The Arthur R. Marshall Loxahatchee National Wildlife Refuge

The Arthur R. Marshall Loxahatchee National Wildlife Refuge Water Budget and Water Quality Models 1

Jeanne C. Arceneaux Ehab A. Meselhe Michael G. Waldon

Prepared for the US Fish and Wildlife Service, Department of Interior by Institute of Coastal Ecology and Engineering University of Louisiana-Lafayette

Report #LOXA-07-004 June 2007

1

Modified from Arceneaux (2007)

i

Table of Contents

List of Tables ................................................................................................................ v List of Figures ............................................................................................................. viii Chapter 1: Introduction .................................................................................................. 1 1.1 Background ................................................................................................. 1 1.2 Refuge Water Management .......................................................................... 5 1.3 Site Description ........................................................................................... 9 1.3.1 Vegetation....................................................................................... 9 1.3.2 Geology ........................................................................................ 11 1.3.3 Marsh Topography ........................................................................ 11 1.3.4 Canals ........................................................................................... 14 1.4 Objective of Study ..................................................................................... 15 Chapter 2: Literature Review ........................................................................................ 17 2.1 Introduction ................................................................................................. 17 2.2 Everglades Water Budget Modeling............................................................. 17 2.2.1 Lin (1979) ..................................................................................... 17 2.2.2 MacVicar et al. (1984)................................................................... 18 2.2.3 Richardson et al. (1990)................................................................. 19 2.2.4 Welter (2002) ................................................................................ 21 2.3 Previous Modeling Completed in Similar Wetlands ..................................... 22 2.3.1 Kadlec and Hammer (1982) and Kadlec and Knight (1996)........... 22 2.3.2 Mitsch (1988) and Mitsch and Reeder (1991) ................................ 24 2.3.3 Wang and Mitsch (2000) ............................................................... 25 2.4 Everglades Water Quality Modeling ............................................................ 25 2.4.1 Raghunathan et al. (2001).............................................................. 26 2.4.2 Munson et al. (2002) ..................................................................... 27 2.4.3 Fitz et al. (2002a) .......................................................................... 28 2.4.4 Walker (1995) ............................................................................... 29 2.4.5 Walker and Kadlec (2006)............................................................. 29 Chapter 3: Data Collection and Analysis ....................................................................... 31 3.1 Introduction ................................................................................................. 31 3.2 Precipitation................................................................................................. 32 3.3 Evapotranspiration....................................................................................... 37 3.4 Flows........................................................................................................... 39 3.5 Water Levels................................................................................................ 44 3.6 Water Quality .............................................................................................. 45 3.6.1 EVPA Monitoring Sites................................................................. 46 3.6.2 XYZ Monitoring Sites ................................................................... 47 3.6.3 Hydraulic Structures...................................................................... 48

ii

Chapter 4: Water Budget Model.................................................................................... 50 4.1 Introduction ................................................................................................. 50 4.2 Modeling Assumptions ................................................................................ 51 4.3 Model Predictions ........................................................................................ 52 4.4 Observed Parameters ................................................................................... 54 4.4.1 Precipitation .................................................................................. 54 4.4.2 Evapotranspiration ........................................................................ 56 4.4.3 Inflows and Outflows .................................................................... 57 4.5 Estimated Parameters................................................................................... 59 4.5.1 Exchange Flow.............................................................................. 59 4.5.2 Groundwater Recharge .................................................................. 60 4.6 Calibration................................................................................................... 61 4.6.1 Calibration Parameters .................................................................. 62 4.6.2 Calibration Results ........................................................................ 63 4.6.3 Calibration Performance Measures ................................................ 65 4.7 Validation.................................................................................................... 69 4.7.1 Validation Results ......................................................................... 69 4.7.2 Validation Performance Measures ................................................. 71 4.8 Results for Period of Record ........................................................................ 72 4.9 Regulation Schedule Analysis...................................................................... 72 4.10 Discussion of Results................................................................................. 75 4.11 Case Study of Model Application............................................................... 76 CHAPTER 5: Water Quality Constituents, Model Selection, and Modeling Approach .. 84 5.1 Introduction ................................................................................................. 84 5.2 Constituents to be Modeled.......................................................................... 85 5.2.1 Chloride ........................................................................................ 85 5.2.2 Phosphorus.................................................................................... 88 5.3 Model Selection........................................................................................... 93 5.4 Water Quality Modeling Approach .............................................................. 96 CHAPTER 6: Chloride Water Quality Modeling........................................................... 99 6.1 Introduction ................................................................................................. 99 6.2 Chloride Excel Model.................................................................................. 99 6.2.1 Excel Model Setup ...................................................................... 100 6.2.2 Calibration .................................................................................. 102 6.2.3 Calibration Results ...................................................................... 104 6.2.4 Validation Results ....................................................................... 107 6.2.5 Discussion of the Chloride Excel Model...................................... 111 6.3 Chloride WASP Model .............................................................................. 111 6.3.1 