee region, the model was applied to several isolated pastures at the Buck Island Ranch in the Lake Okeechobee Basin. Figure 4-4. Location of M acArthur Agro-ecology Res earch Center (MAERC, 2004). The experimental site at the MacArthur Agro-ecology Research Center (MAERC) at Buck Island Ranch, Highlands County, Florid a (see Figure 4-4) consists of 16 pastures, eight 16.72-ha improved summer pastures with stocking from May to October and eight 26.76-ha semi-improved winter pastures with stocking from November to April. Four stocking rate treatments, including 0, 15, 20 and 35 cow-calf pair s per pasture that represent non-grazing, low, medium, and high-grazing, resp ectively, were implemented

PAGE 133

112 for this experiment. Each of the stocking densities had two replicates in each of 16 winter and summer pastures. Figure 4-5 shows the layout of pastures and the instrument locations. Each of the pastures was indivi dually fenced and ditched to capture surface runoff through a single trapezoidal flume. Nutrient concentrati ons, including total phosphorus (TP), soluble-reactive phosphorus (SRP), total N (TKN), nitrate + nitrite (NOx) and ammonium ( 4NH) were measured from water samples collected using ISCO automatic samples installed in each flume. Data collection started on May 19, 1998 for the summer pastures and on May 21, 1998 for the winter pastures, and continued through December 2003, interrupted only by occas ional equipment malfunctions (MAERC, 2004). The winter pastures ha ve two meteorological stati ons and the summer pastures have one station. These meteorological stat ions were installed on May 21, 1998 to record rainfall, windspeed and directi on, air temperature, relative humidity, and solar radiation using Campbell Scientific CR10X dataloggers. In addition, there were 5 tipping bucket rain gages installed thro ughout the pastures to record rain fall. A main weather station is located at the Ranch headquarters, whic h operated throughout the experiment, where manual rainfall readings are taken daily and a datalogger records air temperature, relative humidity, windspeed and direction at 3.0 m and 9.1 m, solar radiation, and soil temperature. For the purpose of model application, four pasture sites, including W6, W7, S1 and S4, were selected for use. Considering the similarity in soil characteristics, and the limited measurements of water table and nutri ent loads throughout the 6-year simulation period from each site, one site each from the summer and winter pastures was used for calibration and the other one for verification. This allows calibration over more varied

PAGE 134

113 weather conditions with more observations at one site than if both calibration and verification were performed at each site. W6 and S4 were used for model calibration and W7 and S1 for verification. Figure 4-5. General layout of the project field (S1-S8 indi cates eight 16.72-ha summer pastures and W1-W8 represent eight 26.76-ha winter pastures) (MAERC, 2004). Tables 4-2, 4-3, and 4-4 list the management practices applied at the four selected pasture sites. Table 4-2 shows that the tw o calibration sites W6 and S4 had a stocking rate of 15 cow-calf pairs a nd the two verification sites W7 and S1 had no stocking throughout the simulation period. Manure applic ation rates were init ially set at 0.0176 tons/ha and 0.0277 tons/ha of beef solid ma nure for W6 and S4, respectively, for each day when cattle were in the pastures. Thes e rates were based on an estimated bodyweight for the Braford cattle of 1,200lbs and a daily excretion of 63 lbs of manure (as excreted) /1,000-lb animal unit (USDA, SCS, 1996). Bu ilt-in values for the composition of manure code 8-beef slurry were used in the model to calculate daily N and P applications.

PAGE 135

114 Ammonium nitrate fertilizer was applied on summer pastures S1 and S4 three times during the simulation period for a total of 56 kg/ha nitrogen on each application (see Table 4-3). Prescribed burns a nd accidental fires occurred in all four pastures at different times (see Table 4-4). In the model, when burns or fires occur, 95% of the biomass of the surface residue is removed. The areas of the four selected pasture site s and the percentages of soil series within each pasture are listed in Table 4-5. Alt hough multiple soil series exist in each selected pasture site, for this study it is assumed that W6 and W7 had the same soil characteristics, and S4 and S1 also had the same soil char acteristics because both winter and summer pasture pairs have the same dominant soil t ypes (see Table 4-5). Pineda fine sand was assumed as the base soil series for W6 a nd W7 and Felda fine sa nd was assumed for S4 and S1. In the model, W6 and W7 were each divided into 18 land segments with surface areas ranging from 0.6 ha to 2.6 ha. S4 a nd S1 were each divided into 12 land segments with surface areas ranging from 1.6 ha to 1.8 ha. This spatial discretization was determined based on the available topographic maps that show the contours of surface elevation and wetlands. Using this informati on, an elevation value was assigned to each land segment. Setting up the initial input data and paramete rs is critical to guarantee the accuracy of model performance. However, due to the un availability of spatiall y distributed data in these pasture sites, all initial physical and bi ochemical data were assumed to be the same for each land segment in the same pasture site except a few parameters including surface elevation, land segment area, initial water table depth, and boundary conditions (i.e. boundary width and hydraulic structure inform ation). Model input parameters are

PAGE 136

115 summarized in Table 4-6 for winter pastures W6 and W7 and Table 4-7 for summer pastures S4 and S1. Initializing the nutrient pool parameters (see Tables 4-6 and 4-7) was difficult because there was little nutrient information available for soils and surface residues in any of these pastures. The literature data obtained from laborat ory analyses of soil samples do not give a breakdown of all the nutrien t pools in all soil layers of interest, and these soil analyses are typically limited to th e plow layer since that may be the depth of most concern to researchers. Therefore, th e procedure for setting up initial nutrient pools introduced in the FHANTM manual (Fra isse and Campbell, 1997) was adopted. For each soil layer, nitrogen including a ll forms of nitrogen, i.e. mineralizable, stable organic, fresh organic, humus, ammonia, etc. was required model inputs. The total nitrogen (TN) was estimated for each computa tional soil layer using the data in Table 1 (Fraisse and Campbell, 1997) Nitrate-nitrogen (NO3-N) was estimated using a concentration of 10 g/g of soil in all soil layers Ammonium nitrogen (NH4-N) was estimated using a concentration of 2 g/g of soil. Root residue from previous biomass productivity was estimated at the beginning of the simulation as 40 kg/ha and distributed vertically in the root zone. Stable mi neral nitrogen was initialized by taking the difference between the TN and the rest of nitrogen pools except nitrate-N. Phosphorus pools in the soil that must be in itialized include fres h organic P in crop residue, organic humus P, labile P, and active and stable mineral P. According to Fraisse and Campbell (1997) soil organic humus can be estimated using the following relationships: TN 35 147 24 9 SORGP For the plow layer (4-9)

PAGE 137

116 TN 58 148 25 11 SORGP For other horizons (4-10) where SORGP is soil organic humus [ g/g], and TN is total nitrogen [%]. The labile P was estimated to be 8.7% of SORGP for sli ghtly weathered soils (Fraisse and Campbell, 1997). The active mineral P pool was estimated using the following relationship derived from Equation (12) (Fraisse and Campbell, 1997): PSP / ) PSP 1 ( PLAB PMINP (4-11) where PMINP is the active mineral P [kg/ha ] and PSP is the P sorption coefficient that can be estimated according to Sharpley and Williams (1990) 73 0 PH 116 0 BAST 0054 0 PSP (4-12) where BAST is base saturation [%] and PH is the soil pH for slightly weathered soils. The stable mineral P was set as four times active mineral P (Fraisse and Campbell, 1997). The fresh organic P is estimated as 10 kg/ha a nd distributed in the ro ot zone (Fraisse and Campbell, 1997). Using the above discussed relationships, th e initial nutrient pools were set up for both N and P. Tables 4-6 and 4-7 list th e calibrated N and P parameters based on the initial nutrient values. Although multiple ve getation species exist in each of selected sites, only one species, bahiagrass (Paspalum notatum Flgge) was assumed to exist for simplicity. Model calibration was based on personal judgm ent and statistical analyses for four major output variables, i.e., groundwater table depth, runoff depth, and N and P loads. Calibration was performed for these output va riables sequentially in the order listed above. During calibration, hydrologic parame ters, including porosity, wilting point, saturated hydraulic conduc tivity, restrictive layer hydr aulic conductivity, Manning’s

PAGE 138

117 roughness coefficient, surface maximum depre ssional storage, and crop coefficient; and nutrient parameters, including rainfall P and N concentrations, active P, stable P, humus P, active N, and stable N, were varied w ithin their physically r easonable ranges until the output variables were judged to be acceptable. It should be noted that the soil porosity and wilting point values for both W6 and S4 (see Tables 46 and 4-7) are smaller than typical values in flatwoods soils. Thes e data were determined through the model calibration process. Due to lack of actual soil information and other data (such as crop coefficient), these calibrated parameters could be biased to a certain point. They can be improved when sufficient soil data become ava ilable in the future. Upon completion of model calibration, the resulting data sets were used for independent verification simulations. Statistical and visualization t ools introduced in Chapter 2 were used to evaluate the model calibration a nd verification performance. Furthermore, the calculated statistics were compared with those from FHANTM by Sims (2004) fo r winter pastures. No comparison was made for summer pastur es since no previous modeling has been performed on any of the summer pastures. Sensitivity analysis Sensitivity analyses were conducted by changing a given input parameter by a predetermined amount, with the other inputs held at their calibrated values, and running a simulation for the period of January 1, 1998 through December 31, 2003. The sensitivity analyses were made for the calibration sites W6 and S4. Hydrologic inputs (see Tables 48 and 4-10) selected for sensit ivity analysis include soil porosity, wilting point, saturated hydraulic conductivity, restrictive layer hydraulic conductivity, Manning’s roughness coefficient, surface maximum depressional stor age, and crop coefficient, and nutrient inputs (see Tables 4-9 and 4-11) selected for sensitivity analysis include rainfall P and N

PAGE 139

118 concentrations, active P, stable P, organic hum us P, active N, stable N, and stocking and fertilizer rates. These choices were made as a result of the calibration process, reference material, and intuition gained during the pr ocess of developing this model. The sensitivity analyses were focused on two ma jor hydrologic outputs including the total surface runoff from the site and maximum wate r table depth at the measurement location, and two major nutrient outputs including the to tal P and N loads from the site throughout the simulation period. Only th e selected hydrologic input parameters were tested for the hydrologic outputs and all selected parameters were tested for the nutrient outputs. For each input parameter, simulations were conducted for the base (calibrated) value(s) and increments of +50%, +10%, -10% and -50%, one at a time. If a selected parameter could have different values by so il layer, then this parameter value was increased or decreased by a gi ven percentage simultaneously fo r all soil layers. Tables 48 to 4-11 list the ranges of values for th e selected hydrologic and nutrient parameters used in the sensitivity analyses for both cal ibration sites W6 and S4. Tables 4-12 to 4-15 show the sensitivity analysis results for W6 and Tables 4-16 to 4-19 show the same results for S4. As discussed in Chapter 2, the relative sensitivity is unitless, which provides a straightforward way to compar e sensitivities over the selected input parameters that vary in their magnitudes in response to the same output variable. The following discussions are based on the relativ e sensitivity analysis results. Figure 4-6 shows the relative sensitiviti es of total surface runoff to the six hydrologic input parameters for the 6-year si mulation period for the calibration sites W6 and S4. The sensitivity of surface runoff to soil porosity at its incr ement of -50% from the base for site W6 is not provided in Fi gure 4-6 because the model run was terminated

PAGE 140

119 since the simulated water table dropped belo w the bottom of the soil profile. This indicates that these soil porosity values are out of the “reasonable range”. Results show that the two most sensitive parameters for surface runoff at W6 and S4 are crop coefficient and soil porosity. Runoff is more se nsitive to crop coefficient for W6 while it is more sensitive to soil porosity for S4. More over, the third most sensitive parameter for surface runoff is restrictive layer hydraulic condu ctivity for both sites. The sensitivities of runoff to the rest of the selected hydrologic parameters are not significantly different from one another. It should be noted that the sensitivity of surface runoff to both crop coefficient and soil porosity are asymmetric for both sites compared with those of the rest of the parameters. In genera l, surface runoff shows similar sensitivity to these hydrologic parameters at both sites. Figure 4-7 shows the relative sensitivities of maximum water table depth to the same selected hydrologic parame ters at both sites. Maximu m depressional storage is not shown since maximum water table depth is not sensitive to this parameter. It should be noted that the magnitude of sensitivities of maximum water table depth is significantly smaller than those for total surface runoff for all parameters at both sites. This is likely due to the fact that response times for subs urface water movement are much slower than for surface water movement. Results indicate that maximum water table depth is most sensitive to soil porosity for W6, while it is most sensitive to crop coefficient for S4. Note that the sensitivity of maximum water table depth to the crop coefficient is quite asymmetric for S4. Also, it is very inte resting that maximum water table depth is significantly more sensitive to the wilting point for S4 than for W6. This is likely because S4 has relatively smaller soil porosit y values than W6 (see Tables 4-6 and 4-7),

PAGE 141

120 which implies less water availability in the soil profile of S4 for hydrological processes such as ET. An increase or decrease in wilt ing point could affect the amount of water available for ET. The remaining paramete rs, saturated hydrau lic conductivity and restrictive layer hydraulic conductivity show quite similar symmetric sensitivity patterns for both sites. The sensitivity tests of total surface runo ff and maximum water table depth to the hydrologic input parameters demonstrate th at the most sensit ive hydrologic input parameters are those that represent physical pr operties that directly affect the soil water budget in a storage-limited hydrol ogic system, i.e., crop coefficient, soil porosity, and restrictive layer hydraulic conductivity. Fu rthermore, the sensitivities of both total surface runoff and maximum water table dept h to the changes in these hydrologic parameters do not have the same magnitudes. Together with the hydrologic input parame ters, several nutrient parameters were selected to test the sensitivity of the total P and N load responses to the changes in these parameters. As shown in Figure 4-8, the magn itude of sensitivities of the P load response to the selected hydrologic parameters is si milar to that for tota l surface runoff for both sites. The crop coefficient and soil porosity appear to be the most sensitive parameters for the P loads for sites W6 and S4. Furthermore, the sensitivity of the P load to restrictive layer hydr aulic conductivity is more significan t than the rest of the hydrologic parameters for both sites. Among all of the selected nutrient parameters, the rainfall P concentration is the most sensitive nutrient pa rameter, followed by the active P, to the P load response for W6. On the contrary, there is little sensitivity of the P load response to the rainfall P concentration for S4. Instead, th e stable P is the most sensitive parameter,

PAGE 142

121 followed by the active P, to the P load res ponse for site S4. Such a difference in sensitivity of rainfall P concentration very likely results from th e extremely different background P values for the semi-improved pa sture site W6 and the improved pasture site S4. Generally, for both sites the P load response shows similar sensitivity patterns to hydrologic parameters as total surface runoff. Figure 4-9 displays the relati ve sensitivities of the N lo ad response to the selected parameters for both calibration sites. For s ite W6, the crop coefficient and restrictive layer conductivity appear to be the most sensitive hydrol ogic parameters and the porosity is also sensitive compared w ith the rest of the hydrologic parameters. Among all selected nutrient parameters, the N load response is mo st sensitive to the changes in rainfall N concentration. However, for site S4 th e surface maximum depressional storage is the most sensitive parameter of all selected parameters although the crop coefficient, restrictive layer c onductivity, and porosity are also sens itive parameters. The reason that N load shows high sensitivity to the surface maximum depressional storage for S4 is most likely from the combination of two aspect s, reducing/increasing the surface maximum depressional storage leads to more/less water available for surface runoff and the nutrient-rich pasture S4 has relatively higher surface runoff N concentration. Unlike for W6, the N load response for S4 is not very se nsitive to the rainfall N concentration. As discussed above for the P load response, this ma y be due to the fact that S4 is a nutrientrich site and almost all nutrie nt pools are much larger than th ose for W6. As a result, the rainfall nutrient concentrations may not be ma jor factors in changing the N load response. In general, it is hard to draw consistent conclusions from both calibration sites for these selected parameters because their sensi tivities appear to be site-specific. This

PAGE 143

122 implies that nonlinear relationships exist between these selected parameters and output variables, and that there may be inte ractions among the input parameters. Results and discussion The simulation results to be discussed include the annual water budgets conducted for the four selected pasture sites; an d the continuous simulation of major output variables, including water table depth, su rface runoff, P and N loads, and P and N concentrations for the two pairs of calibra tion/verification sites: W6/W7 and S4/S1. Daily time series plots as well as monthly and annual linear plot s, and daily duration curves of these major output variables were evaluated. Annual and monthly statistics comparisons of the simulated versus observed results were calculated and compared with those calculated by FHANTM (Sims, 2004). Annual summaries for rainfall, observed a nd simulated runoff, simulated lateral ground water flow, simulated ET, and simulated deep seepage are listed in Tables 4-20, 4-21, 4-22, and 4-23 for the four selected past ure sites. There was less rainfall for year 2000 than for the rest of the years during the simulation period over al l selected sites. There was relatively more rainfall for year s 2002 and 2003 for the wi nter pasture sites and for year 2002 for the summer pasture sites. The percentages in columns for rainfall, ET and deep seepage indicate the percent of annual rainfall th at these hydrological components account for. These data shows th at about 1-34% of annual rainfall converted to runoff, 9-13% to deep seepage, and the remainder to evapotranspiration. These statistics are quite close to the previous studies of pi ne-cypress flatwoods in the southeastern United States according to Riekerk and Korhnak (2000). Figure 4-10 and Figure 4-11 display the continuous simulation of groundwater table depths for the calibration site W6 a nd verification site W7 from September 2000

PAGE 144

123 through December 2003. Very good agreement on timing and trends are found between the predicted and observed wate r table depths for both sites W6 and W7, with slightly larger discrepancies during a long dry peri od approximately from October 2000 through March 2001. During this dry period, the observe d water table is very sensitive to small rainfall inputs so that observed water table fl uctuated more rapidly to the rainfall inputs than the model predictions. Additionally, it is obvious that the measured water tables did not reach the ground surface throughout the si mulation period even during the heavy rainfall inputs when surface runoff occurred This error may come from erroneous measuring of the ground surface elevations of wells, which may have been installed in low spots in the field and thus the surface elev ations measured at such locations may not be an appropriate estimate of the surface el evations for the entire land segment. The conclusions drawn from the winter past ure sites also apply to the simulation of groundwater table depths for the summer past ure sites S4 and S1 (see Figures 4-12 and 413). However, there is a better agreement between the predicted and observed water table depths when the soils get saturated for the site S4 than for the other three sites. The predicted water tables agree with the observed data better in the two calibration sites than in the two verification sites. Figures 4-14 and 4-15 show the continuous simulation of surface runoff from July 1998 to December 2003 for the calibration site W6 and verification site W7. For both sites, the model responded fairly well to th e rainfall events on both timing and trends, and also was able to capture some of the back flows that were caused when the water elevation in the downstream canal was higher than that in the pasture sites. Nevertheless, the model generally did not capture the peak runoff very well. A few possible reasons

PAGE 145

124 may explain this. First, the ditches located in side the pastures are deeper than the area’s ground surface elevation. Thus they can coll ect the surface water and result in runoff at the outlet of the pasture even when no sheet flow occurs. The current version of the model does not consider water movement in these ditches. In the model surface runoff occurs only when surface water is ponded on the entire land segment. Second, the observed runoff is instantaneously measured da ta while the predicted runoff is the daily average value. Using a daily time step may computationally delay the transport of runoff to a certain extent. For exam ple, if rainfall events last only a few hours, the resulting overland flow could reach the downstream flume in less than a day, especially if rainfall occurs near the flume where the runoff was measured. In the m odel, runoff is only transported between land segments once a day. Therefore, the simulated overland flow could be delayed with a resu lting flattened hydrograph. Similar conclusions can be drawn for th e summer pasture sites S4 and S1 (see Figures 4-16 and 4-17). However, the magnitude of surface runoff in the summer pastures is generally smaller than for the wi nter pastures, which re duces the discrepancy in peak runoffs to a certain extent. Figures 4-18 and 4-19 show the continuous simulation of P loads from July 1998 through December 2003 for W6 and W7, respect ively. Similar to the surface runoff simulation, the predicted P loads agree well with the observed P loads on timing and trend, but the model underestimates the peak P loads. Since the load is the product of the volume of surface runoff and P concentrati on, the underestimation of surface peak runoffs, as discussed earlier, di rectly contributes to the unde r-predictions of P loads. However, Figures 4-26 and 4-27 show that underestimation of P concentrations,

PAGE 146

125 especially for those times when peak r unoff was underestimated, is likely another contributor to the underestimation of peak P loads for winter pa stures W6 and W7. Figures 4-20 and 4-21 display the conti nuous simulation of P loads from July 1998 through December 2003 for the two summer past ure sites S4 and S1, respectively. The conclusions from these figures are similar to t hose obtained for the wint er pastures sites. However, the magnitude of the difference in the prediction of P peak load in these summer pastures is more significant than for th e winter pastures due to the fact that the summer pastures have high background nutrients. Figures 4-22 and 4-23 show the continuous simulation of N load from July 1998 through December 2003 for W6 and W7, respectivel y. The predicted N load for sites W6 and W7 agree fairly well with the observed N loads, but the model still underestimates the peak N loads as it did the P loads for thes e sites. Compared with the simuled N loads, the observed N loads show higher peaks and more variability. Likewise, the model underestimates, as seen in Figures 4-24 and 425, most of the peak observed N loads for the two summer pasture sites S4 and S1, respectively. Figures 4-26 through 4-33 show the con tinuous simulation of surface runoff P and N concentrations for the four se lected pasture sites. These figures were prepared to help understand the P and N load prediction errors. For winter site W6 both observed P and N concentrations have higher peaks and more variability than simulated as shown in Figures 4-26 and 4-28. However, the magnit ude of both P and N concentrations for W7 as seen in Figures 4-27 and 4-29 is smaller than that for W6, which leads to less discrepancy. Compared with the winter past ure sites, the model pe rformed much better in predicting the P concentrations for S4 and S1 with a better agreement between the

PAGE 147

126 predictions and observations as shown in Figures 4-30 and 431. Nevertheless, the model generally overestimated the N concentrations for both summer pasture sites S4 and S1 as shown in Figures 4-32 and 4-33. As discussed above, the underestimati ons or overestimations in P and N concentrations directly contribute to the prediction errors of the P and N loads. It should be noted that the measured concentration va lues were from water samples collected at certain moments and locations near the flumes of pasture sites, while the simulated concentration values are daily values which represent the surface water P and N concentrations over the entire land segment. In addition, lack of the nutrient input information may also contribute to the suboptim al predictions of P and N concentrations. Nevertheless, those simulation results imply th at nutrient cycling algorithms need further improvement when sufficient and reliab le nutrient data are available. Judging the model performance solely on the visual observation of the time series simulation can be subjective, especially when the simulation period spans multiple years. In order to quantify the mode l performance, a few statisti cs, including bias, RE, RMSE, CV, R2, and NS as introduced in Chapter 2, were calculated using the predicted and observed data. Tables 4-24, 4-25, 4-26, a nd 4-27 show the comparisons between the statistics obtained from the simulation results of the modified ACRU2000 model and the lumped FHANTM model performed by Sims (2004) for both the winter calibration site W6 and verification site W7. As shown in Tables 424 and 4-25, the statisti cs indicate that the modified ACRU2000 model is mo re accurate than FHANTM in predicting monthly and annual surface runoff and N load with smalle r bias, RE, RMSE, and CV values obtained

PAGE 148

127 from the modified ACRU2000 model than from FHANTM. All sta tistics except RE for the P load obtained from the modified model are also smaller than FHANTM and indicate a better prediction in P load. Bo th models overestimated monthly and annual surface runoff and N load with the positive bias and RE values. The modified model overestimated the P load while FHANTM unde restimated the P load. Both models overestimated the N load. Smaller RMSE and CV values over all predicted output variables from the modified ACRU2000 indi cate that the modifi ed model generally performed better than FHANTM. For W6, the monthly CV values are generally greater than 1.0, which indicates that the model is adequate for screening purpose for monthly predictions (Hedden, 1996). The annual CV va lues are all less th an one indicating the model is accurate enough to be used for site-specific applic ation for annual predictions. Similar conclusions regarding the monthly and annual statistics can be applied to surface runoff and P load (see Tables 4-26 and 4-27) for the verification site W7 as well drawn for calibration site W6. Both models performed simila rly in predicting the N load with quite good statistics. Unlike for W6 the modified model underestimated surface runoff, and P and N loads. FHANTM overes timated the runoff but underestimated the P and N loads. The monthly and annual CV va lues for runoff indicate the model can be used for site specific monthly and annual pred ictions of runoff. The CV value for the P and N loads indicate that the model can only be used for screening application at the monthly level, but can be used for site specif ic application at the a nnual level. Overall, the comparison of statistics between the two models indicates that the modified ACRU2000 performed better than FHANTM fo r both winter sites according to the monthly and annual predictions.

PAGE 149

128 No statistics for the water table depths were obtained for the FHANTM simulation. For the modified ACRU2000 model, the mont hly NS value of 0.93 and annual NS of 0.96 for the calibration site W6, and the mont hly NS of 0.81 and annual NS of 0.86 for the verification site indicate a very good overall performan ce in predicting water table depths. CV values for both sites for both m onthly and annual predicti ons indicate that the model can be used for site-specific applic ation in predicting th e water table depth. Comparing the statistics between the winter calibration and verification sites, Tables 4-20 through 4-23 show that the model performed s lightly better in predicting surface runoff for the verificat ion site with smaller bias RE, RMSE, CV values and higher R2 and NS value for the verification site than the calibration site. For the other three output variables, the model performed better for the calibration site than for the verification site with smaller bias RE, RMSE, CV values and higher R2 and NS values for the calibration site. Since RE, CV, R2 and NS are unitless, it is possible to compare the statistics over the output va riables. Generally, the statistics indicate that the model performed better in predicting hydrologic variables than nutrient variables. There are no statistics available from FHANTM for the summer pastures, thus the evaluation of model performance in the su mmer pastures was ma de by comparing the statistics (see Tables 4-28, 429, 4-30, and 4-31) with those from the winter pastures. As seen in Tables 4-28 and 4-29, the compar isons of monthly st atistics for the two calibration sites W6 and S4 indicate the predic tions of four output variables are slightly more biased for S4 than for W6 but the ge neral model performan ce for both calibration sites are close. For both sites the model pe rformed better in pr edicting the monthly P loads than in the monthly N loads. When comparing the annual st atistics for these two

PAGE 150

129 calibration sites, the bias for all predicted variables is more apparent for S4 than for W6. CV values for runoff, water table depth, and P load indicate that the model can be used for site-specific monthly and a nnual prediction of these variab les. The annual CV values for all variables indicate the model can be us ed for site-specific annual prediction for all variables. For the two verification sites, W7 and S1, the comparison of monthly statistics (see Tables 4-26 and 4-27) indi cates that the model performe d better for runoff and water table depth for W7, but it underestimated bot h P and N loads for W7 and overestimated them for S1. The same conclusion can be dr awn for both sites by comparing their annual statistics for these variables. Overall, the model performed better for the calibration sites than for the verification sites, better for hydrologic out puts than for nutrient outputs, and better for winter pasture sites than for summer pasture sites. Accord ing to the CV values, the model can be used for site-specific annual predic tion of both hydrologic and nutrien t transport and should be used primarily for screening appl ication at the monthly level. Linear plots provide a good tool to examine the agreem ent between the simulated and observed data and to determine over what ranges these data agree more closely with each other. The monthly and annual linear plot s for the four major output variables from the four selected pasture site s are provided in Figures 434 through 4-49. It should be noted that the R2 values in the linear plots indicate the quality of the linear fit between data points, but do not necessa rily indicate how well the simu lated values agree with the observed values (i.e. the quali ty of the 1:1 line fit). Excellent agreement between the monthly and annual surface runoff is shown in the linear plots for the winter past ure sites W6 and W7 in Figures 4-34 and 4-35. These data

PAGE 151

130 fall fairly close to the 1:1 line that indicates a perf ect fit. The monthly and annual linear plots of water table depths for W6 (see Figure 4-36) also indicate excellent model performance. However, slight underestimations of water table depths in the lower range were observed for the verifi cation site W7 (see Figure 437). A fairly good agreement between the simulated and measured P loads are shown in Figure 4-38 for site W6, while significant underestimations of P load in the high range are shown in Figure 4-39 for site W7. This bias is primarily caused by two ex treme points and most of the other points are scattered around the 1:1 line. These extrem e points may result from the underestimations of the peak P loads as discussed earlier. As shown in Figure 4-40, the model predicted the monthly and annual N loads quite well for s ite W6 with some overestimations in the higher ranges. However, larger underestimati on of N loads throughout the entire range of values was observed for site W7 as shown in Figure 4-41. Observed N loads for the verification W7 were obviously hi gher than those for the calib ration site W6. This fact increased the difficulty of model calibration. For the summer calibration site S4, the linear plot (see Figure 4-42) indicates a fairly good match between the simulated a nd observed monthly runoff and a general overestimation of annual runoff. A similar c onclusion can be drawn from the linear plot (see Figure 4-43) for the verifi cation site S1, but the bias is a slightly more pronounced than for S4. A good match is shown in the lin ear plot (see Figure 4-44) of monthly and annual water table depth for the calibration site S4, but larger underestimation of monthly and annual water table depth in the lower rang e were observed from the plot for S1 (see Figure 4-45). A fairly good agreement betw een the simulated and observed P loads is observed in the linear plot for site S4 (see Figure 4-46), whil e a slightly larger

PAGE 152

131 underestimation in the upper range of the P loads was found for S1 (see Figure 4-47). For the N loads, the match between the simula ted and observed data for S4 are quite good with a little underestimation at its upper range (see Figure 448). A larger bias in both upper and lower ranges was observed in the linea r plot of N loads for S1 (see Figure 449). Daily duration curves offer an alternative way to investigate the agreement between daily predicted and measured values, and part icularly to determin e over what range of values the simulated output variables of in terest match their corresponding observations more closely. Figure 4-50 displays the daily duration curves of surface runoff and water table depth for the winter calibration site W6. The duration curve of surface runoff indicates that the simulated daily runoff agrees very well with the observed runoff throughout the simulation period. The duration curve of water table depths indicates that the simulated water table depths are greater than the observed values at the higher and middle range of water table depths, and less th an the observed values in the lower middle range of water table depths. The durati on curves for both P and N loads for the calibration site W6 (see Figure 4-51) are quite consistent except for high P and N loads when the model underpredicts the response. Figure 4-52 displays the daily duration curves of surface runoff and water table depth for the winter verification site W7. The duration curves for surface runoff for the site W7 show a very good match. However, the duration curve of water table depths for W7 sh ows significant discrepa ncies at the higher range of water table depth. As for W6, the duration curve for P and N load for W7 show that the model signif icantly underpredicts th e observed load.

PAGE 153

132 Figure 4-54 displays the daily duration cu rves of surface runoff and water table for the summer calibration site S4. The diffe rences between the predicted runoff and observed runoff primarily exis t at the higher and middle ra nge of values. The model underestimates the higher range and overestimates the middl e range but generally these differences are not very significant. Interest ingly, the duration curve of the water table depth for the site S4 indicates that the m odel overpredicted the observed data from the middle to upper range and underp redicted the rest of the range, and the differences between the simulated and observed values are generally more significant in the overestimation range than in the underestima tion range. Figure 4-55 shows the duration curves of P and N loads for the calibration site S4. The trends and timing of the duration curve for the P loads are similar to the prev ious ones for the winter pasture sites. However, the model obviously performed be tter in predicting the duration curve for N loads with those for P loads. Figure 4-56 il lustrates the daily durat ion curves of surface runoff and water table depth for the summer ve rification site S1. Compared with the previous duration curves for surface runoff, the model performed be tter for S1, with a close match between the predicted surface runoff and measured runoffs throughout the entire range of values. The duration curve for the water ta ble depth indica tes that the model overestimated the water table thr ough the entire simulation period with the predicted duration curve consistently above the observed one. Figure 4-57 shows the duration curves of P and N loads for the veri fication site S1. The conclusions for the P loads drawn from its duration curve (see Figure 4-57) are similar to the previous ones for the other three sites. However, a better ma tch between the prediction and observation is observed at the lower range of P loads. Comp ared with the duration curve of N loads for

PAGE 154

133 S4, the one for S1 is fairly good, with much closer agreement betw een the predicted and observed N loads throughout the entire simulation period. In general, the model performance in pr edicting surface runoff and water table is better than for P and N loads for all the select ed pasture sites. Furthermore, the model did a better job in predicting output variables for the two calib ration sites W6 and S4 than for the two verification sites. This difference is particularly appare nt in P and N loads, and could be due to several factors: first, nut rient predictions inheri t the errors resulting from the hydrologic model; second, the number of nutrient observations are quite limited during the 6-year simulati on period compared with th e hydrologic observations, and therefore provide a smaller range of data for ca librating the nutrient parameters; third, the generally higher P and N loads observed in the verification sites than in the calibration sites increased the difficulty in model calibra tion; finally, insufficiency and unavailability of input data, such as limited N and P observations throughout the 6-year simulation period, may have led to biased simulations. Concluding Remarks In this chapter, a nutrient model capable of multi-directional spatial transport and transformations of N, P and conservative solute was developed based on the existing nutrient module ACRU-NP (Campbell et al ., 2001) in the ACRU 2000 modeling system. In order to investigate the accuracy of th e conservative solute transport component, a test using the hypothetical sc enario introduced in Chapter 3 together with new solute input was conducted by comparing with the particle tracking m odel PMPATH (Chiang and Kinzelbach, 2005). The test ing results indicate the so lute transport predicted by ACRU2000 is in reasonable agr eement with the predictions by PMPATH. Due to the constraints of PMPATH in simulating particle s instead of solute concentrations, it is

PAGE 155

134 difficult to quantitatively judge the performa nce of the conservative solute transport model in the modified ACRU2000. To evaluate the complete performance of the nutrient model within the framework of the previously modified hydrologic model in Chapter 3, an applic ation to beef cattle pastures in Buck Island Ranch, Lake Ok eechobee Basin, Florida was conducted. A complete model testing pr ocedure including model calib ration, verification, and sensitivity analysis was conducted. The wate r budget for each sele cted site was also determined to investigate whether the si mulated major hydrologic components fall in reasonable ranges by comparing with the literatu re values for similar regions. Statistics were calculated to quantitatively evalua te the model performance for both model calibration and verification. Two pairs of pasture sites we re selected, which differ in stocking rotation schedule, so il texture, land cover, a nd management activities. Generally, the predicted monthly and annual surface runoff, water table depth, and P and N loads are fairly promising in all these past ures. The model performance is better in predicting runoff and water tabl e depth than P and N loads, better for the calibration sites than for the verification sites, and better for the winter pasture sites than for the summer sites. The sensitivity analyses demonstrate that crop coefficient, porosity and restrictive layer hydraulic conductivity are the most sensitive input parameters for hydrologic responses among all selected hydro logic inputs, and they are al so sensitive parameters for both P and N load responses. However, the sensitivities of nutrient parameters are sitespecific due to different soil texture and bac kground nutrients. The winter pasture sites have low background nutrients in soils, and th erefore P and N loads are most sensitive to

PAGE 156

135 rainfall P and N concentrations, respectively. On the contrary, the summer pasture sites have high background soil nutrient concentrations and the anal yses results show that P and N loads are not sensitive to rainfall P and N concentrations, nutrient pools including active P and stable P for P loads and active N for N loads. Surprisingly, stocking and fertilizer rates are not sensitive parameters for either P or N responses. This seems to imply that the residual so il nutrient mass/concentrations are responsible for runoff nutrient outputs from these pasture sites, rath er than agricultural activities such as stocking and fertilization. However, this ex periment only lasted for a period of 6 years and the long-term impacts from stocking and fe rtilization may need to be investigated. Additionally, one issue rising from the application to Buck Island is the insufficient nutrient data that was available to calib rate the nutrient parameters. Using instantaneously measured nutrient concentratio ns to calculate the nut rient loads may have caused bias of the observed loads. Furthe rmore, the limited number of nutrient observations may not adequately cover the en tire range of load occurrences, which further challenges the calibration of nutrien t parameters. Moreove r, a common problem existed in the measured nutrient data for both pairs of pasture sites, i.e., for both calibration sites with cattle stocking, the obs erved nutrient loads were lower than those for their paired verification sites without cattle stocking. This makes the model calibration even more difficult when in itial nutrient pools are not well known. Methods for partitioning nutrients between the soil and water phases so that the soil nutrients can be transported between ponde d surface water and soil water may need further improvement with respec t to flatwoods soils. In or der to do this, more accurate measured nutrient data are required. In a ddition, the nutrient load s carried by sediments

PAGE 157

136 are not well considered in this multi-dire ctional spatial simulation model since the sediment transport in Florida flatwoods soils is not as significant as flow transport. Nevertheless, it would enhan ce the model capabilities if im proved sediment transport capacities were added in the fu ture. Finally, the method for dealing with the nutrients released from the waste decomposition, the spatial distribution of cattle waste, and assuming only one plant species on each land segment is oversimplified. Future work should be done to represent these processes more accurately. In general, it has been shown that th is model is capable of simulating both hydrology and nutrient dynamics in field-scal e catchments particularly for screening purposes. Thus, it can be used as a basi s for coupling with a vegetation dynamic simulation model to form a more comp lete ecohydrological modeling system.

PAGE 158

137 Table 4-2. List of stocking activit ies for pastures S1, S4, W6 and W7*. Stocking Rates Timing Management Practices S1 S4 W6 W7 October 21, 1998 Moved cattle into winter pastures 0 15 15 0 November 4, 1998 Moved cattle from winter into summer pastures 0 15 15 0 February 3, 1999 Moved cattle from summer into winter pastures 0 15 15 0 April 13, 1999 Moved cattle from winter into summer pastures 0 15 15 0 December 2-3, 1999 Moved cattle from summer into winter pastures 0 15 15 0 April 6, 2000 Moved cattle from winter into summer pastures 0 15 15 0 November 29, 2000 Moved cattle from summer into winter pastures 0 15 15 0 May 10, 2001 Moved cattle from winter into summer pastures 0 15 15 0 October 23, 2001 Moved cattle from summer into winter pastures 0 15 15 0 May 28, 2002 Moved cattle from winter into summer pastures 0 15 15 0 December 2, 2002 Moved cattle from summer into winter pastures 0 15 15 0 April 15, 2003 Moved cattle from winter into summer pastures 0 15 15 0 October 31, 2003 Moved cattle from summer into winter pastures 0 15 15 0 *Stocking information including dates, activities, and rates were obtained from the report (MAERC, 2004). Table 4-3. List of fertilization activities for pastures S1 and S4*. Timing Management Practices Application May 1-9, 2000 Applied ammonium nitrogen fertilizer 56kg/ha April 24, 2001 Applied ammonium nitrogen fertilizer 56kg/ha March 28-29, 2003 Applied ammonium nitrogen fertilizer 56kg/ha *Fertilization activity information was obtained from the report (MAERC, 2004). Table 4-4. List of burn activities for pastures S1, S4, W6 and W7*. Timing Management Practices November 24-24, 1998 Prescribed burn in all winter pastures February 3, 1999 Prescribed burn in all summer pastures March 1, 2000 Accidental fire occurred in W6 April 5, 2000 Accidental fire occurred in W7 February 11-12, 2002 Prescribed burn in all winter pastures April 15-18, 2002 Prescribed burn in all summer pastures *Burn activity information was obtained from the report (MAERC, 2004). Table 4-5. Percent of area occupied by di fferent soil series and wetlands in selected summer and winter pastures*. Pasture Area Area Mapped Soils Felda Fine Sand Felda Fine Sand Bradenton Fine Sand Gator Muck Pineda Fine Sand Pineda Fine Sand Tequesta Muck Percentage of wetlands based on area of muck soils (unit) (ha) (ha) (%) (%) (%) (%) (%) (%) (%) (%) S1 22.0 21.6 67.0 11.3 0.7 3.1 17.9 20.9 S4 20.5 20.2 92.9 7.1 7.1 W6 32.1 32.1 38.6 59.7 1.7 1.7 W7 30.2 30.2 1.2 94.5 4.3 4.3 *Table is cited from MAERC (2004).

PAGE 159

138 Table 4-6. Model input pa rameters for the winter pastures W6 and W7. Value Parameters Unit layer 1 layer 2 layer 3 layer 4 laye r 5 layer 6 layer 7 layer 8 layer 9 Soil layer deptha m 0.01 0.09 0.26 0.40 0.18 0.18 0.10 0.43 0.38 Bulk densitya g/cc 1.22 1.22 1.56 1.59 1.61 1.63 1.62 1.74 1.81 Base saturationa % 67 67 66 82 46 45 47 56 63 Porosityb m/m 0.32 0.32 0.12 0.14 0.24 0.22 0.25 0.35 0.37 Wilting pointb m/m 0.02 0.02 0.01 0.02 0.04 0.04 0.03 0.06 0.07 Field capacityb m/m 0.11 0.11 0.05 0.06 0.11 0.10 0.11 0.16 0.17 pHa 6.1 6.1 5.9 6.0 5. 7 5.8 5.5 4.2 4.3 Saturated hydraulic conductivityb m/s 3.3E04 3.3E04 4.3E04 4.4E04 1.3E04 1.3E04 1.6E04 2.3E05 6.7E06 Res. conductivityd m/s 8.5E-10 Manning’s coefficientb m1/3/s 0.1 Brooks-Corey hc cm 1.63 1.63 4.28 2.45 17.54 9.69 1.02 17.74 34.16 Brooks-Corey c 0.54 0.54 0.80 0.56 0. 95 0.63 0.37 0.51 0.46 Brooks-Corey c m/m 0.02 0.02 0.02 0.03 0.04 0.04 0.02 0.05 0.07 Exponent for upward fluxc 3.62 3.62 4.4 3.68 4. 85 3.89 3.11 3.53 3.38 Surface maximum depression storage m 0.005 Crop coefficientb 0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 1.0, 0.95, 0.90, 0.85, 0.80, 0.75 for 12 months Silt contenta % 0.9 0.9 0.8 3.0 3.5 4.8 3.0 4.1 4.0 Clay contenta % 2.0 2.0 0.6 0.5 3.7 3.7 2.4 12.3 18.6 Mg contenta mg/kg 49 49 5 4 4 4 4 4 4 AL contenta mg/kg 466 466 213 1465 1465 1465 1465 1465 1465 Organic mattera % 2.17 2.17 0.14 0.14 0.24 0.24 0.19 0.14 0.09 Rainfall P & N concentrationb mg/l 0.020 for nitrate N and 0.028 for labile P Stream nutrient concentrationc mg/l 0.013 for nitrate, 0.006 for ammonium, and 0.008 for labile P Labile Pc kg/ha 0.009 0.008 0.025 0. 039 0.018 0.018 0.010 0.045 0.042 Active Pc kg/ha 0.99 8.93 2.59 2. 60 3.70 3.55 2.33 30.53 13.98 Stable Pc kg/ha 2.65 23.82 6.92 6. 94 9.87 9.47 6.22 81.42 37.27 Organic humus Pc kg/ha 2.21 19.85 40.47 76. 27 34.76 35.19 19.43 89.73 82.49 Fresh organic Pc kg/ha 0.005 0.045 0.130 Nitrate-Nc kg/ha 0.001 0.011 0.041 0. 064 0.029 0.029 0.016 0.075 0.069 Ammonium-Nc kg/ha 0.002 0.022 0.081 0. 127 0.006 0.006 0.003 0.150 0.138 Active Nc kg/ha 7 66 20 32 14 15 8 37 34 Stable Nc kg/ha 293 2635 811 1272 580 587 324 1496 1376 Fresh organic Nc kg/ha 0.05 0.45 1.3 Residue biomassc kg/ha 10 90 260 Manure app. ratee tons/ha 0.003475 tons/ha Fertilizer ratee kg/ha aValues are referred to Pineda fine sand in Carlisle et al. (1989); bValues are calibrated; cValues are initially estimated using the procedures described by Fraisse and Campbell (1997) and further calibrated; dValue is the calibrated restrictive layer hydraulic conductivity; and evalue is estimated accord ing to MAERC (2004).

PAGE 160

139 Table 4-7. Model input pa rameters for the summer pastures S4 and S1. Value Parameters Unit layer 1 layer 2 layer 3 layer 4 layer 5 layer 6 layer 7 layer 8 layer 9 layer 10 Soil layer deptha m 0.01 0.08 0.27 0.17 0. 08 0.30 0.26 0.28 0.28 0.30 Bulk densitya g/cc 1.29 1.29 1.55 1.59 1. 64 1.52 1.66 1.66 1.66 1.66 Base saturationa % 14 14 27 38 69 86 82 99 99 99 Porosityb m/m 0.31 0.31 0.12 0.12 0. 15 0.12 0.14 0.18 0.29 0.30 Wilting pointa m/m 0.03 0.03 0.02 0.02 0. 02 0.02 0.02 0.02 0.02 0.03 Field capacityb m/m 0.08 0.08 0.05 0.05 0. 06 0.05 0.06 0.06 0.06 0.07 pHa 3.8 3.8 4.6 5.1 5.1 5.7 7.6 7.8 7.8 7.6 Saturated hydraulic conductivityb m/s 1.27E -03 1.27E -03 1.44E -03 1.81E -03 2.35E -06 3.10E -06 3.25E -06 3.90E -06 3.90E -06 2.15E05 Res. conductivityd m/s 8.5E-10 Manning’s coefficientb m1/3/s 0.1 Brooks-Corey hc cm 1.43 1.43 2.14 1.02 9. 89 13.6 4.59 3.47 3.47 10.2 Brooks-Corey c 0.56 0.56 0.56 0.46 0.63 0.48 0.44 0.45 0.47 0.63 Brooks-Corey c m/m 0.02 0.02 0.02 0.02 0. 04 0.07 0.03 0.02 0.02 0.04 Exponent for upward fluxc 3.68 3.68 3.68 3.38 3.89 3.44 3.32 3.35 3.41 3.89 Surface maximum depression storage m 0.005 Crop coefficientb 0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 1.0, 0.95, 0.90, 0.85, 0.80, 0.75 for 12 months Silt contenta % 1.3 1.3 1.6 2.0 3. 7 5.9 4.6 3.2 3.2 3.0 Clay contenta % 1.1 1.1 0.3 0.6 5. 2 15.5 4.6 3.1 3.1 5.6 Mg contenta mg/kg 151 151 62 62 62 1302 1302 1302 1302 1302 AL contenta mg/kg 119 119 6 6 6 62 62 62 62 62 Organic mattera % 2.29 2.29 0.17 0.12 0. 31 0.14 0.05 0.12 0.12 0.10 Rainfall P & N concentrationsb mg/l 0.020 for nitrate N and 0.028 for labile P Stream nutrient concentrationsc mg/l 0.013 for nitrate N, 0.006 for a mmonium N, and 0.008 for labile P Labile Pc kg/ha 0.636 10.81 0.90 0.85 0.40 1.506 1.305 1.406 1.406 1.506 Active Pc kg/ha 140 2379 198 2612 288 506 196 127 127 152 Stable Pc kg/ha 636 1081 2 900 1187 5 1308 2301 891 576 576 689 Organic humus Pc kg/ha 689 1172 0 5630 5450 2650 8517 8061 8681 8681 9302 Fresh organic Pc kg/ha 1.87 3.36 6.72 Nitrate-Nc kg/ha 0.026 0.044 0.056 0.054 0.026 0.091 0.086 0.093 0.093 0.099 Ammonium-Nc kg/ha 0.001 0.022 0.028 0.027 0.013 0.046 0.043 0.047 0.047 0.050 Active Nc kg/ha 290 4934 1255 1216 590 1710 1619 1743 1743 1868 Stable Nc kg/ha 1857 3957 9 8036 7785 3779 1094 4 1035 9 1115 6 1115 6 11952 Fresh organic Nc kg/ha 1.33 22.6 17.8 Residue biomassc kg/ha 100 1700 1800 Manure app. ratee tons/ha 0.003475 tons/ha Fertilizer ratee kg/ha 9.24 kg/ha for nitrate and 9.24 kg/ha for iN aValues are referred to Felda fine sand in Carlisle et al. (1989); bValues are calibrated; cValues are initially estimated using the procedures described by Fraisse and Campbell (1997) and further calibrated; dValue is the calibrated restrictive layer hydraulic conductivity; and evalue is estimated accord ing to MAERC (2004).

PAGE 161

140 Table 4-8. Selected hydrologic paramete rs for sensitivity analysis for W6. Testing Range Parameters Unit -50% -10% Base 10% 50% Layer 1 0.160 0.288 0.320 0.352 0.480 Layer 2 0.160 0.288 0.320 0.352 0.480 Layer 3 0.060 0.108 0.120 0.132 0.180 Layer 4 0.070 0.126 0.140 0.154 0.210 Layer 5 0.120 0.216 0.240 0.264 0.360 Layer 6 0.110 0.198 0.220 0.242 0.330 Layer 7 0.125 0.225 0.250 0.275 0.375 Layer 8 0.175 0.315 0.350 0.385 0.525 Porosity Layer 9 m/m 0.185 0.333 0.370 0.407 0.555 Layer 1 0.010 0.018 0.020 0.022 0.030 Layer 2 0.010 0.018 0.020 0.022 0.030 Layer 3 0.005 0.009 0.010 0.011 0.015 Layer 4 0.010 0.018 0.020 0.022 0.030 Layer 5 0.020 0.036 0.040 0.044 0.060 Layer 6 0.020 0.036 0.040 0.044 0.060 Layer 7 0.015 0.027 0.030 0.033 0.045 Layer 8 0.030 0.054 0.060 0.066 0.090 Wilting point Layer 9 m/m 0.035 0.063 0.070 0.077 0.105 Layer 1 1.65 2.97 3.30 3.63 4.95 Layer 2 1.65 2.97 3.30 3.63 4.95 Layer 3 2.15 3.87 4.30 4.73 6.45 Layer 4 2.20 3.96 4.40 4.84 6.60 Layer 5 0.65 1.17 1.30 1.43 1.95 Layer 6 0.65 1.17 1.30 1.43 1.95 Layer 7 0.80 1.44 1.60 1.76 2.40 Layer 8 0.115 0.207 0.230 0.253 0.345 Saturated hydraulic conductivity Layer 9 m/s 10-4 0.034 0.060 0.067 0.074 0.101 Restrictive layer hydraulic conductivity m/s 10-9 0.425 0.765 0.850 0.935 1.28 Manning’s roughness coefficient m1/3/s 0.05 0.09 0.10 0.11 0.15 Surface maximum depressional storage m 10-2 0.25 0.45 0.50 0.55 0.75 January 0.35 0.63 0.70 0.77 1.05 February 0.38 0.68 0.75 0.83 1.13 March 0.40 0.72 0.80 0.88 1.20 April 0.43 0.77 0.85 0.94 1.28 May 0.45 0.81 0.90 0.99 1.35 June 0.48 0.86 0.95 1.05 1.43 July 0.50 0.90 1.00 1.10 1.50 August 0.48 0.86 0.95 1.05 1.43 September 0.45 0.81 0.90 0.99 1.35 October 0.43 0.77 0.85 0.94 1.28 November 0.40 0.72 0.80 0.88 1.20 Crop coefficient December 0.38 0.68 0.75 0.83 1.13

PAGE 162

141 Table 4-9. Selected nutrient paramete rs for sensitivity analysis for W6. Testing Range Parameters Unit -50%-10%Base 10% 50% Rainfall P concentration mg/l 0.014 0.025 0.028 0.031 0.042 Rainfall N concentration mg/l 0.010 0.018 0.020 0.022 0.030 Layer 1 0.500.890.99 1.09 1.49 Layer 2 4.47 8.04 8.93 9.83 13.40 Layer 3 1.30 2.34 2.59 2.85 3.89 Layer 4 1.30 2.34 2.60 2.86 3.90 Layer 5 1.85 3.33 3.70 4.07 5.55 Layer 6 1.77 3.19 3.55 3.90 5.32 Layer 7 1.17 2.10 2.33 2.56 3.50 Layer 8 15.27 27.48 30.53 33.58 45.80 Active P Layer 9 kg/ha 6.99 12.58 13.98 15.38 20.97 Layer 1 1.322.382.65 2.91 3.97 Layer 2 11.91 21.44 23.82 26.21 35.74 Layer 3 3.46 6.23 6.92 7.61 10.38 Layer 4 3.47 6.25 6.94 7.64 10.41 Layer 5 4.93 8.88 9.87 10.85 14.80 Layer 6 4.73 8.52 9.47 10.41 14.20 Layer 7 3.11 5.59 6.22 6.84 9.32 Layer 8 40.71 73.28 81.42 89.56 122.13 Stable P Layer 9 kg/ha 18.64 33.55 37.27 41.00 55.91 Layer 1 1.101.992.21 2.43 3.31 Layer 2 9.9317.8719.85 21.84 29.78Layer 3 20.2336.4240.47 44.51 60.70Layer 4 38.1468.6576.27 83.90 114.41Layer 5 17.3831.2834.76 38.23 52.13Layer 6 17.5931.6735.19 38.71 52.78Layer 7 9.7117.4919.43 21.37 29.14Layer 8 44.8780.7689.73 98.70 134.60 Humus P Layer 9 kg/ha 41.2474.2482.49 90.74 123.73Layer 1 3.666.597.32 8.05 10.98 Layer 2 32.94 59.29 65.88 72.47 98.82 Layer 3 10.14 18.25 20.28 22.31 30.42 Layer 4 15.90 28.62 31.80 34.98 47.70 Layer 5 7.25 13.04 14.49 15.94 21.74 Layer 6 7.34 13.20 14.67 16.14 22.01 Layer 7 4.05 7.29 8.10 8.91 12.15 Layer 8 18.71 33.67 37.41 41.15 56.12 Active N Layer 9 kg/ha 17.20 30.95 34.39 37.83 51.59 Layer 1 146.5263.7293.0 322.3 439.5 Layer 2 1317.5 2371.5 2635.0 2898.5 3952.5 Layer 3 405.5 729.9 811.0 892.1 1216.5 Layer 4 636.0 1144.8 1272.0 1399.2 1908.0 Layer 5 290.0 522.0 580.0 638.0 870.0 Layer 6 293.5 528.3 587.0 645.7 880.5 Layer 7 162.0 291.6 324.0 356.4 486.0 Layer 8 748.0 1346.4 1496.0 1645.6 2244.0 Stable N Layer 9 kg/ha 688.0 1238.4 1376.0 1513.6 2064.0 Manure application rate tons/ha 0.00174 0.00313 0.00348 0.00382 0.00521

PAGE 163

142 Table 4-10. Selected hydrologic paramete rs for sensitivity analysis for S4. Testing Range Parameters Unit -50% -10% Base 10% 50% Layer 1 0.155 0.279 0.310 0.341 0.465 Layer 2 0.155 0.279 0.310 0.341 0.465 Layer 3 0.060 0.108 0.120 0.132 0.180 Layer 4 0.060 0.108 0.120 0.132 0.180 Layer 5 0.075 0.135 0.150 0.165 0.225 Layer 6 0.060 0.108 0.120 0.132 0.180 Layer 7 0.070 0.126 0.140 0.154 0.210 Layer 8 0.090 0.162 0.180 0.198 0.270 Layer 9 0.145 0.261 0.290 0.319 0.435 Porosity Layer 10 m/m 0.152 0.273 0.303 0.334 0.455 Layer 1 0.015 0.027 0.030 0.033 0.045 Layer 2 0.015 0.027 0.030 0.033 0.045 Layer 3 0.010 0.018 0.020 0.022 0.030 Layer 4 0.010 0.018 0.020 0.022 0.030 Layer 5 0.010 0.018 0.020 0.022 0.030 Layer 6 0.010 0.018 0.020 0.022 0.030 Layer 7 0.010 0.018 0.020 0.022 0.030 Layer 8 0.010 0.018 0.020 0.022 0.030 Layer 9 0.010 0.018 0.020 0.022 0.030 Wilting point Layer 10 m/m 0.015 0.027 0.030 0.033 0.045 Layer 1 0.64 1.14 1.27 1.40 1.91 Layer 2 0.64 1.14 1.27 1.40 1.91 Layer 3 0.72 1.30 1.44 1.58 2.16 Layer 4 0.91 1.63 1.81 1.99 2.72 Layer 5 0.0012 0.0021 0.0024 0.0026 0.0035 Layer 6 0.0016 0.0028 0.0031 0.0034 0.0047 Layer 7 0.0016 0.0029 0.0033 0.0036 0.0049 Layer 8 0.0019 0.0035 0.0039 0.0043 0.0059 Layer 9 0.0019 0.0035 0.0039 0.0043 0.0059 Saturated hydraulic conductivity Layer 10 m/s 10-3 0.0108 0.0194 0.0215 0.0237 0.0323 Restrictive layer hydraulic dii m/s 10-9 0.425 0.765 0.850 0.935 1.28 Manning’s roughness coefficient m1/3/s 0.05 0.09 0.10 0.11 0.15 Surface maximum depressional storage m 10-2 0.25 0.45 0.50 0.55 0.75 January 0.35 0.63 0.70 0.77 1.05 February 0.38 0.68 0.75 0.83 1.13 March 0.40 0.72 0.80 0.88 1.20 April 0.43 0.77 0.85 0.94 1.28 May 0.45 0.81 0.90 0.99 1.35 June 0.48 0.86 0.95 1.05 1.43 July 0.50 0.90 1.00 1.10 1.50 August 0.48 0.86 0.95 1.05 1.43 September 0.45 0.81 0.90 0.99 1.35 October 0.43 0.77 0.85 0.94 1.28 November 0.40 0.72 0.80 0.88 1.20 Crop coefficient December 0.38 0.68 0.75 0.83 1.13

PAGE 164

143 Table 4-11. Selected nutrient paramete rs for sensitivity analysis for S4. Testing Range Parameters Unit -50%-10%Base 10% 50% Rainfall P concentration mg/l 0.014 0.025 0.028 0.031 0.042 Rainfall N concentration mg/l 0.010 0.018 0.020 0.022 0.030 Layer 1 70.0125.9139.9 153.9 209.9 L a y er 2 1189.32140.82378.6 2616.5 3568.0 L a y er 3 99.0178.2198.0 217.8 297.0 L a y er 4 1306.22351.22612.4 2873.6 3918.6 L a y er 5 143.9259.2287.7 316.5 431.6 L a y er 6 253.1455.6506.2 556.8 759.3 L a y er 7 98.0176.3195.9 215.5 293.9 L a y er 8 63.3114.0126.7 139.3 190.0 L a y er 9 63.3114.0126.7 139.3 190.0 Active P L a y er 10 kg/ha 75.8136.4151.5 166.7 227.3Layer 1 318.0572.4636.0 699.6 954.0 Layer 2 5406.09730.810812.011893.2 16218.0Layer 3 450.0810.0900.0 990.0 1350.0Layer 4 5937.310687.111874.6 13062.0 17811.8Layer 5 654.01177.11307.9 1438.7 1961.9Layer 6 1150.02070.82300.9 2531.0 3451.3Layer 7 445.3801.6890.6 979.7 1335.9Layer 8 287.9518.2575.8 633.4 863.7Layer 9 287.9518.2575.8 633.4 863.7 Stable P Layer 10 kg/ha 344.3619.8688.7 757.6 1033.0Layer 1 344.5620.1689.0 757.9 1033.5 L a y er 2 5860.010548.011720.0 12892.0 17580.0 L a y er 3 2815.05067.05630.0 6193.0 8445.0 L a y er 4 2725.04905.05450.0 5995.0 8175.0 L a y er 5 1325.02385.02650.0 2915.0 3975.0 L a y er 6 4258.57665.38517.0 9368.7 12775.5 L a y er 7 4030.57254.98061.0 8867.1 12091.5 L a y er 8 4340.57812.98681.0 9549.1 13021.5 L a y er 9 4340.57812.98681.0 9549.1 13021.5 Humus P L a y er 10 kg/ha 4651.08371.89302.0 10232.2 13953.0Layer 1 145.1261.2290.3 319.3 435.4 Layer 2 2467.14440.84934.3 5427.7 7401.4Layer 3 627.81130.01255.3 1381.1 1883.3Layer 4 608.21094.71216.4 1338.0 1824.5Layer 5 295.2531.4590.4 649.4 885.6Layer 6 855.01539.01710.0 1881.0 2565.0Layer 7 809.31456.71618.5 1780.4 2427.8Layer 8 871.51568.71743.0 1917.3 2614.5Layer 9 871.51568.71743.0 1917.3 2614.5 Active N Layer 10 kg/ha 933.81680.81867.5 2054.3 2801.3Layer 1 928.51671.31857.0 2042.7 2785.5 L a y er 2 15789.828421.631579.5 34737.5 47369.3 L a y er 3 4017.87232.08035.5 8839.1 12053.3 L a y er 4 3892.57006.57785.0 8563.5 11677.5 L a y er 5 1889.33400.73778.5 4156.4 5667.8 L a y er 6 5472.09849.610944.0 12038.4 16416.0 L a y er 7 5179.59323.110359.0 11394.9 15538.5 L a y er 8 5577.810040.011155.5 12271.1 16733.3 L a y er 9 5577.810040.011155.5 12271.1 16733.3 Stable N L a y er 10 kg/ha 5976.010756.811952.0 13147.2 17928.0 Stocking rate tons/ha 0.00174 0.00313 0.00348 0.00382 0.00521 Fertilizer rate kg/ha 4.62 8.32 9.24 10.16 13.86

PAGE 165

144 Table 4-12. Sensitivities of total surface r unoff of the selected input parameters throughout the simulation period fo r the calibration site W6. Sensitivity [m3/day] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y ** 49087453594201932171 **0.82 0.00 -0.74-0.58 Wiltin g p oin t 4321944920453594582447705-0.09-0.10 0.00 0.100.10 SatK* 47330456744535945085441940.090.07 0.00 -0.06-0.05 ResK* 52953468164535943958387340.330.32 0.00 -0.31-0.29 Cro p coefficien t 79106509394535940121357491.491.23 0.00 -1.15-0.42 SurfStora g e* 46497455494535945191444910.050.04 0.00 -0.04-0.04 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. ** Not applicable. Table 4-13. Sensitivities of maximum water tabl e to the selected input parameters depth throughout the simulation period fo r the calibration site W6. Sensitivity [m] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y ** 1.72011.69031.66891.6071 **0.18 0.00 -0.13-0.10 Wiltin g p oin t 1.7060 1.69191.69031.69031.68820.020.01 0.00 0.000.00 SatK* 1.6120 1.67811.69031.70141.7387-0.09-0.07 0.00 0.070.06 ResK* 1.6409 1.68081.69031.69941.7377-0.06-0.06 0.00 0.050.06 Cro p coefficien t 1.5832 1.67851.69031.70011.7194-0.13-0.07 0.00 0.060.03 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. ** Not applicable. Table 4-14. Sensitivities of total P loads to the selected input parameters throughout the simulation period for the calibration site W6. Sensitivity [kg/ha] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y ** 3.78893.51173.26282.5305 **0.79 0.00 -0.71-0.56 Wiltin g p oin t 3.3566 3.48083.51173.54693.6669-0.09-0.09 0.00 0.100.09 SatK* 3.6739 3.53763.51173.49053.41090.090.07 0.00 -0.06-0.06 ResK* 4.1091 3.62833.51173.40272.99130.340.33 0.00 -0.31-0.30 Cro p coefficien t 6.1475 3.95783.51173.09572.75871.501.27 0.00 -1.18-0.43 SurfStora g e** 3.6710 3.54983.51173.48803.44510.090.11 0.00 -0.07-0.04 Rainfall P 2.3878 3.27043.51173.75374.6369-0.64-0.69 0.00 0.690.64 Active P 2.8997 3.38633.51173.63664.1358-0.35-0.36 0.00 0.360.36 Stable P 3.5016 3.50973.51173.51383.5223-0.01-0.01 0.00 0.010.01 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. ** Not applicable. Table 4-15. Sensitivities of total N loads to the selected input parameters throughout the simulation period for the calibration site W6. Sensitivity [kg/ha] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y ** 5.64735.70185.42784.6585 **-0.10 0.00 -0.48-0.37 Wiltin g p oin t 5.3777 5.60445.70185.80015.8486-0.11-0.17 0.00 0.170.05 SatK* 5.9961 5.76815.70185.62895.41400.100.12 0.00 -0.13-0.10 ResK* 8.1360 6.10485.70185.31734.07730.850.71 0.00 -0.67-0.57 Cro p coefficien t 10.6956.47865.70184.86974.20781.751.36 0.00 -1.46-0.52 SurfStora g e** 6.3471 5.79395.70185.66345.62350.230.16 0.00 -0.07-0.03 Rainfall N 3.4234 5.24525.70186.15647.9735-0.80-0.80 0.00 0.800.80 Active N 5.3280 5.64735.70185.80016.0918-0.13-0.10 0.00 0.170.14 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. ** Not applicable.

PAGE 166

145 Table 4-16. Sensitivities of total surface r unoff to the selected input parameters throughout the simulation period for the calibration site S4. Sensitivity [m3/day] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y 514486 4012213770713561192907790.730.64 0.00 -0.56-0.46 Wiltin g p oin t 353052 372690377071382299382299-0.13-0.12 0.00 0.140.03 SatK* 392162 3795683770713748243653200.080.07 0.00 -0.06-0.06 ResK* 413582 3845273770713701103429640.190.20 0.00 -0.18-0.18 Cro p coefficien t 511687 3988753770713588743377790.710.58 0.00 -0.48-0.21 SurfStora g e* 389419 3798323770713747723674640.070.07 0.00 -0.06-0.05 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. Table 4-17. Sensitivities of maximum water tabl e to the selected input parameters depth throughout the simulation period fo r the calibration site S4. Sensitivity [m] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y 1.6549 1.58891.57511.55981.53430.100.09 0.00 -0.10-0.05 Wiltin g p oin t 1.6732 1.59051.57511.55751.55750.120.10 0.00 -0.11-0.02 SatK* 1.5188 1.56621.57511.58181.6172-0.07-0.06 0.00 0.040.05 ResK* 1.5295 1.56131.57511.58351.6142-0.06-0.09 0.00 0.050.05 Cro p coefficien t 1.4023 1.56501.57511.57841.5844-0.22-0.06 0.00 0.020.01 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. Table 4-18. Sensitivities of total P loads to the selected input parameters throughout the simulation period for the calibration site S4. Sensitivity [kg/ha] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y 21.61116.85615.84714.98612.2150.730.64 0.00 -0.54-0.46 Wiltin g p oin t 14.71715.67415.84716.05416.054-0.14-0.11 0.00 0.130.03 SatK* 16.43315.94915.84715.76115.3720.070.06 0.00 -0.05-0.06 ResK* 17.32516.15315.84715.57114.4570.190.19 0.00 -0.17-0.18 Cro p coefficien t 21.18916.71215.84715.13314.3700.670.55 0.00 -0.45-0.19 SurfStora g e* 16.63016.06215.84715.69615.2980.100.14 0.00 -0.09-0.07 Rainfall P 15.79015.83515.84715.85815.902-0.01-0.01 0.00 0.010.01 Active P 15.46915.45715.84716.23717.803-0.05-0.25 0.00 0.250.25 Stable P 10.15214.70715.84716.98621.542-0.72-0.72 0.00 0.720.72 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage. Table 4-19. Sensitivities of total N loads to the selected input parameters throughout the simulation period for the calibration site S4. Sensitivity [kg/ha] Relative sensitivit y [ ] Parameters -50% -10% Base 10% 50% -50% -10% Base 10% 50% Porosit y 9.9295 8.31768.00157.73737.34970.480.40 0.00 -0.33-0.16 Wiltin g p oin t 7.6516 7.94318.00158.06808.0680-0.09-0.07 0.00 0.080.02 SatK* 8.3318 8.06388.00157.95147.77820.080.08 0.00 -0.06-0.06 ResK* 9.8029 8.32138.00157.70536.63860.450.40 0.00 -0.37-0.34 Cro p coefficien t 14.0248.68448.00157.44416.58351.510.85 0.00 -0.70-0.35 SurfStora g e* 14.6928.78988.00157.47726.50611.670.99 0.00 -0.66-0.37 Rainfall N 7.6872 7.93788.00158.06448.3241-0.08-0.08 0.00 0.080.08 Active N 7.0743 7.86518.00158.11778.4498-0.23-0.17 0.00 0.150.11 *SatK= saturated hydraulic condu ctivity; ResK= restrictive layer hydraulic c onductivity; surfStorage= surface storage.

PAGE 167

146 Table 4-20. Annual water budget for the calibration site W6. Year Obs. R1 Obs. RO2 Sim. RO3 Sim. LG4 Sim. ET5 Sim. DS6 1998 138.05 20.90 35.24 (25.5%)7 0.20 91.92 (62.0%)7 17.19 (12.5%)7 1999 114.82 15.12 7.32 (6.4%) -0.22 96.80 (81.3%) 14.17 (12.3%) 2000 75.63 0.57 1.09 (1.4%) -0.24 73.75 (85.9%) 9.58 (12.7%) 2001 135.14 24.17 31.14 (23.0%) -0.22 83.46 (67.2%) 13.17 (9.7%) 2002 151.08 25.29 31.88 (21.1%) -0.03 90.69 (67.9%) 16.58 (11.0%) 2003 156.52 46.58 34.70 (22.2%) -0.08 113.64 (66.3%) 17.98 (11.0%) 1Observed rainfall [cm]; 2Observed runoff [cm]; 3Simulated runoff [cm]; 4Simulated lateral groundwater flow [cm]; 5Evapotranspiration [cm]; 6Deep seepage [cm]; and 7Percentages indicate the shar e each hydrologic component accounts for in rainfa ll input, respectively. Table 4-21. Annual water budget fo r the verification site W7. Year Obs. R1 Obs. RO2 Sim. RO3 Sim. LG4 Sim. ET5 Sim. DS6 1998 138.05 24.33 35.90 (26.0%)7 0.04 91.69 (61.6%)7 17.13 (12.4%)7 1999 114.82 14.77 7.65 (6.7%) -0.81 96.83 (81.0%) 14.11 (12.3%) 2000 75.63 0.98 1.21 (1.6%) -0.78 74.26 (85.7%) 9.63 (12.7%) 2001 135.14 32.06 30.93 (22.9%) -0.78 83.80 (67.4%) 13.17 (9.7%) 2002 151.08 33.72 32.27 (21.4%) -0.32 91.25 (67.7%) 16.53 (10.9%) 2003 156.52 45.65 35.75 (22.8%) -0.61 113.15 (65.8%) 17.81 (11.4%) 1Observed rainfall [cm]; 2Observed runoff [cm]; 3Simulated runoff [cm]; 4Simulated lateral groundwater flow [cm]; 5Evapotranspiration [cm]; 6Deep seepage [cm]; and 7Percentages indicate the shar e each hydrologic component accounts for in rainfa ll input, respectively. Table 4-22. Annual water budget for the calibration site S4. Year Obs. R1 Obs. RO2 Sim. RO3 Sim. LG4 Sim. ET5 Sim. DS6 1998 138.24 9.93 37.46 (27.1%)7 0.93 92.63 (62.7%)7 10.04 (10.2%)7 1999 122.00 15.02 27.13 (22.2%) 0.36 84.03 (67.9%) 12.04 (9.9%) 2000 75.17 0.43 1.25 (1.7%) -0.79 68.51 (85.0%)) 10.03 (13.3%) 2001 127.34 23.78 42.48 (33.4%) 0.45 69.65 (58.0%) 10.96 (8.6%) 2002 143.53 26.51 45.62 (31.8%) 1.17 74.56 (59.3%) 12.79 (8.9%) 2003 134.30 19.95 31.10 (23.2%) 0.41 98.02 (66.9%) 13.32 (9.9%) 1Observed rainfall [cm]; 2Observed runoff [cm]; 3Simulated runoff [cm]; 4Simulated lateral groundwater flow [cm]; 5Evapotranspiration [cm]; 6Deep seepage [cm]; and 7Percentages indicate the shar e each hydrologic component accounts for in rainfa ll input, respectively. Table 4-23. Annual water budget fo r the verification site S1. Year Obs. R1 Obs. RO2 Sim. RO3 Sim. LG4 Sim. ET5 Sim. DS6 1998 138.24 8.11 37.66 (27.2%)7 0.46 93.47 (62.7%)7 13.91 (10%)7 1999 122.00 8.16 27.53 (22.6%) 0.18 83.62 (67.7%) 11.92 (10%) 2000 75.17 -1.08 0.59 (0.8%) -0.29 68.69 (86.0%) 9.92 (13%) 2001 127.34 25.60 43.26 (34.0%) 0.34 69.49 (57.4%) 10.93 (9%) 2002 143.53 27.35 46.28 (32.2%) 0.63 74.28 (59.0%) 12.61 (9%) 2003 134.30 14.75 31.37 (23.4%) 0.36 98.15 (66.9%) 13.15 (10%) 1Observed rainfall [cm]; 2Observed runoff [cm]; 3Simulated runoff [cm]; 4Simulated lateral groundwater flow [cm]; 5Evapotranspiration [cm]; 6Deep seepage [cm]; and 7Percentages indicate the shar e each hydrologic component accounts for in rainfa ll input, respectively.

PAGE 168

147 Table 4-24. Monthly statistics for the calibration site W6 Runoff Water Table Depth P Load N Load Statistics* ACRU FHANTM ACRU FHANTM ACRU FHANTM ACRU FHANTM Bias 0.05 0.86 78.36 ** 0.00 0.00 0.01 0.05 RE 0.03 0.43 -0.06 0.11 -0.04 0.20 0.96 RMSE 1.86 4.27 391.06 0.03 0.07 0.05 0.25 CV 1.01 2.14 -0.31 1.41 2.53 1.43 4.84 R2 0.81 0.94 0.77 0.83 NS 0.75 0.93 0.72 0.61 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of statistics are unitless for all variables. **Indicate not calculated by Sims (2004). Table 4-25. Annual statistics for the calibration site W6 Runoff Water Table Depth P Load N Load Statistics* ACRU FHANTM ACRU FHANTM ACRU FHANTM ACRU FHANTM Bias 0.56 10.30 940.32 ** 0.03 -0.01 0.08 0.58 RE 0.03 0.43 -0.06 0.11 -0.04 0.20 0.96 RMSE 8.66 14.80 2487.80 0.10 0.23 0.19 0.86 CV 0.39 0.62 -0.17 0.37 0.73 0.46 1.42 R2 0.69 0.97 0.83 0.78 NS 0.67 0.96 0.81 0.44 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of statistics are unitless for all variables. **Indicate not calculated by Sims (2004). Table 4-26. Monthly statistics for the verification site W7 Runoff Water Table Depth P Load N Load Statistics* ACRU FHANTM ACRU FHANTM ACRU FHANTM ACRU FHANTM Bias -0.18 0.54 291.11 ** -0.00 -0.05 -0.01 -0.01 RE -0.09 0.24 -0.20 -0.16 -0.80 -0.38 -0.31 RMSE 1.59 4.26 677.99 0.03 0.27 0.04 0.09 CV 0.76 2.13 -0.46 1.74 10.31 1.23 1.73 R2 0.87 0.85 0.74 0.79 NS 0.86 0.81 0.71 0.72 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of statistics are unitless for all variables. **Indicate not calculated by Sims (2004). Table 4-27. Annual statistics fo r the verification site W7 Runoff Water Table Depth P Load N Load Statistics* ACRU FHANTM ACRU FHANTM ACRU FHANTM ACRU FHANTM Bias -2.16 6.53 3493.32 ** -0.04 -0.66 -0.16 -0.17 RE -0.09 0.24 -0.20 -0.16 -0.80 -0.38 -0.31 RMSE 5.68 14.75 5331.74 0.12 0.93 0.21 0.30 CV 0.22 0.61 -0.30 0.54 1.14 0.48 0.55 R2 0.87 0.93 0.61 0.76 NS 0.84 0.86 0.55 0.24 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of statistics are unitless for all variables. **Indicate not calculated by Sims (2004).

PAGE 169

148 Table 4-28. Monthly statistics for the calibration site S4 Statistics* Runoff Water Table P Load N Load Bias 0.21 159.04 0.02 0.01 RE 0.16 -0.12 0.23 0.36 RMSE 0.99 472.46 0.08 0.04 CV 0.75 -0.36 0.91 1.32 R2 0.87 0.92 0.85 0.73 NS 0.85 0.89 0.82 0.68 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of sta tistics are unitless for all variables. Table 4-29. Annual statistics for the calibration site S4 Statistics* Runoff Water Table P Load N Load Bias 2.55 1908.44 0.24 0.14 RE 0.16 -0.12 0.23 0.36 RMSE 3.73 2969.41 0.31 0.18 CV 0.23 -0.19 0.30 0.45 R2 0.91 0.97 0.93 0.88 NS 0.82 0.94 0.80 0.64 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of sta tistics are unitless for all variables. Table 4-30. Monthly statistics for the verification site S1 Statistics* Runoff Water Table P Load N Load Bias 0.46 432.09 0.00 0.01 RE 0.40 -0.27 0.03 0.20 RMSE 1.48 735.87 0.09 0.07 CV 1.28 -0.45 0.87 1.81 R2 0.76 0.88 0.85 0.61 NS 0.67 0.82 0.85 0.60 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of sta tistics are unitless for all variables. Table 4-31. Annual statistics fo r the verification site S1 Statistics* Runoff Water Table P Load N Load Bias 5.48 5185.17 0.03 0.09 RE 0.40 -0.27 0.03 0.20 RMSE 6.27 6709.44 0.28 0.28 CV 0.45 -0.34 0.23 0.65 R2 0.91 0.14 0.94 0.63 NS 0.61 0.81 0.92 0.58 *Units of bias and RMSE for runoff and water table depth are cm and for P and N loads are kg/ha. The rest of sta tistics are unitless for all variables.

PAGE 170

149 -1.5 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5Surface Runoff for W 6 PorosityWilting pointSatKResKCrop CoefficientSurface storage Percentage (%) -50% -10% Base 10% 50% -1.5 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5Surface Runoff for S 4 PorosityWilting pointSatKResKCrop CoefficientSurface storage Percentage (%) -50% -10% Base 10% 50% Figure 4-6. Relative sensitivities of the tota l surface runoff of 6-year simulation period over the selected input parameters for the pasture sites W6 and S4 (SatK = saturated hydraulic conductivity; Re sK = restrictive layer hydraulic conductivity; realtive sensitivities are unitless).

PAGE 171

150 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3Maximum Water Table Depth for W6PorosityWilting pointSatKResKCrop coefficient Percentage (%) -50% -10% Base 10% 50% -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3Maximum Water Table Depth for S4PorosityWilting pointSatKResKCrop coefficient Percentage (%) -50% -10% Base 10% 50% Figure 4-7. Relative sensitivities of the maxi mum water table depth of 6-year simulation period over the selected input paramete rs for the pasture sites W6 and S4 (SatK = saturated hydraulic conductivity; ResK = rest rictive laye r hydraulic conductivity; relative sensitivities are unitless).

PAGE 172

151 -1.5 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5P Load for W6PorosityWilting point SatKResKCrop coefficient Surface storage Rainfall PActive PStable P Percentage (%) -50% -10% Base 10% 50% -1.5 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5P Load for S4PorosityWilting point SatKResKCrop coefficient Surface storage Rainfall PActive PStable P Percentage (%) -50% -10% Base 10% 50% Figure 4-8. Relative sensitivities of the total P load of 6-year simulation period over the selected input parameters and variables for the pasture sites W6 and S4 (SatK = saturated hydraulic conductivity; Re sK = restrictive layer hydraulic conductivity; relative sensitivities are unitless).

PAGE 173

152 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0N Load for W6PorosityWilting pointSatKResKCrop coefficient Surface storage Rainfall NActive N Percentage (%) -50% -10% Base 10% 50% -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0N Load for S4PorosityWilting pointSatKResKCrop coefficient Surface storage Rainfall NActive N Percentage (%) -50% -10% Base 10% 50% Figure 4-9. Relative sensitivities of the total N load of 6-year simulation period over the selected input parameters for the pasture sites W6 and S4 (SatK = saturated hydraulic conductivity; ResK = restri ctive layer hydrau lic conductivity; relative sensitivities are unitless).

PAGE 174

153 -200 -160 -120 -80 -40 0 40 9/1/003/1/019/1/013/1/029/1/023/1/039/1/03 Time (day)Water Table (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-10. Continuous simu lation of groundwater table depth from September 2000 through December 2003 for the calibration site W6. -200 -160 -120 -80 -40 0 40 9/1/003/1/019/1/013/1/029/1/023/1/039/1/03 Time (day)Water Table (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-11. Continuous simu lation of groundwater table depth from September 2000 through December 2003 for th e verification site W7.

PAGE 175

154 -200 -150 -100 -50 0 50 9/1/001/1/015/1/019/1/011/1/025/1/029/1/021/1/035/1/039/1/03 Time (day)Water Table (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-12. Continuous simu lation of groundwater table depth from September 2000 through December 2003 for the calibration site S4. -250 -200 -150 -100 -50 0 50 9/1/003/1/019/1/013/1/029/1/023/1/039/1/03 Time (day)Water Table (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-13. Continuous simu lation of groundwater table depth from September 2000 through December 2003 for th e verification site S1.

PAGE 176

155 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-14. Continuous simulation of surf ace runoff from July 1998 through December 2003 for the calibration site W6. -1.0 0.0 1.0 2.0 3.0 4.0 5.0 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-15. Continuous simulation of surf ace runoff from July 1998 through December 2003 for the verification site W7.

PAGE 177

156 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-16. Continuous simulation of surf ace runoff from July 1998 through December 2003 for the calibration site S4. -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (days)Runoff (cm)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-17. Continuous simulation of surf ace runoff from July 1998 through December 2003 for the verification site S1.

PAGE 178

157 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P load (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-18. Continuous simu lation of P loads from July 1998 through December 2003 for the calibration site W6. -0.02 0.00 0.02 0.04 0.06 0.08 0.10 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-19. Continuous simu lation of P loads from July 1998 through December 2003 for the verification site W7.

PAGE 179

158 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-20. Continuous simu lation of P loads from July 1998 through December 2003 for the calibration site S4. -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-21. Continuous simu lation of P loads from July 1998 through December 2003 for the verification site S1.

PAGE 180

159 -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N load(kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-22. Continuous simu lation of N loads from July 1998 through December 2003 for the calibration site W6. -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-23. Continuous simu lation of N loads from July 1998 through December 2003 for the verification site W7.

PAGE 181

160 -0.05 0.00 0.05 0.10 0.15 0.20 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-24. Continuous simu lation of N loads from July 1998 through December 2003 for the calibration site S4. -0.05 0.00 0.05 0.10 0.15 0.20 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Loads (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-25. Continuous simu lation of N loads from July 1998 through December 2003 for the verification site S1.

PAGE 182

161 0.00 0.50 1.00 1.50 2.00 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-26. Continuous simulation of surf ace runoff P concentrations from July 1998 through December 2003 for the calibration site W6. 0.00 0.50 1.00 1.50 2.00 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-27. Continuous simulation of surf ace runoff P concentrations from July 1998 through December 2003 for th e verification site W7.

PAGE 183

162 0.00 0.60 1.20 1.80 2.40 3.00 3.60 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-28. Continuous simulation of surf ace runoff N concentrations from July 1998 through December 2003 for the calibration site W6. 0.00 0.60 1.20 1.80 2.40 3.00 3.60 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-29. Continuous simulation of surf ace runoff N concentrations from July 1998 through December 2003 for th e verification site W7.

PAGE 184

163 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-30. Continuous simulation of surf ace runoff P concentrations from July 1998 through December 2003 for the calibration site S4. 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-31. Continuous simulation of surf ace runoff P concentrations from July 1998 through December 2003 for the calibration site S1.

PAGE 185

164 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-32. Continuous simulation of surf ace runoff N concentrations from July 1998 through December 2003 for th e verification site S4. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 7/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall Observed ACRU2000 Figure 4-33. Continuous simulation of surf ace runoff N concentrations from July 1998 through December 2003 for th e verification site S1.

PAGE 186

165 Observed vs. Simulated Monthly Runoff for W6 y = 1.028x 0.0047 R2 = 0.8104 -5 0 5 10 15 20 -505101520 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Runoff for W6 y = 0.8024x + 4.9308 R2 = 0.6883 0 10 20 30 40 50 60 0102030405060 Observed (cm)Simulated (cm) Figure 4-34. Linear plots of monthly and annual surface runoff from 1998 to 2003 for the calibration site W6.

PAGE 187

166 Observed vs. Simulated Monthly Runoff for W7 y = 0.9776x 0.1326 R2 = 0.8721 -5 0 5 10 15 20 25 30 -5051015202530 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Runoff for W7 y = 0.8711x + 1.0992 R2 = 0.8658 0 10 20 30 40 50 60 0102030405060 Observed (cm)Simulated (cm) Figure 4-35. Linear plots of monthly and annual surface runoff from 1998 to 2003 for the verification site W7.

PAGE 188

167 Observed vs. Simulated Monthly Water Table for W6 y = 1.0203x + 103.8 R2 = 0.9376 -6000 -5000 -4000 -3000 -2000 -1000 0 -6000-5000-4000-3000-2000-10000 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Water Table for W6 y = 0.9847x + 711.02 R2 = 0.9707 -40000 -30000 -20000 -10000 0 -40000-30000-20000-100000 Observed (cm)Simulated (cm) Figure 4-36. Linear plots of monthly and annual water ta ble depth from 1998 to 2003 for the calibration site W6.

PAGE 189

168 Observed vs. Simulated Monthly Water Table for W7 y = 0.9543x + 223.89 R2 = 0.853 -6000 -5000 -4000 -3000 -2000 -1000 0 -6000-5000-4000-3000-2000-10000 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Water Table for W7 y = 0.8434x + 729.4 R2 = 0.9252 -40000 -30000 -20000 -10000 0 -40000-30000-20000-100000 Observed (cm)Simulated (cm) Figure 4-37. Linear plots of monthly and annual water ta ble depth from 1998 to 2003 for the verification site W7.

PAGE 190

169 Observed vs. Simulated Monthly P Load for W6 y = 0.9572x + 0.0035 R2 = 0.7656 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.60 -0.100.000.100.200.300.400.500.60 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual P Load for W6 y = 0.8731x + 0.0654 R2 = 0.8334 0.00 0.20 0.40 0.60 0.80 0.000.200.400.600.80 Observed (kg/ha)Simulated (kg/ha) Figure 4-38. Linear plots of monthly and annual P load from 1998 to 2003 for the calibration site W6.

PAGE 191

170 Observed vs. Simulated Monthly P Load for W7 y = 0.5837x + 0.0046 R2 = 0.7437 -0.05 0.06 0.17 0.28 0.39 0.50 -0.050.060.170.280.390.50 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual P Load for W7 y = 0.4965x + 0.0734 R2 = 0.6071 0.00 0.20 0.40 0.60 0.000.200.400.60 Observed (kg/ha)Simulated (kg/ha) Figure 4-39. Linear plots of monthly and annual P load from 1998 to 2003 for the verification site W7.

PAGE 192

171 Observed vs. Simulated Monthly N Load for W6 y = 1.2497x 0.0017 R2 = 0.8315 -0.10 0.10 0.30 0.50 0.70 0.90 -0.100.100.300.500.700.90 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual N Load for W6 y = 1.1913x + 0.0041 R2 = 0.7833 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0.000.200.400.600.801.001.20 Observed (kg/ha)Simulated (kg/ha) Figure 4-40. Linear plots of monthly and annual N load from 1998 to 2003 for the calibration site W6.

PAGE 193

172 Observed vs. Simulated Monthly N Load for W7 y = 0.6094x + 0.0005 R2 = 0.7927 -0.05 0.08 0.21 0.34 0.47 0.60 -0.050.080.210.340.470.60 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual N Load for W7 y = 0.5677x + 0.0244 R2 = 0.7613 0.00 0.20 0.40 0.60 0.80 0.000.200.400.600.80 Observed (kg/ha)Simulated (kg/ha) Figure 4-41. Linear plots of monthly and annual N load from 1998 to 2003 for the verification site W7.

PAGE 194

173 Observed vs. Simulated Monthly Runoff for S4 y = 0.989x + 0.2255 R2 = 0.8732 -4 -2 0 2 4 6 8 10 -4-20246810 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Runoff for S4 y = 0.9902x + 2.6862 R2 = 0.9122 0 5 10 15 20 25 30 051015202530 Observed (cm)Simulated (cm) Figure 4-42. Linear plots of monthly and annual surface runoff from 1998 to 2003 for the calibration site S4.

PAGE 195

174 Observed vs. Simulated Monthly Runoff for S1 y = 0.9742x + 0.4863 R2 = 0.764 -4 -2 0 2 4 6 8 10 12 -4-2024681012 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Runoff for S1 y = 0.9612x + 6.0156 R2 = 0.9119 -5 0 5 10 15 20 25 30 -5051015202530 Observed (cm)Simulated (cm) Figure 4-43. Linear plots of monthly and annual surface runoff from 1998 to 2003 for the verification site S1.

PAGE 196

175 Observed vs. Simulated Monthly Water Table for S4 y = 1.0191x + 183.46 R2 = 0.9175 -6000 -5000 -4000 -3000 -2000 -1000 0 -6000-5000-4000-3000-2000-10000 Observed (cm)Simulated (cm) Observed vs. Simulated Annual Water Table for S4 y = 0.8975x + 281.62 R2 = 0.9737 -40000 -30000 -20000 -10000 0 -40000-35000-30000-25000-20000-15000-10000-50000 Observed (cm)Simulated (cm) Figure 4-44. Linear plots of monthly and annual water ta ble depth from 1998 to 2003 for the calibration site S4.

PAGE 197

176 Obseved vs. Simulated Monthly Water Table for S1 y = 0.8694x + 219.46 R2 = 0.8798 -6000 -5000 -4000 -3000 -2000 -1000 0 -6000-5000-4000-3000-2000-10000 Observed (cm)Simulated (cm) Obseved vs. Simulated Annual Water Table for S1 y = 0.7533x + 364.25 R2 = 0.9749 -40000 -30000 -20000 -10000 0 -40000-35000-30000-25000-20000-15000-10000-50000 Observed (cm)Simulated (cm) Figure 4-45. Linear plots of monthly and annual water ta ble depth from 1998 to 2003 for the verification site S1.

PAGE 198

177 Observed vs. Simulated Monthly P Load for S4 y = 0.9657x + 0.0229 R2 = 0.85 -0.25 0.00 0.25 0.50 0.75 1.00 1.25 -0.250.000.250.500.751.001.25 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual P Load for S4 y = 1.0312x + 0.2074 R2 = 0.9311 0.00 0.50 1.00 1.50 2.00 2.50 0.000.501.001.502.002.50 Observed (kg/ha)Simulated (kg/ha) Figure 4-46. Linear plots of monthly and annual P load from 1998 to 2003 for the calibration site S4.

PAGE 199

178 Obseved vs. Simulated Monthly P Load for S1 y = 0.8137x + 0.0222 R2 = 0.85 -0.25 0.10 0.45 0.80 1.15 1.50 -0.250.100.450.801.151.50 Observed (kg/ha)Simulated (kg/ha) Obseved vs. Simulated Annual P Load for S1 y = 0.8104x + 0.2704 R2 = 0.936 -0.50 0.50 1.50 2.50 -0.500.000.501.001.502.002.503.00 Observed (kg/ha)Simulated (kg/ha) Figure 4-47. Linear plots of monthly and annual P load from 1998 to 2003 for the verification site S1.

PAGE 200

179 Observed vs. Simulated Monthly N Load for S4 y = 0.8677x + 0.0161 R2 = 0.7319 -0.15 -0.05 0.05 0.15 0.25 0.35 -0.15-0.050.050.150.250.35 Observed (kg/ha)Simulated (kg/ha) Observed vs. Simulated Annual N Load for S4 y = 0.9604x + 0.1571 R2 = 0.8782 0.00 0.25 0.50 0.75 1.00 0.000.250.500.751.00 Observed (kg/ha)Simulated (kg/ha) Figure 4-48. Linear plots of monthly and annual N load from 1998 to 2003 for the calibration site S4.

PAGE 201

180 Obseved vs. Simulated Monthly N Load for S1 y = 0.5746x + 0.0224 R2 = 0.6075 -0.20 0.00 0.20 0.40 0.60 0.80 -0.200.000.200.400.600.80 Observed (cm)Simulated (cm) Obseved vs. Simulated Annual N Load for S1 y = 0.5352x + 0.286 R2 = 0.6347 -0.25 0.10 0.45 0.80 1.15 1.50 -0.250.100.450.801.151.50 Observed (kg/ha)Simulated (kg/ha) Figure 4-49. Linear plots of monthly and annual N load from 1998 to 2003 for the verification site S1.

PAGE 202

181 Duration Curve of Daily Surface Runoff for W6 -1.0 0.0 1.0 2.0 3.0 4.0 0102030405060708090100 Percent of Days Runoff Equaled or Exceeded (%)Surface Runoff (cm ) Observed Simulated Duration Curve of Daily Water Table Depth for W6 -175 -150 -125 -100 -75 -50 -25 0 25 0102030405060708090100 Percent of Days Water Table Equaled or Exceeded (%) Water Table (cm) Observed Simulated Figure 4-50. Duration curves of daily surface runoff and water table depth for the calibration site W6.

PAGE 203

182 Duration Curve of P Load for W6 -0.02 0.01 0.04 0.07 0.10 0102030405060708090100 Percent of Days P Load Equaled or Exceeded (%)P Load (kg/ha) Observed Simulated N Load Duration Curve for W6 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0102030405060708090100 Percent of days Equaled or Exceeded (%)N Load (kg/ha) Observed Simulated Figure 4-51. Duration curves of daily P and N loads for the calibration site W6.

PAGE 204

183 Duration Curve of Daily Surface Runoff for W7 -1.00 0.00 1.00 2.00 3.00 0102030405060708090100 Percent of Days Runoff Equaled or Exceeded (%)Surface Runoff (cm ) Observed Simulated Duration Curve of Daily Water Table Depth for W7 -200 -175 -150 -125 -100 -75 -50 -25 0 25 0102030405060708090100 Percent of Days Water Table Equaled or Exceeded (%) Water Table (cm) Observed Simulated Figure 4-52. Duration curves of daily surface runoff and water table depth for the verification site W7.

PAGE 205

184 Duration Curve of P Load for W7 -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0102030405060708090100 Percent of Days P Load Equaled or Exceeded (%)P Load (kg/ha) Observed Simulated Duration Curve of N Load for W7 -0.12 -0.08 -0.04 0.00 0.04 0.08 0.12 0102030405060708090100 Percent of Days N Load Equaled or Exceeded (%)N Load (kg/ha) Observed Simulated Figure 4-53. Duration curves of daily P a nd N loads for the verification site W7.

PAGE 206

185 Duration Curve for Surface Runoff for S4 -0.10 0.30 0.70 1.10 1.50 1.90 2.30 0102030405060708090100 Percent of Days Equaled or Ecxceeded (%)Surface Runoff (cm ) Observed Simulated Duration Curve for Water Table Depth for S4 -175 -150 -125 -100 -75 -50 -25 0 0102030405060708090100 Precent of Days Equaled or Exceeded (%)Water Table Depth (cm) Observed Simulated Figure 4-54. Duration curves of daily surface runoff and water table depth from 1998 to 2003 for the calibration site S4.

PAGE 207

186 Duration Curve of P Load for S4 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0102030405060708090100 Percent of Days Equaled or Exceeded (%)P Load (kg/ha) Observed Simulated Duration Curve of N Load for S4 -0.03 0.02 0.07 0.12 0102030405060708090100 Percent of Days Equaled or Exceeded (%)N Load (kg/ha) Observed Simulated Figure 4-55. Duration curves of daily N and P loads from 1998 to 2003 for the calibration site S4.

PAGE 208

187 Duration Curve of Surface Runoff for S1 -1.00 0.50 2.00 3.50 0102030405060708090100 Percent of Days Equaled or Exceeded (%)Surface Runoff (cm ) Observed Simulated Duration Curve of Water Table Depth for S1 -200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0 20 0102030405060708090100 Percent of Days Equaled or Exceeded (%)Water Table Depth (cm) Observed Simulated Figure 4-56. Duration curves of daily surface runoff and water table depth from 1998 to 2003 for the verification site S1.

PAGE 209

188 Duration Curve of P Load for S1 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0102030405060708090100 Percent of Days Equaled or Exceeded (%)P Load (kg/ha) Observed Simulated Duration Curve of N Load for S1 -0.04 0.00 0.04 0.08 0.12 0.16 0102030405060708090100 Percent of Days Equaled or Exceeded (%)N Load (kg/ha) Observed Simulated Figure 4-57. Duration curves of daily N and P loads from 1998 to 2003 for the verification site S1.

PAGE 210

189 CHAPTER 5 VEGETATION DYNAMICS SIMULATION MODEL Introduction A vegetation dynamics simulation model was developed to predic t plant growth in flatwoods watersheds where various landscapes such as wetlands, uplands and transition zones coexist. Best management practices (BMPs) intended to improve water quality in Lake Okeechobee (for example, increased wate r levels and reduced cattle stocking rates) impact hydrology and nutrient cycling in thes e watersheds and thus affect vegetation dynamics. Previous studies have indicated that vegetation composition patterns can change over time and space when either hydrol ogy or nutrient dynamics are changed. Some species may expand spatially to encro ach the habitats wher e other species were previously established, while other may declin e in their original habitats because of the adverse growth environment caused by these changes. Therefore, the vegetation composition pattern, as an integrated ecohydr ologic indicator, may help to identify changes in soil, water, and nutrient conditi ons in a watershed where BMPs are being implemented if the feedback relation ships between vegetation and hydrology and nutrients are well understood. Additionally well-calibrated nu merical vegetation simulation models can be very useful in reduc ing the cost of field experiments, gaining insights into relevant processe s, and predicting long-term tren ds. Therefore, a vegetation dynamics simulation model was developed in th is Chapter to couple with the hydrologic and nutrient models discussed in the previous Chapters.

PAGE 211

190 The available data to establish and calib rate a vegetation model for pasture and wetland plants in south Florida flatwoods watersheds is very limited. Therefore a simple model that avoids overwhelming data requireme nts, but is still capable of capturing the vegetation dynamics in a logical manner is required. When sufficient data become available in the future, the model can be further improved and calibrated. Consequently, the objectives of this st udy are to develop a dynamic vegetation model to simulate Daily biomass productivity and leaf area index. Plant growth dynamics limited by wa ter and nutrient availabilities. Evapotranspiration with consideration of multiple species. Nutrient uptake with consideration of multiple species. Species composition pattern over time and space as impacted by water and nutrient availability. A vegetation model, built on a spatially heterogeneous lands cape through a land segment system in which the number, size and shape of the land segments can be defined according to land cover, soil and vegetation t ypes, and topography, was developed. Each land segment is initialized with one or multiple species, which compete for light, water and nutrients. For each time step, plant growth is simulated, driven by climate variables including solar radiation and temperature. The growth reduction caused by suboptimal water and nutrient availability is calculated daily and applied to re duce the potential plant growth. Losses in plant biomass due to fire harvest, and senescence are withdrawn from the biomass pool and affect plant growth and competition in the next time step. Within a land segment, homogeneity of plant distribut ion, plant biomass, leaf area, and soil conditions is assumed. The population dynami cs of the herbivores and their grazing

PAGE 212

191 impacts to plant growth are not included in this study, but these impacts can be incorporated into the model once an herbi vore movement model is developed. Land segments are dependent on each other in th e sense that water and nutrient flows are transferred between land segments, and ther e are feedback relationships between the water, nutrient and vegetation models. Theref ore, the vegetation development within one land segment indirectly affect s the vegetation development in another land segment. The main processes described in this model are light interception, dry matter production and partitioning, and leaf area deve lopment. The driving variables of the model are climate data (radiation and temperat ure) and the time step is daily. The state variables are aboveground and belowground biomass and their corresponding N and P pools. Water stress reflecting water limitati on, water logging reflect ing anoxic soil water conditions, and nitrogen (N) a nd phosphorus (P) stress reflec ting nutrient defi ciencies are calculated on a daily basis in the model, a nd their combination defi nes a growth reduction factor affecting the plant growth. The following assumptions were made in the development of the vegetation model: Light interception occurs in one canopy layer that contai ns all plant species of interest. Potential plant growth is driven primaril y by solar radiation and temperature, but it is reduced by the water and nutrient limitations. Within land segments, functional groups of species are simulated instead of individual plants. Thus all plants within a species are treated as a single entity. Total dry matter for each species is si mply partitioned into aboveground and belowground biomass. All species within land segments are assu med to have the same root lengths and have the same distribution fraction of r oots in the computational soil layers. Alive plant N and P concentrations are uniform throughout the entire plant.

PAGE 213

192 Plant senescence starts when the sum of leaf area index (LAI) over all species is greater than the critical LAI. When long-term inundation causes the death of plants, 25% of the aboveground and belowground biomass and the corresponding N and P are removed per day from live biomass pools to dead biomass pools. When fire occurs, 95% of the abovegr ound biomass and the corresponding N and P in the aboveground biomass are removed fo r all species in the land segment. Methodology Model Structure The model for simulating the biomass produc tivity of pasture and wetland grasses in order to understand how spatial biomass distribution and composition pattern are impacted by altered hydrology and nutrient concentr ations is depicted in Figure 5-1. This model is designed based on the previous studies by Spitters and Schapendonk (1990), Ivens (1992), Kooman (1995), Jones and L uyten (1998), Van Oene et al. (1999), and Jones et al. (2000). The basic processes in this model are light interception, conversion of light into dry matter production, and allocation of dry matter between aboveground and belowground dry matter. The impacts from the changes in hydrology and nutrient concentrations are expressed in growth limiting fact ors. Water stress or logging indicate water deficiency or water logged soil conditions, and N and P stre ss factors indicate N and P deficiency. These stress factors are combined to define a growth reduction factor that is used in the model to reflect the adverse growth conditions causing the reduction of the potential dry matter production. Different species have diffe rent growth niches (Whittaker and Levin, 1975) defined by the plant’s physiological a nd phenological characteristics, which determine whether a species can adapt to the changing environment. The resulting plant

PAGE 214

193 composition pattern responding to the new enviro nment can be different and can thus be an indicator of how and to what extent the ecology of a watershed is impacted. Figure 5-1. Diagram for daily pl ant growth in relation to we ather and water and nutrient availabilities (DM = dr y matter; N = nitrogen; P = phosphorus; and SLA = specific leaf area). Plant Growth Potential growth If a plant is free from weed competition, pests and disease, and well-supplied with water and nutrients, the growth rate is determined by incident light, ambient temperature and plant characteristics. This growth rate is described as the potential growth rate (Ivens, 1992). In the model develope d here, the potential growth rate ( Wpot,i,t [kg/m2/day]) for each species i on day t is ca lculated through a linear function of the Fractional interception Radiation Water stress/ lo gg in g N and P stresses Light use efficiency Partitioning factor SLA Daily growth Leaf area index Temperature Hydrologic model Nutrient model Reduction facto r Plant senescence Belowground DM N and P Belowground DM N and P

PAGE 215

194 absorbed light and a mean radiation-use efficiency parameter (Cannell et al., 1987; Spitters and Schapendonk, 1990) as shown in Equa tion (5-1). This function is modified by a temperature factor (Van Oene et al., 1999). ) T ( F I RUE Wt t i t i abs i t i pot (5-1) where RUEi is the average radiation use efficiency of species i [kg/MJ(PAR)] and is a summary variable for all processes deali ng with photosynthesis a nd respiration. Iabs,i,t is the light interception by species i on day t, which is calculated using the Lambert-Beer equation (Monsi and Saeki, 1953). Each specie s absorbs an amount of light proportional to its share in the total weighted leaf area multiplied by the species specific extinction coefficient times the total light absorption of the vegetation as shown in Equation (5-2): ) e 1 ( I LAI k LAI k Ii 1 t i i extLAI k t 0 i 1 t i i ext 1 t i i ext t i abs (5-2) where Iabs,i,t is the absorbed radiati on by species i [MJ(PAR)/m2/day]; kext,i is the light extinction coefficient of species i [-]; LAIi,t-1 is the leaf area index of species i [m2 leaf /m2 ground] on day t-1; I0,t is the incoming photosynthetic ac tive radiation (PAR) on day t [MJ/ m2/day]. For simplicity, the difference in light interception at various positions in the canopy is neglected; instead it is a ssumed only one canopy layer exists, which contains all plant species. The distributi on of the plant species over this layer is determined by the leaf biomass distribution of the species. Leaf area per species is calculated from biomass a nd specific leaf area (SLAi) (see Equation (5-12)).

PAGE 216

195 Figure 5-2. Diagram of temper ature function for species i. Fi,t(Tt), in Equation (5-1), is a temperature factor for assimilation for species i on day t [-], which is written as i max, t i 2 opt i 2 opt i max, t i max, i 2 opt t i 1 opt i 1 opt t i min, i min, i 1 opt i min, t i max, t i min, t t t iT T T ) T T /( ) T T ( T T T 1 T T T ) T T /( ) T T ( T T or T T 0 ) T ( F (5-3) in which, Tt is the mean air temperature on day t [ C], obtained by taking the average of daily maximum and minimum air temperatures, which are inputs to the model; Tmin,i is the base/minimum temperature for assimilation for species i [ C]; Topt1,i is the optimum temperature 1 for assimilation for species i [ C]; Topt2,i is the optimum temperature 2 for assimilation for species i [ C]; Tmax,i is the maximum temperature for assimilation for species i [ C]. An example diagram showing the relationship between the temperature function and daily air temperat ure is shown in Figure 5-2. For this model, only one temperature function is assumed for each sp ecies throughout the entire growing season. 1.0 0.0 Tmin,i Topt1,i Topt2,i Tmax,i Fi,t(Tt) [-] Mean air temperature [ C]

PAGE 217

196 Reduced growth The plant growth rate may be limited by N or P deficiency, water shortage or water logging during different parts of the growing season: t i t i pot t i redRF W W (5-4) t i t i t i pot t iCN RF W N (5-5) t i t i t i pot t iCP RF W P (5-6) where Wred,i,t is the reduced dry matter produc tion rate of species i [kg/m2/day]; Ni,t and Pi,t are the N and P uptake rates corresponding to Wred,i,t [kg/ha], respectively; RFi,t is a growth reduction factor [] of species i, which integrates the limiting factors from water (see details in the following growth reduction factor secti on), N and P; and CNi,t and CPi,t are the biomass N and P percents, resp ectively (see Equations (5-31) and (538)) [%]. Applying the Euler’s method to Equations (5-4) to (5-6), the total dry matter production and dry matter N and P can be calculated as t RF ) T ( F I RUE W Wt i t t i t i abs i 1 t i red t i red (5-7) t CN RF ) T ( F I RUE N Nt i t i t t i t i abs i 1 t i t i (5-8) t CP RF ) T ( F I RUE P Pt i t i t t i t i abs i 1 t i t i (5-9) where Wred,i,t and Wred,i,t-1 are the dry matter production on day t and t-1, respectively [kg/m2]; Ni,t and Ni,t-1 are the total dry matter N on day t and t-1 [kg/ha]; Pi,t and Pi,t-1 are the total biomass P on day t and t-1 [kg/ha], re spectively; and t is the time step, which is one day in this model.

PAGE 218

197 Dry matter partitioning The total dry matter producti on is distributed among various plant organs. A first partitioning is that between aboveground and belowground plant organs, characterized by the shoot/root ratio, fa,i, which is generally a function of gr owth stage, but in this model is assumed to be a constant value. Dry matter may be further partitioned into economic products (e.g. grain, fruits) a nd crop residue (e.g. straw, st ubble). However, for this model, the total dry matter is only parti tioned into the abovegr ound and belowground dry matter using the equations: ) f 1 ( W Wi a t i red t i a (5-10) i a t i red t i bf W W (5-11) where Wa,i,t is the aboveground dry matter at the time step of t [kg/m2]; Wb,i,t is the belowground dry matter at the time step of t [kg/m2]; fa,i is the dry matter partitioning coefficient [-]. Leaf Area Index Changes in the leaf area index (LAIi,t) for each species on day t may be caused by grazing, fire, and harvesting in addition to plant growth and senescence. Equation (5-12) shows the accumulated daily leaf area can be calculated by multiplying the aboveground biomass by the specific leaf area on the same day. i t i a t iSLA W LAI (LAImin,i LAIi,t LAImax,i) (5-12) where LAIi,t is the leaf area index for each sp ecies i at the time step of t [m2 leaf /m2 ground]; Wa,i,t is the daily aboveground dry matter [kg DM/m2 ground] for species i at the time step of t; SLAi is the specific leaf area to convert the leaf dry matter into leaf area index [m2 leaf/kg DM leaf]. Howe ver, the calculated LAIi,t should not exceed a pre-

PAGE 219

198 defined maximum LAImax,i to represent the steady state a mature species stand reaches associated with the long-term climate, nor should it fall below a pre-specified minimum LAImin,i to account for the leaf area or stubble that is inaccessible to cattle due to their bite structure (Kiker, 1998). When long-term water inundation causes the death of a species, discussed in the following section, a thre shold LAI value of 0.001 was assumed in the model to allow regrowth once the so il conditions become favorable. Plant Senescence Plant senescence is assumed to start when the daily sum of leaf areas (i t i sumLAI LAI[m2 leaf/m2 ground]) of all species on one land segment exceeds the critical leaf area (LAIcr [m2 leaf/m2 ground]), an input to the model. This approach accounts for senescence associated with insuffici ent light availability to maintain all of the biomass that has accumulated and is now shaded. All species are assumed to be affected in the same way, i.e., when LAIsum exceeds this critical value, each species senesces LAI, biomass, N and P in proportion to its leaf areas. Th is process leads to a reduction in both aboveground and belowground biomass and corresponding N and P for all species on that land segment for each day when LAIsum exceeds LAIcr. The daily total senesced biomass (Ws,i,t [kg/ha]) and the corresponding N (Ns,i,t [kg/ha]) and P and (Ps,i,t [kg/ha]) removed through the senesced biomass for each species on that land segment can be calculated (Jones JW, pe rsonal communication, 2006): i cr cr sum bs t i sSLA / ) LAI LAI LAI ( Frac W (5-13) t i red t i s t i t i sW / W N N (5-14) t i red t i s t i t i sW / W P P (5-15)

PAGE 220

199 where Fracbs is the fraction of biomass above the cr itical level to senesce per day [-], which is assumed to be a constant value in the model for all species. The senesced biomass and biomass N and P decrease the amount of live biomass and its corresponding N and P pools: t i s t i red t i redW W W (5-16) t i s t i t iN N N (5-17) t i s t i t iP P P (5-18) Accordingly, plant residue and soil biom ass are increased by the senesced biomass and N and P: ) f 1 ( W M Mi a t i s 1 t r t r (5-19) i a t i s 1 t s t sf W M M (5-20) ) f 1 ( N M Ma t i s 1 t rn t rn (5-21) a t i s 1 t sn t snf N M M (5-22) ) f 1 ( P M Ma t i s 1 t i rp t i rp (5-23) a t i s 1 t i sp t i spf P M M (5-24) where Mr,t and Mr,t-1 are the residue biomass on day t and t-1 [kg/ha]; Ms,t and Ms,t-1 are the soil biomass on day t and t-1 [kg/ha]; Mrn,t and Mrn,t-1 are the residue biomass N on day t and t-1 [kg/ha]; Msn,t and Msn,t-1 are the soil biomass N on day t and t-1 [kg/ha]; Mrp,t and Mrp,t-1 are the residue biomass P on day t and t-1 [kg/ha]; Msn,t and Msn,t-1 are the soil biomass P on day t and t-1 [kg/ha].

PAGE 221

200 Evapotranspiration Evapotranspiration (ET) descri bed in Chapter 3 does not consider the transpiration partitioning among multiple plant species, instead it was designed for one plant species within a land segment. To account for the transpiration by multiple plant species with different crop coefficients, the method for es timating ET in Chapter 3 was modified. The ET process is important in the context of the vegetation simulation in that transpiration is the driving force for nutrient uptake, and in sufficient ET leads to reduced plant growth due to water stress discussed in the following section In this process, first, a lumped potential evapotranspiration (PET) for all species in a land segment is calculated by multiplying re ference evapotranspiration (RET, input or calculated by the model) by a crop coefficient that is weighted by the LAI of each species within one land segment: i 1 t i i t i 1 t i tLAI ) CAY 1 V LAI ( CAY 1 WV (5-25) where WV1CAYt is the weighted crop coefficient on day t [-], which represents all existing plant species in the land segment. LAIi,t-1 is the leaf area of species i on day t-1 [m2 leaf/m2 ground], and V1CAYi,t is the crop coefficient of species i [-]. Using the lumped crop coefficient, a partitioning fraction (Fract [-]), with a range of values from 0 to 0.95, can be defined as: 2 0 CAY 1 WV 0 2 0 CAY 1 WV 8 0 ) 2 0 CAY 1 WV ( 95 0 Fract t t t (5-26)

PAGE 222

201 Thus the potential pl ant transpiration (Tpot,t), which is lumped over species and is required later for calculating the water stress factor (see Equation (5-44)), can be obtained: t t t potPET Frac T (5-27) Next, the lumped PET is applied from th e top soil layer downward to the deeper soil layers in the root zone. The layer actual evapotranspiration (LETj,t) is the minimum between the PETt and the available water storage (be yond the wilting point) for a specific soil layer j. If there is sti ll some PET left after applying in the upper soil layer, then the remaining PET will be applied to the next soil layer and so forth until all PET is used up or the bottom soil layer in the root zone is reached. Using the sa me method in Equation (5-27), the layer LETj,t can be split between layer soil water evaporation (AETS,j,t) and layer plant transpiration (AETT,j,t), the latter is required for later use in estimating nutrient uptake. t t j t j TFrac LET AET (5-28) t j T t j SAET 1 AET (5-29) Furthermore, the layer actual transpira tion can be split among species existing in one land segment using the leaf area rati os (Jones JW, private communication, 2006): i LAI k LAI k t j T t j i) e 1 ( e 1 AET T1 t i i ext 1 t i i ext (5-30) where Ti,j,t is the speciesspecific transpiration [ mm] and the index j indicates the number of soil layer. Nitrogen Uptake N uptake in GLEAMS (Knisel and David, 1999) was patterned after that in the EPIC model (Sharpley and W illiams, 1990) for estimation of nitrogen demand, but it was

PAGE 223

202 modified to consider the difference in n itrate and ammonium uptake by plants. All vegetation species differ in their affinity fo r nitrate or ammonia, but in the GLEAMS model nitrate and ammonia uptake are treat ed equally in those soil layers where transpiration occurs. In the vegetation m odel developed, the N uptake approach in GLEAMS was modified to account for the N uptake by multiple plant species in one land segment. 0.00 1.50 3.00 4.50 6.00 7.50 0.00.10.20.30.40.50.60.70.80.91.0 Growth ratio [-]Biomass N concentration [%] Figure 5-3. An example relationship between plant biomass nitrogen concentration and growth ratio (using empirical coefficients C1 = 1.25 and C2 = -0.278 in Fraisse and Campbell (1997). LAI was a ssumed to range from 0.01 to 5.2). In the N uptake process, the nitrogen concentration, CNi,t, of the plant biomass for species i on day t [%], is calculated using the equation: i2 C t i i t iGRT 1 C CN (5-31) where C1i and C2i are constant empirical input coefficients [-] and GRTi,t is the growth ratio [-], which is defined as: i t i t iPOTLAI LAI GRT (5-32) where LAIi,t is the leaf area [m2 leaf/m2 ground] for species at time t, and POTLAIi is the potential leaf area [m2 leaf/m2 ground] for species i, which represents the maximum leaf

PAGE 224

203 area without the constraints of water and nutrient limitations over the growing season. The relationship between the plant nitrogen concentration and growth ratio can be illustrated through the example diagram as seen in Figure 5-3. Using Equation (5-33) the total dry matter N, TDMNi,t, of species i at time t [kg/ha] can be estimated: t i red t i t iW CN TDMN (5-33) where Wred,i,t is the total biomass of species i esti mated in Equation (5-7). The daily N demand, DEMNi,t [kg/ha], for species i at time step t is then calculated as the difference between the dry matter N at two successive time steps, which represents the optimum N required to maintain plant growth without N limitation. 1 t i t i t iTDMN TDMN DEMN (5-34) The N supply, determined by the N availa bility in soils, can be calculated by multiplying the N concentration with transpir ation from the respective layers. In the model, N uptake is treated equally for both ni trate and ammonia. The amount of N taken up by species i from each soil layer j where tran spiration occurs, including nitrate uptake (UPNHi,j,t [kg/ha]) and ammonia uptake (UPNOi,j,t [kg/ha]) can be calculated: t j i j d s t j iT W 3 CNO UPNO (5-35) t j i j d s t j iT W 4 CNH UPNH (5-36) where CNO3Ws,d,j and CNH4Ws,d,j are the nitrate and ammonium concentration for the jth soil layer, respectively, determined fr om Equations (4-2) a nd (4-9) described in Chapter 4, and Ti,j,t is the transpiration that occurred fr om that soil layer for species i at time t determined from Equation (5-25). Th e total uptake of N is summed for each plant

PAGE 225

204 species over the number of soil layers where transpiration occurs and it increases the alive biomass N pool by ij t j i ij t j i t upUPNH UPNO N (5-37) where Nup,t is the total N uptake per day [kg/ha]. The N demand can be greater or less than the N supply. If demand is less than supply, then the uptake occurs in the amount calculated in Equations (5-35) and (5-36). Otherwise, a demand factor, th e ratio between the demand and availability, is calculated to reduce the uptake to the demand for both nitrate and ammonia at soil layers where transpiration occurs. 0.00 0.20 0.40 0.60 0.80 1.00 1.20 0.00.10.20.30.40.50.60.70.80.91.0 Growth ratio [-]Biomass P concentration [%] Figure 5-4. An example relationship between plant biomass phosphous concentration and growth ratio (using empirical coefficients C1 = 1.25 and C2 = -0.278 in Fraisse and Campbell (1997). LAI was a ssumed to range from 0.01 to 5.2). Phosphorus Uptake The algorithm for P uptake is similar to that for N uptake. The optimum P content of the plant biomass [%] is estimated from the nitrogen content (CNi,t in Equation (5-31)) and the N:P ratio, NPRi [-], an input to the model, as: i t i t iNPR CN CP (5-38)

PAGE 226

205 Figure 5-4 shows an example relationship between the plan t P concentration and the growth ratio. The total dry matter P, TDMPi,t [kg/ha] for species i at time t, is calculated using the equation t i red t i t iW CP TDMP (5-39) and the P demand for species i at time t, DEMPi,t [kg/ha], is determined by the difference between the dry matter P on two successive days as 1 t i t i t iTDMP TDMP DEMP (5-40) Uptake of labile P, UPLPi,t [kg/ha], is estimated for each layer where transpiration, AET,t, occurs using t j i j d s t j iT CPLABW UPLP (5-41) where the concentration of labile P, CPLABWs,d,t, is determined by Equation (4-15) in Chapter 4. The total uptake of P is the su m over all species i and all layers j where transpiration occurs. The P taken up is converted into the plant biomass P: ij t j i t upUPLP P (5-42) where Pup,t is the plant biomass P [kg/ha], whic h will be used to calculate plant P concentration for determining the P stress in the following section. Similar to N, a demand factor for P that is a ratio between th e demand and availability of P is calculated to reduce the uptake of P when supply exceeds demand. Growth Reduction Factor The growth reduction factor, RFi,t, used in Equation (5-5) is a unitless, speciesspecific growth reduction factor with a value ranging from 0 to 1, which is obtained by

PAGE 227

206 taking the minimum value of water stress, wa ter logging, and N and P stress factors as shown in the following equation: ) F F F F min( RFt i P t i N t i WL t i WS t i (5-43) where FWS,i,t is the water stress fact or for species i [-]; FWL,i,t is the water logging factor for species i [-]; FN,i,t is the N stress factor [-]; and FP,i,t is the P stress factor [-]. The details regarding these factors are discussed below. Water stress and logging factors Water is one of the limiting growth factor s due to its functi on in transporting nutrients and its assimilation in to the plants. The basis for plant production is the uptake of atmospheric carbon dioxide (CO2) for assimilation through the stomata in the epidermis of leaves. As a consequence, wa ter moves in the oppos ite direction through the stomata, a process referred to as transpir ation. If the water supply cannot meet the demand for water, it consequently limits the intake of CO2 and results in the decline of the assimilation rate. In this way, water sh ortages curtail plant pr oduction (Ivens et al., 1992). On the other hand, transpiration can be hampered in soils with moisture content above field capacity, as a resu lt of restricted water/oxygen uptake by the roots. This situation, referred to as wa ter logging (Ivens et al., 1992), is common characteristic of wetland soils. When soil moisture is above field capacity (Fc), soil tends to be saturated and water logging can occur. Under water logging c onditions, the growth of most plants is seriously hampered since root respiration, a prerequisite for the uptake of both water and nutrients by the roots, is suppressed due to oxygen shortage. On the other hand, when soil moisture is below the critical soil moisture (SWP) and before it reaches the

PAGE 228

207 permanent wilting point (PWP), moisture extr action is hampered, which leads to drought stress in plants. This causes a reduction in stomatal aperture and hence less water loss and impaired CO2 assimilation (Ivens, 1992). Both water stress and logging constrain pl ant growth. However, different plant species can tolerate water stress or logging to a varying extent and thus these factors are species-specific. Therefore, these two fact ors are very important in the context of investigating the impact of water retention on ranches. The water st ress factor of species i, FWS,i,t can be expressed as a ratio between th e actual transpiration and the potential transpiration for species i: t pot j t j i j i WST AET F (1 F 0t i WS ) (5-44) where AETi,j,t is the layer actual transpiration [mm] fo r species i in layer j at time t; the index j represents the number of soil layers where transpiration occurs; and Tpot,t is the potential transpiration [mm]. When the soil moisture goes above the fi eld capacity, the water logging factor of species i, FWL,i,t [-], can be expressed as 3 0 W W W W 7 0 Ffc po s po t i WL ( 1 F 0j i WL ) (5-45) where Wpo is the soil porosity [-], Ws is the current soil moisture content [-] at time t, and Wfc is the field capacity [-]. For soil layers where transpiration occurs, these soil parameters may be different. To account fo r these differences, thei r weighted values by the root fraction are used in Equation (5-45) to calculate the wa ter logging factor.

PAGE 229

208 Duration of inundation is another important f actor considered in this model because prolonged standing water may eventually kill some plants that cannot tolerate inundation for a long period. To account for the inundati on duration factor, the successive days of inundation (it is assumed that surface ponded wa ter has to exceed 2 cm to be considered inundation) are recorded in the model for each species. Before the number of accumulated days of inundation reaches th e maximum length of inundation tolerance (species specific input to the model), plants are assumed to grow at a reduced rate defined in Equation (5-45). Once the maximum inunda tion tolerance is reached, it is assumed 25% of aboveground and belowground biomass di es daily while inundation continues. During this time it is assumed that no growth for this species occurs. Nitrogen stress factor The N stress, FN,i,t, a factor that determines the growth reduction for species i at time t caused by N deficiency, was modele d according to Seligman and Van Keulen (1981), and is dependent on the actual N con centration of the shoot for species i, CNshoot,i,t [g N/g DM], which is calculated by dividing the plant biomass Ns,i,t [kg N/ha] by the plant biomass for species i [kg DM/ha]: t i red t i t i shootW N CN (5-46) where Wred,i,t is the plant biomass for species i at time t obtained in Equation (5-7); and Ni,t is the biomass N for species i at time t, which is obtained from Equation (5-8). If CNshoot,i,t exceeds the critical value CNcr,i [g N/g DM], plant gr owth is not reduced. Below this N concentration, growth rate decreases linearly with concentration: i min, i cr i min, t i shoot t i NCN CN CN CN F (1 F 0i N ) (5-47)

PAGE 230

209 where Nmin,i is the minimum shoot N concentration [g N/g DM] required to maintain the plant growth input to the model by users. Phosphorus stress factor The P stress, FP,i,t, is a factor that determines the growth reduction for species i at time t caused by P deficiency. Similar to the N stress factor, FP,i,t can be calculated as: i min, i cr i min, t i shoot t i PCP CP CP CP F (1 F 0t i P ) (5-48) where CPshoot,i,t is the shoot biomass P concentration for species i at time t, which can be obtained by dividing the biomass P (Pi,t [kg N/ha]) by the aboveground biomass [kg DM/ha]: t i red t i t i shootW P CP (5-49) If CPshoot,i,t exceeds the critical value CPcr,i [g N/g DM], plant growth is not reduced. Below this P concentration, growth rate decreases linearly with concentration. Hypothetical Scenario Model Testing Scenario Description Sufficient vegetation data ar e not currently available fr om south Florida flatwoods watersheds to quantitatively te st the vegetation model. Ther efore, a hypothetical test was designed to evaluate the algor ithms of the vegetation model, coupled with the hydrologic and nutrient models developed in Chapters 3 and 4 to determine whether the model behaves as expected when the hydrology and nutrient concentrations are manipulated.

PAGE 231

210 Figure 5-5. Aerial photo show ing the layout of the improved summer pasture site S4 and location of associated instrumentatio n. S1 to S6 indicate the individual summer pasture sites and LS1 to LS12 i ndicate the land segments divided for the site S4. The dotted lines were ma de to show the boundary between land segments (modified from MAERC, 2004). A simple scenario was designed based on the summer pasture S4 (see Figure 5-5) at Buck Island Ranch (described in Chapter 4). To simulate the changes in vegetation dynamics, a water retention BMP was assumed to have been applied on this site, i.e. the water level at the pasture drainage outlet (lo cated in the southwest corner as shown in Figure 5-5) was controlled so that discharges did not occur until the water level inside the pasture reached the design water level which wa s above the pre-BMP discharge level. To realize this scenario in the model, a weir with an elevation of 8.33 m was set up on the boundary between land segment 11 (hereaf ter LS11), which has a ground surface elevation of 7.98 m, near th e outlet and the downstream. T hus after implementation of S3S4S2 S1 S5 S6 LS3 LS4 LS1 LS2 LS5 LS6 LS7 LS8 LS9 LS10 LS11 LS12

PAGE 232

211 the BMP discharges will occur only when the water elevation of the outlet land segment exceeds 8.33 m. Without the application of water retention BMP, discharge would occur when the water elevation reached 7.985 m (m aximum depressional storage in depth = 0.5 cm). Identical hydrologic and nutrient inputs, as described in Chapter 4, were used for running the coupled model to investigate the vegetation dynamics due to the application of the hypothetical BMP for the pasture site S4. Agricultural activities including fertilization, burning, and stocking were not included, so that the impact due to water retention on vegetation dynamics could be anal yzed independently. By holding the water levels higher at the outlet, th e soil inundation period is prolon ged in the pasture and thus the length of anoxic soil conditi ons is extended as well. Th is type of hydrologic change may affect nutrient cycling processes that ma y alter the concentrations of N and P. Changes in both hydrologic regi me and nutrient cycling may impact plant growth. Vegetation species that do not adapt well to wet soil conditions for prolonged periods may experience growth reduction and eventually vanish from the site while others that adapt well to such conditions may expand their areas. Over time, the spatial composition patterns of vegetation species may change, and th ese patterns may be used as an indicator of soil and water conditions. To compare the changes in species composition pattern over time and space, the model was run twice to obtain simulation results before and after holding the water at a higher level (h ereafter called pre-BMP and post-BMP, respectively). Multiple plant species exist in the pastur e site S4 as seen in Table 5-1, among which bahiagrass (Paspalum notatum Flgge) is the most dominant species accounting

PAGE 233

212 for up to 93% of the land cover, while the remaining species account for about 7% together. However, for this test three pere nnial species including bahiagrass, floralta (Hemarthria altissima), and panicum (panicum rigidulum) were assumed to be the major species for the pasture site S4. These vegeta tion species represent those adapted to soil conditions of uplands, transition zo nes and wetlands, respectively. S4 was divided into 12 land segments (see Fi gure 5-5), as discussed in Chapter 4. Each land segment was assumed to have th e same amount of initial total biomass (aboveground + belowground) over all selected sp ecies, but different percentages of these species (see Table 5-3). The hypothetical assignment of spec ies and percentage of each species was made up according to the locations of land segments. Land segments with higher topographical elevations, far from wetland s were assumed to be relatively drier in their soil conditions than those located near or in wetlands and thus to contain more upland plant species. With the initia l biomass, shoot to root ratio (fa), and specific leaf area (SLA) (see Table 5-2), the model can initialize the aboveg round and belowground biomass and green LAI for each species in each land segment. To run the vegetation model, inputs includ ing the time series inputs (see Table 5-2) such as daily solar radiation, maximum and minimum ambient temperatures and the model parameter values must be provided. Other time series data such as grazing, harvest, burning, and tillage must be specifi ed as well if applicable. The required model parameter values were obtained from the literature if they were available. However, the data for floralta and panicum we re especially difficult to fi nd and thus they were assumed to have the same values for some parameters as the bahiagrass. Collecting and setting up these species-specific parameters was one of the challenges encountered during

PAGE 234

213 development of the vegetation model. Table 5-4 lists those variables calculated through the hydrologic and nutrient models and used in the vegetation mode l and Table 5-5 lists major daily output variables from the vegetation model. Results and Discussion Water and nutrient responses to water retention BMP Figures 5-6 to 5-11 show the comparison of surface runoff, water table depth, and N and P loads and concentrations predicte d by the integrated hydrology, nutrient and vegetation dynamics simulation model before and after BMP for site S4 for a 6-year simulation period. It is noted that after holdi ng the water level at 8.33 m at the outlet, no surface runoff (see Figure 5-6) and hence no N and P loads (see Figures 5-8 and 5-9) are released downstream throughout the entire si mulation period. Figure 5-7 shows that the ponded water remains above the ground surface for much of the wet season after the water retention BMP is implemented. A period from approximately September 2000 to February 2001 was drier than the rest of simu lation period, thus during this time the water tables stayed fairly deep even with the water retention BMP. Several periods between the late spring and early summer in years 1999, 200 0 and 2002 were also relatively drier with long-term water table drawdown (see Figure 57). After the BMP, the groundwater table depths were much shallower than those before BMP except during the driest period discussed above. Meanwhile, several long pe riods occurred during which soils became saturated and the water table remained on th e ground surface for approximately 6 months. This occurred primarily from the middle of summer to the early spring between years 1998 and 1999, 1999 and 2000, 2001 and 2002, and 2002 and 2003. The ponded water depth during these periods (see Figure 5-7) indicates that the soils experience prolonged saturation and had standing water on the ground surface. Note that Figure 5-7 was made

PAGE 235

214 using the predicted water tables for LS9 wh ere the observation well was installed and the observations from Figure 5-7 may not a pply to the entire pasture site. The simulated results of P and N concentrations in surface water in Figures 5-10 and 5-11, respectively indicate that both P and N concentrations were dramatically decreased after the water retention BMP. The lower P and N concentrations likely result from changes in biochemical processes due to prolonged anoxic soil conditions both in soils and on surface residue. Overall, it was observed that after the water retention BMP, the hydrologic and nutrient dynamics were signi ficantly changed in pasture site S4. The pasture soils become water logged for prol onged periods and nutrient concentrations become significantly lower. Influence of differences in temperature sensitivities among species The sensitivity of biomass response to th e temperature function was recognized during the development of this vegetation m odel. To investigate the influence of different temperature functions on the predictions of bioma ss, three scenarios shown in Table 5-6 were designed. In Scenario 1, an identical temperature function was applied for all three species; in Scenario 2, the te mperature function for floralta was different from those for other two species, bahia and panicum, which were assumed to have the same temperature function as specified in Scenario 1. The temperature function for floralta in Scenario 2 was modified by shifting -7 C from the one used for the other two species, allowing floralta to grow in a lower temperature range from 0 C to 38 C. In the last scenario, the temperature functions fo r bahia and panicum were assumed to be the same as those for them in Scenarios 1 and 2, but the temperature f unction for floralta was further adjusted to allow this species to grow primarily during winter time, in a narrower

PAGE 236

215 temperature range from 0 C to 30 C. In order to investigate the sensitivities of biomass productivity specifically to temperature factor s, all these scenario tests were performed without simulating the growth reduction, that is, potential biomass production was compared. Figure 5-12 shows that the predicted aboveground biomass on LS11 varied dramatically among the three sc enarios corresponding to differe nt temperature functions. When the identical temperature function was used for all three species in Scenario 1, the predicted potential biomass had similar trends and patte rns throughout the simulation period but was different in magnitude among spec ies due to the difference in their initial conditions. This may imply that in this scen ario the three species grow in a relatively stable situation without competition with one another. In Scenario 2, the predicted potential biomass for floralta increased tremendously over time whereas the biomass for bahia and panicum decreased and these speci es tended to disappear from this land segment. This behavior can be explained th rough the temperature factors for Scenario 2 shown in Figure 5-13. Compared with ba hia and panicum, floralta apparently experienced less growth reduction during the winter time in S cenario 2. Even during the summer time, floralta did not suffer much grow th reduction. This situation allows more time for floralta to grow in compar ison to the other two species. In Scenario 3, an opposite response from Scenario 2 occurred for predicted biomass for all species. The temperature factors us ed for Scenario 3 and shown in Figure 5-13 allow both bahia and panicum to grow faster dur ing the summer time while floralta is in its lowest growth time. The resulting incr eased biomass for bahia and panicum leads to senescence, which reduces the existing biomass for floralta in the same proportion.

PAGE 237

216 Although floralta grows better in winter time in this scenario, the amount of biomass produced during the winter time did not sufficiently balance the loss due to the senescence during the summer time. Over tim e the production of floralta decreased and gradually this species vanished from this land segment. These analyses, based on hypothetical temp erature functions, demonstrated that a reasonable, species-specific temperature functi on is very important for reliable simulation results. Temperature sensitivities of differe nt species are needed in order to correctly simulate the changes in competition of speci es over time and space and in response to BMPs. Nevertheless, such data are not yet available for these species. Therefore, for the remaining analyses in this study, the same temperature function (see Table 5-6) was assumed for all three species for simplicity. In addition to sensitivity to temperature, these species may differ in their responses to photoperiod as well. Phot operiod effects were not in cluded in the current method because of lack of information. However, differences in photoperiod sensitivity could also have a big impact on seasonal growth and species compositions (Jones JW, private communication, 2006). Species composition dynamics due to water retention BMP Figure 5-14 shows relative plant distri bution (calculated using the aboveground biomass for each species at the end of each year of the simulation period) over land segments throughout the entire simulation pe riod before and after BMP implementation in the summer pasture S4. The distribution in each land segment is expressed as the percentage of total aboveground biomass (s ince the belowground biomass follows the same patterns, it is not discussed). Simulation results from the pre-BMP scenario case indicate that the basic composition pattern s over land segments are approximately the

PAGE 238

217 same over the entire simulation period, and no species vanished from any land segment where it was initially assigned. Bahia account ed for the most significant percentage of biomass over the entire pasture throughout the 6-year simulation period, followed by panicum, and then floralta. The results befo re the water retention BMP indicate that preBMP soil conditions were relatively favorable for the coexistence of three species. After the water retention BMP was applied, the composition patterns of species over the land segments throughout the simulation period changed. In the first three years, the difference of biomass distribution among species over land segments was not significant, and the composition patterns of speci es were very close to those for the preBMP conditions. Although there was some rain fall variability over the first three years (as discussed in Chapter 4) this variability was not enough to produce changes in biomass productivity among these species. However, af ter year 2001, a dramatic change in the biomass distribution pattern occurred in the la nd segments close to the drainage outlet, including LS5, LS7, LS8, LS9, LS10, LS 11, and LS12. Panicum grew significantly faster in those land segments, and floralta showed a similar trend but with a smaller magnitude. Meanwhile, the bahia biomass decr eased dramatically and almost vanished from LS12, LS11, LS10, and LS9. Additionally floralta decreased significantly in LS11 where the drainage outlet is located. Due to the lack of a mechanism for new species to grow in land segments where they are not originally present, there was no noticeable change observed in LS6. The distribution pa ttern in 2001 remained similar in years 2002 and 2003, very likely due to the consecutive we t weather conditions. Overall, changes in biomass distribution for the last three y ears under the water rete ntion BMP scenario demonstrate that the soil conditions became more favorable for the development of

PAGE 239

218 panicum, the wetland species; less favorable fo r floralta, the transi tion zone species, and least favorable for baiha, the upland species. Figure 5-14 provides insight into the re lative percentages of plant biomass distribution, but it lacks information on the absolute biomass of each species over land segments on a daily basis. Figures 5-15 a nd 5-16 show the continuous simulation of aboveground biomass for bahia, floralta, and panicum thr oughout the 6-year simulation period for the preand post-BMP scenarios in three selected land segments (LS8, LS11 and LS12), respectively. As seen in Figur e 5-15, the aboveground biomass patterns for all species in these selected land segments are basically similar over time for the preBMP scenario, and the magnitude of bioma ss for each species stayed stable over the simulation period. However, Figure 5-16 s hows a significant change occurring late during year 2001 for the post-BMP scenario as observed in Figure 5-14. Bahia biomass was reduced significantly in LS 8 around the end of year 2001, but began to grow again in year 2002 and 2003 while panicum and floralta in creased their growth in all years. In LS11, bahia died out around the end of 2001, followed by floralta in 2001, while at the same time panicum maintained aggressive gr owth through the entire simulation period. Since LS11 is the land segment immediately conn ected to the outlet of pasture, after the water retention BMP, it became the most inundated location in the pasture, which eventually caused the die-off of bahia and floralta. The growth dynamics in LS12 was similar to that in LS8. The difference was that bahia was not able to grow back again. This can be explained by the difference in soil conditions in LS8 and LS12. The latter apparently had longer saturated soil conditi ons than the former. Comparison of the continuous biomass simulation results for the preand post-BMP scenarios further

PAGE 240

219 support the conclusion that the application of the water retention BMP caused significant changes in vegetation dynamics. Figures 5-19 and 5-20 show the growth limiting factors for bahia in LS11 for the preand post-BMP scenarios. Consistently, in both scenarios, water stress and water logging were more influential factors than the N and P stress factors. This is reasonable for a pasture site like S4 that has a hi gh background nutrient concentration. The comparison of water stress and logging factors for preand post-BMP scenarios indicates that the frequency of water stress was decreased significantly, while that for water logging was increased, after the water retenti on BMP. This result demonstrates that the model is capable of capturi ng water dynamics due to the water retention BMP. Concluding Remarks To simulate vegetation dynamics cause d by changes in hydrologic and nutrient dynamics in south Florida flatwoods waters heds, a vegetation model was developed and coupled with the hydrologic and nutrient models developed earlier. In this model, plant biomass production is mainly driven by climat e (solar radiation a nd temperature) and affected by growth reduction factors, includ ing water stress, water logging, and N and P stresses. Total production is simply partitioned between aboveground and belowground biomass and the leaf area is obtained by multiplying the aboveground biomass by the specific leaf area. Three pere nnial plant species, bahia, floralta, and panicum, were selected as representative of plants adap ted to upland, transition zone and wetlands growth conditions, respectively. An analysis of the influence of difference temperature sensitivities among species i ndicate that reasonable, sp ecies-specific temperature functions are needed to correc tly simulate the plant competition over time and specie and simulate response to BMPs.

PAGE 241

220 A hypothetical BMP scenario based on the su mmer pasture site S4 at Buck Island Ranch, Lake Okeechobee Basin, Florida wa s developed to take advantage of the calibrated hydrologic and nutrient parameters computed in Ch apter 4. In this scenario, it is assumed that a water retention BMP that retained all water below an elevation of 8.33 m was implemented in pasture S4. Due to wa ter retention, the soil in the pasture was inundated much longer than when the water was allowed to flow freely off the pasture. The changes in hydrology and hence in nutrient dynamics were demonstrated to have the potential to change vegetation. The hypothetical scenario test showed that after implementing the water retention BMP, all surface runoff and nutrients were retain ed inside the pasture site and the periods of water logging or inundation increased. With these changes in hydrology and nutrients, wetland species like panicum survived better th an bahia and floralta while the transition zone species floralta did better than upla nd species bahia. The spatial distribution patterns of biomass were distinct over time for the preand post-BMP scenarios and significant dynamics in the magnitude of biomass and the number of species were observed through the predicted results. Consequently, it was concluded that this vegetation model is capable of simulating vegetation dynamics caused by changes in hydrology and nutrient cycling due to water retention BMPs. However, the model input parameters, mostly assumed or taken from the literature, may not be sufficient to accurately represent physiologi cal and phenological characteristics of the selected species. Furthermore, the model performance was evaluated in a hypothetical scenar io, not against actual field da ta. In order to use this model for prediction, future efforts in colle cting vegetation data a nd evaluating the model

PAGE 242

221 with observed data and experimentally-deriv ed parameters should be undertaken. In addition, the model needs to be further devel oped to enable simulation of root length and root biomass distribution to more accura tely capture the nutrient uptake and evapotranspiration differences among species and an algorithm fo r simulating plant height should be added to enable the fee dback relationship between surface runoff and a dynamic surface roughness, which is a function of vegetative cover. A start-up mechanism for new species initiation in land segments should be added to more accurately simulate plant competition and spat ial evolution. Finally, an herbivore movement model should be developed to simu late the impacts of cattle behavior and grazing performance on vegetation growth.

PAGE 243

222 Table 5-1. Percent cover of vegetatio n on summer pastures S4 (MAERC, 2004). Scientific Name Common Name S4 Scientific Name Common Name S4 Paspalum notatum bahiagrass 93% Paspalum urvillei vasey grass 1% Axonopus affinis carpet grass 1% Juncus effuses softrush 2% Cynodon dactylon bermuda grass Cyperaceae spp. sedges 1% Setaria geniculata Foxtail Eupatorium apillifolium dog fennel Paspalum dilitatum dallis grass Phyla nodiflora Lippia Centella asiatica Centella 1% Hydrocotyle umbellata pennywort Sporobolus indicus smut grass Polygonum sp. smartweed <1% Andropogon glomeratus Bluestem 1% Table 5-2. Parameter values for the different plant species for th e vegetation model. Symbol Description Unit bahia floralta panicum I0 Incoming photosynthetic active radiation MJ/m2/day Variable over timea Tmin Daily Minimum ambient temperature C Variable over timea Tmax Daily Maximum ambient temperature C Variable over timea LAImin Minimum green LAI m2 leaf/m2 land 0.10c 0.10c 0.10c LAImax Potential green LAI m2 leaf/m2 land 8.0c 8.0c 8.0c LAIcr Critical LAI m2 leaf/m2 land 5.0c 5.0c 5.0c fa Dry matter partitioning coefficient 0.33c 0.33c 0.33c k Light extinction coef ficient of species 0.5d 0.5d 0.5d Tmin Base temperature for growth C 7g 7g 7g Topt1 Optimum temperature 1 below which light interception increases C 28g 28g 28g Topt2 Optimum temperature 2 above which light interception decreases C 35g 35g 35g Tmax Base temperature for growth C 45g 45g 45g RUE Radiation use efficiency g MJ/PAR 3.0d 3.0d 3.0d SLA Specific leaf area cm2 leaf/g leaf 64.8b 64.8b 64.8b L Maximum water logging limit day 28c 55c 120c C1 Plant NC1 1.25e 1.25e 1.25e C2 Plant NC2 -0.278e -0.278e -0.278e C:N Carbon: nitrogen ratio 80e 80e 80e N:P Nitrogen Phosphorus ratio 6.7e 6.7e 6.7e CNmin Shoot minimum N concentration g N/g DM 0.005f 0.005f 0.005f CNcr Shoot critical N concentration g N/g DM 0.0135c 0.0135c 0.0135c CPmin Shoot minimum P concentration g N/g DM 0.005d 0.005d 0.005d CPcr Shoot critical P concentration g N/g DM 0.0135d 0.0135d 0.0135d fbs Declining rate due to plant senescence 0.0714c 0.0714c 0.714c aMeasured through Buck Is land project (MAERC, 2004); bFor bahiagrass by Stuart (2004); cAssumed parameter values; dParameter values for Agrostis stolonifera Phragmites australis and Salix alba by Van Oene et al.(1998); eParameter values for bahiagrass by Fraisse and Campbell (1997); fParameters values for Holcus Anthoxanthum and Festuca by Schippers and Kropff (2001); gTemperature function curve for the vegetative phase by Jones et al. (2000).

PAGE 244

223 Table 5-3. Initial composition percentage a nd amount for species in each land segment. Composition in percentage [%] Composition in biomass [kg/ha] Land segment Bahia Floralta Panicum Bahia Floralta Panicum LS1 100 5000 0 0 LS2 80 20 4000 1000 0 LS3 20 10 70 1000 500 3500 LS4 85 10 5 4250 500 250 LS5 40 20 40 2000 1000 2000 LS6 90 10 4500 500 0 LS7 60 20 20 3000 1000 1000 LS8 70 20 10 3500 1000 500 LS9 20 20 60 1000 1000 3000 LS10 60 20 20 3000 1000 1000 LS11 30 20 50 1500 1000 2500 LS12 70 20 10 3500 1000 500 Species does not exist in this land segment. Table 5-4. Variables calculated from other models. Symbol Description Unit Source CNH4W Soil layer nitrate concentration mg/l Calculated CNO3W Soil layer ammonium concentration mg /l Calculated CPLABW Soil layer labile P concentration mg/l Calculated Tpot Potential transpiration mm Calculated Tact Actual transpiration mm Calculated Table 5-5. Output variables from the vegetation model. Symbol Description Unit Source W Total dry matter kg/ha Calculated Wa Aboveground dry matter kg/ha Calculated Wb Belowground dry matter kg/ha Calculated LAI Green leaf area index m2 leaf/m2 ground Calculated Table 5-6. Testing scenarios for temperature functions. Scenario Species Tmin [ C] Topt1 [ C] Topt2 [ C] Tmax [ C] Bahia 7 28 35 45 Floralta 7 28 35 45 Scenario 1 Panicum 7 28 35 45 Bahia 7 28 35 45 Floralta 0 21 28 38 Scenario 2 Panicum 7 28 35 45 Bahia 7 28 35 45 Floralta 0 15 20 30 Scenario 3 Panicum 7 28 35 45

PAGE 245

224 -1.00 -0.50 0.00 0.50 1.00 1.50 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Surface runoff (cm ) 0 10 20 30 40 50 60 70Rainfall (cm) Rainfall pre-BMP post-BMP Figure 5-6. Continuous simulation of su rface runoff throughout the simulation period from 1998 to 2003. -175 -155 -135 -115 -95 -75 -55 -35 -15 5 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Water table depth (cm)-175 -155 -135 -115 -95 -75 -55 -35 -15 5Rainfall (cm) pre-BMP WTD post-BMP PWD post-BMP WTD pre-BMP PWD Figure 5-7. Continuous simulati on of water table depths at land segment 9 throughout the simulation period from 1998 to 2003 (W TD = water table depth; PWD = ponded surface water depth).

PAGE 246

225 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P load (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall pre-BMP post-BMP Figure 5-8. Continuous simula tion of P loads throughout the simulation period from 1998 to 2003. -5 0 5 10 15 20 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N load (kg/ha)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall pre-BMP post-BMP Figure 5-9. Continuous simu lation of N loads throughout the simulation period from 1998 to 2003.

PAGE 247

226 0 50 100 150 200 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall pre-BMP post-BMP Figure 5-10. Continuous simulation of P con centrations throughout the simulation period from 1998 to 2003. 0 300 600 900 1200 1500 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N Concentration (mg/l)0 10 20 30 40 50 60 70Rainfall (cm) Rainfall pre-BMP post-BMP Figure 5-11. Continuous simulation of N con centrations throughout the simulation period from 1998 to 2003.

PAGE 248

227 Scenario 1 0 2000 4000 6000 8000 10000 12000 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Scenario 2 0 2000 4000 6000 8000 10000 12000 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Scenario 3 0 2000 4000 6000 8000 10000 12000 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Figure 5-12. Comparison of predicted potenti al aboveground biomass for species in land segment 11 with different temperature functions.

PAGE 249

228 Scenario 3 0.00 0.20 0.40 0.60 0.80 1.00 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Temperature factor (-) Bahia Floralta Panicum Scenario 2 0.00 0.20 0.40 0.60 0.80 1.00 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Temperature factor (-) Bahia Floralta Panicum Scenario 1 0.00 0.20 0.40 0.60 0.80 1.00 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Temperature factor (-) Bahia Floralta Panicum Figure 5-13. Comparison of temperature factor s among the three selected species in land segment 11.

PAGE 250

229 Inital (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 1998 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 1999 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Initial (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 1998 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 1999 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Figure 5-14. Species distribution (percent of total aboveground biomass at the end of each year) over land segments for bahi a, floralta and panicum throughout the simulation period from 1998 to 2003.

PAGE 251

230 Year 2000 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2001 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2002 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2003 (before BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2000 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2001 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2002 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Year 2003 (after BMP)0 10 20 30 40 50 60 70 80 90 100 123456789101112 Land segment (-)Distribution of total abovegroun d biomass (kg/ha) Bahia Floralta Panicum Figure 5-14. Continued

PAGE 252

231 Land Segment 8 (pre-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Land Segment 11 (pre-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Land Segment 12 (pre-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Figure 5-15. Comparisons of continuous simulation of aboveground biomass for all species in the selcted land segments 8, 11, and 12 before water retention BMP.

PAGE 253

232 Land Segment 12 (post-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Land Segment 11 (post-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Land Segment 8 (post-BMP) 0 2000 4000 6000 8000 10000 1/1/981/1/991/1/001/1/011/1/021/1/03 Time (day)Aboveground biomass (kg/ha) Bahia Floralta Panicum Figure 5-16. Comparisons of continuous simulation of aboveground biomass for all species in the selcted land segments 8, 11, and 12 after water retention BMP.

PAGE 254

233 Comparsion of N Stress Factors before and after BMP 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)N stress factor (-) pre-BMP N stress post-BMP N stress Comparsion of P Stress Factors before and after BMP 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)P stress factor (-) pre-BMP P stress post-BMP P stress Comparsion of Water Logging Factors before and after BMP 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Water logging factor (-) pre-BMP water logging post-BMP water logging Comparsion of Water Stress Factors before and after BMP 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1/1/987/1/981/1/997/1/991/1/007/1/001/1/017/1/011/1/027/1/021/1/037/1/03 Time (day)Water stress factor (-) pre-BMP water stress post-BMP water stress Figure 5-17. Growth limiting factors for ba hiagrass in land segm ent 8 before water retention BMP was applied.

PAGE 255

234 CHAPTER 6 SUMMARY AND CONCLUSIONS Lake Okeechobee, located at the center of the Kissimmee-Okeechobee-Everglades aquatic ecosystem in south Florida, is experi encing water quality degradation. Non-point agricultural runoff from dairies and cow-calf operations in the northern watershed of the lake is considered to be the primary sour ce of excess phosphorus (P ) loading discharged into the lake. In order to protect th e water quality of Lake Okeechobee and reach environmental restoration goals, a variety of alternative land management practices have been implemented in the Lake Okeechobee wate rsheds. To evaluate the effectiveness of those practices, a coupled ecohydrological model integrating hydrology, nutrient and vegetation dynamics was developed. The coupled model was developed with in the Java-based object-oriented framework of the existing ACRU2000 model (Campbell et al., 2001; Clark et al., 2001; Kiker and Clark, 2001). The primary objective of this study was to develop a coupled modeling system capable of simulating hydr ology, nutrient and vegetation dynamics simultaneously for south Florida flatwoods wate rsheds that incorporate wetlands, uplands and transition zones. In this effort, a hydrologic simulation m odel capable of multidirectional spatial simulati on of surface and subsurface water movement was first developed (Chapter 3). The existing nutrient components in ACRU2000 were modified to enable multi-directional spatial transport of N, P, and conservative solute (Chapter 4). Finally, a vegetation dynamics simulation m odel was developed to couple with the

PAGE 256

235 hydrologic and nutrient models to enable the simulation of the plant growth dynamics under the impacts of changes in water and nutrient availa bility (Chapter 5). Hydrologic Simulation Model The hydrologic model capable of mul ti-directional spatial simulation was developed based on the existing hydro logic model in ACRU2000 by adding new hydrologic components and modifying the existing components. New components include an overland flow component capable of spatial surface wa ter exchange between land segments and a lateral groundwater co mponent capable of spatial soil water exchange between saturated soils of adj acent land segments. These components are important in that they link different land se gments together as an entity through lateral water movement and provide the pathways capable of spatial nutrient transport. A simulation sequence analysis was conducte d to analyze the effect of simulation sequence on model predictions, and also to compare with well-known physically based models (MIKE SHE and MODFLOW). Results indicate that different simulation sequences do lead to slightly different simula tion results, but they are all consistent when compared with MIKE SHE (DHI, 2004). The results also indicate that the simulation sequence following the expected flow dir ections based on topogra phy should be chosen when there are multiple combinations of simulation sequence to choose from. A test was conducted for the Dry Lake Dair y #1 site in the Kissimmee River Basin, Florida to compare the ca pacity of the modified ACRU2000 hydrologic model in simulating surface runoff and groundwater ta ble to those predicted by the lumped FHANTM model. From the comparisons and statistical analyses, it can be concluded that the modified ACRU2000 hydrologic model is capable of adequately simulating overland flow and lateral groundwater flow at the Dry Lake Dairy #1 site, but the

PAGE 257

236 modified ACRU2000 model is no t better than FHANTM. It is possible that ACRU2000 would perform better if it were calibrated to the observed data, particularity if sufficient data were available to calibrate spatially distributed parameters. Nutrient Simulation Model The nutrient model capable of multi-directional spatial transport of N, P and conservative solute was developed base d on the existing nutrient module ACRU-NP (Campbell et al., 2001) in the ACRU2000 m odeling system by adding new lateral transport nutrient components. In order to investigate the accuracy of th e conservative solute transport component, the hypothetical scenario test was conducted by comparing with a par ticle tracking model PMPATH (Chiang and Kinzelbach, 2005). The test results indicat e that the solute transport predicted by the modified ACRU2000 is in reasonable agreement with those by PMPATH. Due to the constraints of PMPATH in simulating solute concentration, the model performance in predicting conser vative solute is qualitative. To evaluate the complete performan ce of the coupled hydrologic and nutrient model, an application was conducted for the be ef cattle pastures at Buck Island Ranch, Lake Okeechobee Basin, Florida. A comple te model testing procedure was performed including model calibration, verification, and sensitivity analysis using the measured data over a 6-year period. The sensitivity analys es indicated that the hydrologic parameters including crop coefficient, por osity, and restrictive layer hydraulic conduc tivity are the most sensitive parameters to the hydrologi c response. The sensitivities of nutrient parameters tend to be more site-specific. Neither P nor N responses was found to be sensitive to either stopcking or fertilizer rate for any site in this application. Rainfall N and P are the two most sensitive parameters to the N and P load responses, respectively

PAGE 258

237 for the low background nutrient winter pastur e site, while and the active P and active N are the two most senstivite parameters for the high nutrient background summer pasture site. Statistics were calculated to evalua te the model performance for calibration and verification. The comparison between statis tics obtained using the predictions from ACRU2000 and FHANTM indicate that the modi fied ACRU2000 did a much better job in simulating hydrologic and nutrient processes. Results indicate that the coupled model can predict hydrologic variables (surface runo ff and groundwater elevations) with high accuracy, but that it predicts nutrient loads (N and P) with relatively low accuracy. However, it is difficult to evaluate the model performance on its nutrient simulation capacities based on the limited nutrient observa tions available. Future work may be needed to sufficiently judge the model performance, especially with regards to nutrient simulation. Vegetation Dynamics Simulation Model The vegetation dynamics simulation model was developed, within the framework of previously developed hydrologic and nutrien t models, to simulate plant growth for pasture and wetlands species. In the mode l, solar radiation and temperature are considered to be the major drivers to predic t plant growth while e nvironmental impacts, expressed in stress factors (nutrient deficiency and water availability), determine if the plant can grow at its potential rate. Due to the lack of vegetation data available for flatwoods soils, the testing of this model was based on a hypot hetical scenario of water retention BMP scenario at summer pasture S4 at the Buck Island Ranch. The simulated results indicate that temperature sensitivities among species are needed to correctly simulate the plant growth. Th e hypothetical scenario demonstr ated that wetlands species, represented by panicum, can tolerate prolonge d water logged soil c onditions much better

PAGE 259

238 than transition zone and upland species represen ted by floralta and bahi a, and that floralta is more tolerant in those conditions than bahi a. However, further testing is needed when sufficient reliable vegetation data become available. Implications of the Research As the first working version, the mu lti-directional spa tial, ecohydrological simulation model marks a significant step in adapting the existing ACRU2000 model to better represent hydrology, nutri ents and vegetation dynamics for south Florida flatwoods watersheds. Viable mechanisms have been added to enable simulation of some of the unique aspects of flatwoods hydrology. Adapta tion to other watersheds or soil conditions at different temporal and spatia l scales should be simple using adjusted parameters or by adding relevant process components into this version. However, additional testing and calibration are still needed to improve the r obustness of the model fo r general use in the future. Future Research Recommendations The following recommendations are made for those who may be interested in continuing this work in the future. Model Preand Post-processing Capacity A user graphical interface (GUI) program capable of model preand postprocessing should be added into the modified ACRU2000. This is especially necessary when the model runs in larger watersheds wh ere more land segments must be defined and parametertized. In the current version, each land segment requires its own land segment input file that includes all required initial, boundary and parameter data. The data required in this input file may confuse us ers without detailed know ledge of the model structure. In addition, in creasing the number of land se gments leads to a tremendous

PAGE 260

239 workload in preparing these land segment input files, as well as for adjusting parameters required in model calibration and sensitivity analysis. Pre-processing capacity providing a simple interface would make this process more efficient. Furthermore, a post-processing program is necessary as well. The current version of the model generates output files by la nd segments. When users need to compare output variables over land segments, this must be done through spreadsheet software. It is difficult to observe the simulation result s immediately without additional tedious and troublesome work to process the outputs. A dditionally, the current output procedures do not allow output of integer variables. This deficiency should be corrected. Use Consistent Units for Parameters and Variables Use of consistent units throughout the modeli ng system is necessary to make use of the model easier and to minimize input errors. Unfortunately, different unit systems still exist in the current ve rsions of ACRU2000. Documentation A complete documentation regarding the modifications and additions since the version of ACRU (v3.00) should be complete d to provide the potential users an updated manual. Potential Changes to Existing Objects A redesign for the control object ( AAcru2000StandardProcesses) should be considered. This object is used to determ ine what processes are to be added in the execution lists (both vertical and horizontal). Currently, by switching the option variable, a specific process can be added on or take n off from a list for each land segment. However, this object is becoming complicat ed and it may be confusing for users to customize their own process lists.

PAGE 261

240 Another model input control object (AOldNewAcruVariable Reference) is used to establish a reference/mapping relationship between the DData object with the CComponent object for all relationships in the model. The huge collection of hard-coded relationships in this object has made it very cumbersome to understand. Use of an input file, instead of hard-coded rela tionships in this object, coul d be one way to improve the situation. Sub-Daily Time Step An incompatibility in the spatial an d temporal resolution is a significant impediment to a robust coupled modeling syst em that involves diffe rent processes having varied temporal scales. The capacity to ope rate with flexible time steps is necessary, especially when the model is applied in much smaller/larger catchments. A capacity that would allow the model to run in a pre-specifie d global time step for the entire model, but also allows specific model processes to run in their own local time st eps, would make the entire model more efficient and physically reasonable. Herbivore Movement Module A cattle roaming component should be adde d to better represen t the impacts of grazing on nutrient cycling and plant growth dynami cs in this model. This is especially important if the model will be used to investigate the effectiveness of stocking-related BMPs. Hydrologic Model Hydrologic model should be tested for a wi der range of conditions, including more physical locations, larger watersheds and a wider range of runoff conditions.

PAGE 262

241 Nutrient Model The biased predictions of nutrient concentr ations in surface runoff in this research demonstrate that the mechanisms for si mulating nutrient cycling need further improvement. Methods for partitioning nutri ents between the soil and water phases and methods for dealing with the nutrients re leased from waste decomposition may require further development. Vegetation Model The entire vegetation model should be furt her evaluated before it is used in a predictive capacity. Components for root le ngth and root biomass distribution dynamics should be added so that species-specific ev apotranspiration and nutrient uptake can be more accurately simulated. An algorithm for plant height determination should also be added to enable a feedback relationship between surface runoff and a dynamic roughness coefficient. Moreover, a start-up mechanism for new species initiation in land segments is especially necessary so that spatial e xpansion across land segments can be modeled and plant competition can be more thoroughly co nsidered. Finally, a herviore movement model should be developed to simulate th e impacts of cattle behavior and grazing performance on vegetation growth.

PAGE 263

242 APPENDIX A FLOW CHART FOR THE COUPLED MODELING SYSTEM AND MODEL PROCESSES Open I/O Files Input Model Data Last Day of Simulation Loop Last Land Segment Last Land Segment End Yes Lateral Hydrolog ic Processes Lateral Nutrient Processes No E D Yes Vertical Hydrologic Processes Vertical Vegetation Processes Vertical Nutrient Processes A B C START Yes No No YesNote: The upper-case letters indicate expandable tree flow charts. (The proce ss objects included are those used for the coupled multi-directional spa tial simulation and do not include all existing processes in the original ACRU2000 model.)

PAGE 264

243 epth ndedWaterD PFindNewPo on distributi Re mited PStorageLi pth terTableDe PFindNewWa tration mitedInfil PStorageLi all inf PLandSegRa eption PVegInterc ) ion Transpirat PMDSSEvapo ( nspiration leEvapoTra PSuperSimp ction ibutionFun PRootDistr ) ent opCoeffici PMDSSSerCr ( efficient PSetCropCo ion erEvaporat PPondedWat eptionEvap PVegInterc rrection fPotEvapCo Re P PanEvap PObsDailyA ge PDeepSeepa PUpwrdFlux all inf eRa PInitialis ionallCorrect inf PRa ateTemp lim PMeanC rBalance PCheckWate epth ndedWaterD PFindNewPo on distributi Re mited PStorageLi ionHWT eSoilUFOpt PInitialis ndex PLeafAreaI scence PPlantSene oductivity Pr PDryMatter tor ductionFac Re PGrowth ss PMDSSPStre ss PMDSSNStre Logging PMDSSWater A B

PAGE 265

244 eModel oneExchang iveMixingZ PConservat rasnport ubsurfaceT iveSoluteS PConservat Transport vaporation iveSoluteE PConservat nputs iveSoluteI PConservat ification PHWTDenitr odel eExchangeM PMixingZon ort faceTransp PHWTSubsur PTillage ation PVolatiliz PHarvest on mobilizati Im PHWT lization PHWTMinera ication PHWTNitrif cation PHWTAmmoni emp PMeanSoilT t onTranspor PEvaporati ke PPlantUpta n akeFixatio PNPlantUpt ecationPAnimalDef estion PAnimalDig nsfer utrientTra egetationN PConsumedV Transfter ssNutrient PDeadBioma entInputs PMDSSNutri ents alizeNutri PMDSSIniti s essureHead Pr eLayer min PDeter cay icMatterDe PMDSSOrgan terFlow alGroundwa PCalcLater andFlow PCalcOverl C D

PAGE 266

245 rt aceTranspo uteSubsurf rvativeSol PMDSSConse Transport uteSurface rvativeSol PMDSSConse port rfaceTrans PMDSSSubsu t ceTranspor PMDSSSurfa E

PAGE 267

246 APPENDIX B FLOW CHART OF OVERL AND FLOW SIMULATION Start Initialization of Current Adjacent Land Segment Ponded Water Depth Maximum Depressional Storage Water Slope 0.0 End Storage Apportionment Check Overland Flow Transfer Information Initialization Overland Flow Calculation Last Land Segment? Yes No No Yes

PAGE 268

247 APPENDIX C FLOW CHART OF LATERAL GROU NDWATER FLOW SIMULATION Start End Update Water Table for Source and Destination Land Segments Information Initialization for Adjacnet Land Segment Lumped Groundwater Flow Calculation Information Initialization Lumped Lateral Flow > 0.0? Last Land Segment? Last Land Segment? Stoarge Apportoinment Check Split Lumped Lateral Flow into Layer Flow Transfer Layer Flow Update Air Volume for Source and Destination Land Segments No Yes No No Yes Yes

PAGE 269

248 APPENDIX D LISTS OF NEW AND MODIFIED OBJECTS New objects listed below mean they did not ex ist but were added through this research; modified objects mean they existed in the previous model or were created by other modelers but were changed to fit the multidirectional spatial simulation model. Also, hereafter the multi-directional spatial simulation option is called MDSS for short. New Model Objects MAcru2000ModifiedStandard This model object is the starti ng point for the model with a main method inside. It is created to decide how model input data, com ponents, and processes are structured and in what mode the model will be executed. This object allows the model to run either with a lumped mode (MDSS= 0) or a multidirectional distributed mode (MDSS = 1). New Process Objects PAdjustOverlandFlow This process object is used when the MDSS option is on (MDSS = 1). The major function of this process object is to compare calculated po tential overland flow for each adjacent land segment with available amount of surface water storage in the source land segment to guarantee the conservation of mass. If potential overland fl ow is greater than available amount, then adjustments have to be conducted. PAdjacentSpatialUnits This process object is used when the MDSS option is on (MDSS = 1). The major function of this process object is to es tablish the relationships among adjacent land segments for one source land segment according to the boundary conditions, which is used to support the simulation of overland flow. PCalcLateralGroundwaterFlow This process object is invoked when the MD SS option is on (MDSS = 1). The major function of this object is to calculate the latera l groundwater flows be tween the saturated soils of two neighbouing land segments and tran sfer the lateral flows. Darcy's Law is used to estimate the lumped lateral groundwat er flow when the hydraulic gradient exists between two saturated soils. Then, the lumped flow is split into the layer flows to those soil layers that contains and is below the cu rrent water table of source land segment. The layer lateral flow is transferred layer by layer.

PAGE 270

249 PCalcOverlandFlow This process object is used when the MDSS op tion (MDSS = 1) is on. It is designed to calculate overland flows and then transfer the flows to the adjacent land segments. PCalcWaterBalance This process is designed to check the daily water balance for both surface and subsurface water for each land segment. PDryMatterProductivity This process is invoked when the vegetati on simulation option (VEGOPT = 1) is on. This process object is designed to calculate daily biomass productivity for each species in one land segment on the assumption of the plant growth is driven by the climate variables (radiation and temperature). The actual biomass is estimated by multiplying the potential biomass with the growth reduction factor (the minimum among water stress, water logging, and P and N stress factors). Additi onally, the actual bi omass is partitioned between the aboveand belowground biomass fo r the use of other vegetation processes. PFindNewPondedWaterDepth This process object is designed to find out the current surface ponded water depth. It is called twice in the vertical process list, at the beginning and the end of a given day, before the horizontal processes are executed. PGrowthReductionFactor This process is invoked when the VEGOPT option is on. This process object is used to calculate the dry matter growth reduction fact or that is a minimu m among water stress, water logging, and N and P stress factors. It will be invoked when the VEGOPT option is on. PLeafAreaIndex This process object is used to calculate the daily green leaf area index due to new growth, harvest, grazing, burn and senescence for each vegetation species in each land segment. It will be invoked when the VEGOPT option is on. PManningRoughnessCoefficient This process object is used to calculate the manning roughness coefficient according to an approximate method (Fitz et al., 1996) that assume the coefficient is a function of sediment type, plant height and surface water depth. PMDSSSurfaceTransport This process object is used when the MDSS op tion (MDSS = 1) is on. It is used to calculate and transport ammonium N, nitrate N, and labile P carried by surface runoff. PMDSSSubsurfaceTransport

PAGE 271

250 This process object is used when the MDSS op tion (MDSS = 1) is on. It is used to calculate and transport ammoni um N, nitrate N, and labile P carried by groundwater flow. PMDSSConservativeSoluteSurfaceTransport This process object is used when the MDSS op tion (MDSS = 1) is on. It is used to calculate and transport conservative solute carried by surface runoff. PMDSSConservativeSoluteSubsurfaceTransport This process object is used when the MDSS op tion (MDSS = 1) is on. It is used to calculate and transport conservative so lute carried by subsurface groundwater. PMDSSEvapoTranspiration This process is used when the VEGOPT option is selected. It is designed to calculate evapotranspiration by multiple species from one land segment by using a very simple method that is identical to that used in DRAINMOD/FHANTM. Potential ET is applied from the top down, reducing soil layers to the wilting point before moving to the next layer. The catual ET is equal to potentia l ET as long as there is water available. PMDSSInitalizeNutrients This process object is used when the MDSS option (MDSS = 1) is on. It was modified according the PInitalizeNutrient. This process handles initialization of nutrient flux records, plant characteristics and other nutrient-related parameters as defaults when me asured values are not available. If soil nutrient data are available for local conditions (soils), the m odel user should input those values and they will override the generalized estimates generated by the model. Also, the modified object enables to update the N fraction daily. PMDSSNutrientInputs This process object is used when the MDSS op tion (MDSS = 1) is on. It was modified from PHWTNutrientInputs to make it more applicable for multi-directional spatial simulation and enable to consider the time se ries inputs. This pr ocess handles nutrient inputs to the watershed system through ra infall, fertilizer, and animal wastes. This process was remodified from PHWTNutr ientInputs and PNutrientInputs to make it more applicable for multi-directional spatial simluation and consider time series events, such as fertilizer, animal wast e, etc. Major changes include: (1) Runoff is included only when the model runs in a lumped mode since at this stage runoff have not been simuleted when the MDSS option is on. (2) Modify the design in PNutrientsIn puts where fertilizer and animal waste applications can not simutanluously occur. The modification enables consideration of both. (3) Enable the application depth for fertil izer and animal waste to be different. The original related data objects were modified into time-series data objects to handle the potential events.

PAGE 272

251 PMDSSNPlantUptakeFixation This process object is used when the MDSS option (MDSS = 1) is on. This process calculates the N uptake by multiple plant species of ammonium and nitrate from the soil water by transpiration and/or assimilation (fix ation) of nitrogen from the atmosphere by leguminous plants. Fixation is not assumed to add nitrogen to the soil until harvest and/or tillage. This process is modifed according to the existing process PNPlantUptakeFixation. PMDSSNstress This process is invoked when the VEGOPT option is on. This process calculates the soil water N stress factor from 0.0 to 1.0 according to the method by Seligman and Van Keulen (1981). PMDSSPPlantUptake This process object is used when the MDSS option (MDSS = 1) is on. This process calculates the uptake by multiple plant species of labile phosphorus from the soil water by transpiration. PMDSSPStress This process is invoked when the VEGOPT option is on. This process calculates the soil water P stress factor from 0.0 to 1.0 accord ing to according to the similar method as the N stress. PMDSSSetCropCoefficient This process is invoked when the VEGOPT opt ion is on. This process was modified according to PSetCropCoefficient to meet the needs when multiple plant species exist. The crop coefficient for each species needs to be set up. PMDSSTillage This process object is used when the MDSS option (MDSS = 1) is on. This process calculates the effects of soil tillage on N and P forms and their movement. It was modified form PTillage. PMDSSWaterLogging This process is invoked when the VEGOPT opti on is on. This process is designed to calculate the water logging factor (0~1) and th e death indicator that indicate the plant will die caused by the long-term water logging. POverlandFlowTransfer This process object is used when the MDSS is on (MDSS = 1). Its major function is to transfer the actual overland flow s to neighboring land segments. PPlantSenescence

PAGE 273

252 This process is invoked when the VEGOPT option is on. This process object is designed to calculate the reduction of live bioma ss both on surface and subsurface by plant sensecense. It is assumed that the decline of live biomass into dead biomass is a linear process during the period of plant senes cemce. The corresponding N and P also are transferred from live biomass pools to dead biomass pools. PSoilParameters This process is used to estimate the so il mass, phosphorus and ammonium partitioning coefficients for each soil layer when the nutri ent option (NUTRI = 1) is turned on. PStorageApportionment This process object is used when the MDSS option is on (MDSS = 1). Its major function is to calculate the share of available water storage for each potential overland flow from source land segment. PSortAdjacentSpatialUnits This process object is used when the MDSS option is on (MDSS = 1). The major function of this process object is to arrange the sequence of adjacent land segments that receive potential overland flows in desce nding order of their boundary heights. New Data Objects DAboveGroundBiomassFluxRecord Double flux record data object is to store the aboveground dry matter. DACRUVegOption Integer data object holds the option to determine if the model enables the vegetation dynamic simulation. VEGOPT = 1, enable simulation of vegetation dynamics; VEGOPT = 0, otherwise. DAmmoniumPartitioningCoefficient Double data object holds the ammo inium partitioning coefficient. DBelowGroundBiomassFluxRecord Double flux record data object is to store the belowground dry matter. DBoundaryInflow Double data object holding the flow input from outside of simulation domain. DBoundaryOption Integer array data object holds the boundary option including: BOPT = 0, without weir, w ithout downstream stage; BOPT = 1, without weir, with downstream stage;

PAGE 274

253 BOPT = 2, with weir, without downstream stage; BOPT = 3, with weir, with downstream stage; BOPT = 4, inflow boundary; DBoundaryOutflow Double data object holding the outflows though the domain boundary. DBoundaryWidth Double array data object hol ding the widths of boundaries for one land segment. DCriticalLeafAreaIndex This daily data class is to store the degree days for each plant species. DDaysofWaterlogging Integer data object hold the number of succe ssive days of water logged a plant suffers. DDaysofWaterloggingLimits Integer object holds the maximum number of da ys that a plant species still grow under the inundated soil conditions. DDryMatterPartitioningCoefficient Double data object holds the dry matter part itioning coefficient for a specific plant. DExtinctionCoefficient Double data object holds th e extinction coefficient for a specific plant. DGrowthReductionFactor Double data object is to store the growth reduction factor. DHorizontalSaturatedHydraulicConductivity Double data object holding th e horizontal saturated hydraulic conductivity for soil layers. DIncrementOfAboveGroundBiomass Double data object is to store the daily increment of aboveground biomass. DIncrementOfBelowGroundBiomass Double data object is to store the daily increment of belowground biomass. DISoilHorizon Integer data object hold the integer ID of th e soil horizon to indicat e which soil horizon a computational soil layer belong to. It is us ed to choose different formula when calculate the P partitioning coefficient for sandy soils.

PAGE 275

254 DLabilePPartitioningCoefficient Double data object is to store th e labile P partitioning coefficient. DLayerAmmNUptake Double data object is to store the layer ammonium N uptaken by each plant species. DLayerLabPUptake Double data object is to st ore the layer labile P upta ken by each plant species. DLayerNitNUptake Double data object is to st ore the layer nitrate N uptak en by each plant species. DLayerSoilMass Double data object is to store the layer soil mass. DLateralGroundwaterAmmNLoad Double data object holds the ammonium N lo ad levaing out of the simultaion domain through the lateral groundwater flow. DLateralGroundwaterConservativeSoluteConc Double data object holds the c oncentration of conservative solute in lateral groundwater discharge through the watershed outlet. DLateralGroundwaterConservativeSoluteLoad Double data object holds the load of conservative solute in lateral groundwater discharge through the watershed outlet. DLateralGroundwaterLabPLoad Double data object holds the la bile P load levaing out of the simultaion domain through the lateral groundwater flow. DLateralGroundwaterNitNLoad Double data object holds the ni trate N load levaing out of the simultaion domain through the lateral groundwater flow. DLateralGroundwaterOutflow Double data object holds the total groundwater discharg e through the outlet of the simulation domain. DMaximumGrowthTemperature Double data object holds the maximum temperatur e for a specific plan t to enable growth. DMaxManningRoughness

PAGE 276

255 Double data object holds the maximum Manning’s roughness coefficient of a land segment. DMaxSurfaceDepressionStorage Double data object holds the maximum de pressional surface storage on the ground surface of a land segment. DMDSSConservativeSoluteOption Integer data object holds the option to determine whether or not the model simulates the multi-directional spatial transport of conser vative solute (MDSSCS = 0, turn on multidirectional spatial simulation of conservative solute transport; MDSSCS = 1, turn off). DMDSpatialSimulationOption Integer data object holds th e option to determine if the model enables the multidirectional spatial simulation (MDSS = 1, turn on multi-directional spatial simulation; MDSS = 0, turn off). DMinManningRoughness Double data object holds the minimum Manning’s roughness coefficient of a land segment. DMinimumGrowthTemperature Double data object holds the minimum temperatur e for a specific plan t to enable growth. DMinimumLeafAreaIndex Double data object holds the minimum l eaf area index for a specific plant. DNumofVegetationSpecies Integer data object holds the number of vegetation species to be simulated in one land segment. DNeighbourWidth Double array data object holds the widths of boundaries between one land segment and its adjacent land segments or outside of the simulation domain. DOptimum1GrowthTemperature Double data object holds the fisrt optimum growth temperature for a specific plant. DOptimum2GrowthTemperature Double data object holds the second optimum growth temperature for a specific plant. DOptimumPConcentration Double data object holds the optimum P con centration for a speci fic plant to grow.

PAGE 277

256 DOutletLength Double data object holds the distance from th e center of the land segment to downstream or to the external boundary which sepa rates the land segment and downstream. DOverlandFlow Double array data object holds the overland fl ows occurred from source land segment to adjacent land segments. DOverlandOutflow Double data object holds the overland flow leaving the simulation domain through the watershed outlet. DPlantDeathIndicator Integer dat object indicates if a plant will die off the current land segment due to the water logging (= 1 to die; = 0 not to die). DPlantNPercent Double data object holds the plant N percent. DPlantPPercent Double data object holds the plant P percent. DPlantSenescenceRate Double data object holds the fraction of remaining biomass after plant senescence. DPondedWaterDepth Double data object holds daily ponded wate r depth on the ground surface of a land segment. DPStressFactor Double data object holds the pl ant growth stress factor re sulting from a shortage of phosphorus in the root zone. DRUE Double data object holds the average light use efficiency for a specific plant. DSenescedAboveGroundBiomass Double data object holds the senscened a boveground biomass for a specific plant. DSenescedBelowGroundBiomass Double data object holds the senscened belo wground biomass for a specific plant. DShootCriticalNConc

PAGE 278

257 Double data class is to store the critical shoot N concentration. DShootCriticalPConc Double data class is to store the critical shoot P concentration. DShootMinimumGrowthNConc Double data object holds the minimum shoot N concentration for maintaining plant growth. DShootMinimumGrowthPConc Double data object holds the minimum shoot P concentration for maintaining plant growth. DShootPConcentration Double data class is to stor e the shoot P concentration. DSpecificLeafArea Double data object holds th e specific leaf area rati o for a specific plant. DSoilAluminumContent Double data object holds the soil aluminum content. DSoilMagnesiumContent Double data object holds the soil aluminum content. DStockingOption Integer data object stores the stocking option. DStreamLabPConc Double data object holds the str eam labile P concentration. DStreamNitNConc Double data object holds the str eam nitrate N concentration. DStreamAmmNConc Double data object holds the str eam ammonium N concentration. DTemperatureFactor Double data class is to store the temp erature factor for a specific plant. DTotalDryMatter Double data class is to store the to tal dry matter for a specific plant.

PAGE 279

258 DTotalOutgoingOverlandFlow Double data object holds the total outgoing overland flows form one land segment. DTotalReverseOverlandFlow Double data object holds the to tal incoming overland flows fo rm outside of a simulation domain. DTotalSurfaceConservativeSoluteStorage Double data object holds the total conservative solute mass in the surface water of a land segment. DTotalSubSurfaceConservativeSoluteStorage Double data object holds the total conservative solute ma ss within the soil water of a land segment. DTotalSurfaceWaterStorage Double data object holds the total water storage on the surface of a land segment. DTotalSubSurfaceWaterStorage Double data object holds the to tal water storage within th e soil of a land segment. DXcoord Double data object holds the x-coordinate of the centroid of a land segment. DYcood Double data object holds the y-coordinate of the centroid of a land segment. DWaterStressFactor Double data class is to st ore the water stress factor. DWaterLoggingFactor Double data class is to st ore the water logging factor. DWeirCoefficient Double array data object is to hold the weir coefficients. DWeirElevation Double array data object is to hold the el evation of weir crust above the datum. DWeirOption Integer array data object is to hold the weir options. DWeirWidth

PAGE 280

259 Double array data object is to hol d the widths of weir crust. New Module ACRU_Veg To contain all plant growth related proce sses and collect data objects required by the vegetation dynamic simulation model. Modified Component Objects CComponent To enable to create two separate vertical a nd horizontal process lists when the model runs with the MDSS (=1) mode, this existing component object was modified by adding instance variables and methods, including fi ve instance variables: firstVerProcess, firstHorProcess, verProcessesList, horProcesse dList, and landSegHorProcessList, and ten methods: addVerticalProcesses(PProcess) addHorizontalProcesses(PProcess), setFirstVerticalProcess(PProcess), se tFirstHorizontalProcess(PProcess), getDIntegerArray(String), getFirstVertical Process(), getFirstHorizontalProcess(), getLandSegHorProcessesList(), runVerticalPr ocesses(), and runHor izontalProcesses(). Modified Control Objects AAcru2000ModelInput Created one instance variable horProce ssesList and added two methods including addToHorizontalProcessList(St ring, String) and getHorizon talProcessesList() to help with creating a horizontal process list. AAcru2000StandardModelCreator Modified the createProcessObjects() and in itialiseProcess() met hods to enable two separate vertical and horizontal process list to be created. AAcru2000StandardCComponents Modified the addLandSegmentSubComponents() to enable additions of multiple plant species components. AAcru2000StandardProcesses Modified addLandSegmentProcesses() method to decide which pro cesses objects will be put on the vertical and horizontal pro cess lists when the MDSS option is on. AOldNewAcruVariableReference Hard coded statements to establish the ma pping between input and output variables and the component objects which they belong to. AProcessCreator

PAGE 281

260 Created one overloading method, createProcesses(), to create two separate vertical and horizontal process processes lists, and this method will be invoked only when the MDSS option is on. Modified Data Objects DDeepSeepageHeadBoundary Modified the data object type from DDailyDatae to DDouble. DDryMatterRatio Modified the data object type from DDailyData into DDouble. DPotentialYield Modified the data object type from DDailyData into DDouble. DPlantID Modified the data object type from DDailyInteger into DInteger. DPlantNC1 Modified the data object type from DDailyData into DDouble. DPlantNC2 Modified the data object type from DDailyData into DDouble. DPlantNitrogenPhosphorusRatio Modified the data object type from DDailyData into DDouble. DPlantBiomassFluxRecord Added several biomass transfer methods. Modified Process Objects PAnimalDigestion Deduct the assimilated consumed biomass N and P. PConservativeMixingZoneExchangeModel Changed startments in the codes to account for the surface runoff when the MDSS option is on. And corrected th e unit conversion error. PHWTDenitrification Added two more formula for the underestimation problems. PHWTNutrientInputs

PAGE 282

261 Changed statements in the codes to account for the surface runoff when the MDSS option is on. PHWTSubsurfaceTransport Correctyed the unit conversion error. PLabilePPartitioningCoefficient Added one overloading method to calculate the phosphorus partitioning coefficient for the specified soil layer using the empirica l formula (Nair et al., 1997) for Florida flatwoods sandy soils. PMeanSoilTemperature Changed the process object to account for th e biomass produced by all plants species. PMixingZoneExchangeModel Changed some statements in the codes to account for the surface runoff when the MDSS option is on and corrected the unit conversion error in total amount of surface water. PNPlantUptakeFixation Added the codes to account fo r the mass balance check to avoide the rounding errors. PPPlantUptake Added the codes to account fo r the mass balance check to avoide the rounding errors. POrganicMatterDecay Corrected the errors in destroying the biomass for the residue layer.

PAGE 283

262 APPENDIX E LISTS OF NEW INPUT AND OUTPUT VARIABLES Table E-1. New input and output variables required towards the multi-directional spatial simulation beyond the existing variables in ACRU2000. Input Definition Unit ABGDM1* = Species specific aboveground biomass kg/ha ALSS = Aluminum content for top soil layer mg/kg AL1*** = Aluminum content for soil layer 1 mg/kg BDWD(01)** = Boundary width m BDOPT(01)** = Boundary option BEGDM1* = Species specific belowground biomass kg/ha BOF = Surface outflow out of the domain through the stream m3/day BIF = Boundary inflows m3/day CRITILAI = Critical leaf area index m2 leaf/m2 ground CONSERV = Conservative solute transport option DCS = Initial conservative solute load applied kg/ha DPOND = Ponded surface water depth mm EXTCOF1* = Extinction coefficient FA1* = Dry matter partitioning coefficient GCS = Lateral groundwater conservative solute concentration mg/l GCSL = Lateral groundwater conservative solute load kg/ha INABGBIO1* = Daily aboveground biomass increment kg/ha INBEGBIO1* = Daily belowground biomass increment kg/ha INLIMIT1* = Water logging limit day ISHSS = Soil horizon code for top soil layer ISH1*** = Soil horizon code for soil layer 1 KHSATSS = Horizontal saturated hydraulic conductivity for top layer m/s KHSAT1*** = Horizontal saturated hydraulic conductivity for soil layer 1 m/s LGWF = Lateral groundwater outflow out of the domain m3/day LGNILOD = Lateral groundwater nitrate nitrogen load kg/ha LGAMLOD = Lateral groundwater ammonium nitrogen load kg/ha LGLPLOD = Lateral groundwater labile phosphorus load kg/ha LGCSLOD = Lateral groundwater conservative solute load kg/ha MSDS = Maximum surface depressional storage m

PAGE 284

263 MDSS = Multi-directional spatial simulation option MDSSCS = multi-directional spatial conservative solute simulation option MAXROUGH = Maximum Manning’s roughness coefficient m1/3/day MAXTEMP1* = Maximum plant growth temperature C MGSS = Magnesium content for top soil layer mg/kg MG1*** = Magnesium content for soil layer 1 mg/kg MINLAI1* = Minimum leaf area index m2/m2 MINROUGH = Minimum Manning’s roughness coefficient m1/3/day MINTEMP1* = Minimum plant growth temperature C NUMVEG = Number of plant species in one land segment OMSS = Organic matter for top soil layer % OM1*** = Organic matter for layer 1 % OPT1TEMP1* = First optimum growth temperature C OPT2TEMP1* = Second optimum growth temperature C OUTL = Distance from the centroid of land segment to the external boundary. m PLRSPS = Plant residue P flux record for top soil layer PLRSP1*** = Plant residue P flux record for layer 1 PLRSNS = Plant residue N flux record for top soil layer PLRSN1*** = Plant residue N flux record for layer 1 PLBMASS = Plant residue biomass flux record for top soil layer PLBMAS1*** = Plant residue biomass flux record for layer 1 RAMLOD = Surface runo ff ammonium nitrogen load kg/ha RAMCON = Surface runoff ammonium nitrogen concentration mg/l RCSC = Surface runoff conservative solute concentration mg/l RCSL = Surface runoff conservative solute load kg/ha RF1* = Productivity reduction factor RUE1* = Average light use efficiency g MJ/PAR RLPLOD = Surface runoff labile P load kg/ha PLPCON = Surface runoff lab ile P concentration mg/l RNILOD = Surface runoff nitr ate nitrogen load kg/ha PNICON = Surface runoff nitrate nitrogen concentration mg/l SAMMCONC = Stream ammonium nitrogen concentration mg/l SCRINCONC1* = Shoot critical N concentration g N/g DM SCRIPCONC1* = Shoot critical P concentration g P/g DM SENCERATE1* = Plant senescence rate SENEAGBIOM1* = Senesced aboveground biomass kg/ha SENEBGBIOM1* = Senesced belowground biomass kg/ha SLA1* = Specific leaf area cm2 leaf/g leaf SLABCONC = Stream labile phosphorus concentration mg/l

PAGE 285

264 SLR = Specific leaf area ratio m2/g leaf SMINGNCONC1* = Shoot minimum growth N concentration g N/g DM SMINGPCONC1* = Shoot minimum growth P concentration g P/g DM SNITCONC = Stream nitrate n itrogen concentration mg/l SRL = Specific root length m/g root SSH = Specific shoot height m3/g shoot SWS = Stream water stages m TSCSS == Total surface conservative solute storage kg/ha TSSCSS = Total subsurface conser vative solute storage kg/ha TSSWS = Total subsurface water storage m3 TSWS = Total surface water storage m3 TEMPFAC1** = Temperature factor V1PLNTID* = Plant ID code V1LEG* = Legume code for plant species V1PRNNL* = Plant perennial code V1POTLAI* = Potential leaf area index m2 leaf/ m2 ground V1DMR* = Dry matter ratio V1POTYLD* = Potential yield kg/ha V1CNP* = Plant C:N ratio V1NPR* = Plant N:P ratio V1NC1* = Plant NC1 V1NC2* = Plant NC2 V1BMAS* = Initial total specie-specific plant biomass kg/ha V1LAI* = Daily species specific leaf area index m2 leaf/ m2 ground V1WSTFAC* = Daily species specific water stress factor V1WLTFAC* = Daily species specific water logging factor V1PSTFAC* = Daily species specific P stress factor V1NSTFAC* = Daily species specific N stress factor VEGOPT = Vegetation dynamic simulation option WCOEFF(01)** = Weir coefficient WELEV(01)** = Weir elevation m WTDEP = Water table depth for land segment m WWD(01)** = Weir width m XCOOD = X coordinate of the centroid of the land segment m YCOOD = Y coordinate of the centroid of the land segment m *Species-specific variables. The digit in the variable names indicates the corresponding species type. **Land segment related variables. The digit in the variable names indicates the corresponding boundary. ***Soil layer related variables. The digit in the variable names indicates the corresponding soil layer.

PAGE 286

265 APPENDIX F LISTS OF MODEL INPUT FILES Overland Flow along a Flat Rectangular Plane cell.csv Control.men iroder LandSeg_1.men through LandSeg_20.men Overland and Groundwater Flow over an Axisymmetric Domain cell.csv Control.men iroder LandSeg_1.men through LandSeg_36.men Application at Dry Lake Dairy #1, Kissimmee River Basin, Florida cell.csv Control.men iroder LandSeg_1.men through LandSeg_4.men Buck Island Application, La ke Okeechobee Basin, Florida For winter pastures 6 and 7 cell.csv Control.men iroder LandSeg_1.men through LandSeg_18.men For summer pastures 4 and 1 cell.csv Control.men iroder LandSeg_1.men through LandSeg_12.men

PAGE 287

266 APPENDIX G ALGORITHMS OF WATER STORAGE APPORTIONMENT Case 1: Single Adjacent Land Segment The basic idea for deciding the available amount for single adjacent neighbor is moving water until the water levels in both si des are equilibrated in case of sufficient water storage available in source land segmen t. Otherwise, the available amount could be equal to the total available storage. As a matter of fact, the si ngle neighbor could be either land segment or river type spatial unit. In Figure G-1, As and Ad are the surface areas of source and adjacent land segment, respectively; Hs and Hd are their corresponding water elevation, respectively; Es is the surface elev ation of source land segment, which includes the surface depressiona l storage (in depth). avaAmount is the maximum available amount from source land se gment to its neighbor. The formula used to determine the amount for different situations are listed below: 1. If Single Neighbor is of Land Segment Type A. There is no weir between source land segment and single neighbor Figure G-1. Single land segment type neighbor receiving overland flow: (A) Hd Es and (B) Hd < Es. Hd As Ad Hs Es (A) As Ad Hu Es Hd (B)

PAGE 288

267 Hd Es totAvaStorage = (Hs Hd ) As avaAmount = totAvaStorage Ad / ( As + Ad) Hd < Es totAvaStorage = (Hs Es ) As receivingVol = (Es Hd ) Ad If (totAvaStorage receivingVol) avaAmount = receivingVol + [totAvaStorage receivingVol] Ad / (As + Ad) If (totAvaStorage < receivingVol) avaAmount = totAvaStorage B. There is a weir between source land segment and single neighbor Assume the elevation of weir crust is higher than the surface elevation of source land segment. In Figure G-2, Ew represents the elevation of the weir crust. Hd is the water elevation in the river type spatial unit. Figure G-2. Single river type neighbor receiving overland flow: (A) Hd Ew and (B) Hd < Ew. Hd Ew avaAmount = (Hs Hd) As Ad / ( As + Ad) Hd < Ew totAvaStorage = (Hs Ew) As receivingVol = (Ew Hd ) Ad Hd As Ad Hs Es (A) Ew Hd As Ad Hs Es (B) Ew

PAGE 289

268 If (totAvaStorage receivingVol) avaAmount = receivingVol + [totAvaStorage receivingVol] Ad / (As + Ad) If (totAvaStorage receivingVol) avaAmount = totAvaStorage 2. If Single Neighbor is of River Type A. If there is no weir in between and no downstrean water stage data available avaAmount = (Hs Es) As B. If there is a weir in between and no downstrean water stage data available If Es Ew avaAmount = (Hs Es) As If Es Ew avaAmount = (Hs Ew) As C. If there is no weir in between and downstrean water stage data available avaAmount = (Hs Hwater stage) As D. If there is a weir in between and downstrean water stage data available If Hwater stage Ew avaAmount = (Hs Hwater stage) As If Hwater stage < Ew avaAmount = (Hs Ew) As Case 2: Multiple Adjacent Land Segments Figure G-2 displays the relationship be tween a source land segment and its 4 adjacent land segments. Assume all these ne ighbors receive outflows from the source land segment and the order of water elevati onss of these 4 neighbors follows: neighbour 1 > neighbour 2 > neighbour 3 > neighbour 4. Apparently, the water levels of these neighbors are lower than that of source land se gment. This diagram only displays one of potential scenarios of the relationships be tween source land segmen t and its neighbors.

PAGE 290

269 The actual relationship patterns could be diffe rent. The algorithm should be able to handle all these situations. In Figure G-3 and the following formula, As and A1~A4 represent the areas of source land segment and its co rresponding neighbors. Hs and H1~H4 are their corresponding water elevations. Es is the surface elevation of source land segment. avaAmount 1~avaAmount 4 is the available amount from source land segment to its corresponding neighbors. totAva Storage is the total availabl e amount from this unit. receivingVol is the volume in a neighbor, wh ich is above its elevation and below the elevation of this unit. potWLIncrease is the potential increase in the water depth in a neighbor. remainingStorage represents the rema ining total available amount in this unit. The formula used to determine each available amount for different situations are listed below: Figure G-3. Configuration of multiple directional overland flows from source land segment to adjacent land segments. Neighbor 3 Neighbor 4 Neighbor 2 As A3 A4 A2 A1 H2 H1 Hs H3 H4 Es Neighbor 1

PAGE 291

270 Loop 1 neighbour 4 1. If neighbour 4 is of land segment type A. If there is no weir between source land segment and neighbour 4 If H4 Es totAvaStorage = (Hs H4) As potWLIncrease = totAvaStorage / (As + A4) If potWLIncrease boundaryHeightDiff3,4 avaAmount 4 = boundaryHeightDiff3,4 A4 remainingStorage = totAvaStorage avaAmount 4 If potWLIncrease < boundaryHeightDiff3,4 avaAmount 4 = potWLIncrease A4 Break up the loop. If H4 < Es totAvaStorage = (Hs Es ) As receivingVol = (Es H4 ) A4 If totAvaStorage receivingVol potWLIncrease = [receivingVol + (totAvaStorage receivingVol) A4 / (As +A4)] / A4 If potWLIncrease boundaryHeightDiff3,4 avaAmount 4 = boundaryHeightDiff3,4 A4 If NBH3 Es remainingStorage = totAvaStorage avaAmount 4 (NBH3 Es) As If NBH3 < Es remainingStorage = totAvaStorage avaAmount 4 If potWLIncrease < boundaryHeightDiff3,4 avaAmount 4 = potWLIncrease A4 Break up the loop. If totAvaStorage < receivingVol potWLIncrease = totAvaStorage / A4 If potWLIncrease boundaryHeightDiff3,4 avaAmount 4 = boundaryHeightDiff3,4 A4 remainingStorage = totAvaStorage avaAmount 4 If potWLIncrease < boundaryHeightDiff3,4 avaAmount 4 = potWLIncrease A4

PAGE 292

271 Break up the loop. B. If there is a weir between source land segment and neighbour 4 If H4 Ew4 totAvaStorage = (Hs H4 ) As potWLIncrease = totAvaStorage / (As +A4) If potWLIncrease boundaryHeightDiff3,4 avaAmount 4 = boundaryHeightDiff3,4 A4 remainingStorage = totAvaStorage avaAmount 4 If potWLIncrease < boundaryHeightDiff3,4 avaAmount 4 = potWLIncrease A4 Break up the loop. If H4 < Ew4 totAvaStorage = (Hs Ew4 ) As receivingVol = (Ew4 H4) A4 If totAvaStorage receivingVol potWLIncrease = (totAvaStorage receivingVol) / (As + A4) If potWLIncrease boundaryHeightDiff3,4 avaAmount 4 = receivingVol + boundaryHeightDiff3,4 A4 remainingStorage = totAvaStorage avaAmount 4 If potWLIncrease < boundaryHeightDiff3,4 avaAmount 4 = receivingVol + potWLIncrease A4 Break up the loop. If totAvaStorage < receivingVol avaAmount 4 = totAvaStorage Break up the loop. 2. If neighbour 4 is of river type A. If there is no weir in between and no downstrean water stage data available avaAmount 4 = (Hs Es) As Break up the loop. B. If there is a weir in between and no downstrean water stage data available If Es Ew4 avaAmount 4 = (Hs Es) As Break up the loop

PAGE 293

272 If Es Ew4 avaAmount 4 = (Hs – Ew4) As Break up the loop C. If there is no weir in between and downstrean water stage data available avaAmount 4 = (Hs – Hwater stage) As D. If there is a weir in between and downstrean water stage data available If Hwater stage Ew4 avaAmount 4 = (Hs – Hwater stage) As Break up the loop If Hwater stage < Ew4 avaAmount 4 = (Hs Ew4) As Break up the loop Loop 2 neighbour 3 1. If neighbour 3 is of land segment type A. If there is no weir between source land segment and neighbour 3 If H3 Es potWLIncrease = remainingStorage / (As + A3 + A4) If potWLIncrease boundaryHeightDiff2,3 avaAmount 3 = boundaryHeightDiff2,3 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff2,3 A4 remainingStorage = remainingStorage boundaryHeightDiff2,3 (As + A3 + A4) If potWLIncrease < boundaryHeightDiff2,3 avaAmount 3 = potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If H3 < Es receivingVol = (Es-H3) (A3+A4) If remainingStorage receivingVol potWLIncrease = [receivingVol + (remainingStorage receivingVol) (A3 + A4) / (As + A3 + A4)]/(A3 + A4) If potWLIncrease boundaryHeightDiff2,3 avaAmount 3 = boundaryHeightDiff2,3 A3

PAGE 294

273 avaAmount 4 = avaAmount 4 + boundaryHeightDiff2,3 A4 If NBH2 Es remainingStorage = remainingStorage boundaryHeightDiff2,3 (A3 + A4) (NBH2 Es) As If NBH2 < Es remainingStorage = remainingStorage boundaryHeightDiff2,3 (A3 +A4) If potWLIncrease < boundaryHeightDiff2,3 avaAmount 3 = potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If remainingStorage < receivingVol potWLIncrease = remainingStorage / (A3 + A4) If potWLIncrease boundaryHeightDiff2,3 avaAmount 3 = boundaryHeightDiff2,3 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff2,3 A4 remainingStorage = remainingStorage boundaryHeightDiff2,3 (A3 + A4) If potWLIncrease < boundaryHeightDiff2,3 avaAmount 3 = potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. B. If there is a weir between source land segment and neighbour 3 If H3 Ew3 potWLIncrease = remainingStorage / (As + A3 + A4) If potWLIncrease boundaryHeightDiff2,3 avaAmount 3 = boundaryHeightDiff2,3 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff2,3 A4 remainingStorage = remainingStorage – boundaryHeightDiff2,3 (As + A3 + A4) If potWLIncrease < boundaryHeightDiff2,3 avaAmount 3 = potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If H3 < Ew3 receivingVol = (Ew3 H3) A3 If remainingStorage receivingVol

PAGE 295

274 potWLIncrease = (remainingStorage receivingVol) / (As + A3 + A4) If potWLIncrease boundaryHeightDiff2,3 avaAmount 3 = receivingVol + boundaryHeightDiff2,3 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff2,3 A4 remainingStorage = remainingStorage receivingVol boundaryHeightDiff2,3 (A3+A4) If potWLIncrease < boundaryHeightDiff2,3 avaAmount 3 = receivingVol + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If remainingStorage < receivingVol avaAmount 3 = remainingStorage Break up the loop. 2. If neighbour 3 is of river type A. If there is no weir in between and no downstrean water stage data available avaAmount 3 = remainingStorage Break up the loop. B. If there is no weir in between and downstrean water stage data available maxStorage = (Hs – Hwater stage) As If maxStorage remainingStorage avaAmount 3 = remainingStorage Break up the loop If maxStorage < remainingStorage avaAmount 3 = maxStorage remainingStorage = remainingStorageavaAmount 3 C. If there is a weir in between and no downstrean water stage data available If Es Ew3 avaAmount 3 = (Hs Es) As Break up the loop If Es < Ew3 avaAmount 3 = (Hs Ew3) As Break up the loop

PAGE 296

275 D. If there is a weir in between and downstrean water stage data available If Hwater stage Ew3 maxStorage = (Hs – Hwater stage) As if maxStorage remainingStorage avaAmount 3 = remainingStorage if maxStorage < remainingStorage avaAmount 3 = maxStorage remainingStorage = remainingStorage avaAmount 3 If Hwater stage < Ew3 maxStorage = (Hs – Ew3) As if maxStorage remainingStorage avaAmount 3 = remainingStorage if maxStorage < remainingStorage avaAmount 3 = maxStorage Break up the loop Loop 3 neighbour 2 1. If neighbour 2 is of land segment type A. If there is no weir between s ource land segment and neighbour 2 If H2 Es potWLIncrease = remainingStorage / (As + A2 + A3 + A4) If potWLIncrease boundaryHeightDiff1,2 avaAmount 2 = boundaryHeightDiff1,2 A2 avaAmount 3 = avaAmount 3 + boundaryHeightDiff1,2 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff1,2 A4 remainingStorage = remainingStorage boundaryHeightDiff1,2 (As + A2 + A3 + A4) If potWLIncrease < boundaryHeightDiff1,2 avaAmount 2 = potWLIncrease A2 avaAmount 3 = avaAmount 3 + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If H2 < Es receivingVol = (Es H2) (A2 + A3 + A4)

PAGE 297

276 If remainingStorage receivingVol potWLIncrease = [receivingVol + (remainingStorage receivingVol) (A2 +A3 +A4)/ (As + A2 + A3 + A4)] / (A2 + A3 + A4) If potWLIncrease boundaryHeightDiff1,2 avaAmount 2 = boundaryHeightDiff1,2 A2 avaAmount 3 = avaAmount 3 + boundaryHeightDiff1,2 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff1,2 A4 If NBH1 Es remainingStorage = remainingStorage boundaryHeightDiff1,2 (A2 + A3 + A4) (NBH1 Es) As If NBH1 < Es remainingStorage = remainingStorage boundaryHeightDiff1,2 (A2 + A3 + A4) If potWLIncrease < boundaryHeightDiff1,2 avaAmount 2 = potWLIncrease A2 avaAmount 3 = avaAmount 3 + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If remainingStorage < receivingVol potWLIncrease = remainingStorage / (A2 + A3 + A4) If potWLIncrease boundaryHeightDiff1,2 avaAmount 2 = boundaryHeightDiff1,2 A2 avaAmount 3 = avaAmount 3 + boundaryHeightDiff1,2 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff1,2 A4 remainingStorage = remainingStorage boundaryHeightDiff1,2 (A2 + A3 + A4) If potWLIncrease < boundaryHeightDiff1,2 avaAmount 2 = potWLIncrease A2 avaAmount 3 = avaAmount 3 + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. B. If there is a weir between source land segment and neighbour 2 If H2 Ew2 potWLIncrease = remainingStorage / (As + A2 + A3 + A4) If potWLIncrease boundaryHeightDiff1,2 avaAmount 2 = boundaryHeightDiff1,2 A2

PAGE 298

277 avaAmount 3 = avaAmount 3 + boundaryHeightDiff1,2 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff1,2 A4 remainingStorage = remainingStorage boundaryHeightDiff1,2 (As + A2 + A3 + A4) If potWLIncrease < boundaryHeightDiff1,2 avaAmount 2 = potWLIncrease A2 avaAmount 3 = avaAmount 3 + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If H2 < Ew2 receivingVol = (Ew2 H2) A2 if remainingStorage receivingVol potWLIncrease = (remainingStorage receivingVol) / (As + A2 + A3 + A4) if potWLIncrease boundaryHeightDiff1,2 avaAmount 2 = receivingVol + boundaryHeightDiff1,2 A2 avaAmount 3 = avaAmount 3 + boundaryHeightDiff1,2 A3 avaAmount 4 = avaAmount 4 + boundaryHeightDiff1,2 A4 remainingStorage = remainingStorage receivingVol boundaryHeightDiff1,2 (A2 + A3 + A4) if potWLIncrease < boundaryHeightDiff1,2 avaAmount 2 = receivingVol + potWLIncrease A2 avaAmount 3 = avaAmount 3 + potWLIncrease A3 avaAmount 4 = avaAmount 4 + potWLIncrease A4 Break up the loop. If remainingStorage < receivingVol avaAmount 2 = remainingStorage Break up the loop. 2. If neighbour 2 is of river type A. If there is no weir in between and no downstrean water stage data available avaAmount 2 = remainingStorage Break up the loop. B. If there is no weir in between and downstrean water stage data available maxStorage = (Hs Hwater stage) As If maxStorage remainingStorage

PAGE 299

278 avaAmount 2 = remainingStorage Break up the loop If maxStorage < remainingStorage avaAmount 2 = maxStorage remainingStorage = remainingStorage avaAmount 2 C. If there is a weir in between and no downstrean water stage data available If Es Ew2 avaAmount 2 = (Hs Es) As Break up the loop If Es < Ew2 avaAmount 2 = (Hs – Ew2) As Break up the loop D. If there is a weir in between and downstrean water stage data available If Hwater stage Ew2 maxStorage = (Hs Hwater stage) As If maxStorage remainingStorage avaAmount 2 = remainingStorage If maxStorage < remainingStorage avaAmount 2 = maxStorage remainingStorage = remainingStorage avaAmount 2 If Hwater stage < Ew2 maxStorage = (Hs Ew2) As If maxStorage remainingStorage avaAmount 2 = remainingStorage If totAvaStorage < remainingStorage avaAmount 2 = maxStorage Break up the loop Loop 4 neighbour 1 1. If neighbor 1 is of land segment type A. If there is no weir in between If H1 Es avaAmount 1 = remainingStorage A1 / (As + A1 + A2 + A3 + A4)

PAGE 300

279 avaAmount 2 = avaAmount 2 + remainingStorage A2 / (As + A1 + A2 + A3 + A4) avaAmount 3 = avaAmount 3 + remainingStorage A3 / (As + A1 + A2 + A3 + A4) avaAmount 4 = avaAmount 4 + remainingStorage A4 / (As + A1 + A2 + A3 + A4) If H1 < Es receivingVol = (Es H1 ) ( A1 + A2 + A3 + A4 ) If remainingStorage receivingVol amountToAllNeigh = receivingVol + (remainingStorage receivingVol) (A1 + A2 + A3 + A4) / ( As + A1 + A2 + A3 + A4) avaAmount 1 = amountToAllNeigh A1 / (A1 + A2 + A3 + A4) avaAmount 2 = avaAmount 2 + amountToAllNeigh A2 /(A1 + A2 + A3 + A4) avaAmount 3 = avaAmount 3 + amountToAllNeigh A3 / (A1 + A2 + A3 + A4) avaAmount 4 = avaAmount 4 + amountToAllNeigh A4 / (A1 + A2 + A3 + A4) If remainingStorage < receivingVol avaAmount 1 = remainingStorage A1 / (A1 + A2 + A3 + A4) avaAmount 2 = avaAmount 2 + remainingStorage A2 / (A1 + A2 + A3 + A4) avaAmount 3 = avaAmount 3 + remainingStorage A3 / (A1 + A2 + A3 + A4) avaAmount 4 = avaAmount 4 + remainingStorage A4 / (A1 + A2 + A3 + A4) B. If there is a weir in between If H1 Ew1 avaAmount 1 = remainingStorage A1 / (As + A1 + A2 + A3 + A4) avaAmount 2 = avaAmount 2 + remainingStorage A2 / (As + A1 + A2 + A3 + A4) avaAmount 3 = avaAmount 3 + remainingStorage A3 / (As + A1 + A2 + A3 + A4) avaAmount 4 = avaAmount 4 + remainingStorage A4 / (As + A1 + A2 + A3 + A4) If H1 < Ew1 receivingVol = (Ew H1) A1 If remainingStorage receivingVol amountToAllNeigh = (remainingStorage receivingVol) (A1 + A2 + A3 + A4) / ( As + A1 + A2 + A3 + A4) avaAmount 1 = receivingVol + amountToAllNeigh A1 / (A1 + A2 + A3 + A4) avaAmount 2 = avaAmount 2 + amountToAllNeigh A2 / (A1 + A2 + A3 + A4) avaAmount 3 = avaAmount 3 + amountToAllNeigh A3 / (A1 + A2 + A3 + A4) avaAmount 4 = avaAmount 4 + amountToAllNeigh A4 / (A1 + A2 + A3 + A4) If remainingStorage < receivingVol avaAmount 1 = remainingStorage

PAGE 301

280 2. If neighbour 1 is of river type A. If there is no weir in between and no downstrean water stage data available avaAmount 1 = remainingStorage B. If there is a weir in between and no downstrean water stage data available If Es Ew1 avaAmount 1 = (Hs Es) As If Es < Ew1 avaAmount 1 = (Hs Ew1) As C. If there is no weir in between and downstrean water stage data available maxStorage = (Hs – Hwater stage) As If maxStorage >= remainingStorage avaAmount 1 = remainingStorage If maxStorage < remainingStorage avaAmount 1 = maxStorage D. If there is a weir in between and downstrean water stage data available If Hwater stage Ew1 maxStorage = (Hs Hwater stage) As If maxStorage remainingStorage avaAmount 1 = remainingStorage If maxStorage < remainingStorage avaAmount 1 = maxStorage If Hwater stage < Ew1 maxStorage = (Hs – Ew1) As If maxStorage remainingStorage avaAmount 1 = remainingStorage If maxStorage < remainingStorage avaAmount 1 = maxStorage

PAGE 302

281 Nomenclature As Area of source land segment Ad Area of destination land segment A1 Area of adjacent land segment 1 A2 Area of adjacent land segment 2 A3 Area of adjacent land segment 3 A4 Area of adjacent land segment 4 amountToAllNeigh Amount of water for all adjacent land segments avaAmount Allocated amount of storage for single adjacent land segment avaAmount 1 Accumulated amount of storage for adjacent land segment 1 avaAmount 2 Accumulated amount of storage for adjacent land segment 2 avaAmount 3 Accumulated amount of storage for adjacent land segment 3 avaAmount 4 Accumulated amount of storage for adjacent land segment 4 boundaryHeightDiff1,2 Difference of boundary heights between land segment 1 and 2 boundaryHeightDiff2,3 Difference of boundary heights between land segment 2 and 3 boundaryHeightDiff3,4 Difference of boundary heights between land segment 3 and 4 Hs Water level of source land segment Hd Water level of destination land segment H1 Water level of adjacent land segment 1 H2 Water level of adjacent land segment 2 H3 Water level of adjacent land segment 3 H4 Water level of adjacent land segment 4 Hwater stage Downstream water stage Es Elevation of source land segment Ew1 Weir elevation between source land segment and land segment 1 Ew2 Weir elevation between source land segment and land segment 2 Ew3 Weir elevation between source land segment and land segment 3 Ew4 Weir elevation between source land segment and land segment 4 maxStorage Maximum storage can be transferred NBH1 Boundary height between land se gment 1 and source land segment NBH2 Boundary height between land se gment 2 and source land segment NBH3 Boundary height between land se gment 3 and source land segment receivingVol Portion of water volume is allocated to an adjacent land segment remainingStorage Water storage has no been allocated potWLIncrease Potential water level increase w ith a given amount of water input totAvaStorage Total available amount if water storage in source land segment

PAGE 303

282 LIST OF REFERENCES Acutis M and Donatelli M. 2003. SOILPAR 2. 00: software to estimate soil hydrological parameters and functions. European Journal of Agronomy 18: 373-377. Adams B, White A, and Lenton TM. 2003. An an alysis of some diverse approaches to modeling terrestrial net primary productiv ity. Ecological Modelling 177: 354-391. Allen, R.G., Pereira, L.S., Raes, D., and M. Smith. 1998. Crop evapotranspiration: Guidelines for computing crop water requirements. FAO Irrigation and Drainage Paper Number 56, Food and Agricultura l Organization of the United Nations, Rome. 300 pp. Anat A, Duke HR, and Corey AT. 1965. Steady upward flow from water tables. Hydrology Paper No. 7, Colorado State University, Fort Collins, CO. Armstrong W. 1964. Oxygen diffusion from the roots of British bog plants. Nature 204: 801-802. Arnold JG, Williams JR, Nicks AD, Samm ons NB. 1990. SWRRB: A Basin Scale Simulation Model for Soil and Water Resources Management. Texas A&M University Press: College Station, TX. Arnold JG, Allen PM, and Bernhardt G. 1993. A comprehensive surface–groundwater flow model. Journal of Hydrology 142:47-69. Arnold JG and Fohrer N. 2005. SWAT2000: current capabilities and research opportunities in applied watershed mode ling. Hydrological Processes 19:564-572. American Society of Civil Engineers (ASC E) Task Committee on Definition of Criteria for Evaluation of Watershed Models of the Watershed Management Committee. 1993. Criteria for evaluation of watershe d models. Journal of Irrigation and Drainage Engineering 119(3): 429–442. Beasley DB and Huggins LF. 1980. ANSWERS (Area Non-point Source Watershed Environment Response Simulation), User’s manual. Purdue University, West Lafayette. Beven KJ and Kirkby MJ. 1979. A physically-bas ed variable contri buting area model of basin hydrology. Hydrology Scie nce Bulletin 24(1):43-69.

PAGE 304

283 Bhat KKS, Flowers TH, and O’Callagham JR 1981. Nitrogen losses and surface runoff from landspreading of manures. In: J.C. brogan (Eds.). Proceedings of a workshop in the European Economic Community Progr amme of Coordination of Research on Effluents from Livestoc k. Wexford, Ireland, May 20-22, 1980. pp. 222-246. Bouraoui F and Dillaha TA. 1996. ANSWER S-2000: Runoff and sediment transport model. Journal of Environment Engineering, ASCE 122(6): 493-502. Bottcher AB, Tremwel TK, and Campbell KL. 1995. Best management practices for water quality improvement in the la ke Okeechobee watershed. Ecological Engineering 5: 341-356. Bottcher D, Pickering N, and Cooper AB. 1998. EAAMOD-FIELD: A flow and phosphorus model for high water tables. In Drainage in the 21st Century: Food Production and the Environment, ed. L. C. Brown, 599-606. St. Joseph, MI.: ASAE. Bradley C and Gilvear DJ. 2000. Saturated and unsaturated flow dynamics in a floodplain wetland. Hydrological Processes 14: 2945-2958. Broadbent FE and Clark FE. 1965. Soil nitr ogen, W.V. Bartholomew and F.E. Clark (Eds.). Agron. 10, 344. Brooks RH and Corey AT. 1964. Hydraulic prop erties of porous media. Hydrology Paper No. 3, Colorado State University, Fort Collins, CO. Burnash RJC, Ferral RL, and McGuire RA 1973. A General Streamflow Simulation System—Conceptual Modeling for Digital Co mputers. Report by the Joliet Federal State River Forecasts Center, Sacramento, CA. Busnardo MJ, Gersberg RM, Langis R, Sinicr ope TL, and Zedler JB. 1992. Nitrogen and phosphorus removal by wetland mesocosms s ubjected to different hydroperiods. Ecological Engineering 1: 287-307. Capece JC. 1994. Hydrology and contaminant transport on flatwoods watersheds. Ph.D. dissertation. Gainesville, FL: University of Florida. Campbell KL, Capece JC, and Tremwel TK. 1995. Surface/subsurface hydrology and phosphorus transport in the Kissimmee River Basin, Florida, Ecological Engineering 5: 301-330. Campbell KL, Kiker GA, and Clark DJ. 2001. De velopment and testing of a nitrogen and phosphorus process model for Southern Af rican water quality issues. 2001 ASAE Annual International meeting. Pape r No. 012085, St. Joseph, MI.: ASAE. Cannell MGR, Milne R, Sheppard LJ, and Un sworth MH. 1987. Radiation interception and productivity of willow. J Appl Ecol 24: 261-78.

PAGE 305

284 Carlisle VW, Sodek F, Coillins III ME, Hammond LC, and Harris WG. 1989. Characterization data for selected Florida soils. University of Florida, IFAS Soil Science Department soil Characterizati on laboratory in cooperation with USDA SCS. Chen DX, and Coughenour MB. 1994. GEMT a general model for energy and mass transfer of land surfaces and its application at the Fife s ites. Agricultural and Forest Meteorology 68: 145171. Chiang C, Craft CB, Rogers DW, and Richards on CJ. 2000. Effects of 4 years nitrogen and phosphorus additions on Everglades plant communities. Aquatic Botany 68: 61-78. Chiang WH and Kinzelbach W. Accessed at April 15, 2005. Processing MODFLOW for Windows (PMWIN), Version 5.3.0. Ava ilable http://www.pmwin.net/pmwin5.htm. Childs SW and Hanks JR. 1975. Model of soil salinity effects on crop growth. Soil Science Society of American Journal 39: 617-622. Clark DJ, Kiker GA, and Schulze RE. Septembe r 2001. Object-oriented restructuring of the ACRU agrohydrological modeling sy stem, Tenth South African National Hydrology Symposium 26-28. Crawford NH and Linsley RS. 1966. Digita l Simulation in Hydrology: The Stanford Watershed Model IV. Technical Report no. 39, Department of Civil Engineering, Stanford University, Palo Alto, CA. Cunge JA. 1975. Two-dimensional modeling of flood plains, Unsteady Flow in Open Channels. Volume II, Chapter 17. Water Resources Publications, P.O. Box 303, Fort Collins, Colorado 80533, USA: 705-762. Danish Hydraulic Institute (DHI). 2004. MI KE SHE Water Movement User Guide and Technical Reference Manual. Agern All 5 DK-2970 Hrsholm, Denmark. Davis SM. 1991. Growth, decomposition, and nutrient retention of Cladium Jamaicense Crantz Typha Domingensis Pers. In the Florida Everglades. Aquatic Botany 40: 204-224. Davis JA and Froend R. 1999. Loss and Degradat ion of wetlands in southern Australia: underlying causes, consequences an d solutions. Wetlands Ecology and Management 7: 14-23. Evans BM, Sheeder SA, and Lehning DW. 2003. A spatial technique for estimating streambank erosion based on watershed ch aracteristics. Journal of Spatial Hydrology, Vol.3, No.1.

PAGE 306

285 Ferreira VA, Weesies GA, Yoder DC, Foster GR, and Renard KG. 1995. The site and condition specific nature of sensitivity analysis. Journal of Soil and Water Conservation 50(5): 493-497. Fisher J and Acreman MC. 2004. Wetland nutrie nt removal: a review of the evidence. Hydrology and Earth System Sciences 8(4): 673-685. Fitz HC, DeBellevue EB, Costanza R, Boumans R, Maxwell T, Wainger L, and Sklar FH. 1996. Development of a general ecosyst em model for a range of scales and ecosystems, Ecological Modelling 88: 264-295. Fitz HC, Wang N, Godin J, Sklar FH, Tr imble B, and Rutchey K. 2002. Calibration Performance of ELM v2.1a: 1979-1995 Wate r Quality and Hydrology. Report to the RECOVER Model Refinement Team, available at http://www.sfwmd.gov/org/wrp/elm/results /cal_ver/elm2.1/ELMcalibAnalysis_draf t.pdf, South Florida Water Management District, West Palm Beach, FL. Flaig EG and Reddy KR. 1995. Fate of phosphor us in the lake Okeechobee watershed, Florida, USA: overview and recommendati ons. Ecological Engineering 5: 127-142. Flaig EG and Havens KE. 1995. Historical tr ends in the Lake Okeechobee ecosystems. I. Land use and nutrient loadings. Arch Hydrobiol. (Suppl.), 107: 1-24. Fraisse CW and Campbell KL. 1996. Modifi ed FHANTM for water quality modeling of beef cow/calf operations in south Flor ida. Paper No. 962057, St. Joseph, MI.: ASAE. Fraisse CW and Campbell KL. 1997. FHAN TM (Field Hydrologic And Nutrient Transport Model) Version 2.0 User's Manua l. Gainesville, FL: Agricultural and Biological Engineering Dept., IF AS, University of Florida. Gurnell AM, Hupp CR, and Gregory SV 2000. Linking hydrology and ecology. Hydrological Processes 14: 2814-2815. Gunsalus B, Flaig EG, and Ritter GJ. 1992. Effectiveness of agricultural best management practices implemented in the Taylor Creek/Nubbin Slough and lower Kissimmee River basins. Proc. Natl. Rural Clean Waters Program Syrup, 1992, pp. 161-172. Hann CT and Zhang J. 1996. Impact of un certain knowledge of model parameters on estimated runoff and phosphorus loads in the Lake Okeechobee Basin. Trans. ASAE 39(2): 511-516. Hedden KF. 1986. Example field testing of soil fate and transport model, PRZM, Dougherty Plain, Georgia. In Vadose Zone Modeling of Organic Pollutants, eds. S. C. Hern and S. M. Melancon, 81-101, Mich.: Lewis Publishers Inc.

PAGE 307

286 Hendricks GS. 2003. Performance evaluation of two hydrologic/water quality models on beef pastures at Buck Island Ranch. Master Thesis. Gainesville, FL: University of Florida. Heatwole CD. 1986. Field and basin scal e water quality models for evaluating agricultural nonpoint pollution abatement programs in a south Florida Modified CREAMS hydrology model for Coastal Plai n watersheds. Trans. ASAE 30(4): 1014-1022. Heatwole CD, Campbell KL, and Bottche r AB. 1987. Modified CREAMS hydrology model for Coastal Plain watershe ds. Trans. ASAE 30(4): 1014-1022. Hession WC, Shanholtz VO, Mostaghimi S, and Dillaha TA. 1994. Uncalibrated performance of the finite element stor m hydrograph model. Trans. ASAE 37(3): 777-783. Hunt HW, Trlica MJ, Redente EF, Moore JC, Detling JK, Kittel TGF, Walter DE, Fowler MC, Klein DA, and Elliot ET. 1991. Simulation model for the effects of climate change on temperate grassland ecosy stems. Ecological Modelling 53:205-246. Hydrologic Engineering Center. 1981. HE C-1, Flood Hydrograph Package—User’s Manual. US Army Corps of Engineers: Davis, CA. Innis GS. 1978. Grassland simulation model. Ecological Studies 26. New York: SpringerVerlag. Ivens WPMF. 1992. World Food Production: Biophysical Factors in Agricultural Production, Open universiteit, Heerlen, the Netherlands. Jones CA, Cole CV, Sharpley AN, and Williams JR. 1984. A simplified soil and plant phosphorus model: I. Documentation. Soil Sci. Soc. Am. J. 48: 800-805. Jones JW, and Luyten J. 1998. Simulation of biological processes. In Agricultural Systems Modeling and Simulation, eds. R. M. Peart and R.B. Curry. Marcel Dekker, Inc. pp 19-62. Jones JW, Hoogenboom G, Porter CH, Boote KJ, Batchelor WD, H unt LA, Wilkens PW, Singh U, Gijsman AJ, and Ritchie JT. 2003. The DSSAT cropping system model. Europ. J. Agronomy 18: 235-265. Jones JW, Boote KJ, White J, Hoogenboom G, and Porter CH. 2000. Phenology submodule for CROPGRO, in DSSAT v.4.0 Do cumentation and source code listing. Agricultural and Biological Engineeri ng Department Research report No. 20001201. University of Florida, Gainesville, Florida. Kiker GA. 1998. Development and comparison of savanna ecosystem models to explore the concept of carrying capacity. Ph D Dissertation. Cornell University.

PAGE 308

287 Kiker GA and Clark DJ. 2001. The developm ent of a Java-based, object-oriented modeling system for simulation of S outhern African hydrology. 2001 ASAE Annual International Meeting. ASAE Pa per no. 012030. St. Joseph, MI.:ASAE. Knisel WG (ed.). 1980. CREAMS: A field scal e model for chemicals, runoff, and erosion from agricultural management systems. USDA Conservation Research Report 26: 643. Knisel WG, Yates P, Sheridan JM, Woody TK, Allen III LH, and Asmussen LE. 1985. Hydrology and hydrogeology of Upper Taylor Creek watershed, Okeechobee County, Florida: data and analysis. ARS25, USDA-Agricultural Research Service, Washington, D.C. Knisel WG, Leonard RA, and Davis FM. 1993. GLEAMS version 2.10 Part I: Nutrient Component Documentation. Unpublis hed, USDA/ARS. Tifton, GA. 44p. Knisel WG and Davis FM. 1999. GLEAMS (Groundwater Loading Effects of agricultural Management Systems) Version 3.0 User Manual. SEWRLWGK/FMD-050199. Tifton, GA. 167p. Krysanova V, Meiner A, Roosaare J, and Va silyev A. 1989. Simulation modelling of the coastal waters pollution from agricultural watershed. Ecological Modelling 49: 729. Krysanova V, Mller-Wohlfeil DI, Becker A. 1998. Development and test of a spatially distributed hydrological/wat er quality model for mesoscale watersheds. Ecological Modelling 106: 261–289. Krysanova V, Hattermann F, and Wechsung F. 2005. Development of the ecohydrological model SWIM for regiona l impact studies and vulnerability assessment. Hydrologic Processes 19: 763-783. Lal AMW. 1998. Weighted implicit finite-vol ume model for overland flow. Journal of Hydraulic Engineering. 124: 941-950. Lauenroth WK, Urban DL, Co ffin DP, Parton WJK, Shugart HH, Kirchner TB, and Smith TM. 1993, Modeling vegetation struct ure-ecosystem processes interactions across sites and ecosystem. Ec ological Modeling 67: 49-80. Leith H. 1975a. Modeling the primary productivi ty of the world. In: Leith H, Whittalcher, RH. (Eds); Primary Productivity of the Biosphere. Springer-Verlag, New York. Leonard RA, Knisel WG, and Still DA. 1987. GLEAMS: Groundwater loading effects of agricultural management systems. Transact ions of the ASAE 30 (5): 1404-1418. Martinez C. 2006. Object-oriented hydrologi c and water-quality model for high-watertable environments. Ph.D. dissertation. Ga inesville, FL: University of Florida.

PAGE 309

288 MacArthur Ecology Research Center (MAE RC). Optimization of best management practices (BMPs) for beef cattle ranching in the Lake Okeechobee Basin. SFWMD contact C-13414. Final report. Prepared for the South Florida Water Management District. July 2004. McCaffery PM, Hinkley WM, MacGill R and Ch err GD. 1976. Report of investigations in the Kissimmee-Lake Okeechobee Watershe d. Fla. Depart. Environ. Reg. Tech. Series, Vol. 2, No. 2. Tallahassee. Florida. McDonald MG and Harbaugh AW. 1988. A Modular Three-dimensional Finitedifference Ground-water Flow Model. US Geological Survey, Techniques of Water Resources Investigation Book 6, Chapter A1; 586 pp. McDonald MG and Harbaugh AW. 1996. MODF LOW: the modular three dimensional finite-difference ground water flow model. USGS. Monsi M and Saeki T. 1953. Uber den Lictfakt or in den Pflanzengesell -schaften und sein Bedeutung fur die Stoffproduktion. Japan J. Bot. 14: 22-52. Mortimer C. 1941. The exchange of dissolved substances between mud and water in lakes. Journal of Ecology 30: 147-201. Nair VD, Graetz DA, and Portier KM. 1995. Forms of phosphorus in soil profiles from dairies of south Florida. Soil Sc i. Soc. Am. Journal 59: 1244-1249. Nalluri C and Judy ND. 1989. Factors affec ting roughness in coefficient in vegetated channels. In: B.C. Yen (Editors), Channe l Flow and Catchment Runoff: Centennial of Manning's Formula and Kuichling's Rati onal Formula. Univer sity of Virginia, Charlottesville, VA, pp. 589. Nash JE and Sutcliffe JV. 1970. River flow fo recasting through conceptual models part I a discussion of principles Journal of Hydrology 10: 282–290. New MG and Schulze RE. 1996. Hydrologic se nsitivity to climate change in the Langriver catchment, Stellenbosch, Sout h Africa and some implications for geomorphic processes. Zeitschrift fur Geomorephologie, Supplementband 107: 1134. Newman S, Grace JB, and Koebel JW. 1996. Effects of nutrients and hydroperiod on Typha, Cladium, and Eleocharis: implicati on for Everglades restoration. Ecological Applications 6(3): 774-783. Nichols DS. 1983. Capacity of natural wetla nds top remove nutrients from wastewater. Journal of the Water Polluti on Control Federa tion 55: 495-505. Parton WJ, Schimel DS, Cole CV, and Ojima DS. 1987. Analysis of factors controlling soil organic matter levels in Great Plai ns Grasslands. Soil Society of America Journal 51: 11731179.

PAGE 310

289 Parton WJ, Coughenour MB, Scurlock JMO, Ojima DS, Gilmanov TG, Scholes RJ, Schimel DS, Kirchner TB, Menaut JC, S eastedt TR, Garcia ME, Kamnalrut A, Kinyamario JI, and Hall DO. 1996. Globa l grassland ecosystem modelling: development and test of ecosystem models for grassland systems. In: Breymeyer AI, Hall DO, Melillo JM, Agren GI, eds. Global change: effects on coniferous forests and grasslands. (SCOPE 56). Ch ichester: John Wiley and Sons, 229267. Petryk S, Asce AM and III B, 1975. Analysis of flow through vegetation. J. Hydraul. Div., HY7: 871-883. Refsgaard JC. 1997. Parameterization, calib ration and validation of distributed hydrological models. Journal of Hydrology 198: 69-97. Refsgaard JC and Storm B. 1995. MIKE SHE. In Computer Models of Watersheds Hydrology, Singh V (ed.). Water Resources P ublication: Highlands Ranch. CO; 809-846. Richardson CJ. 1985. Mechanisms control ling phosphorus retention capacity in freshwater wetlands. Sc ience 228: 1424-1427. Ritchie JT. 1972. Model for predicting evapor ation from a row crop with incomplete cover. Water Resources Research 8(5): 1204-1213. Rockwood DM, Davis ED, and Anderson JA. 1972. User Manual for COSSARR Model. US Army Engineering Division, North Pacific: Portland, OR. Rogers JS. 1985. Water management model ev aluation for shallow sandy soils. Trans. ASAE 28(3): 785-790. Rymph SJ. 2004. Modeling growth and compositi on of perennial tropical forage grasses. Ph.D. dissertation. Gainesville, FL: University of Florida. Schippers P and Kropff MJ. 2001. Competition for light and nitrogen among grassland species: a simulation analysis Function Ecology 15: 155-164. Schulze RE. 1989. ACRU: Background, Concepts and Theory. WRC Report No. 154/1/89. ACRU Report No.35. Schulze RE. 1995. Hydrology and Agrohydrology: a text to accompany the ACRU 3.00 Agrohydrological modelling System. WRC Report No. TT69/95. ACRU Report No.43. Schulze RE, Kiker GA, and Kunz RP. 1993. Global climate-change and agricultural productivity in southern Africa. Gl obal environmental Change 3:330-349.

PAGE 311

290 Seligman NG and Van Keulen H. 1981. PAPR AN: A simulation model of annual pasture production limited by rainfall and nitrogen. p. 192-220. In M.J. Frissel and J.A. van Veen (ed.) Simulation of nitrogen be haviour of soil-plant systems. Pudoc. Wageningen, the Netherlands. Sharpley AN and Williams JR (Eds). 1990. EP IC-Erosion/Productivity impact calculator: 1. Model Documentation. U.S. Department of Agriculture, Agricultural Research Service, technical Bulletin No. 1768. 235p. South Florida Water Management Di strict (SFWMD). 1997a. Surface water improvement and management (SWIM) pl an update for Lake Okeechobee. West Palm Beach, Florida. South Florida Water Management Distri ct (SFWMD). 1997b. SFWMD wetlands model: calibration of the coupled periphyton/ve getation model to WCA-2A. HydroQual, Inc. October 21. West Palm Beach, Florid a (material via private communication). South Florida Water Management District (SFWMD). 1998. Natu ral system model version 4.5 documentation. Planning de partment. West Palm Beach, Florida. February 1998. South Florida Water Management District (SFWMD). 2005. Documentation of the South Florida Water Management Model, version 5.5. Final Draft. West Palm Beach, Florida (material via private communication). South Florida Water Management Distri ct (SFWMD). Accesses April 28, 2005. Hydrologic Systems Modeling. Available at: http://www.sfwmd.gov/org/pld/hsm/hsm.html. Spitters CJT and Schapendonk AHCM. 1990. Ev aluation of breeding strategies for drought tolerance in potato by means of crop growth simulation. Plant and Soil 123: 193-203. Sims R. 2004. Performance evaluation of the field hydrologic and nutrient transport model (FHANTM 2.0) on beef pastures at Buck Island Ranch, Fl orida. A technical paper presented to the gradua te school of the University Of Florida in partial fulfillment of the requirements for the degree of master of interdisciplinary ecology. University of Florida. Skaggs RW. 1980. DRAINMOD reference repor t: methods for design and evaluation of drainage-water management systems fo r soils with high water tables. Soil Conservation Service, United States Depa rtment of Agriculture, Fort Worth, TX. Sklar FH, Fitz HC, Wu Y, Zee RV, and McVoy C. 2001, The design of ecological landscape models for Everglad es restoration. Ecologica l Economics 37: 379-401. Smithers JC and Caldecott RE. 1993. Devel opment and verification of hydrograph routing in a daily simulationmodel. Water SA, 19, 263-267.

PAGE 312

291 Smith RE. 1992. Opus, An integrated simula tion model for transpor t of nonpoint-source pollutants at the field scale, Vols. 1 & 2. US Department of Agriculture, ARS-98. Sugawara M, Ozaki E, Wantanabe I, a nd Katsuyama Y. 1976. Tank model and its application to Bird Creek, Wollombi Brook, Bihin River, Sanaga River, and Nam Mune. Research Note 11, National Center for Disaster Prevention, Tokyo, Japan. Sun G, Riekerk H, and Comerford NB. 1998. Modeling the forest hydrology of wetlandupland ecosystems in Florida, Journal of the American Water Resources Association, 34(4): 827-841. Thornley JHM and Cannell MGR. 1997. Temp erate grassland response to climate change: an analysis using the Hurley Pasture Model. Annals of Botany 80: 205221. Thomann RV. 1982. Verification of water qua lity models. Journal of Environmental Engineering. Div. ASCE 108(5): 923-939. Tremwel TK. 1992. Field hydrologic and nut rient transport model, FHANTM. Ph.D. dissertation. Gainesville, FL: University of Florida. Urban NH, Davis SM, and Au men NG. 1993. Fluctuations in Sawgrass and Cattail densities in Everglades Water Conser vation area 2A under varying nutrient, hydrologic and fire Regimes. Aquatic Botany 46: 204-223. United States Department of Agriculture (USDA). 1990. State soil geographic database for Florida. U.S. Department of Agri culture, Soil Conservation Service, Gainesville, Florida. United States Department of Agricultur e (USDA), SCS. 1996. Agricultural waste management field handbook. United St ates Department of Agriculture. Washington, D.C. Uusi-Kamppa J, Turtola E and Ylaranta T. 1997. The interactions of buffer zones and phosphorous runoff. In: Buffer Zones: Their pro cess and potential in Water Protection, N.E. Haycock, T.P. Burt, K.W.T. Goulding and G. Pinay. (Eds) Quest Envronmental, Harpenden, UK. Van Ittersum MK, Leffelaar PA, Van Keulen H, Kropff MJ, Bastiaans L, and Goudriaan J. 2003. On approaches and applicat ions of the Wageningen crop models. European Journal of Agronomy 18:201-234. Verberne ELJ. 1992. Simulation of the nitr ogen and water balan ce in a system of grassland and soil. Nota 258. DLO-Ins titut voor Bodemvruchtbaarheid, Oosterweg 92, 9750 RA, Haren, Netherlands.

PAGE 313

292 White JR. 1987. ERHYM-II: model description and user guide for the BASIC version. US Department of Agriculture, Agricultural Research Service, ARS-59, Washington, D. C. Whittaker RH and Levin SA. 1975. Niche: theo ry and application. Stroudsburg, Pa.: Dowden, Hutchinson & Ross. New York : distributed by Halsted Press, William JR, Jones CA, and Dyke PT. 1984. A modeling approach to determining the relationship between erosion and soil pr oductivity. Transactions of the ASAE 27:129-144. Williams JR and Hann RW. 1983. HYMO: Pr oblem-oriented language for hydrologic modeling—User’s Manual. USDA: ARS-S-9. Williams JR, Renard KG, and Dyke PT. 1983. EPIC–a new model for assessing erosion’s effect on soil productivity. Journal of So il and Water Conservation 38(5): 381-383. Wu Y, Sklar FH, and Rutchey K. 1997. An alysis and Simulation of Fragmentation Patterns in the Everglades. Ecol ogical Application 7: 268-276. Young GKK and Alward CL. 1983. Calibration and testing of nutrient and pesticide transport models. In Agricultural Management and Water Quality. F. W. Schaller and G. W. Bailey (eds.), 263-277, Ames: Iowa Sate Univ. Press. Young RA, Onstad CA, Bosch DD, and Anderson WP. 1989. AGNPS: A nonpoint source pollution model for evaluating agricu ltural watersheds. Journal of Soil and Water Conservation 44(2): 168–173. Zhang J, Haan CT, Tremwel TK, and Kike r GA. 1995. Evaluation of phosphorus loading models for south Florida. Trans. ASAE 38(3): 767-773. Zhang J and Gornak SI. 1999. Evaluation of fiel d-scale water quality models for the Lake Okeechobee regulatory program. Applied E ngineering in Agriculture 15(5): 441447.

PAGE 314

293 BIOGRAPHICAL SKETCH Lei Yang was born in Henan province, People’s Republic of China, in a family led by fairly open-minded parents who always en courage their kids to pursue what they dreamed in their lives. Lei graduated from Hohai University, China, with a BS in 1991 and a MS in 1994 majoring in hydrology and wa ter resources engineering. Upon the completion of her MS, she was then employed as a civil engineer in th e China Institute of Water Resources and Hydropower Research fo r close to three years. Excelling in research projects developing flood inundati on and mitigation simulation models, in October 1997 she was recommended to join a research student program in Kyoto University, Japan, financed with a Ja panese MOBUSHO scholarship. With the experience gained during that period, Lei star ted to pursue one of her dreams that she wanted to reach a higher professional level in the field of hydrol ogy and water resources through advanced study. In August 2000, she joined the Agricultural and Biological Engineering Department at the University of Florida as a graduate research assistant and a PhD student.


Permanent Link: http://ufdc.ufl.edu/UFE0013090/00001

Material Information

Title: Coupled Simulation Modeling of Flatwoods Hydrology, and Nutrient and Vegetation Dynamids
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0013090:00001

Permanent Link: http://ufdc.ufl.edu/UFE0013090/00001

Material Information

Title: Coupled Simulation Modeling of Flatwoods Hydrology, and Nutrient and Vegetation Dynamids
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0013090:00001


This item has the following downloads:


Full Text












COUPLED SIMULATION MODELING OF FLATWOODS HYDROLOGY,
NUTRIENT AND VEGETATION DYNAMICS


















By

LEI YANG


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

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Lei Yang




























This dissertation is dedicated to

my parents,
my high school teacher Zhiting Wang,
and my friend Joseph S. Smithl

I recognize and appreciate the life-long influence they brought to me at different stages of
my life.















ACKNOWLEDGMENTS

I am greatly indebted to my supervisor, Dr. Wendy D. Graham, for her constant

guidance, insight, encouragement, and continuous support as well as confidence in my

research over the past five years. Her thorough and thoughtful coaching with all aspects

of my research was unselfishly tireless, and her enthusiasm for research and quest for

excellence have left me an everlasting impression. I would like to express sincere

appreciation to Dr. Kenneth L. Campbell for his invaluable advice on addressing each of

my technical problems and concerns. Without his constant supervision and guidance

throughout the model development, the completion of this model would have been

impossible. I am grateful to Dr. James W. Jones for his insight and invaluable advice on

the methodology of the vegetation dynamic model and support in offering important crop

growth model related literature; to Dr. Mark W. Clark for introducing me to the niche

theory and the interactive relationship between wetland hydrology and vegetation

dynamics and his help in identifying pasture vegetation species for simulation; to Dr.

Gregory A. Kiker for introducing me to the Java programming language with his great

enthusiasm and his technical support regarding the ACRU2000 modeling system during

the model development. I deeply benefited from many hours of precious discussions

with each of these committee members on a multitude of perspectives regarding my

research. Without their combined supervision of each step throughout my research, this

study would not have been possible.









I would like to acknowledge my debt to Chris Martinez for his great cooperation as

a teammate throughout the development of hydrologic and nutrient models. I wish also

to thank Dr. Michael D. Annable for his generous comments on specific technical

problems during many brown-bag group meetings; Dr. Patrick Bohlen in MacArthur

Agro-ecology Research Center, Florida, for introducing me to the pasture sites at Buck

Island Ranch and offering useful documents; Mr. Gregory S. Hendricks for offerings of

data for Buck Island Ranch and nutrient related documents; Dr. Stuart J. Rymph from the

University of Wisconsin for providing important bahiagrass related documents; Ms.

Cheryl H. Porter for offering crop modeling documents; and Dr. William Wise and his

graduate student Min Joong Hyuk for helping with DHI software.

Also, special thanks go to a few friends including Joseph S. Smith, Tricia G. Smith,

Donna L. Miller, Paul Miller, and David R. Murphy for their friendship and support

throughout the past few years, especially during some tough times. I would like to

acknowledge all the good friends for their friendship. I am also grateful to several

labmates for their friendship and encouragement, and faculty, staff and students in the

Agricultural and Biological Engineering Department for the quality academic

environment.

Finally I particularly appreciate my parents and siblings for their unconditional

love, understanding, patience and encouragement. Without their affection, it would have

been even more difficult to complete this research.















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST O F TA B LE S ............... .. ......................... .......... .... ............... x

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

A B S T R A C T .......................................... ..................................................x x

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

Study B background ...................... ...... ...... ........ .......... ................. .. 1
Overview of the Coupled Modeling System .............. ............................................8
Stu dy O bjectiv es .................................................. ................ .. 9

2 LITERATURE REVIEW ........................................................................... 12

Overview of Previous Modeling Efforts................................................................12
ACRU2000 Modeling System ......... ............... .................... 20
M odel Testing Procedures ................................................. ............................. 24
M odel C alibration .......... .......................................................... .. .... .... ..... 25
M odel V alidation .......... ..... ........................................................... ... .... ... ... 25
Sensitivity A analysis ...... .. .... .................................................. .. .... .... ... ..26
M odel E valuation ...................... .................... ................. ..... .... 28
S ta tistic s ................................................................................................. 2 8
G raphic representation ........................................ ........................... 31

3 HYDROLOGIC SIMULATION MODEL.....................................................34

Introduction ............................................................................................ ........ 34
Vertical Hydrologic Com ponents .....................................................................38
R a in fa ll ...............................................................3 8
C anopy Interception ....................... ............................ .. ....... .......... .. .. 38
Evapotranspiration .................. .. ...... .................. ....... .................... 39
In filtratio n ................................ ............. .. ................ ................ 4 2
Soil W ater Redistribution ........ .......... ... .... ......................................42
Upward Flux ................................ ................................ ......... 43
D eep Seepage ......................................................................44









Horizontal Hydrologic Components ................. ....... .... ......................... ............... 44
Overland Flow ......................................................................... ........ ..................45
O verland flow calculation ........................................ ........................ 46
Surface water storage apportionment................................................49
Ground after Flow ........ .... ............ .... .. ...... .. .... .. .. ............ 51
Lateral groundwater flow calculation .................................. ............... 54
Groundwater storage apportionment...... .... ....................... ..... .. ....... 56
Canal Flow.................. ........ ...................56
Initial and B oundary C onditions.......................................... ........................... 57
Model Testing and Validation ............................................58
Simulation Sequence and Model Performance Accuracy Analysis ..................58
Case 1: Overland flow along a flat rectangular plane ...............................60
Case 2: Overland and groundwater flow over an axisymmetric domain .....65
Application at Dry Lake Dairy #1, Kissimmee River Basin, Florida .................76
Site description ............... ......... .................76
R results and discussion ............................................ ........... ............... 79
Concluding R em arks ........................................ .... ....... .... ....... 82

4 NUTRIENT SIMULATION MODEL............................. .................... 87

In tro d u ctio n ............. ...... ...... ... ................. ................................ 8 7
N utrient Com ponents .................. .............................. ........ .......... .... 89
N itrogen Cycle C om ponents ........................................ .......... ............... 89
M in eralization ...........................................................9 1
Im m ob ilization .............................. ........................ .. ........ .... ............92
D enitrification ............. ......... .... ........ .. ...................... ...... .... 92
Runoff, sediment transport and percolation............. ................................. 93
U ptake, evaporation, and fixation ..................................... ............... 94
R ainfall and fertilizer ............................................................................95
A m m onia volatilization ..... ................................................ ... ... ............... 95
Surface and subsurface lateral nitrate nitrogen transport...........................95
Surface and subsurface lateral ammonium nitrogen transport...................97
Phosphorus Cycle Com ponents.......................................... ......... ... ............... 98
M in eralization ...........................................................99
Im m mobilization .......... .......... ..................... ..... ........ .. ........ .... 100
Runoff, sedim ent, percolation ............. ............... ................... .............. 100
Uptake and evaporation.. .. ..........................................101
R ainfall and fertilizer ...................... ............... .. ............... ... 101
Surface and subsurface lateral labile phosphorus transport .....................102
Conservative Solute Transport Components ...............................................104
Initial and B oundary Conditions......................................... .......................... 105
M odel Testing and Validation ............................................................................ 106
Conservative Solute Test ......................... ................... 106
Scenario description ............................................... ....... ... .. ............ 107
Results and discussion.................... ..... .... ..................... 107
Application at Buck Island Ranch, Lake Okeechobee Basin, Florida .............111
Project description ................................... ........ ........... ...... ........111









Sensitivity analysis ..................................... ................................... 117
R results and discussion ........................................... .......... ............... 122
C including R em arks .......................................... .. .. .... .. ....... ........133

5 VEGETATION DYNAMICS SIMULATION MODEL ......................................189

In tro d u ctio n ...............9..............................................
M methodology ...................................................................................................... 192
M odel Structure ........................... ............................. ...... .. ...... 192
Plant G row th.................................................... 193
P potential grow th .......................................... .. ......... ...............193
R educed grow th ............ .................................... ........ ........ ... ......... 196
Dry matter partitioning....... ........................................ ....... ........ 197
Leaf Area Index ............. ................... .................... .. ................. 197
Plant Senescence ........................................ .. .. .... .......... ....... 198
Evapotranspiration....... ............................................... ..... 200
N nitrogen U take ...................................... ................. .... ....... 201
Phosphorus Uptake ............... ............ ................... .......204
Growth Reduction Factor ........................................................................... 205
W ater stress and logging factors ..................................... ............... ..206
N itrogen stress factor ...........................................................................208
Phosphorus stress factor .............. ....... ......................... ............... 209
Hypothetical Scenario M odel Testing ........................................... ............... 209
Scenario D description ................... ................ ................. ..... ..... 209
Results and Discussion ...... .................... ........................213
Water and nutrient responses to water retention BMP............................213
Influence of differences in temperature sensitivities among species .........214
Species composition dynamics due to water retention BMP ...................216
Concluding R em arks ...................................... ................. .... ....... 219

6 SUMMARY AND CONCLUSIONS ........... ................................. ...............234

Hydrologic Simulation M odel ............................................................................ 235
N utrient Sim ulation M odel ................................................... ........................ 236
Vegetation Dynamics Simulation Model......................................................237
Im plications of the R research ......................................................... ............... 238
Future Research Recom m endations ........................................ ...... ............... 238
M odel Pre- and Post-processing Capacity....................................................... 238
Use Consistent Units for Parameters and Variables.................... .......... 239
Documentation .............................. .......... ...................239
Potential Changes to Existing Objects .............................................................239
Sub -D aily T im e Step .............................................................. .....................24 0
H erbivore M ovem ent M odule ........................................ ....................... 240
H ydrologic M odel ........................... ..... .. .. ...... ...............240
N utrient M odel ........................... ........ .. ......................241
V vegetation M odel .................................... ............... .... ....... 241



viii









APPENDIX

A FLOW CHART FOR THE COUPLED MODELING SYSTEM AND MODEL
PR O C E SSE S .............................................. ..................... ......... 242

B FLOW CHART OF LATERAL GROUNDWATER FLOW SIMULATION.........247

C LISTS OF NEW AND MODIFIED OBJECTS ............... .................. ............248

D LISTS OF NEW INPUT AND OUTPUT VARIABLES ................. ...............262

E LISTS OF M ODEL INPU T FILES ........ .. ............. ................... .....................265

F ALGORITHMS OF WATER STORAGE APPORTIONMENT...........................266

L IST O F R E F E R E N C E S ......... .. ............... ................. ..............................................282

BIOGRAPH ICAL SKETCH ...................................................... 293
















LIST OF TABLES


Table page

3-1 List of strategies for water transmission and water storage updating for
simulation sequence experim ents. .............................................. ............... 59

3-2 Statistics for the surface water depths predicted by the modified ACRU2000
model and MIKE SHE for the three experiments with the same simulation
sequence .............................................................................62

3-3 Statistics for the surface water depths predicted by the modified ACRU2000
model and MIKE SHE for the three experiments................................................69

3-4 Statistics for the simulated groundwater table depths by the modified
ACRU2000 model, MIKE SHE, and MODFLOW for the three experiments
w ith the sam e sim ulation sequence..................................................................... 75

3-5 Summary of annual water budget on Dry Lake Dairy #1 site ..............................78

3-6 Model input parameters for the Dry Lake Dairy #1 site .......................................85

3-7 Statistics for the simulated surface runoff and groundwater table depths by the
modified ACRU2000 and FHANTM. ...................................... ............... 85

4-1 Statistics for the solute mass predicted by the modified ACRU2000................08

4-2 List of stocking activities for pastures S1, S4, W6 and W7. ............................137

4-3 List of fertilization activities for pastures S1 and S4 .........................................137

4-4 List of burn activities for pastures S1, S4, W6 and W7..................................... 137

4-5 Percent of area occupied by different soil series and wetlands in selected
sum m er and w inter pastures.................................................................... ....... 137

4-6 Model input parameters for the winter pastures W6 and W7...........................138

4-7 Model input parameters for the summer pastures S4 and S1............................139

4-8 Selected hydrologic parameters for sensitivity analysis for W6..........................140

4-9 Selected nutrient parameters for sensitivity analysis for W6.............................141









4-10 Selected hydrologic parameters for sensitivity analysis for S4. ........................142

4-11 Selected nutrient parameters for sensitivity analysis for S4.............................143

4-12 Sensitivities of total surface runoff of the selected input parameters throughout
the simulation period for the calibration site W6 ................................................144

4-13 Sensitivities of maximum water table to the selected input parameters depth
throughout the simulation period for the calibration site W6...........................144

4-14 Sensitivities of total P loads to the selected input parameters throughout the
simulation period for the calibration site W6. .............................................. 144

4-15 Sensitivities of total N loads to the selected input parameters throughout the
simulation period for the calibration site W6. .............................................. 144

4-16 Sensitivities of total surface runoff to the selected input parameters throughout
the simulation period for the calibration site S4. ............................................145

4-17 Sensitivities of maximum water table to the selected input parameters depth
throughout the simulation period for the calibration site S4..............................145

4-18 Sensitivities of total P loads to the selected input parameters throughout the
simulation period for the calibration site S4. ..................................................145

4-19 Sensitivities of total N loads to the selected input parameters throughout the
simulation period for the calibration site S4. ........... ......................................145

4-20 Annual water budget for the calibration site W6..........................................146

4-21 Annual water budget for the verification site W7 ..............................................146

4-22 Annual water budget for the calibration site S4......................................... 146

4-23 Annual water budget for the verification site S1 ...........................................146

4-24 Monthly statistics for the calibration site W6.................... ..................................147

4-25 Annual statistics for the calibration site W 6....................................................... 147

4-26 M monthly statistics for the verification site W 7.....................................................147

4-27 Annual statistics for the verification site W 7...................................................... 147

4-28 Monthly statistics for the calibration site S4.................................................... 148

4-29 A annual statistics for the calibration site S4...................................... .................148

4-30 M monthly statistics for the verification site S1 ..................................................... 148









4-31 Annual statistics for the verification site S ................................. ............... 148

5-1 Percent cover of vegetation on summer pastures S4. ........... ...............222

5-2 Parameter values for the different plant species for the vegetation model..........222

5-3 Initial composition percentage and amount for species in each land segment ...223

5-4 Variables calculated from other models. .................................. .................223

5-5 Output variables from the vegetation model ........................... ................223

5-6 Testing scenarios for temperature functions ..............................................223

E-1 New input and output variables required towards the multi-directional spatial
simulation beyond the existing variables in ACRU2000............................. 262
















LIST OF FIGURES


Figure page

1-1 Schematic of the feedback relationships among hydrology, nutrient, and
vegetation dynamics in the coupled model system.................................. .....8

2-1 General structure of the lumped ACRU (v3.00) model..................................21

2-2 Land segment configuration of the ACRU2000 model ................................. 23

3-1 Schematic of an example catchment and its spatial discretization ......................35

3-2 Hydrological processes in the modified ACRU2000 hydrologic model .............36

3-3 Configuration of overland flows between source land segment and adjacent land
segm ents ..................................... ................................. ........... 46

3-4 The relationship between Manning's roughness-coefficient (n) to water depth
and to plant height, both of which change dynamically in the model ...................48

3-5 Structured boundary............................................................... ......49

3-6 Schematic representation of the lateral groundwater flow from the higher head
hs to the lower head hd. (A) and (B) represent the situations without and with
overland flow respectively ......................................................... ............. 52

3-7 Diagram of lateral groundwater flow between two land segments. (A) and (B)
represent the generalized situations without and with overland flow,
resp ectiv ely ...................................................... ................. 53

3-8 A rectangular plane with 20 land segments (arrow indicates water movement
direction and digits assigned in each grid cell indicate the number for each land
segm ent) .............................................................................60

3-9 Comparisons of the simulated surface water depth from the modified
ACRU2000 model on the three experiments between two opposite simulation
sequences. .......................................... ............................ 63

3-10 Comparisons of the simulated surface water depths between the MIKE SHE's
overland flow model and the modified ACRU2000 hydrologic model on the
three experiments with the simulation sequence from LS1 to LS20. ....................64









3-11 Schematics of two-dimensional axisymmetric domain (left) and its three-
dimensional discretization (right). The digits assigned in the left diagram
indicate the number for each land segment ........................................................66

3-12 Comparisons of simulated water depths of the selected land segments between
the modified ACRU2000 model and the MIKE SHE's overland flow model for
the three experim ents. ................................................ ................................ 68

3-13 Comparisons of the simulated water depths of the selected land segments
between the modified ACRU2000 and MIKE SHE's overland flow model for
E x p erim ent 3 ........................... ............ .. .......... ..............................70

3-14 Comparisons of the groundwater table depths at the selected land segments
between and the modified ACRU2000 hydrologic model and MIKE SHE's
groundwater flow model for the three experiments......................................... 74

3-15 Location of Dry Lake D airy #1 site. .......................................... ............... 76

3-16 Dry Lake Dairy # 1 site map, topographic survey and location of well stations
and tracer application compound. Distance scales are in feet and elevation
contours are in feet above m ean sea level................................... ............... 77

3-17 Comparisons of the continuous simulations of runoff from the modified
ACRU2000 and FHANTM against the observed data. ...........................84

3-18 Comparisons of the cumulative simulated runoff from the modified ACRU2000
and FHANTM with the cumulative observed data................................. 84

3-19 Comparisons of the continuous simulation of groundwater table depths from the
modified ACRU2000 and FHANTM against the observed data........................... 85

3-20 Scatterplots of observed vs. simulated surface runoff depths predicted by the
modified ACRU2000 and FHANTM ...................................... ............... 86

3-21 Scatterplots of observed vs. simulated groundwater table depths predicted by the
modified ACRU2000 and FHANTM ...................................... ............... 86

4-1 Schematic representation of the GLEAMS nitrogen cycle................................90

4-2 Schematic representation of the GLEAMS phosphorus cycle.............................98

4-3 Comparisons of the solute mass in the 3rd and 4th soil layers of selected land
segments between the modified ACRU2000 and PMPATH ............................110

4-4 Location of MacArthur Agro-ecology Research Center.............................. 11

4-5 General layout of the project field ............................................ 113









4-6 Relative sensitivities of the total surface runoff of 6-year simulation period over
the selected input parameters for the pasture sites W6 and S4..........................149

4-7 Relative sensitivities of the maximum water table depth of 6-year simulation
period over the selected input parameters for the pasture sites W6 and S4.........150

4-8 Relative sensitivities of the total P load of 6-year simulation period over the
selected input parameters and variables for the pasture sites W6 and S4............151

4-9 Relative sensitivities of the total N load of 6-year simulation period over the
selected input parameters for the pasture sites W6 and S4...............................152

4-10 Continuous simulation of groundwater table depth from September 2000
through December 2003 for the calibration site W6.................. .. ..................153

4-11 Continuous simulation of groundwater table depth from September 2000
through December 2003 for the verification site W7. .......................................153

4-12 Continuous simulation of groundwater table depth from September 2000
through December 2003 for the calibration site S4. .......................................154

4-13 Continuous simulation of groundwater table depth from September 2000
through December 2003 for the verification site S .........................................154

4-14 Continuous simulation of surface runoff from July 1998 through December
2003 for the calibration site W 6.................................... .................................... 155

4-15 Continuous simulation of surface runoff from July 1998 through December
2003 for the verification site W 7. ............................................. ............... 155

4-16 Continuous simulation of surface runoff from July 1998 through December
2003 for the calibration site S4. .........................................................................156

4-17 Continuous simulation of surface runoff from July 1998 through December
2003 for the verification site S ...................................... ........................ 156

4-18 Continuous simulation of P loads from July 1998 through December 2003 for
the calibration site W 6. ................................................ ............ ........ 157

4-19 Continuous simulation of P loads from July 1998 through December 2003 for
the verification site W 7 ................................................................................. ..... 157

4-20 Continuous simulation of P loads from July 1998 through December 2003 for
the calibration site S4................... .. ...... .................. .. ........... ........ .. 158

4-21 Continuous simulation of P loads from July 1998 through December 2003 for
the verification site S ................................... .....................................158









4-22 Continuous simulation of N loads from July 1998 through December 2003 for
the calibration site W 6. ............................................................. ..................... 159

4-23 Continuous simulation of N loads from July 1998 through December 2003 for
the v erification site W 7 .......................................................................... .... ... 159

4-24 Continuous simulation of N loads from July 1998 through December 2003 for
the calibration site S4 ................... .. ...... .................. .. ........... .... .......... 160

4-25 Continuous simulation of N loads from July 1998 through December 2003 for
the verification site S ................................... .....................................160

4-26 Continuous simulation of surface runoff P concentrations from July 1998
through December 2003 for the calibration site W6.................. .. ..................161

4-27 Continuous simulation of surface runoff P concentrations from July 1998
through December 2003 for the verification site W7. .......................................161

4-28 Continuous simulation of surface runoffN concentrations from July 1998
through December 2003 for the calibration site W6.................. .. ..................162

4-29 Continuous simulation of surface runoffN concentrations from July 1998
through December 2003 for the verification site W7. .......................................162

4-30 Continuous simulation of surface runoff P concentrations from July 1998
through December 2003 for the calibration site S4. .......................................163

4-31 Continuous simulation of surface runoff P concentrations from July 1998
through December 2003 for the calibration site S ............................................163

4-32 Continuous simulation of surface runoffN concentrations from July 1998
through December 2003 for the verification site S4.................. ... .................164

4-33 Continuous simulation of surface runoffN concentrations from July 1998
through December 2003 for the verification site S 1.........................................164

4-34 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the
calibration site W 6. ........................ ...... .... ......... ............... 165

4-35 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the
verification site W 7 .................. ........................... .. .. ..... .. ........ .... 166

4-36 Linear plots of monthly and annual water table depth from 1998 to 2003 for the
calibration site W 6. ........................ ...... .... ......... ............... 167

4-37 Linear plots of monthly and annual water table depth from 1998 to 2003 for the
verification site W 7 .................. ........................... .. .. ..... .. ........ .... 168









4-38 Linear plots of monthly and annual P load from 1998 to 2003 for the calibration
site W 6 ........................................................................ 16 9

4-39 Linear plots of monthly and annual P load from 1998 to 2003 for the
verification site W 7 .................. ............................ .... .. .. .. ........ .... 170

4-40 Linear plots of monthly and annual N load from 1998 to 2003 for the calibration
site W 6 ........................................................................ 17 1

4-41 Linear plots of monthly and annual N load from 1998 to 2003 for the
verification site W 7 .................. ............................ .... .. .. .. ........ .... 172

4-42 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the
calib ration site S 4 ............................................................... 17 3

4-43 Linear plots of monthly and annual surface runoff from 1998 to 2003 for the
verification site S 1 .... ....... ................. ............ ...... 174

4-44 Linear plots of monthly and annual water table depth from 1998 to 2003 for the
calib ration site S 4 ............................................................... 17 5

4-45 Linear plots of monthly and annual water table depth from 1998 to 2003 for the
verification site S 1 .... ....... ................. ............ ...... 176

4-46 Linear plots of monthly and annual P load from 1998 to 2003 for the calibration
site S 4 ........................................................................ 1 7 7

4-47 Linear plots of monthly and annual P load from 1998 to 2003 for the
verification site S 1 .... ....... ................. ............ ...... 178

4-48 Linear plots of monthly and annual N load from 1998 to 2003 for the calibration
site S 4 ........................................................................ 1 7 9

4-49 Linear plots of monthly and annual N load from 1998 to 2003 for the
verification site S 1 .... ....... ................. ............ ..... 180

4-50 Duration curves of daily surface runoff and water table depth for the calibration
site W 6 ........................................................................ 18 1

4-51 Duration curves of daily P and N loads for the calibration site W6 .................. 182

4-52 Duration curves of daily surface runoff and water table depth for the verification
site W 7 ........................................................................ 18 3

4-53 Duration curves of daily P and N loads for the verification site W7.................84

4-54 Duration curves of daily surface runoff and water table depth from 1998 to 2003
for the calibration site S4. ...... ........................... .....................................185









4-55 Duration curves of daily N and P loads from 1998 to 2003 for the calibration
site S 4 ........................................................................ 1 8 6

4-56 Duration curves of daily surface runoff and water table depth from 1998 to 2003
for the verification site S1 ........................................... ............................. 187

4-57 Duration curves of daily N and P loads from 1998 to 2003 for the verification
site S 1 ........................................................................... 1 8 8

5-1 Diagram for daily plant growth in relation to weather and water and nutrient
availabilities (DM = dry matter; N = nitrogen; P = phosphorus; and SLA =
specific leaf area). ................ ............................ 193

5-2 Diagram of temperature function for species i. ............................................ 195

5-3 An example relationship between plant biomass nitrogen concentration and
grow th ratio ......................................................................... 202

5-4 An example relationship between plant biomass phosphous concentration and
grow th ratio ......................................................................... 204

5-5 Aerial photo showing the layout of the improved summer pasture site S4 and
location of associated instrumentation. S1 to S6 indicate the individual summer
pasture sites and LS1 to LS12 indicate the land segments divided for the site S4.
The dotted lines were made to show the boundary between land segments........210

5-6 Continuous simulation of surface runoff throughout the simulation period from
1998 to 2003 ........................................................................ .................. ............ 224

5-7 Continuous simulation of water table depths at land segment 9 throughout the
simulation period from 1998 to 2003......... ............ ........... .............. 224

5-8 Continuous simulation of P loads throughout the simulation period from 1998 to
2 0 0 3 ................................................................................................. .. 2 2 5

5-9 Continuous simulation of N loads throughout the simulation period from 1998
to 2003 ........................................... .......................................... .....................225

5-10 Continuous simulation of P concentrations throughout the simulation period
from 1998 to 2003 .................. .......................... .. ...... ................. 226

5-11 Continuous simulation of N concentrations throughout the simulation period
from 1998 to 2003 .................. ............................ .. ...... .. ........ .... 226

5-12 Comparison of predicted potential aboveground biomass for species in land
segment 11 with different temperature functions. .............................................227


xviii









5-13 Comparison of temperature factors among the three selected species in land
segm ent 11 .........................................................................228

5-14 Species distribution (percent of total aboveground biomass at the end of each
year) over land segments for bahia, floralta and panicum throughout the
simulation period from 1998 to 2003............... ........... .. ..... .. ............. 230

5-15 Comparisons of continuous simulation of aboveground biomass for all species
in the selected land segments 8, 11, and 12 before water retention BMP ............231

5-16 Comparisons of continuous simulation of aboveground biomass for all species
in the selected land segments 8, 11, and 12 after water retention BMP ..............232

5-17 Growth limiting factors for bahiagrass in land segment 8 before water retention
B M P w as applied. ........................... ...................... ... ........ .. .... ...... ...... 233

G-1 Single land segment type neighbor receiving overland flow............................266

G-2 Single river type neighbor receiving overland flow ........................................267

G-3 Configuration of multiple directional overland flows from source land segment
to adjacent land segm ents. ........................................... ............................ 269















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

COUPLED SIMULATION MODELING OF FLATWOODS HYDROLOGY,
NUTRIENT AND VEGETATION DYNAMICS

By

Lei Yang

May 2006

Chair: Wendy D. Graham
Cochair: Kenneth L. Campbell
Major Department: Agricultural and Biological Engineering

Lake Okeechobee, located at the center of the Kissimmee-Okeechobee-Everglades

aquatic ecosystem in south Florida, is experiencing water quality degradation. Non-point

agricultural runoff from dairies and cow-calf operations in the northern watershed of the

lake is considered to be the primary source of excess phosphorus (P) loading discharged

into the lake. In order to evaluate alternative land management practices that result in

reduced P loading from the watershed to the lake, a coupled modeling system integrating

hydrology, nutrient and vegetation dynamics simulation was developed.

The coupled modeling system was developed within the Java-based, object-

oriented framework of the ACRU2000 modeling system by adding new hydrologic and

nutrient components and a vegetation model to enable multi-directional spatial simulation

of hydrological, chemical, and biological processes simultaneously in a daily time step.

The coupled model was tested for accuracy by comparing performance with well-

accepted models including MIKE SHE and MODFLOW. Results indicate that the









coupled model is capable of simulating, with reasonable accuracy, hydrological and

solute transport processes for the hypothetical scenarios. Additionally, the model was

tested in the Kissimmee River Basin and Lake Okeechobee Basin by comparing with the

FHANTM model and against measured data. These applications demonstrated that the

coupled model is statistically close to the performance of FHANTM with respect to

hydrologic response in the Kissimmee River Basin, but much better than FHANTM with

regard to hydrologic and nutrient responses in the Lake Okeechobee Basin. From the

testing, it was concluded that the model is able to continuously simulate the surface

runoff and groundwater tables with adequate accuracy. However the model's capacity to

simulate nutrient loading needs further testing after sufficient reliable nutrient data

becomes available. The vegetation model, coupled with the hydrologic and nutrient

models, was tested for a hypothetical scenario based on the conditions in the Lake

Okeechobee Basin. The test results show that the temporal and spatial vegetation

composition pattern can be an indicator of the ecohydrological impacts of alternative land

management practices. However, for actual application of this model, further testing is

required when more vegetation data are available.

Recommended future research includes further development of the coupled model

to enable a user-friendly pre- and post-processing graphical interface, an option for sub-

daily time steps, beef-cattle roaming simulation, and plant competition. Further testing of

the coupled model should be conducted at larger watershed scales and for the nutrient and

vegetation simulations when additional data are available.














CHAPTER 1
INTRODUCTION

Study Background

Lake Okeechobee is a large, multi-functional lake located at the center of the

Kissimmee-Okeechobee-Everglades aquatic ecosystem in south Florida. The lake

provides regional flood protection, water supply for agricultural, urban and natural areas,

and is a critical habitat for fish, birds and other wildlife. Water quality in the lake has

declined over recent decades due to urban development, channelization of the Kissimmee

River and agricultural operations. The 1997 Lake Okeechobee Surface Water

Improvement and Management Plan (South Florida Water Management District

[SFWMD], 1997a) found that excess phosphorus (P) loading is one of the most serious

problems the lake is facing.

The Lake Okeechobee watershed, with an area of 12,000 km2 extending from

Orlando to the Everglades, lies predominately in the southern Florida flatwoods

physiographic region characterized by flat, poorly drained, high-water table, and fine

sandy soils, which consist of Spodosols, Entisols, and Histosols (United States

Department of Agriculture [USDA], 1990). The majority of the soils in the northern

watersheds of the lake are Sposodols, with 8-20 cm thick surface horizons underlain by

spodic horizons at depth of 0.5 m to greater than 2 m (USDA, 1990). These soils, with

greater than 90% sand, are characterized by high infiltration rates and poor internal

drainage due to low permeability of the spodic horizon. When rainfall occurs, the soils

often become saturated in a short time. The water table is commonly within 1 m of the









ground surface during the wet season and may recede to 2 m depth during the dry season

(Knisel et al., 1985). With highly permeable surface soils there is little surface runoff

until the soil pore space is filled with water and the water table reaches the surface

(Heatwole et al., 1987); slow downward or lateral movement of water and solutes occurs

with water table recession.

The land use in the Okeechobee watershed is primarily beef cattle, dairy, and citrus.

Since 1930, ranching has intensified from native pastures to improved pastures with high

quality grasses and legumes, drainage and fertilization. During the 1950s, the dairy

industry first moved to Okeechobee County, and by the mid-1980s, there were 49

milking barns and 50,000 milk cows in the lower Kissimmee River (LKR) and Taylor

Creek/Nubbin Slough (TCNS) regions (Flaig and Reddy, 1995). The primary feed for

dairy cows, including high P containing materials, was imported into the watershed.

Animal waste management was almost non-existent until the 1970s (Flaig and Reddy,

1995). Historically, the majority of P load to the lake was derived from the LKR and the

TCNS basins with 13% contributed by the LKR basin and 22% by the TCNS basin. With

the implementation of improved management practices during recent decades, the P load

from the TCNS basin has decreased by 17% (Gunsalus et al., 1992), and the LKR basin

now provides the greatest P load to the lake. Previous studies have indicated that non-

point agricultural runoff in the northern watershed of the lake is considered to be the

primary source of excess P being discharged into Lake Okeechobee, which typically

exceeds the recommended total maximum daily loading (TMDL = 140 metric ton/year).

P in agricultural runoff mainly originates from one or more of four sources: fertilizers,









animal manures, mineralization of organic materials, and/or atmospheric deposition. The

first three sources can be controlled by agricultural best management practices.

In order to protect the water quality of Lake Okeechobee and reach environmental

restoration goals, a variety of best management practices (BMPs) have been implemented

in the Lake Okeechobee watershed. BMPs are on-farm activities designed to reduce

nutrient losses in drainage waters to an environmentally acceptable level, while

simultaneously maintaining an economically viable farming operation (Bottcher et al.,

1995). To reduce P loading one must either reduce P concentration or water volume.

According to Bottcher et al. (1995), there are three ways to reduce P concentrations in the

runoff water from agriculture: 1) reduce the amount of P on the farm by minimizing P

inputs to the farm and maximizing non-runoff P output from the farm; 2) reduce the

hydrologic mobility of the P that is on the farm by limiting water contact and/or reducing

the solubility or erodibility of phosphatic materials; 3) edge of field/farm pre-discharge

treatment using uptake, adsorption, deposition, or precipitation technologies, such as

wetland and/or chemical additives. There are two methods used to reduce water runoff

volume: 1) increase the evapotranspiration from the farm; 2) decrease off-farm or

groundwater irrigation water inputs to the farm by improved irrigation efficiency or by

using storage runoff as a substitute irrigation supply. From numerous studies in the

Okeechobee basin as well as generally accepted practices from other parts of the country,

Bottcher et al. (1995) summarized the BMPs appropriate for the Lake Okeechobee basin

as follows: 1) fertility BMPs including calibrated soil testing (CST), banding of fertilizer,

prevention of misplaced fertilizer, and split application; 2) animal manure BMPs

including dairy high intensity area (HIA) drainage control, collection and distribution of









barn manure, watering, feed and shade facilities placement, fencing animals from ditches

and streams, grazing management, selecting high P uptake crops from manure application

areas, and composting; 3) general BMPs including crop management, irrigation and

drainage management, maximum flow distance for P control, flow-way buffer strips,

limit drainage of organic and/or wetland soils, and alternative land use; 4) edge-of-

field/farm treatment including runoff retention/detention system, use of wetlands, and

chemical treatment.

Only a few of the above listed BMPs have been field tested, and even those were

tested for only a limited set of conditions (Bottcher et al., 1995). Flaig and Reddy (1995)

indicated that implementation of BMPs has not been sufficient to meet P load reduction

goals, and additional P control practices to further reduce P are needed. Recent BMPs

efforts to reduce nutrient loading from the Lake Okeechobee watershed have focused on

restoration of wetlands for their particular capacities to reduce nutrient loadings, thereby

reducing eutrophication in adjacent water bodies.

Wetlands are an important component of the Lake Okeechobee watershed.

However, many of these wetlands have disappeared or have been degraded due to

hydrologic alteration, urbanization and agricultural practices. Seasonal and year-round

isolated and connected wetlands used to occupy 25% of the watershed area (McCaffery et

al., 1976). Now many isolated wetlands are connected by shallow drainage ditches and

have been converted into pastures. Currently, wetlands represent about 15% of the land

area in the Lake Okeechobee watershed (Flaig and Havens, 1995). Wetland loss

inevitably leads to loss of biological, environmental quality and socio-functions such as









flood storage, groundwater recharge, sediment trapping, retention and removal of

nutrients and pollutants, and wildlife and recreational habitat (Davis and Froend, 1999).

Among all wetland functions, transport and transformation processes including

sedimentation, sediment adsorption, nutrient uptake, microbial assimilation and

transformations, and denitrification may be the most important mechanisms as they are

responsible for nutrient removal or retention. Fisher and Acreman (2004) investigated 57

wetlands from around the world and indicated that the majority of wetlands reduced

nutrient loading and there was little difference in the proportion of wetlands that reduced

nitrogen (N) to those that reduced P loading. However, they also pointed out that some

wetlands increased nutrient loading by increasing the loading of soluble N and P species,

thus potentially driving aquatic eutrophication. Busnardo et al. (1992) researched the

effect of hydroperiod on nutrient removal efficiency in replicate wetland mesocosms and

concluded that alternate draining and flooding of sediments (pulsed discharge) increased

nutrient removal efficiency compared to the continuous-flow "control". Uusi-Kamppa et

al. (1997) indicated that in many wetlands the retention of soluble P is much less efficient

than that of particulate P. Also, waterlogged sediments are known to release P into

overlying waters (Mortimer, 1941), where it is more likely to be exported from the

watershed.

Denitrification, which occurs under anaerobic conditions to release N into the

atmosphere, is believed by many to be the major mechanism of N loss in wetlands.

Denitrification rates can be limited by carbon availability and, in this way, vegetation can

influence denitrification rates indirectly (Broadbent and Clark, 1965). Vegetation may

also influence nitrification and denitrification by influencing the oxygen concentration of









the wetland substrate within the rhizosphere (Armstrong, 1964) or by providing bacteria

which can fix N in root nodules. There is evidence that N removal efficiency is not

affected by the length of time the wetland has received N pollution while, in contrast, the

ability of a wetland to remove P is known to decline with time (Nichols, 1983;

Richardson, 1985).

Wetlands are not stand-alone ecosystems. They impact hydrology, water quality

and vegetation dynamics throughout the watershed. Wetlands are characterized by the

periodic excess of water inflow over outflow that provides a saturated substrate. The

effect of this characteristic is substantial water storage within wetlands and the

development of a readily identifiable wetlands flora and fauna which are adapted to

periodic anoxic conditions (Bradley and Gilvear, 2000). P loading can alter plant

communities through increased plant productivity, tissue P storage, soil P enrichment,

and shifts in plant species composition (Davis, 1991; Urban et al., 1993; Chiang et al.,

2000). Altered hydrologic regime, caused by water management infrastructure and

operations, can also affect the pattern of vegetation communities. Newman et al. (1996)

showed that P concentration and water depth are two important driving forces for cattail

invasion into the Everglades. Urban et al. (1993) also indicated that a combination of

prolonged hydroperiod and P loading stimulates cattail reproduction and encroachment

and thus results in the degradation of vegetative habitats and other ecological

characteristics in wetlands. Conversely, the physical and physiological characteristics of

vegetation influence hydrological response and nutrient cycling. The flora and fauna

impact the hydrology of many wetlands in that their incomplete decomposition leads to

the progressive development of an organic substrate that itself influences the pattern and









direction of water flow through wetlands and the quantity of water storage (Bradley and

Gilvear, 2000). Gurnell et al. (2000) indicated that interception, evapotranspiration, and

infiltration processes are particularly heavily influenced by the characteristics and

dynamics of the vegetation cover. Many ecologists have recognized that changing

vegetation pattern/structure causes feedback that can alter rates of hydrologic processes

and nutrient cycling, which, in turn, can cause additional changes in vegetation structure

(Lauenroth et al., 1993).

From the above discussion, it is clear that it is critical to understand the role of both

BMPs and wetland functions in nutrient removal, retention and storage in order to reduce

P loading into Lake Okeechobee. However, it is impractical to test all BMPs for their

effectiveness through field experiments. The use of computer models is therefore

beneficial because simulation results can not only predict how well a proposed BMP or

combined BMPs will reduce P loads, but can also quantitatively evaluate specific

hydrologic and biogeochemical processes associated with management activities for

BMP design and optimization.

Ecohydrologic modeling that simulates hydrological, biochemical and ecological

processes and their interrelations in soil and water bodies has captured the attention of

hydrologists and other scientists in recent years. Modeling provides a useful tool for

gaining insight into ecohydrological processes and evaluating management practices if

model predictions are accurate. Many models have been developed but they are typically

linked to the regions where they were developed and tested for a specific purpose and are

often limited to specific spatial scales. The unique flatwoods hydrology in the Lake

Okeechobee watershed requires the design of a model that can simulate integrated









multidimensional surface and subsurface water, nutrient, and vegetation dynamics at

multiple temporal and spatial scales.

Overview of the Coupled Modeling System

For this study, a coupled modeling system that enables the distributed simulation of

hydrology, water quality and vegetation dynamics was developed for Okeechobee

flatwoods watersheds that incorporate uplands, wetlands, and transition zones located in

between these landscapes. The proposed coupled modeling system and the feedback

relationships among its submodels are depicted in Figure 1-1. This model system

contains a hydrologic submodel as a critical component which simulates precipitation,

interception, evapotranspiration, infiltration, water movement within unsaturated soil

zones, upward flux, deep seepage, overland flow and groundwater flow. It also includes

a submodel for nutrient cycling to simulate the transport and transformation processes for

nitrogen and phosphorus. Another necessary part of the model is a vegetation submodel

that simulates plant growth dynamics under the combined influence of hydrology,

nutrient availability, and land management practices including beef-cattle stocking rates,

fire, fertilization, etc.

Transport
Hydrologic T Nutrient --
~ Model Model

Water stress/logging
Decomposition

Roughness Uptake
Interception Vegetation Nutrient stress
Transpiration Model (N and P)

Figure 1-1. Schematic of the feedback relationships among hydrology, nutrient, and
vegetation dynamics in the coupled model system.









The three submodels are coupled in that the interactions and feedbacks between

processes of these submodels are simulated together. As outlined in Figure 1-1, plant

communities respond to available nutrients and water, which are drivers for plant growth;

dynamics of live and standing dead vegetation alter surface water runoff through changes

in canopy structure and thus surface roughness. Hydrology in the model responds

directly to the vegetation via linkages such as Manning's roughness coefficient and

transpiration losses. Water losses by plants vary with changes in biomass (leaf area

index) and physical canopy structure. Availability of water in surface, unsaturated and

saturated zone storage is one control on plant growth and mortality. Both surface and

subsurface water transport dissolved nutrients, and the soil water conditions affect the

biogeochemical processes, while nutrient availability and uptake kinetics can control

plant growth. Dead organic matter in different forms is a major source of nutrients.

The coupled model was developed within the Java-based, object-oriented

framework of the ACRU2000 model (Campbell et al., 2001; Clark et al., 2001; Kiker and

Clark, 2001), an agrohydrological modeling system originally created by Schulze (1989,

1995) in South Africa. The coupled model was developed by adding hydrological

components capable of multi-directional spatial simulation of overland flow and lateral

groundwater flow, nutrient components capable of multi-directional spatial simulation of

conservative solute, nitrogen and phosphorus transport, as well as a new vegetation

dynamics simulation model.

Study Objectives

The overall purpose of this study was to develop a coupled model system capable

of simulating hydrology, nutrient and vegetation dynamics simultaneously for south









Florida flatwoods watersheds that incorporate wetlands, uplands and transition zones.

Specific objectives of this research can be summarized as follows:

* Modify the existing ACRU2000 modeling system to enable the multi-directional
spatial simulation of flatwoods hydrology, nutrient and vegetation dynamics.

* Test the accuracy of the modified hydrologic and nutrient models' performance by
comparing with the existing models MIKE SHE and MODFLOW.

* Validate the modified hydrological model by comparing with FHANTM and
against the measured data in Dry Lake Dairy #1, Kissimmee River Basin, Florida.

* Calibrate, validate and evaluate the coupled hydrologic and nutrient model using
the measured data from Buck Island Ranch, Lake Okeechobee Basin, Florida.

* Test the coupled hydrologic, nutrient, and vegetation model using scenarios based
on conditions at Buck Island Ranch, Lake Okeechobee Basin, Florida.

* Investigate the interactions among wetlands hydrology and nutrient dynamics
imposed by alternative land management practices and anthropogenic activities in
south Florida flatwoods watersheds.

This dissertation is organized into 6 chapters. Chapter 1 briefly introduces the

background and objectives of this study, as well as giving an overview of the coupled

modeling system; Chapter 2 reviews previous modeling efforts in hydrology, nutrient

and vegetation dynamic simulation, and discusses model testing procedures including

model calibration, verification, evaluation and sensitivity analysis; Chapter 3 focuses on

the development of the multi-directional spatial simulation of the hydrologic model,

including a description of model components, algorithms, assumptions, and model testing

and validation; Chapter 4 describes the development of the multi-directional spatial

nutrient (nitrogen, phosphorus, and conservative solute) transport and transformation

model, including model components, algorithms, assumptions and modeling testing and

validation; Chapter 5 introduces the vegetation dynamics simulation model, including






11


model algorithms and testing. Finally, Chapter 6 gives a summary of the results found in

this study, conclusions, and recommendations for future work.














CHAPTER 2
LITERATURE REVIEW

Overview of Previous Modeling Efforts

Surface and subsurface water movement serves as a major nutrient transport

mechanism, both delivering essential nutrients to the biota and moving excess nutrients to

receiving water bodies. The importance of hydrologic transport has been long recognized

and considerable effort has been put into creating adequate models for various landscapes

(Beven and Kirkby, 1979; Beasley and Huggins, 1980). The complexity of a specific

watershed simulation model depends on the temporal and spatial resolution, and on the

extent to which important hydrological and biochemical processes are considered

(Krysanova et al., 1998). Over the past several decades many models, ranging from

lumped conceptual models to semi-distributed models to fully distributed physically-

based models, have been developed.

Early examples of lumped hydrologic models are the Stanford Watershed Model

(Crawford and Linsley, 1966), the SSARR (Streamflow Synthesis and Reservoir

Regulation) model (Rockwood et al., 1972), the Sacramento model (Burnash et al.,

1973), the tank model (Sugawara et al., 1976), HEC-1 (Hydrologic Engineering Center,

1981) and the HYMO (Williams and Hann, 1983). In these models, both differential

equations based on simplified hydraulic laws and empirical algebraic equations were

used for different processes. More recent conceptual models have incorporated soil

moisture replenishment, depletion and redistribution for the dynamic variation in areas

contributing to direct runoff (Arnold and Fohrer, 2005).









Progress in developing coupled hydrological/water quality models is more evident

at the field scale or in small homogeneous watersheds than at large watershed and

regional scales. Starting from the early 1970s, non-point source models have been

developed in the USA in response to the Clean Water Act. CREAMS (Chemicals,

Runoff, and Erosion from Agricultural Management Systems) (Knisel, 1980) was

developed to simulate the impact of land management on water, sediment, nutrients and

pesticides leaving the edge of a field. Subsequently, several field-scale models evolved

from the original CREAMS including GLEAMS (Groundwater Loading Effects of

Agricultural Management Systems) (Leonard et al., 1987) to simulate groundwater

pesticide loadings, EPIC (Erosion-Productivity Impact Calculator) (Williams et al., 1984)

to simulate the impact of erosion on crop productivity, and OPUS (Smith, 1992) to

estimate the effects of management practices on non-point source pollution.

Spatially-distributed models in larger watersheds represent a more complicated

problem (Krysanova et al., 1998). Semi-distributed physically-based hydrologic models

have the advantage of a simple model structure and fewer model parameters together

with a realistic representation of the watershed hydrologic process. SWAT (Soil and

Water Assessment Tool) (Arnold et al., 1993) is a continuous-time distributed simulation

watershed model to predict the effects of alternative management decisions on water,

sediment and agricultural chemical yields with reasonable accuracy in watersheds and

large river basins. SWAT was originally developed from CREAMS to a basin-scale

model SWRRB (Arnold et al., 1990), and then combined with the ROTO model (Arnold,

1990) to form the more comprehensive model SWAT. The latest version, SWAT2000,

has several significant enhancements that include: bacteria transport routines; urban









routines; Green and Ampt infiltration equation; improved weather generator; ability to

read in daily solar radiation, relative humidity, wind speed and potential ET; Muskingum

channel routing; and modified dormancy calculations for tropical areas (Arnold and

Fohrer, 2005). SWIM (Soil and Water Integrated Model) (Krysanova et al., 1998, 2005),

a hybrid of SWAT and MATSALU (Krysanova et al., 1989), is a continuous-time

spatially semi-distributed model, which integrates hydrological processes, vegetation

growth (agricultural crops and natural vegetation), nutrient cycling (nitrogen and

phosphorus) and sediment transport at the river basin scale. AGNPS (Agricultural Non-

Point Source pollution model) (Young et al., 1989) was developed to examine water

quality as it is affected by soil erosion from agriculture and urban areas during single

precipitation events at watershed scale. TOPMODEL (a TOPography based hydrological

MODEL) (Beven and Kirkby, 1979) is a variable contributing area conceptual model, in

which the major factors affecting runoff generation are the catchment topography and the

soil transmissivity that diminishes with depth. ANSWERS (Areal Nonpoint Source

Watershed Environment Response Simulation) (Beasley and Huggins, 1980) was

developed to evaluate the effects of BMPs on surface runoff and sediment loss from

agricultural watersheds. The current version of the model, ANSWERS-2000, is a

continuous simulation model that was developed in the mid 1990s (Bouraoui and Dillaha,

1996) with the revised nutrient submodels, improved infiltration (Green and Ampt), and

new soil moisture and plant growth components to permit long-term continuous

simulation.

Another class of models are based on differential equations for conservation of

mass, energy and momentum, and are called fully physically-based distributed models.









The spatial distribution of catchment parameters in such models is achieved by

representing the basin on a grid network. Examples of physically-based distributed

models include MIKE SHE (Refsgaard and Storm, 1995) and MODFLOW (McDonald

and Harbaugh, 1988). MIKE SHE is a physically-based, distributed, integrated

hydrological and water quality modeling system. It simulates the hydrological cycle

including evapotranspiration, overland flow, channel flow, soil water and ground water

movement. Related water quality modules include: 1) advection-dispersion, 2) particle

tracking, 3) sorption and degradation, 4) geochemistry, 5) biodegradation, and 6) crop

yield and nitrogen consumption. This modeling system can be used to predict pollutant

loading and transport, pesticide leaching, and outcomes of alternate BMPs on watersheds

and their underlying aquifers. MODFLOW is a three-dimensional finite difference

groundwater model of the U. S. Geological Survey, which can simulate various stresses

to the system such as wells, rivers, drains, head-dependent boundaries, recharge and

evapotranspiration. MODFLOW can simulate homogeneous/heterogeneous systems,

isotropic/anisotropic media, and steady-state/transient flow.

Model development is often linked to where the model was developed and tested.

To simulate the unique flatwoods ecohydrological issues in south Florida watersheds,

several models have been developed. These models can be approximately sorted into 3

types according to their functionalities, including (1) hydrologic models such as

DRAINMOD (Skaggs, 1980), the weighted implicit finite volume model (Lal et al.,

1998), SFWMM (South Florida Water Management Model) (SFWMD, 2005) and NSM

(Natural System Model) (SFWMD, 1998); (2) water quality models such as CREAMS-

WT (Chemicals, Runoff, and Erosion from Agricultural Management Systems-Water









Table) (Heatwole, 1986), EAAMOD-FIELD (Everglades Agricultural Area MODel-

Field) (Bottcher et al., 1998) and FHANTM (Field Hydrologic And Nutrient Transport

Model) (Tremwel, 1992; Fraisse and Campbell, 1996, 1997); and (3) ecohydrological

models such as ELM (Everglades Landscape Model) (Fitz et al., 2002) and

FLATWOODS (Sun et al., 1998).

Some of these models are field-scale, lumped conceptual models, incapable of

spatially distributed simulation of ecohydrologic variations. For example, DRAINMOD

(Skaggs, 1980) was designed to evaluate the effects of drainage on the water table depth

below agricultural fields in the coastal plain of North Carolina where it performed well

for the silty-clay soils, but required modifications to improve its ability to predict the

outflows from a field with sandy soils (Rogers, 1985). CREAMS-WT (Heatwole, 1986)

was modified from CREAMS (Knisel, 1980) to better represent the low phosphorus

buffering capacity of sandy soils and the hydrology of flat, sandy, high-water-table

flatwoods watersheds. FHANTM (Tremwel, 1992) is based on the DRAINMOD model

but was modified to include simulation of phosphorus movement and routing of overland

flow. The model was further modified and released as FHANTM 2.0 (Fraisse and

Campbell, 1996, 1997) for generalized use in modeling cow/calf operations. This new

version incorporates the nutrient component of GLEAMS (Leonard et al., 1987).

EAAMOD-FIELD (Bottcher et al., 1998) was developed to assess the effects of different

agricultural practices on phosphorus losses from fields in the Everglades Agricultural

Area. Zhang et al. (1995) compared CREAMS-WT and FHANTM and concluded that

there is little difference in performance of these models in estimating runoff, phosphorus

concentration and loads. Zhang et al. (1999) further compared FHANTM, FHANTM









2.0, and EAAMOD-FIELD and concluded EAAMOD-FIELD was the best model to use

for the regulatory program because it has the capability of simulating land use changes in

the middle of a simulation period and the most potential to be enhanced so that it can be

used in the Lake Okeechobee WOD regulatory program. Hendricks (2003) compared

FHANTM and EAAMOD, and concluded that EAAMOD predicted water table levels

more accurately than FHANTM, and EAAMOD was able to remove the depressional

storage water more quickly than FHANTM, allowing EAAMOD's water table level to

respond more quickly when rainfall events subside.

Some of the other models, although they are physically distributed, generally are

designed for regional, long-term applications to watersheds at a scale of thousands of

square kilometers. Although scalable, performance constraints may impose practical

limits on the time and space scales. These models are not intended for local-scale

decision-making support because the details of local-scale watersheds may not be

sufficiently simulated. For instance, SFWMM (SFWMD, 2005) is a regional-scale

computer model that simulates the hydrology and the management of the water resources

system from Lake Okeechobee to Florida Bay using a mesh of 2 mile x 2 mile cells.

NSM (SFWMD, 2005) uses the same climatic inputs, time step, calibrated model

parameters and algorithm as the SFWMM, but it differs from the SFWMM in that it does

not simulate the influence of any man-made features and uses estimates of pre-subsidence

topography and historical vegetation cover. ELM (Fitz et al., 2002) is a regional-scale,

integrated ecological assessment tool designed to understand and predict the landscape

response to different water management scenarios in south Florida, USA. In simulating

changes to habitat distributions, the ELM dynamically integrates hydrology, water









quality, soils, periphyton, and vegetation in the Everglades region. Due to the

computational complexity in the hydrologic modules, the model generally constrains the

spatial resolution to 1.0 km2 grid cells.

Other models either require relatively small time steps to maintain stability and

accuracy, such as the weighted implicit finite-volume model (Lal et al., 1998), or are

specially designed for certain application such as FLATWOODS (Sun et al., 1998),

which was developed for the cypress wetland-pine upland landscape using a distributed,

physically based approach especially useful in the study of forest hydrology.

Vegetation models with different complexities range from simple empirical

formulae to complex physiologically-based models. These models differ as a result of

the objectives of model development, and hence the required scale and degree of detail

and comprehensiveness (Van Ittersum et al., 2003). Over the past decades, a

considerable number of vegetation models have been developed to target many different

aspects of the terrestrial carbon cycle. At the core of most of these models is a net

primary productivity submodel, although many of the mechanisms behind terrestrial

productivity are still not properly understood. As a consequence of this, as well as the

availability of data and computational capacity at the time of development, models

developed to calculate net primary productivity are quite diverse in their approaches

(Adams et al. 2003). At one end of the spectrum is the simple, empirically derived,

correlation of net primary productivity with air temperature and precipitation used, for

example, in the Miami model (Leith, 1975a). At the other end is the detailed simulation

of process-oriented, biochemistry used in the Hurley Pasture Model (Thomley and









Cannell, 1997), DSSAT (Jones et al., 2003) and the Wageningen crop models (Van

Ittersum et al., 2003) .

Several process-based grassland models have been reported such as ELM (Innis,

1978), PAPRAN (Seligman and Van Keulen, 1981), CENTURY (Parton et al., 1987,

1996), ERHYM-II (White, 1987), GEM (Hunt et al., 1991), CCGRASS (Verberne, 1992)

and GEMT (Chen and Cougheour, 1994). Compared with these models, the Hurley

Pasture Model (Thornley and Cannell, 1997) is more comprehensive in simulating

ecosystem processes.

A limited number of previous efforts at modeling vegetation dynamics in south

Florida flatwoods watersheds have been reported in the literature. Fitz et al. (1996)

developed a processed-oriented General Ecosystem Model (GEM) to capture the

response of macrophyte and algae communities to simulated levels of nutrients, water

and environment inputs such as light intensity, temperature and fire. The vegetation

submodel in SFWMD wetlands model (SFWMD, 1997b) was developed for sawgrass

(Cladium) and cattail (Typha) to simulate aboveground (leaves and shoots) and

belowground (roots and rhizomes) biomass and nutrient content. This model includes the

interactions between light, temperature and nutrients on plant growth and decomposition

but it does not consider the effects of hydrology, grazers or other influence on plant

growth or survival. Wu et al. (1997) developed SAWCAT using Markov Chain

probabilities to simulate vegetation dynamics in wetlands in response to levees, water

depth, and phosphorus. The transitional probability model only simulates the number,

mean size, and largest size of patches of each vegetation type, but does not simulate

biomass production.









Process-oriented vegetation models generally require comprehensive physiological

and phenological parameters so that application and testing of these models is difficult in

watersheds where little data are available. In this research a simple and generic

vegetation dynamics model for simulating the growth dynamics of pasture and wetland

grasses found in south Florida watersheds was developed. Details of the vegetation

model development are presented in Chapter 5.

ACRU2000 Modeling System

The ACRU (Agricultural Catchment Research Unit, v3.00) model was originally

developed in FORTRAN by Schulze (1995) in the department of Agricultural

Engineering, University of Natal, Pietermaritzburg, South Africa for simulating

catchment, forest and wetland hydrology, flood routing, and soil erosion, dam design,

irrigation modeling and crop yield modeling within South Africa. Schulze (1995) gives a

comprehensive description of ACRU (v3.00) and particular aspects of the model, such as

hydrologic sensitivity to climate change, global climate-change and agricultural

productivity are described in a range of publications (Schulze et al., 1993; Smithers and

Caldecott, 1993; New and Schulze, 1996). ACRU (v3.00) has been enhanced by many

additions in response to the need for answers to a range of water-related issues in South

Africa and elsewhere (Clark et al., 2001).

ACRU (v3.00) is described as a physical, conceptual model. It is physical in that

the hydrological processes are represented as explicitly as possible and conceptual in that

the components, relationships and processes in the system are idealized. The hydrologic

model in ACRU (v3.00) can operate either as a lumped small catchment model with

relatively homogeneous soil and land cover attributes, or as a distributed cell-type model

where complex catchments are separated into subcatchments or land segments.




















."A:JR41' 1'..-".. --" :'..... .... DUICKFLOW






INTERMEDIATE STORE

BASEFLOW
GROurDWATER STORE


Figure 2-1. General structure of the lumped ACRU (v3.00) model (Schulze, 1995).

In the lumped ACRU (v3.00) model, the core of the model comprises the water

budget routines for multiple soil layers on a catchment, the general structure of which is

illustrated in Figure 2-1. Water enters the subcatchment as precipitation and/or irrigation.

The vegetated or impervious land surface may intercept all or part of the precipitation,

and this intercepted water in turn is evaporated back into the atmosphere. For

precipitation reaching the soil surface, runoff is calculated using the modified SCS

equation, and the balance infiltrates into the topsoil horizon (Clark et al. 2001). Soil

water evaporation takes place in the topsoil horizon, and transpiration takes place in those

soil horizons that contain roots. Soil water movement takes place between soil horizons.

Soil water can percolate from the bottom soil horizon to the groundwater storage.

Baseflow is generated and released from the intermediate and groundwater stores as a

part of daily surface runoff in this subcatchment. Surface runoff is routed to the

subcatchment outlet as a combination of quick flow (storm released into the stream on the









same day as the rainfall events) and delayed storm flow, a surrogate for post storm

interflow.

In the distributed ACRU model, a catchment is divided into a set of subcatchments

where the lumped model is applied individually throughout the entire simulation time

period, then the generated daily runoff from each subcatchment can be convoluted into a

runoff hydrograph using the triangular unit hydrograph approach for that subcatchment.

The runoff hydrograph from each subcatchment is then routed along the pre-specified

pathways to the outlet of the catchment using the Muskingum-Cunge method. No spatial

exchange of water flows is considered among subcatchments in ACRU (v3.00). This

design may be sufficient to present hydrology in watersheds where topographical

gradients dominate water movement and subcatchment hydrology is relatively

independent, but is not desirable for low-relief catchments.

To enable code expansion and functionality to meet different needs, ACRU (v3.00)

was entirely restructured into ACRU2000 (Campbell et al., 2001; Clark et al., 2001;

Kiker and Clark, 2001) within a Java-based, object-oriented framework. In ACRU2000,

the internal hydrological modeling features of ACRU (v3.00) were retained but the

underlying foundation was changed to allow new features to be developed. The

restructured version revised the surface runoff routing algorithm in ACRU (v3.00) by

executing each subcatchment every time step and then routing runoff from each

subcatchment to the outlet of watershed using the pre-specified pathway before

continuing to the next time interval. Additionally, a nutrient module, ACRU-NP

(Campbell et al., 2001), patterned after transformation and transport concepts used in the

GLEAMS model (Knisel et al. 1993), was added into ACRU2000. Campbell et al.









(2001), Clark et al. (2001), and Kiker and Clark (2001) describe the restructured

ACRU2000.



Land Segments Flow Network



Da IL % nII







Reaches
Land Segment Flow Directions

Figure 2-2. Land segment configuration of the ACRU2000 model (Kiker et al., 2001).

Figure 2-2 shows an example of the land segment configuration of a simulation

domain when ACRU2000 runs in a distributed mode. Hydrologic processes, such as

rainfall, canopy interception, evapotranspiration, infiltration, runoff, and percolation, etc.,

simulated in the lumped ACRU model are consistently applied in each land segment

when operating the model in a distributed mode. Land segments are connected to one

another in a pre-specified pathway, which is used to deliver the runoff from upstream

land segments down to outlet of the simulation domain without interactions in land

segments along the way. Groundwater is treated as part of surface runoff from each land

segment. Physical water exchange is ignored between land segments because water

flows bypass downstream land segments. This design may be sufficient to present

hydrology in watersheds where topographical gradients dominate water movement and

subcatchment hydrology is relatively independent. However, it is not adequate in









simulating south Florida flatwoods hydrology where water movement is primarily

dominated by hydraulic gradients and ecohydrologic variation is spatially interactive in a

significant manner. Furthermore, the nutrient model in ACRU2000, patterned after

transformation and transport concepts used in GLEAMS (Leonard et al. 1987), does not

consider the spatial transport of nutrients through lateral subsurface water movement.

In order to make the ACRU2000 model applicable in south Florida flatwoods

watersheds, modifications were needed to enable multi-directional spatial simulation of

water and nutrients. With the object-oriented design of ACRU2000, it is feasible to add

new modules while retaining the existing capacities of the model. The following

Chapters 3, 4 and 5 describe the details of modifications for hydrology and nutrient

transport simulation, and the addition of the vegetation dynamics model, respectively.

Model Testing Procedures

Model testing is a very important part in model development in that it can detect

inappropriate design in model algorithms, increase the model robustness and accuracy

through model calibration and validation, and determine the limitations and constraints of

model application through model evaluation. However, general methodologies related to

model calibration and validation have been subject to considerable discussion and dispute

during the past decades (Refsgaard, 1997).

Distributed models have the capacity to simulate spatial variations. Therefore the

number of parameters and variables are often several orders of magnitude higher than

those required for lumped models of the same area. Lumped and distributed models

should have different requirements with regard to model calibration and validation

procedures. Unfortunately, much attention has been given to specific calibration and

validation procedures for lumped models, whereas very limited attention has so far been









devoted to the more complicated tasks in connection with distributed models (Refsgaard,

1997). The procedure used for model calibration and validation in this study are

summarized in the following sections.

Model Calibration

Calibration is the process by which model parameters are adjusted to give the best

fit between simulated results and observed data at a particulate site. In other words,

calibration involves adjusting certain model parameters by systematically comparing

simulated results with observations while model structure remains the same. Model

calibration is built on the assumption that a well calibrated model enhances its predictive

capability by incorporating the best available data and adjusting calibration parameters to

obtain a close agreement between model output and historical data.

To calibrate a model adequately, the calibration period should include events that

significantly stress the simulation system. Calibration of distributed integrated models is

more challenging, as discussed above, due to complex model structures and large

parameter sets, and requires calibration over both time and space.

Model Validation

Validation is the process by which the calibrated model parameters are used to

predict state variables during periods when comparisons can be made to an independent

set of historical data, which were not previously used in the calibration process. The

purpose of validation is to determine if the model is sufficiently accurate for its

application as defined by objectives of the simulation study. The model is said to be

validated if its accuracy and predictive capacities in the validation period have been

proven to lie within well defined limits that may depend on intended model use.









Sensitivity Analysis

Sensitivity analysis (SA) is the process of varying model input parameters and

evaluating how model output changes with such variation (SFWMD, 2005). If a small

change in a parameter results in relatively large changes in the model output, the model is

said to be sensitive to that parameter. This may mean that the parameter has to be

determined very accurately. Prior to accepting the final set of calibrated model

parameters, a sensitivity analysis should be performed to determine the relative

magnitude of model response to changes in selected parameters.

The most common form of SA is independent parameter perturbation in which

parameters are varied individually by a fixed percentage around the base value (Ferreira

et al., 1995). An example of this approach is first-order analysis (Hann and Zhang,

1996), in which the first order derivatives in the Taylor series approximation of the

output variables are estimated:

Oyj yj(xi +Axi)- yj(xi)
sji= (i = 1, 2, ..., n; j = 1, 2, ..., m) (2-1)
Dx. Ax.

in which, sji is the sensitivity of output yj to the change of parameters xi; yj (j = 1, 2,.--, m)

are the m predictions; xi (i = 1, 2,..-, n) are the n parameters. These sensitivities sji have

units and thus are difficult to compare over parameters of interest directly. Thus one can

employ normalized or dimensionless sensitivities, Jones and Luyten (1998) used the

relative sensitivity, rsji:

59y3/y xj
rs= =si (2-2)
Ji Cxi/xi Yj

where xi is the calibrated value for input parameter xi, and yj is the prediction by the

calibrated model parameters. However, interactions between two or more parameters









(non-linear effects) may affect the model outputs, even if the output is linear in each

parameter. To check for two parameter interactions, second or high order derivatives in

the Taylor series approximation must be evaluated. In practice, the computation load to

find all possible significant derivatives may be too excessive.

Distributed integrated models are structured to enable simulation of spatial

variations in catchment characteristics and thus require more parameters, in principle,

than the lumped models. Thus it is necessary to investigate the spatial and temporal

sensitivities of input parameters. The first order method as shown in Equation (2-1) and

(2-2) was used to conduct the SA on major model outputs including surface runoff,

groundwater table depth, and P and N loads. If the output is a time series, the sensitivity

is a function of time, which may help to detect the influences of input parameters on

extreme values and trends of outputs. If the output is an individual value, an individual

sensitivity is calculated, which may help to detect the influences of input parameters on

the overall model response. The spatial sensitivities of outputs can be calculated using

the same equations when the spatial observation of output variables are available.

The significance of model sensitivity analysis is two-fold in that it provides

information on the behavior of model output to input parameters which, in turn, can be

used in model calibration; also, it gives insight in establishing priorities related to future

data collection efforts. Sensitivity analysis is distinguished from uncertainty analysis in

that it is a measure of the relative importance that each input parameter has on the range

of simulated outputs. Uncertainty analysis quantifies the confidence in particular output

variables given the probability distributions for input parameters. While sensitivity

analysis is often limited to parameter sensitivity, uncertainty may be generated by a









number of factors including parameter uncertainty, boundary and initial condition

uncertainty, model spatial and temporal resolution, availability and quality of data, and

model errors.

Model Evaluation

For sake of the model calibration and validation, and the comparison of models,

one needs quantitative information to measure model performance compared to observed

data or other model predictions. Statistical approaches to quantify the accuracy of model

predictions provide standardized measures of model performance although even these

methods do not provide completely clear-cut conclusions about the accuracy of model

predictions (Ramph, 2004). Given these caveats, the use of several different measures of

performance to evaluate a model may present a more complete picture of model

performance than any single measure and may allow the user to weight individual results

according to their priorities (Ramph, 2004).

Statistics

A number of statistics were used to evaluate model results in this study, including

model bias, average relative error (RE), root mean square error (RMSE), coefficient of

variation (CV), Pearson product-moment correlation coefficient (R2), and Nash-Sutcliffe

(NS) coefficient (Nash and Sutcliffe, 1970). Brief overviews of these statistical measures

are provided below.

Bias, also called the average error, determines the average deviation of the

predicted values from the measured values given N simulated-measured value pairs. As

can be seen from Equation (2-3), where Pi is the observed value, Oi is the model-

simulated value, and N is the number of observations, bias is calculated as the mean









differences between paired observed and simulated values. Bias values closer to zero

indicate better overall model performance.

1 N
bias = (P O) c< bias< +oo (2-3)
N 1-

The relative error is a unitless, normalized parameter that also quantifies the bias of

the predicted values. The arithmetic average relative error (RE) as shown in Equation (2-

4) determines if the model overestimates (positive deviation) or underestimates (negative

deviation) the measured values, which is most useful for comparing models and data sets

(Hession et al., 1994). However, Hession et al. (1994) indicated that the arithmetic

average relative error can be significantly influenced by one or two outliers.

N
SZ(Pi -O)
RE = = o < RE < +o (2-4)
O N

The root mean square error (RMSE), or standard deviation/error, provides a direct

measure of the error between the model and the observed data (Thomann, 1982) as seen

in Equation (2-5). This statistic is used to measure the discrepancy between modeled and

observed values, and indicates the overall predictive accuracy of a model. Due to the

quadratic term, greater weight is given to larger discrepancies. With this measure,

smaller values indicate better model performance (Evans et al., 2003).


RMSE = (PiOi)2 0< RMSE < +oo (2-5)


The RMSE is essentially the overall sum of squares errors normalized to the number of

observations (Hession et al., 1994).









The coefficient of variation (Young and Alward, 1983) defines a normalized error

as shown in Equation (2-6), which is the RMSE normalized to the overall mean.

1 ]1N
CV1= 1 (Pi-O)2 -m OVN i=l
0N

Using the RMSE, along with the CV, Hedden (1986) suggests that for screening

applications a model should be able to replicate observed data within an order of

magnitude, and for site-specific applications the predictions should be within a factor of

two. In the criterion of within "a factor of two" will be satisfied when the CV value is

less than one and the criterion of within "an order of magnitude" will be met when it is

less than nine.

The Pearson product-moment correlation coefficient (R2) value, also called the

goodness of fit, is a measure of the degree of linear association between two variables.

Depending on the strength of the linear relationship, the R2 can vary from 0 to 1. The

closer R2 is to 1, the better the regression explains the relationship between simulated and

observed. However, it does not explain how close the relationship is to the perfect linear

fit (R2 = 1) between observed and predicted values.

N -2
S(oi- O)(Pi -P)

(O O)2 Z(Pi -P)2


Like the R2 measure described above, Nash-Sutcliffe (NS) coefficient (Nash and

Sutcliffe, 1970) is another indicator of goodness-of-fit, and is one of the statistics that

have been recommended by the American Society of Civil Engineers (ASCE, 1993) for

evaluation of the performance of hydrological models. The NS can vary from 0 to 1,









with 1 indicating a perfect fit. Computationally, the NS could be negative but this

becomes rather meaningless as far as interpretation or results are concerned. For NS = 0,

the interpretation can be made that the model is predicting no better than using the

average of the observed data (ASCE, 1993). The statistics works best when the

coefficient of variation for the observed data set is large (ASCE, 1993).

N
(oi -pi)2
NS=1- i=1 -oo< NS<1 (2-8)
N
(Oi -O)2
i=1

Graphic representation

Graphic representation enables visualizing the relationship between predicted (Pi)

and measured (Oi) values and provides a straightforward way to compare the fit between

predicted and measured values. Scatterplots and duration curves are two commonly used

graphic representation techniques in the hydrological literature.

Scatterplot is a simple method to visualize the relationship between Pi and Oi by

fitting a regression line between Pi (Y-axis) and Oi (X-axis), relative to a line designating

a 1:1 relationship. The resulting pattern indicates the type and association/strength of the

relationship between predicted and observed values. A positive association would be

indicated by upward trend (positive slope) and a negative association would be indicated

by a downward trend (negative slope). Or, there might not be any notable association, in

which case a scatterplot would not indicate any trends whatsoever. The slope of the

regression line and its Y-intercept may provide evidence of systematic error in the model,

providing quantities that can be compared across models. A slope of 1.0 with an

intercept equal to 0.0 indicates perfect fit of the model predictions.









A duration curve represents the relationship between magnitude and frequency of a

variable for a particular watershed providing an estimate of the percentage of time a

given variable was equaled or exceeded over a period of time. There are different

approaches to generate duration curves for satisfying different purposes. Among these,

the flow duration curve (FDC) has widespread application in hydrologic studies such as

hydropower, water-supply and irrigation planning. In recent years, duration curves have

also been applied in other areas such as load assessment and TMDL development. For

this study, the flow and load duration curves were compared for simulated and observed

variables.

The basic procedure for generating a duration curve includes: 1) ranking the

daily/weekly/annually simulated and measured data separately from highest to lowest for

each dataset of interest; 2) calculating percent of days/weeks/years these variables were

exceeded (= rank/number of total data points); 3) plotting both simulated and measured

data vs. percent of time interval exceeded on the same plot.

Comparison of duration curves between model predictions and observed data

provides a direct evaluation of the accuracy of particular flow/load ranges. However,

there are some shortcomings of the duration curve method. For example, there is no

commonly accepted method to deal with repeated values. Therefore, using duration

curves altogether with other statistics is advisable in model evaluation.

No one of the abovementioned model evaluation approaches will be best in all

situations. Reviewing several of these measures together will provide a more complete

description of model performance (Ramph, 2004). The results should also be viewed in

the context of the intended use of the model. Users must decide for themselves what









level of performance is acceptable and, likewise, which approach is most appropriate to

their interests.














CHAPTER 3
HYDROLOGIC SIMULATION MODEL

Introduction

In the coupled modeling system, the hydrologic model plays an important role in

that it not only simulates hydrologic processes but also provides a framework for the

nutrient and vegetation models discussed in Chapters 4 and 5. Therefore, the design and

development of the hydrologic model has to take these other models into account.

An example catchment, shown in Figure 3-1A, is discretized into a network (Figure

3-1B) of multiple land segments, each of which is a unique hydrologic unit with varying

shape and size depending on land cover, topography, vegetation types and soil types.

Each land segment is bordered by either the internal boundaries, which are artificially

defined separations between two land segments, and/or the external boundaries, which

are the separation between the simulated catchment and its neighboring catchments. A

continuous unconfined groundwater system is assumed to underlie the discretized model

domain. For each land segment, it is assumed that the soil beneath is laterally

homogenous with heterogeneous vertical soil layers that are divided according to natural

soil horizons and soil properties. Additionally, soil layers in neighboring land segments

have the same number of soil layers of the same thickness, and adjacent soil layers are

laterally connected.

A fixed time step of one day is used in the coupled model. The selection of the

time step is a tradeoff between time scales required for different model processes and

general temporal scales of input data, such as rainfall and evapotranspiration.











External boundary

Internal boundary


Land segment


A. Example Catchment B. Catchment Discretization


Figure 3-1. Schematic of an example catchment and its spatial discretization.

An overall flow chart of the coupled modeling system and major model processes

included in the model are illustrated in Appendix A. In the natural condition,

hydrological processes, together with nutrient and vegetation processes, occur

continuously in a simultaneous manner, but in the model these processes are discretized

and simulated in a sequential manner within the one day time step. In the model these

processes are grouped into vertical and horizontal processes for each land segment

according to the flow direction each process describes and whether a process interacts

with others in a neighboring land segment. Vertical processes deal with the water

movement, nutrient cycling, and vegetation dynamics through the soil profile whereas

horizontal processes simulate lateral water movement, nutrient cycling, and vegetation

dynamics over the domain and require the simulation of water and nutrient exchange

between adjacent land segments. The vertical process group is the basic building block

of the model, simulating the temporal dynamics of important hydrological, chemical, and

growth processes within one land segment. The horizontal process group is the core of

the distributed modeling system in that it spatially connects all land segments in a

watershed together through processes such as lateral water movement, nutrient transport,









and biomass expansion. In the coupled model, the vertical processes are executed

sequentially throughout the simulation domain for each land segment followed by the

horizontal processes across all land segments. Nutrient and vegetation processes are also

separately grouped into vertical and horizontal processes. Discussions of these processes

are presented in detail in Chapters 4 and 5.



Rainfall Evapotranspiration


Canal Flow

Interception
Infiltration
Overland Flow
Upward Fux tso /dsrbto

Deep Seepage -* Groundwater Flow



Figure 3-2. Hydrological processes in the modified ACRU2000 hydrologic model.

Hydrologic processes simulated in the modified ACRU2000 model are depicted in

Figure 3-2, including rainfall, interception, evapotranspiration (ET), infiltration, soil

water redistribution, upward flux, deep seepage, overland flow, canal flow, and

groundwater flow. Although some of these processes including rainfall, interception, ET,

infiltration, and soil water redistribution existed in the original ACRU2000 model,

Martinez (2006) added alternative processes to simulate ET, infiltration, and soil water

redistribution and created new processes for deep seepage and upward flux in order to

better represent the soil water dynamics for flat, poorly-drained sandy, and high-water-

table flatwoods soils in south Florida. Multi-directional spatial simulation of overland









and lateral groundwater flow, and canal flow did not exist in the original ACRU2000

model. Developing algorithms for these processes was the primary purpose of this study.

In the following hydrologic components sections, the methodology used to simulate

each process is described, with more detail given to the new overland flow (including

canal flow) and lateral groundwater flow processes. Schulze (1995) gave a

comprehensive description of the hydrologic processes incorporated into ACRU (v3.00).

The hydrologic model was developed based on the existing hydrologic model in

ACRU2000 (Campbell et al., 2001; Clark et al., 2001; Kiker and Clark, 2001) by adding

new components and modifying existing components needed for multi-directional spatial

simulation, with an emphasis of considering lateral surface and subsurface flow exchange

throughout all soil layers between adjacent land segments. Changes were made so that

the original model features were retained for other potential applications while enabling

new functionality.

The objectives of this Chapter are

* To modify the ACRU2000 hydrologic model to add new hydrologic components
that enable multi-directional spatial simulation of overland flow and lateral
groundwater flow;

* To analyze the influence of the simulation sequence on simulation results from the
modified model;

* To test the ability of the modified model to accurately simulate hydrological
processes by comparing with well-accepted models including MIKE SHE (DHI,
2004) and MODFLOW (McDonald and Harbaugh, 1988);

* To test the reliability of the modified model in predicting major hydrologic
processes by comparing with the FHANTM model simulations of a field
experiment in the Kissimmee River Basin, Florida.

In the following sections, the methodology for each hydrologic component

simulated in the modified version of the hydrologic model is described, followed by the









description of boundary and initial conditions required for the model. Then, simulation

sequence analyses, hypothetical scenario tests and an application in Dry Lake Dairy #1,

Kissimmee River Basin, Florida are presented and discussed and conclusions are drawn.

Appendix D lists and describes the added or modified objects. Appendix E lists all newly

added variables associated with the coupled modeling system.

Vertical Hydrologic Components

Rainfall

Rainfall is the fundamental driving force behind most hydrological processes. The

success of hydrological simulation studies depends to a large extent on the precision with

which the rainfall data are observed temporally and spatially and how they are processed

in the model. Several techniques in ACRU (V3.00) are available for estimating aerial

rainfall for input files, including

* Driver station method
* ACRU-300 trend surface approach
* Thiessen polygon method
* Inverse distance interpolation technique and
* Spline interpolation technique

The design of input files in the model allows each land segment to have its own

daily time series of rainfall depth input. Therefore, the temporal and spatial distribution

of rainfall can be differentiated when such spatial rainfall data are available. But within

one land segment, or when the model is used in the lumped mode, the rainfall is assumed

to have a uniform spatial distribution.

Canopy Interception

Canopy interception is calculated as the minimum of the available rainfall and the

amount of interception storage capacity available on the intercepting vegetation leaf

canopy. The available rainfall first fills available canopy storage, then the remaining









rainfall is assumed to reach the land surface and is available for infiltration or/and surface

runoff. If the rainfall is less than or equal to the available canopy storage, all rainfall is

intercepted.

Evapotranspiration

The calculation of actual evapotranspiration requires three steps including reference

potential evapotranspiration (RET), potential evapotranspiration (PET), and actual

evapotranspiration (AET). The latter is further separated into soil evaporation and plant

transpiration for later use in nutrient uptake discussed in Chapter 4.

The reference evapotranspiration (RET) refers to the maximum evaporation from a

short grass surface, or the maximum evaporation from an actively growing alfalfa crop,

at least 0.3 m tall, standing erect and covering an extensive area. There are many

methods of estimating RET in ACRU (v3.00), ranging from complex physically based

equations to simple measurements and even simpler surrogates based on single variables

such as temperature:

* The daily United States Weather Bureau Class A pan evaporation.
* The 1948 Penman equation
* Temperature based equations in ACRU (v3.00)
* The Linacre suite of equations (1977, 1984, 1991)
* The Hargreaves and Samani methods (1982 & 1985)
* The Blaney and Criddle (1950) equation
* The Thomthwaite (1948) equation

In addition, Martinez (2006) added the Penman-Monteith grass reference potential

evapotranspiration approach (Allen et al. 1998), referred to as FAO56PM in the model.

The RET can be calculated separately and input as a time series or calculated directly by

the model with required input parameters and data.









For all methods, evapotranspiration is calculated in a top-down approach. First the

evaporation demand is applied to intercepted water, next to ponded water on the ground

surface, and then to the soil as evaporation and transpiration. The PET for a crop is

obtained by multiplying RET with a month-by-month crop coefficient which integrates

numerous dynamic processes relating to crop transpiration and soil evaporation.

Two existing AET calculation methods of ACRU (v3.00) have been adapted for

use, which calculate AET either as a lumped quantity or by determining soil evaporation

and transpiration separately using Ritchie's method (Ritchie 1972). In addition, a third

simple actual evapotranspiration method was added by Martinez (2006).

When calculating evapotranspiration as a lumped quantity, the RET is multiplied

by the crop coefficient (VICAY) to get the PET, then the layer AET within the root zone

is calculated as the minimum between the PET and the available water storage (beyond

the wilting point) for a specific soil layer. Using an approximate method (Childs and

Hanks, 1975), the layer AET is split between soil water evaporation (AEs) and plant

transpiration (AEt) for use for plant uptake of solutes and nutrients:

AE, = AET AEt (3-1)

ViCAY- 0.2
0.95AETx( CAY-0.2) V1CAY>0.2
AEt = 0.8 (3-2)
0 V1CAY <0.2


When using Ritchie's method to determine evapotranspiration, the potential soil

evaporation and plant transpiration are separated as a function of the leaf area index

(V1LAI).

PE, =PET-PE, (3-3)









(0.7V1LAIo -0.21)x PET V1LAI<2.7
PEt = (3-4)
0.95PET V1LAI> 2.7

The potential transpiration (PEt) is further adjusted by being multiplied by a crop

coefficient. The actual plant transpiration can be either equal to maximum transpiration

for a given soil layer or be less than maximum transpiration because of a deficiency of

available soil water or an excess of soil water. According to Martinez (2006), if the

FA056PM RET is used, the potential evaporation is multiplied by a factor of 1.15 to

account for the difference in albedo between bare soil and a vegetated surface as

recommended by Allen et al. (1998). The potential soil evaporation (PEs) is further

adjusted for the percent surface cover by mulch. The soil water evaporation takes place

down to a user defined depth of the soil with recommended values ranging from 0.1 to

0.15 m (Allen et al. 1998). According to Ritchie's method actual evaporation from the

soil surface continues at a maximum rate equal to the potential rate (stage 1) until the

accumulated soil water evaporation exceeds the stage 1 upper limit. After that, soil water

evaporation proceeds at a reduced (stage 2) rate.

The new actual evapotranspiration method that has been added to the model is a

simplification to the lumped evapotranspiration method described above. The

simplifications include neglecting water stress and the apportionment of

evapotranspiration to different layers. Potential evapotranspiration is applied to

individual soil layers in the original model according to the fraction of plant roots within

that layer. Similarly, the new method applies potential evapotranspiration from the top-

most soil layer until the wilting point is reached and then applies the remainder to the

layer below until the bottom of the root zone (the last layer containing roots) is reached.









Infiltration

Infiltration is a process by which water on the soil surface enters the soil.

Considering the high infiltrating capacity of flatwoods soils, the rainfall rate is assumed

to be less than the maximum infiltration rate and the water on the ground surface is

allowed to infiltrate until the soil profile becomes completely saturated. Thus in the

modified ACRU2000, saturated-excess ponding and runoff is simulated. The volume of

infiltration is taken as the minimum of the rainfall rate multiplied by grid cell area and

time step and the available void space between the water table and land surface.

Infiltrating water is moved into the top-most soil layer and is further moved into deeper

soil layers through a process discussed below. Water that infiltrates into the soil profile

will cause a rise in the water table after any depleted root zone soil moisture is

replenished.

Soil Water Redistribution

In ACRU (v3.00), unsaturated soil water redistribution downwards can take place

from the top to a subsoil horizon and from the subsoil horizon to groundwater when the

soil moisture content in the upper soil horizon is above the field capacity. Upward

redistribution, which is similar to capillary movement, takes place from the bottom soil

horizon and works up through the soil horizons. If a soil horizon is above porosity then

the excess water is moved to the soil horizon above, or in the case of the top soil horizon

the surface of the land segment, where it is added to ponded water.

Martinez (2006) created an alternative process specifically for the flatwoods soils.

This process redistributes infiltrating water based on the drained-to-equilibrium soil

moisture content of each soil layer. According to Martinez (2006), this process has to be

turned on after infiltration, net infiltration, new water table depth (if any), and new









drained-to-equilibrium soil moisture contents (if any) have been determined. The process

proceeds from the top down, and each soil layer is drained to equilibrium. Then, starting

from the bottom-most layer, excess water is moved upwards from the lower to upper

layers, or to the ground surface if necessary, to ensure that no layer is above its drained-

to-equilibrium water content.

Upward Flux

Evapotranspiration depletes the soil moisture content in the root zone and causes an

upward gradient between the soil below and the root zone, which induces an upward flux

of water within the soil. This depleted root zone has to be replenished first before

infiltration occurs to cause a rise in the water table.

The upward flux algorithm was added by Martinez (2006) using an approximate

method (Anat et al., 1965) to calculate the maximum steady-state upward flux:

1.886 d
quf =Ks 1+ 8] (3-5)
1 2 +1 hb

where quf is the amount of upward flux on a given day [L/T]; Ks is the saturated hydraulic

conductivity of the soil [L/T]; d is the distance between the water table and the depleted

root zone [L]; hb is the bubbling pressure head [L]; and r is a function of the pore size

distribution index, k, of the Brooks and Corey (1964) model, i.e., r = 2 +3X.

This relationship for upward flux assumes that the soil profile is homogenous. For

each soil layer a set of parameters, including Ks, hb, and k, are input. These parameters

can be best obtained by fitting the above equation to a steady-state solution of Richard's

equation for all water table depths below the layer in question. Alternatively, SOILPAR

(v2.00) (Acutis and Donatelli, 2003) can be used to estimate these parameters as well.









The amount of upward flux that actually occurs on a given day is the minimum of

the maximum value calculated and the amount to which the root zone was depleted (and

is now fully replenished by upward flux).

Deep Seepage

Deep seepage can occur through a restrictive layer located below the soil profile

according to Darcy's Law. It is calculated as:

H -H
qds = K w d (3-6)
d

where qds is the amount of daily deep seepage through the restrictive layer [L/T]; Ks the

restrictive layer saturated hydraulic conductivity [L/T]; Hw is the height of water table

above the restrictive layer [L]; Hd is the hydraulic head below the restrictive layer [L];

and d is the thickness of the restrictive layer [L].

Horizontal Hydrologic Components

Horizontal hydrologic components including overland flow, lateral groundwater

flow and canal flow were newly added into the original ACRU2000 model. The

simulation of these components requires water exchange between adjacent land

segments. Lateral flow is simulated sequentially, with the simulation sequence

determined by the topographic elevations of land segments and the sequence of land

segment input files. More details and testing regarding the simulation sequence are

discussed in the following section.

The following assumptions were made in the simulation of multi-directional lateral

flow:

For each land segment, only outgoing overland flows are estimated except when
inflows occur through external boundaries;









* Overland flows happen only when a positive hydraulic gradient occurs between
one land segment and its neighboring land segments, and the ponded water depth in
this land segment exceeds its maximum depression storage;

* Potential overland flows are estimated for all neighboring land segments based on
hydraulic gradients. If there is insufficient storage to supply all potential outflows,
actual outflows are determined through storage apportionment;

* Soil within each land segment is unconfined, thus the hydraulic head is the water
level elevation in this land segment above the datum;

* Soils beneath land segments have the same thickness and they are split into the
same number of computational soil layers;

* Adjacent computational soil layers in neighboring land segments have the same
thickness and are laterally connected.

* Hydraulic gradients are the driving force to transfer groundwater flow laterally in
saturated soils;

* From each land segment only outgoing lateral groundwater flows are estimated
except when lateral inflows occur through external boundaries;

* Lateral groundwater flows only happen when a positive hydraulic gradient occurs
between one land segment and its neighboring land segments;

* Potential groundwater flows are estimated for all neighboring land segments. If
there is insufficient storage to supply all potential groundwater flows, actual
groundwater flows are decided through storage apportionment.

Overland Flow

Appendix B shows a flow chart of overland flow simulation for one source land

segment. The potential outgoing overland flows from the source land segment to

adjacent land segments are calculated as discussed in the following section and inflows

through the external domain boundary are input or extrapolated depending on the actual

boundary conditions. Subsequently, a storage check is conducted to guarantee there is

enough surface water storage in the source land segment to supply the potential flows. If

there is insufficient storage the available water is apportioned according to assumptions

discussed in the following section and the detailed algorithm is documented in Appendix









G. External boundary inflows are introduced into boundary land segments each day. The

outgoing flows are then transferred to adjacent land segments, which causes a decrease of

water storage in the source land segment and an increase of water storage in the adjacent

land segment. The whole process described in Figure 3-3 is applied to each land segment

in the simulation domain following the simulation sequence generated by the model.

Overland flow calculation

The generalized spatial relationship between one land segment and its adjacent land

segments is shown in Figure 3-3. Two types of overland flow over internal boundaries

are defined in this model:



NiNeighbor 1

SQs,,d,2
sE,d,3 I Sourc

Neighbor 3




I / Neighbor 2

Neighbor 4



Figure 3-3. Configuration of overland flows between source land segment and adjacent
land segments. The digit (1-4) of adjacent land segment index indicates the
priority to receive the overland flow from source land segment, which is
determined by the water elevation of that land segment and smaller water
elevations correspond to higher priorities. An adjacent land segment with a
larger index has higher priority to receive the overland flow.

Unstructured boundaries, where no local obstacles exist to the flow (no singular

head losses) and a mean resistance coefficient for a given cross section of the flow can be

used. In this case, Manning's formula is used:









W D R2/3 S1/2 uT
s i s,d,i s,d,i s,d,i s,d,i m (3 7)
n s,d,i


where subscript s and d indicate source and destination land segments and i the specific

adjacent neighbor; Qs,d,i is the volumetric discharge from source to destination land

segments [L3/T]; Ws,d,i is the width of the cross-sectional area between source and

destination land segments [L]; Ds,d,i is the water depth of the cross-sectional area between

source and destination land segments [L], an average of water depths in both source and

destination land segments; Rs,d,i is the hydraulic radius of the flow section between source

and destination land segments and it is essentially the average surface water depths [L],

roughly equivalent to Ds,d,i for wide, shallow cross-sectional area or streams; Ss,d,i is the

water surface slope between source and destination land segments [L/L], it is estimated as


Ss,d,i s Zd (3-4)


in which, Zs and zd are the water levels of source and destination land segments,

respectively [L]; L is the distance between the centers of source and destination land

segments [L]; Um is the unit conversion for the Manning formula [L1/3/T]; ns,d,i is the

average of Manning's roughness coefficients of source and destination land segments,

which is modified to be a function of sediment type and the interaction of the vegetation

height/density and water depth (Fitz et al. 1996). This function creates a dynamic

feedback loop between the physical process of flow and the biological process of plant

growth (Sklar et al. 2001), which is calculated as:























0.0
00 Plant height (m)
(dynamic)
Increasing water depth (m)


Figure 3-4. The relationship between Manning's roughness-coefficient (n) to water
depth and to plant height, both of which change dynamically in the model
(Fitz et al., 1996).

ns,d,i nmax (nmax -min)(21-Dd/H -1) (3-5)


where nmax is the maximum Manning's roughness coefficient associated with the dynamic

vegetation density in the land segment [L1/3/T]; nmin is the minimum Manning's

roughness coefficient for a vegetation-free land segment [L1/3/T]; Dd is the surface water

depth [L] and Hm is the dynamic height of the macrophytes in the land segment [L] (Fitz

et al., 1996). Equation (3-5) returns a positive roughness coefficient that ranges from a

vegetation-free minimum to a maximum at the point of full plant immersion (Petryk et

al., 1975). As water depth increases over that of the macrophyte height, the roughness

decreases to an asymptote at the baseline sediment roughness (Nalluri and Judy, 1989).

Figure 3-4 shows the relationship between Manning's roughness-coefficient, n, to water

depth and to plant height.

Structured boundaries, where berms or dikes form a boundary that can be

represented by a singular head loss between two land segments. In this case, two kinds of









classical discharge formulae for broad crested weirs (Cunge 1975) are applied according

to the following criteria (Cunge 1975):

dike
cell s dike cell d







datum


Figure 3-5. Structured boundary (Cunge, 1975)

Free flow condition (Free flow weir)


Qs,d,i = ltlb 1 (zd Z )Zd Z Zs Z <-~(Zd z) (3-6)

Submerged flow condition (Drowned weir)


Qs,d,i = t2b[2g(z- )z, -z z, -2(Zd Zw) (3-7)

where Qs,d,i is the volumetric discharge between source and destination land segments

over a weir [L3/T]; wl and [t2 are the discharge coefficients; b is the effective width of the

weir [L]; g is the gravitational acceleration [L2/T]; Zs and zd are the water levels of source

and destination land segments, respectively [L]; Zw is the elevation of the weir crest [L].

Figure 3-5 shows the relationship between source and destination land segments when a

weir exists in between.

Surface water storage apportionment

Potential overland flows from one land segment to all its adjacent land segments

are calculated separately without taking into account the storage change due to the fluxes

that have been already computed. Hence, it is possible that the sum of the potential flows









could exceed surface water storage in the land segment. This situation is especially likely

to occur if the time step is large. To avoid this situation, a procedure was developed to

limit the actual overland flow to its maximum available volume of storage, hereafter

simply called the available volume, which is a portion of the current volumetric water

storage on ground surface allocated to a specific neighbor based on the following

assumptions:

* The storage apportionment is allocated only to those neighbors that receive the
overland flows;

* The current water storage in the land segment is the maximum available volumetric
water storage;

* The allocation of the maximum water storage in the land segment is made
according to the priority of its neighbors;

* The priority of the neighbors is set up according to the elevation of the water
surface boundary between the land segment and its specific neighbor.

* The water surface elevation of the boundary between the land segment and its
neighbor is determined by the maximum of the water level or the elevation of the
weir crest, whichever exists;

* If a weir exists in between the land segment and its neighbor, the elevation of the
weir crest has to be higher than the elevation of the land segment;

* The neighbor with the lowest elevation of boundary has the highest priority for
allocation of water storage from the source land segment;

* A neighbor with lower priority is not allocated any water storage until the neighbor
with next higher priority is filled to an equavlant water surface elevation, i.e., the
two neighbours have equal water surface elevation;

* The available volume for a specific neighbor can not be more than required to
equilibrate the water levels of the land segment and this neighbor.

Once the available volumes are determined for all those neighbors receiving

potential overland flows from the land segment, each of them is compared with its

corresponding pre-estimated potential overland flow. If any potential overland flow

exceeds its available volume, then the actual overland flow is set equal to the available









volume. Otherwise, the actual overland flow is equal to its potential value. Details of the

assumptions and algorithm for surface water appointment are described in Appendix G.

Groundwater Flow

Groundwater flow here refers to lateral groundwater flow which occurs in the

saturated soils. Appendix C shows a flow chart of lateral groundwater flow simulation

for one land segment. For a source land segment, the potential outgoing lumped lateral

groundwater flows and inflows from external domain boundary are calculated, followed

by a storage check to guarantee there is sufficient groundwater storage to provide the

potential flows from the source land segment. Lumped groundwater flow available for

transfer is split into layer flows for the saturated zone in proportion to the layer

transmissivity. The layer flows are then transferred to the neighboring land segment.

The transmission of layer groundwater flow increases the void space in the soil of the

source land segment and decreases that in the neighboring land segments. At the end of






52


Ground surface
////////////////////////////////////////

Water Table


s Lateral groundwater flow hd




Impermeable strata
(A)


SWater Table

SOverland flow Ground surface


Lateral groundwater flow
hd






Impermeable strata
(B)


Figure 3-6. Schematic representation of the lateral groundwater flow from the higher
head hs to the lower head hd. (A) and (B) represent the situations without and
with overland flow, respectively.






53



1 1 sd----- - ,di
AI
,*/
'-- I
"" -------Qsi, -------- ---

zs Hs ----- -- Qi,2 Hd Zd
-- Qsi,3 ---- .

40 Qs,i,4

/ (A)
L






L_.. .,





Zs Us /7 [-y ^ Qs'i'2 H{A d Zd
iI ( s1 ------- ---
I 'b
zs H1 Qsi2 ------- Hd Zd
s 3 I i-------



"^'bj---^/(B)




Figure 3-7. Diagram of lateral groundwater flow between two land segments. (A) and (B)
represent the generalized situations without and with overland flow,
respectively.









simulation, the water table depths are updated for all land segments. The complete

process described in Appendix C is applied to each land segment in the simulation

domain following the simulation sequence generated by the model.

Lateral groundwater flow calculation

Lateral groundwater flow is calculated first in a lumped manner between the

saturated zone of source land segment and its adjacent land segments according to

Darcy's law. The lumped flow is then partitioned into the saturated soil layers. Figure 3-

6 shows the generalization of lateral flows between two adjacent land segments, in case

A (Figure 3-6A) the water table is below the ground surface and case B it is above the

ground surface (Figure 3-6B).

The generalization of (A) and (B) in Figure 3-6 is correspondingly illustrated in

Figure 3-7(A) and (B). First of all, using Darcy's Law, the lumped lateral groundwater

flow from the source land segment to the destination land segment is estimated

Qs,d,i =Ks,d,lSs,d,lW s,d,1Hs,d,1 (3-8)

Where subscript s and d indicate source and destination land segments and i the specific

neighboring land segment; Qs,d,i is the lumped lateral groundwater flow from source to

destination land segments [L3/T]; Ss,d,i is the hydraulic gradient between source and

destination land segments [L/L] and is calculated as


SSd d (3-9)
L

in which Zs and zd are the groundwater heads of the source and destination land segments,

respectively [L]; and L is the distance between the centroids of the source and destination

land segments [L]; Ks,d,i is the effective horizontal saturated hydraulic conductivity [L/T],









an average of horizontal saturated hydraulic conductivities of the source and destination

land segments [L/T]. It is estimated as

1 1
K,d, (- Z K b + Kdkbdk)/2 (3-10)
D, j d k

where subscript j and k indicate the number of soil layers below the water table of the

source and destination land segments, respectively; Ks is the saturated horizontal

hydraulic conductivity of soil layer j of source land segment [L/T] and Kdk, that of soil

layer k of destination land segment [L/T]; bs and bdk is the thickness of saturated soil of

layer j in source land segment and layer k in destination land segment [L]; Dsj and Ddk is

the thickness of saturated soils for source and destination and segments [L]; Ws,d,i is the

boundary width of the cross-sectional area between source and destination land segments,

perpendicular to the direction of lateral groundwater flow [L]; Hu,d,i is the average of the

thickness of saturated soil of source and destination land segments [L],

H, + H
Hsdi d (3-11)

where Hs and Hd are the thickness of saturated soils of the source and destination land

segments, respectively [L].

Next, the lateral groundwater flows from source to destination saturated soils are

distributed to individual saturated soil layers by the formula

Ki~jbi~j
Qs,i,j = Qs,d,i N ( = 1,....,N) (3-12)
YKi,jbi,j
j=1

where Qs,ij is the layer lateral flow from the saturated soil layer j of source land segment

to the layer j of destination land segment [L3/T]; Kij is the saturated horizontal hydraulic

conductivity of the layer j of source land segment [L/T]; bij is the thickness of saturated









soil of the layer j of source land segment [L]; N indicates the number of saturated soil

layers. In case the water table is located somewhere within a soil layer, the thickness of

saturated soil in this layer is the depth from the water table to the bottom of this soil

layer.

Groundwater storage apportionment

Similar to the procedure of surface water storage apportionment for overland flow,

a procedure to check if the sum of the lumped groundwater flows exceeds the available

saturated groundwater storage in the source land segment is needed. The available

groundwater storage is assumed to be the portion of volumetric water above the field

capacity. If the sum of outgoing lumped flows is greater than the available groundwater

storage, each lumped groundwater flow is adjusted using a ratio, which is the difference

between the sum of outgoing lumped flows and the available volumetric storage to the

available volumetric storage. By doing this adjustment, the actual lumped flows are

limited to the available storage. Since the groundwater flow is much slower than

overland flow this procedure is rarely invoked when the model runs at a daily time step.

Canal Flow

The interaction of canals with the simulation domain can occur both through

overland flow and lateral groundwater flow depending on the stage of canal and the water

level in the domain. When the canal stage is higher than the water level in an adjacent

land segment, back flow from the canal into the domain occurs. When the canal stage is

lower than the water level in an adjacent land segment, discharge from the domain is

released into the canal. Flow between the canal and domain is estimated using the

following equations:









WR5/3S1/2
q = W Overland flow (3-13)
n

q = KWSD Lateral groundwater flow (3-14)

where q represents the amount of overland flow to/from the canal in Equation (3-13) and

lateral groundwater flow to/from the canal in Equation (3-14) [L3/T]; W is the width of

the land segment-canal boundary [L]; R is the water depth above the ground surface [L];

S is the water slope that is determined by the difference of water level between canal and

the adjacent land segment divided by the distance between the centroid of the canal and

the centroid of the land segment [-]; n is the Manning's roughness coefficient, which is

assumed to be the same as the one for the adjacent land segment for this study [L13/T];

K is the hydraulic conductivity between the canal and the saturated soils of the adjacent

land segment, which is assumed to be the same as the saturated horizontal hydraulic

conductivity for the adjacent land segment for this study; and D is the thickness of the

saturated soil profile [L]. Currently canal flow is not an individual process in the model

but is a built-in routine inside of overland flow and lateral groundwater flow processes.

Initial and Boundary Conditions

To ensure the accuracy of the solution of the simulation model, both initial and

boundary conditions must be prescribed for the domain being simulated. Initial

conditions in the model include water table depth in the soil, ponded water depth if any,

and depressional water storage for each land segment. Boundary conditions include

boundary inflow, outflow, or fixed head conditions and the description of physical

hydraulic structures, levees, fences, and berms, etc. If physical structures exist,

geometric information and relevant empirical formula for flows through these structures

must be prescribed. Time series of atmospheric variables such as rainfall, temperature,









and potential ET, etc. must also be specified for the desired simulation period. Appendix

F shows the input files, required for applications in this dissertation, which contain the

initial and boundary conditions.

Model Testing and Validation

Simulation Sequence and Model Performance Accuracy Analysis

In the modified ACRU2000 hydrologic model, the simulation sequence determines

the execution order of land segments. The simulation sequence is generated according to

the spatial water flow configuration and the sequence of land segment input files

specified in the control menu file (Control.men in Appendix F). The spatial water flow

configuration is arranged according to the topographical elevation of each land segment,

which specifies the directions of outflow from each land segment and is deterministic

once the topography in a simulation domain is specified. However, the sequence of land

segment input files in Control.men is arbitrary, for instance, one can specify the sequence

from the first land segment to the last one, or its opposite, or other orders. Different

simulation sequences can produce different simulation results but the extent to which

simulated results, resulting from different simulation sequences, differ from one another

needs to be investigated. With this consideration, influence of simulation sequence on

model predictions was analyzed. Moreover, the model accuracy is the most important

indicator of whether a model is successfully developed. Simulation sequences directly

affect the accuracy of the model. Therefore it is necessary to evaluate the model

performance accuracy in terms of appropriately selected simulation sequence by

comparing the modified model with well-known models.

To investigate the influence of simulation sequence, it is important to understand

why the simulated results can be different when different simulation sequences are









specified. As discussed in the introduction section of this Chapter, the vertical and

horizontal hydrologic processes are executed in two separate loops in the modified

ACRU2000 model. During the vertical processes water is distributed vertically in the

soil profile, which results in variation of hydraulic heads between land segments, but

each land segment is simulated independently. Horizontal processes transfer surface and

subsurface water between neighboring land segments making them interdependent. The

simulation of horizontal processes on a particular land segment causes a change of water

levels in adjacent land segments. These influences can be propagated to downstream

land segments that are simulated later in the simulation sequence. Therefore, it is

essential to investigate the difference between different simulation sequence strategies

before a decision is made on what would be the most appropriate sequence for the model.

Table 3-1. List of strategies for water transmission and water storage updating for
simulation sequence experiments.
Experiments Strategies to transfer water and update water storage

Water transferred in lateral processes is not available for transmission to
Experiment 1 downstream land segments and also does not cause head changes until the
end of the day when all land segments have been simulated.
Water transferred in lateral processes is available for transmission to
Experiment 2 downstream land segments but does not cause head changes until the end
of the day when all land segments have been simulated.

Experiment 3 Water transferred in lateral processes is available for transmission to
downstream land segments and also causes heads changes immediately.
* Water storage either on the ground surface or in the saturated soil layers.

Three experiments, corresponding to different strategies for water transmission and

storage updating, are listed in Table 3-1. These experiments were conducted based on

two hypothetical scenarios, a flat rectangular plane and an axisymmetric domain. To

further evaluate the model performance accuracy, the simulated results from the first

scenario were compared with the two-dimensional overland flow model in MIKE SHE









(DHI, 2004) and the results from the second scenario were compared with the three-

dimensional groundwater flow models in both MIKE SHE (DHI, 2004) and MODFLOW

(McDonald and Harbaugh, 1996).

Case 1: Overland flow along a flat rectangular plane

A flat rectangular plane with a size of 100 m x 2000 m was used to test the

influence of simulation sequence on overland flow in the modified ACRU2000

hydrologic model. The plane was divided into 20 land segments, each with a size of 100

m x 100 m as shown in Figure 3-8. An initial head of 0.5 m was assigned for land

segment 1 (hereafter LS1), 0.4 m for LS2, 0.3 m for LS3, 0.2 m for LS4, 0.1 m for LS5,

and 0.0 m for the rest of land segments so that all with a mild water slope (= 0.001 m/m)

was created from LS1 through LS5 to drive water propagation from the left to the right of

the plane. A closed boundary condition was assumed and a 30-day period was simulated.

The modified ACRU2000 model was first run for the three experiments (Table 3-1)

from LS 1 to LS20, and then the model was run again for the same experiments but with

an opposite sequence, from LS20 to LS1.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Figure 3-8. A rectangular plane with 20 land segments (arrow indicates water movement
direction and digits assigned in each grid cell indicate the number for each
land segment).

The simulated surface water depths from the three experiments each using the two

simulation sequences are compared in Figure 3-9. For Experiment 1, the model produced

identical results for both simulation sequences. This is because the hydraulic gradients

over the domain do not change during one time step for this experiment and water









transmission and water storage updates do not happen until the end of each time step

when all land segments have been simulated. Significant difference in surface water

depths between simulation sequences is observed in Experiment 2. Water moves faster

when the simulation order is consistent with the flow direction (from LS1 to LS20) and it

moves slower when the simulation order is against the flow direction (from LS20 to

LS1). In this experiment the water gradient remains constant during one time step,

however the water storage are updated within the time step which causes more water to

move downstream during the time step if it is available. Differences in simulated water

depths between simulation sequences also occur in Experiment 3 but they are not as

significant as those in Experiment 2. When the water slopes within the domain get

smaller (i.e. on days 20 and 30), the differences between both simulation sequences

diminish. These comparisons show that the simulation order does cause a difference in

surface water depths but the magnitude of the difference depends on the gradient and

storage update strategies.

Figure 3-10 compares the simulated surface water depths along the domain from

the three experiments with those from MIKE SHE's overland flow model for the

simulation sequence from LS1 to LS20. On Day 2, there is a good agreement between

these two models for all three experiments. On day 10 and 20, the modified ACRU2000

hydrologic model moves water faster than MIKE SHE's overland flow model for all

three experiments, although the models agree with each other fairly well on timing and

trends. Statistics calculated by using the predictions from MIKE SHE's model as

observed values, and the predictions from the modified ACRU2000 model as the

simulated values quantify the differences. As shown in Table 3-2, all statistics including










bias, RE, RMSE, CV, R2 and NS are very close to one another among the three

experiments, especially those for bias and RE. For all selected days, Experiment 1 is

better than Experiments 2 and 3 for the statistics RMSE, CV, R2 and NS although the

differences are not that significant. These graphic and statistical comparisons indicate

that for all three experiments the ACRU2000 model compares reasonably well with

MIKE SHE even though where significantly different methodologies and time steps are

used.

Table 3-2. Statistics for the surface water depths predicted by the modified ACRU2000
model and MIKE SHE for the three experiments with the same simulation
sequence.
Statistics* bias RE RMSE CV R2 NS
(units) (m) (-) (m) (-) (-) (-)
Experiment 1 0.0000 0.0000 0.0057 0.0171 0.9984 0.9984
Day 2 Experiment 2 0.0000 0.0001 0.0148 0.0443 0.9899 0.9893
Experiment 3 0.0000 0.0001 0.0087 0.0261 0.9976 0.9963
Experiment 1 0.0000 0.0001 0.0133 0.0396 0.9935 0.9874
Day 10 Experiment 2 0.0000 0.0001 0.0283 0.0845 0.9610 0.9425
Experiment 3 0.0000 0.0000 0.0372 0.1108 0.9604 0.9011
Experiment 1 0.0000 -0.0001 0.0239 0.0714 0.9749 0.9490
Day 20 Experiment 2 0.0000 0.0000 0.0402 0.1197 0.9399 0.8564
Experiment 3 0.0000 -0.0001 0.0495 0.1476 0.9182 0.7818
* Statistics were calculated by assuming the predictions from MIKE SHE as the observed values while the
predictions from the modified ACRU2000 as the predicted values. All three experiments were conducted
with the same simulation sequence from LS1 to LS20.

However, this rectangular plane domain is quite special in that each land segment

only has water exchange with either upstream or downstream land segments and the

simulation sequence either is consistent with or against the flow direction. To further

investigate the influence of simulation sequences, a more complicated domain was used

in the following experiment.







63



Comparison of Surface Water Depths over Land Segments for Experiment 1

0.45
Day2 LS1-LS20
0.36 ....... Day10 LS1-LS20
E -----Day20 LS1-LS20
0.27
.27 n ----- Day30 LS1-LS20
S0.18 A-- o Day2 LS20-LS1
a Day10 LS20-LS1
0.09 x Day20 LS20-LS1
3:: a Day30 LS20-LS1
0.00.
LS1 LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19
Land Segment


Comparison of Surface Water Depths over Land Segments for Experiment 2

0.45
o Day2 LS1-LS20
0.36 ------- Day10 LS1-LS20
E -----Day20 LS1-LS20
5 0.27 -
S.O2 ----- Day30 LS1-LS20

0.18 o Day2 LS20-LS1
'' Day10 LS20-LS1
0.09 x Day20 LS20-LS1
So .'-. .. '-'-"a -----.. a Day30 LS20-LS1
0.00
LS1 LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19
Land Segment


Comparison of Surface Water Depths over Land Segments for Experiment 3

0.45
0o -- Day2 LS1-LS20
0.36 ------- Day10 LS1-LS20
E \ -----Day20 LS1-LS20
E 0.27
a. --- Day30 LS1-LS20
S0.18 o Day2 LS20-LS1
0.18 1-]--'- a
S.-A-'. .. oa Day10 LS20-LS1
0.09 "": x Day20 LS20-LS1
0.00 a_-' --- Day30 LS20-LS1

LS1 LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19
Land Segment



Figure 3-9. Comparisons of the simulated surface water depth from the modified
ACRU2000 model on the three experiments between two opposite simulation
sequences.








64





Comparison of Surface Water Depths over Land Segments on Day 2


0.45

0.36

E 0.27
CL
0.18

0.09-

0.00 L L L L L L
LS1 LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19


, D
0
A o
A \

A'\l~a


MIKE SHE
o Experiment 1 ACRU2000
o Experiment 2 ACRU2000
A Experiment 3 ACRU2000


Land Segment


Comparison of Surface Water Depths over Land Segments on Day 10


0.36

S0.27

0.18

0.09

0.00
LS


0.27

0.18

0.09


0.00
LS1


-MIKE SHE
* Experiment 1 ACRU2000
o Experiment 2 ACRU2000
a Experiment 3 ACRU2000


LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19
Land Segment


Comparison of Surface Water Depths over Land Segments on Day 20





-MIKE SHE
a Experiment 1 ACRU2000
So o Experiment 2 ACRU2000
A Experiment 3 ACRU2000


LS3 LS5 LS7 LS9 LS11 LS13 LS15 LS17 LS19
Land Segment


Figure 3-10. Comparisons of the simulated surface water depths between the MIKE
SHE's overland flow model and the modified ACRU2000 hydrologic model
on the three experiments with the simulation sequence from LS1 to LS20.


-1









Case 2: Overland and groundwater flow over an axisymmetric domain

An axisymmetric domain with its 2D and 3D gridded discretization as shown in

Figure 3-11 was used to test the influence of simulation sequence on both overland and

groundwater flows in the modified ACRU2000 model. The domain, with a size of 600 m

x 600 m, is composed of 100 meter square grid cells, each of which is 2 m in depth. Four

equal-thickness computational soil layers with homogenous soil properties for each soil

layer were assumed for each land segment. The surface elevation for each land segment

was set to 10 m. Homogeneous hydrologic and physical characteristics were assumed for

each land segment including the Manning's roughness coefficient (= 0.1 m1 3/s), soil

layer thickness (= 0.5 m), horizontal/vertical hydraulic conductivity (= 0.0000444 m/s),

specific yield (= 0.25), and storage coefficient (= 0.00005 m-1). An initial water depth of

0.5 m was assigned to the central four land segments, 0.25 m to the land segments

surrounding the central ones, and 0.0 m to the rest of land segments for the overland flow

simulation in this scenario. For the groundwater flow simulations, the previous setting

for initial water levels in each land segment was lowered by a depth of 1.5 m so that no

surface water movement would occur. A closed boundary condition was assumed. A 30-

day simulation period was used for the overland flow simulation and a 2-year simulation

period for the groundwater flow simulation.

The objectives for this case study were to investigate the difference in simulated

results for the three experiments (Table 3-1) in a more complicated domain where land

segments may interact spatially with one another in multiple directions, and compare the

simulated surface water depths with those from MIKE SHE (DHI, 2004) and

groundwater table depths with those from MIKE SHE (DHI, 2004) and MODFLOW

(McDonald and Harbaugh, 1988). A simulation sequence, which transfers water from the









center to boundary of the domain symmetrically following the flow direction, was

specified. An hourly time step was used for the 2D overland flow model in MIKE SHE,

a varying time step ranging from 2 to 12 hours for the 3D groundwater flow model in

MIKE SHE, and a daily time step for the groundwater flow model in MODFLOW and

the modified ACRU2000 model. The model was run separately with different initial

water level settings so that both surface overland flow and groundwater flow can be

investigated individually.


J LO .3 L 3U ] Z -f \
31 32 33 34 35 36 0 i W


Figure 3-11. Schematics of two-dimensional axisymmetric domain (left) and its three-
dimensional discretization (right). The digits assigned in the left diagram
indicate the number for each land segment.

For the overland flow simulation, the simulated surface water depths from LS 19,

LS20, LS21, LS25, LS26, and LS31 were selected for analysis because of the

axisymmetric distribution of water depths over the domain. For each selected land

segment, the simulated water depths from the three experiments were compared with

those from MIKE SHE's overland flow model as shown in Figure 3-12. Similar to the

conclusions drawn from Case 1, the comparisons from all three experiments in all

selected land segments indicate that water moves slightly faster in the modified

ACRU2000 model than in MIKE SHE. The results from Experiments 1 and 2 show


1 2 3 4 5 6
7 8 9 10 11 12

13 14 15 16 17 18
19 20 21 22 23 24
nr -11- "IT no "n -in








67




Surface Water Depths over Time at Land Segment 19


_C3 13 BB B E BBB BB B




3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)


Surface Water Depth over Time at Land Segment 20


-- MIKE SHE
- -a -Experiment 1
-e Experiment 2
-- Experiment 3


-- MIKE SHE
- -Experiment 1
-o -Experiment 2
-A Experiment 3


1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)


Surface Water Depths over Time at Land Segment 21


0.50 a


0.40


B 0.30 -


0.20


0.10


-- MIKE SHE
---E--- Experiment 1
-- -e-- Experiment 2
-- ---- Experiment 3


1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)



Figure 3-12. Comparisons of simulated water depths of the selected land segments
between the modified ACRU2000 model and the MIKE SHE's overland flow
model for the three experiments.


0.40

0.30
0)
c)

0.20

0.10

0.00


E
_ 0.30
0)
0.20

0.10

0.00








68



Surface Water Depths over Time at Land Segment 25


0.P~
4kar-SP-E- 0f-4nnf--U-l-- Kf


-- MIKE SHE
---5--- Experiment 1
-- -e-- Experiment 2
-- ---- Experiment 3


1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)


Surface Water Depths over Time at Land Segment 26


-- MIKE SHE
---s--- Experiment 1
- ----- Experiment 2
- -- -- Experiment 3


1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)


Surface Water Depths over Time at Land Segment 31


-- MIKE SHE
---.--- Experiment 1
-- -e-- Experiment 2
-- ---- Experiment 3


1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Time (day)


Figure 3-12. Continued.


0.40
-

0.30
Q.

0.20

0.10

0.00


0.40

_ 0.30
a

0.20

0.10

0.00


". 3 : _


0.50

0.40

0.30
Q.

0.20

0.10

0.00


A '
Lj










Table 3-3. Statistics for the surface water depths predicted by the modified ACRU2000
model and MIKE SHE for the three experiments.
Statistics* bias RE RMSE CV R2 NS
(units) (m) (-) (m) (-) (-) (-)
Experiment 1 -0.0015 -0.0111 0.0078 0.0589 0.9139 0.9047
LS19 Experiment 2 0.0000 0.0002 0.0121 0.0909 0.8028 0.7726
Experiment 3 0.0011 0.0082 0.0068 0.0512 0.9315 0.9278
Experiment 1 -0.0138 -0.0861 0.0175 0.1091 0.8224 0.3598
LS20 Experiment 2 -0.0164 -0.1022 0.0205 0.1281 0.7208 0.1187
Experiment 3 -0.0130 -0.0814 0.0161 0.1008 0.8827 0.4535
Experiment 1 -0.0225 -0.1207 0.0292 0.1563 0.9524 0.8059
LS21 Experiment 2 -0.0253 -0.1353 0.0322 0.1725 0.9393 0.7637
Experiment 3 -0.0265 -0.1419 0.0335 0.1794 0.9319 0.7445
Experiment 1 0.0128 0.1092 0.0214 0.1832 0.7289 0.3180
LS25 Experiment 2 0.0141 0.1205 0.0232 0.1986 0.6855 0.1983
Experiment 3 0.0141 0.1204 0.0159 0.1358 0.9322 0.6249
Experiment 1 0.0015 0.0105 0.0221 0.1546 0.4450 -0.1851
LS26 Experiment 2 0.0015 0.0102 0.0210 0.1470 0.4703 -0.0715
Experiment 3 -0.0028 -0.0193 0.0092 0.0642 0.8471 0.7955
Experiment 1 0.0256 0.2555 0.0328 0.3271 0.7377 -0.1002
LS31 Experiment 2 0.0276 0.2759 0.0355 0.3542 0.6569 -0.2900
Experiment 3 0.0290 0.2895 0.0331 0.3307 0.7684 -0.1246
* Statistics were calculated by assuming the predictions from MIKE SHE as the observed values while the
predictions from the modified ACRU2000 model as the simulated values. All three experiments were
performed with the same simulation sequence.

fluctuations at early times of simulation but eventually trend to the results from

Experiment 3 in which the curve of predicted water depths appears smooth. However,

the differences among the three experiments seem to be not that significant. Both models

agree with each other fairly well on timing and trends.

Table 3-3 shows the statistics that were calculated by using the predictions from

MIKE SHE as the "observed" values. The statistics indicate that the model generally

produced better predictions in Experiment 3 than in the other two experiments with

smaller RMSE and CV values and larger R2 and NS values in four selected land segments

(LS19, LS20, LS25 and LS26) and similar predictions in LS21 and LS31 with the other











two Experiments. It should be noted that slightly larger values of bias and RE are

observed in Experiment 3 in some of these land segments. Overall, it was concluded that

the modified ACRU2000 model performed better in Experiment 3 for this scenario and

thus Experiment 3 is the most appropriate strategy for producing reasonable surface water

depths for this axisymmetric domain. To further visualize the comparison of simulated

results for Experiment 3, Figure 3-13 displays a time series comparison of surface water

depths for all selected land segments for these two models.



0.5

-LS19 MIKE SHE -- LS20 MIKE SHE
LS21 MIKE SHE LS25 MIKE SHE
---LS26 MIKE SHE -- LS31 MIKE SHE
0.4 LS19ACRU2000 LS20ACRU2000
S- LS21 ACRU2000 LS25 ACRU2000
I LS26 ACRU2000 LS31 ACRU2000

0.3





0.2
=i i \~."- -- -- - - -
v V"'"--------^'"------------ ---------------



0 /





0.0
1 4 7 10 13 16 19 22 25 28
Time (day)


Figure 3-13. Comparisons of the simulated water depths of the selected land segments
between the modified ACRU2000 and MIKE SHE's overland flow model for
Experiment 3.









A groundwater simulation test was made by changing initial water levels as

discussed earlier. Similar to the above overland flow simulation, the simulated

groundwater table depths also display an axisymmetric distribution pattern over the

domain. Therefore the simulated water table depths from LS19, LS20, LS21, LS25,

LS26, and LS31 were used to compare with the MIKE SHE's predictions for all three

experiments as shown in Figure 3-14. From this figure, it is fairly difficult to tell the

difference among the simulated results from the three experiments for they all have very

similar predictions for all selected land segments. Interestingly, the ACRU results for the

three experiments are consistently located in between the predictions from MIKE SHE

and MODFLOW for all selected land segments, although these results agree with MIKE

SHE's predictions better in LS19, LS20, LS25, LS26, and LS31 and with MODFLOW's

predictions better in LS21. The comparisons also reveal that the ACRU2000

groundwater movement is slightly faster than MIKE SHE and a little bit slower than

MODFLOW. The difference in predicted groundwater table depths between the

modified ACRU200 model and the other two models seems larger in the central land

segments (LS20, LS21, and LS26) than it appears in the land segments along the

boundary (LS19, LS25, and LS31). This could result from the larger hydraulic gradient

variations between the central land segments and milder gradients between the boundary

land segments. In general, three models agree with one another on timing and tendency

quite well.

Table 3-4 shows the statistics that were calculated by using the predictions from

MIKE SHE and MODFLOW as the "observed" values for the three experiments for all

selected land segments. Over all selected land segments, there are no significant









differences among the bias, RE, RMSE, CV, and R2 values. However, some systematic

differences in the NS values are noticed between the predictions from MIKE SHE and

MODFLOW as the "observed" values: excellent NS values close to 1 for LS19, LS25,

and LS31, fairly good NS values for LS21 and LS26 and poor NS values were obtained

for LS20 when comparing to MIKE SHE's predictions. On the other hand, the NS values

vary gradually for all selected land segment from excellent to fairly good when

comparing to MODFLOW's predictions. Thus it is difficult to differentiate the model

performance among the three experiments and thus conclude unequivocally which

experiment provides the most appropriate strategy.

From the previous case studies for two different hypothetical scenarios, it can be

concluded that the simulation sequence affects the simulation results, but not very

significantly. Thus picking a simulation sequence that follows the flow directions

determined by the topography on the ground surface is reasonable. Moreover, among the

three alternative water transmission and water storage updating strategies, Experiment 3

appears to be the most appropriate one for reasonable simulation of overland flow but no

significant difference was observed in simulation of groundwater flow. By comparing to

results from MIKE SHE and MODFLOW, the modified ACRU2000 model has been

shown to simulate hydrology reasonably well.








73




Groundwater Table Depths over Time at Land Segment 19


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709


- MIKE SHE
- MODFLOW
--Experiment 1
--Experiment 2
- Experiment 3


Groundwater Table Depths over Time at Land Segment 20


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709


- MIKE SHE
- MODFLOW
- Experiment 1
--Experiment 2
- Experiment 3


Groundwater Table Depths over Time at Land Segment 21


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709


- MIKE SHE
- MODFLOW
--Experiment 1
Experiment 2
- Experiment 3


Figure 3-14. Comparisons of the groundwater table depths at the selected land segments
between and the modified ACRU2000 hydrologic model and MIKE SHE's
groundwater flow model for the three experiments.


E 1.10

0 1.20
a)
1.30
0-

u 1.40

1.50


E 1.10

) 1.20
a)
S1.30
I-

u 1.40

1.50


1.00

1.10
-c

) 1.20
-

S1.30

S1.40
-:








74





Groundwater Table Depths over Time at Land Segment 25


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709
1.00

E 1.10
t-
I 1.20

S1.30
I-

' 1.40

1.50


Groundwater Table Depths over Time at Land Segment 26


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709


E 1.10-
a-

e 1.20
0-)
u 1.30
I-

1.40

1.50


- MIKE SHE
-MODFLOW
--Experiment 1
Experiment 2
Experiment 3


- MIKE SHE
-MODFLOW
SExperiment 1
SExperiment 2
SExperiment 3


Groundwater Table Depths over Time at Land Segment 31


Time (day)
1 60 119 178 237 296 355 414 473 532 591 650 709


- MIKE SHE
-MODFLOW
--Experiment 1
--Experiment 2
Experiment 3


Figure 3-14. Continued.


E 1.10

g 1.20

c 1.30

1.40

1.50


~ ~45C5=F


- -














Table 3-4. Statistics for the simulated groundwater table depths by the modified ACRU2000 model, MIKE SHE, and MODFLOW for
the three experiments with the same simulation sequence.
Bias RE RMSE CV R2 NS
Statistics
[1]" [2]b [1] [2] [1] [2] [1] [2] [2] [1] [1] [2]

(units) (m) (m) (-) (-) (m) (m) (-) (-) (-) (-) (-) (-)
Experiment 1 -0.0010 0.0118 -0.0007 0.0084 0.0019 0.0123 0.0014 0.0020 0.9982 0.9932 0.9968 0.8774
LS19 Experiment 2 -0.0012 0.0116 -0.0009 0.0082 0.0021 0.0120 0.0015 0.0019 0.9982 0.9932 0.9963 0.8825
Experiment 3 -0.0012 0.0116 -0.0008 0.0082 0.0021 0.0121 0.0014 0.0020 0.9982 0.9932 0.9964 0.8808
Experiment 1 0.0113 -0.0163 0.0090 -0.0127 0.0140 0.0173 0.0112 0.0031 0.9979 0.9774 -1.3373 0.3915
LS20 Experiment 2 0.0111 -0.0166 0.0088 -0.0129 0.0139 0.0175 0.0110 0.0031 0.9979 0.9778 -1.2802 0.3763
Experiment 3 0.0115 -0.0162 0.0091 -0.0126 0.0142 0.0171 0.0113 0.0031 0.9977 0.9782 -1.3765 0.4006
Experiment 1 0.0289 -0.0181 0.0258 -0.0155 0.0323 0.0205 0.0289 0.0040 0.9988 0.9998 0.5957 0.9247
LS21 Experiment 2 0.0288 -0.0182 0.0257 -0.0156 0.0322 0.0206 0.0287 0.0041 0.9987 0.9998 0.5990 0.9237
Experiment 3 0.0293 -0.0177 0.0261 -0.0152 0.0326 0.0203 0.0291 0.0040 0.9987 0.9998 0.5890 0.9265
Experiment 1 -0.0004 0.0101 -0.0003 0.0070 0.0010 0.0104 0.0007 0.0017 0.9992 0.9987 0.9984 0.8414
LS25 Experiment 2 -0.0007 0.0098 -0.0005 0.0068 0.0011 0.0102 0.0008 0.0016 0.9992 0.9987 0.9980 0.8482
Experiment 3 -0.0006 0.0099 -0.0004 0.0069 0.0011 0.0103 0.0007 0.0016 0.9993 0.9987 0.9981 0.8456
Experiment 1 0.0104 -0.0155 0.0080 -0.0116 0.0113 0.0160 0.0086 0.0027 0.9986 0.9808 0.7156 0.6420
LS26 Experiment 2 0.0102 -0.0157 0.0078 -0.0117 0.0111 0.0162 0.0085 0.0028 0.9986 0.9806 0.7255 0.6328
Experiment 3 0.0106 -0.0153 0.0081 -0.0115 0.0114 0.0158 0.0087 0.0027 0.9986 0.9805 0.7110 0.6473
Experiment 1 -0.0015 0.0130 -0.0010 0.0089 0.0017 0.0150 0.0012 0.0023 0.9996 0.9937 0.9840 0.5285
LS31 Experiment 2 -0.0016 0.0129 -0.0011 0.0088 0.0019 0.0149 0.0013 0.0023 0.9995 0.9940 0.9810 0.5382
Experiment 3 -0.0016 0.0129 -0.0011 0.0088 0.0019 0.0149 0.0013 0.0023 0.9995 0.9940 0.9810 0.5383
a statistics were calculated using the predictions by MIKE SHE as the observed values and the predictions by the modified model as the simulated values.
b statistics were calculate using the predictions by MODFLOW as the observed values and the predictions by the modified model as the simulated values.










Application at Dry Lake Dairy #1, Kissimmee River Basin, Florida

Site description

The Dry Lake Dairy #1 is an agricultural pasture site located within the lower

Kissimmee River and Taylor Creek-Nubbin Slough Basins (Figure 3-15). The site is

located on Immokalee soil with surface depressions and little slope, thus the pasture often

becomes flooded during the rainy season. The spodic horizon for Immokalee fine sand

typically occurs at approximately 90 cm and averages 50 cm in thickness. The upper

zone of the spodic horizon is black and weakly-cemented, while the lower zone is a

mottled, dark reddish brown and is even more weakly cemented than the upper layer.

During the wet season, the water table stands near or at the surface for short periods and

recedes to below 120 cm during the dry season (Capece, 1994).




LOCATION#
MAP




LOWER
i KISIMMEE
PO. BASIN
i C C C ECEO 4



SW.F. Rucks Da
L R kTAYLOR CREEK-
NUBBIN SLOUGH

Williamson Ranch


C- Dry* Lak Dairy #
CL LarsS Dairy #6


LAKE 14
OKEECHOBEE


Figure 3-15. Location of Dry Lake Dairy #1 site (Capece, 1994).







77



-a00 -100 0 100 200 300 400 500 600 700 8o00 00 1000 1100
1100 1 0100

10oo 1000

o 00



700 00



6 0 0 -
S4


- fence e boced dtch
400 [00



100















2- ditch weather station
.. major ditch or stream 0 tree or dense vegetation



Figure 3-16. Dry Lake Dairy # 1 site map, topographic survey and location of well
stations and tracer application compound. Distance scales are in feet and
elevation contours are in feet above mean sea level (Capece, 1994).

A topographic map of the Dry Lake Dairy #1 site is shown in Figure 3-16. As

observed, a low, broad berm isolated the study area from the rest of the pasture. The

berm varied in height from 45 to 60 cm above ground surface and was approximately 5 m

across at its base. The total area contained within this berm was 5.9 ha (Campbell et al.,

1995). The flume station was located approximately 200 m from the weather station.

Two supplemental wells were constructed at the weather station for continuous water-









table monitoring. Water-table depth in one well near the flume also was continuously

monitored by the flume datalogger (Campbell et al., 1995). A shallow collection ditch

paralleled the perimeter berm for approximately one third of its length nearest the flume

station. This V-shaped collection ditch was grassed and was less than 30 cm deep and

approximately 3 m wide (Campbell et al., 1995). The constructed collection ditch

discharged into a trapezoidal flow measurement flume with a maximum capacity of 7.1

cfs (200 l/s). A total of 69 well stations were established on this site. Each well station

was composed of two or three wells of different depths.

Weather, runoff and groundwater quality measurements were obtained at the Dry

Lake Dairy #1 site to provide information concerning water and P movement in flat,

sandy, high-water-table soils. Summary data for annual rainfall, ET, and observed runoff

are listed in Table 3-5. These data indicate that years 1989 and 1990 were dry years

when compared to year 1991 with less rainfall and more ET. Year 1991 was relatively

wet with relatively more rainfall. A simple water budget in the last column of Table 3-5

shows year 1989 is the driest, year 1990 ranks the second, and year 1991 is the wettest

among the three years.

Table 3-5. Summary of annual water budget on Dry Lake Dairy #1 site".
Years Rainfall (R) Runoff (RO) ET 100x(ET+RO-R)/R
(cm) (cm) (cm) (o%)
1989 (April-December) 88.66 5.15 91.37 8.87
1990 117.47 25.71 100.30 7.27
1991 133.00 44.29 92.18 2.61
Total 339.13 75.14 283.85 5.86
aData from Tremwel (1992).

The purpose of this application was to validate the modified ACRU2000 model by

comparing with the lumped FHANTM model (Tremwel, 1992), which was applied in the

Dry Lake Dairy #1 site in 1992. For the modified ACRU2000 simulation, the Dry Lake









Dairy #1 site was divided into 4 land segments, each with identical hydrologic properties

and soil layers since no spatially distributed data were available. The discharge from the

site through the flume was estimated using the empirical formula (Tremwel, 1992)

Q= 3.282156h2.5 +0.508387h15 + 0.121409h05 (3-10)

in which Q is the discharge in cubic feet per second and h is the surface water depth in

feet. The measured rainfall, pre-calculated potential ET, and temperature from Tremwel

(1992) were used as the time series input and were applied uniformly for each land

segment. Some input such as soil properties (field capacity, soil porosity, wilting point,

and hydraulic conductivity) as shown in Table 3-6 were adapted from the calibrated

parameters by Tremwel (1992) for FHANTM and the rest of input were estimated. No

deep seepage process was considered in the test. No calibration was performed for the

modified ACRU2000 model.

The simulation period for this validation test was from April 1989 through

December 1991. The model was run throughout the whole simulation period and the

simulated results for surface runoff and groundwater table were compared with those

from FHANTM, for both the calibration period (April 1, 1989 to August 31, 1990) and

the verification period (September 1, 1990 to December 31, 1991). Model prediction

statistics were calculated using the results from both models to quantitatively evaluate the

model performance. Scatterplots of observed vs. simulated values were prepared to

visualize the predicted results from both models.

Results and discussion

The major output variables for this test were the surface runoff through the outlet of

the Dry Lake site and the groundwater table depth, which was taken as the field average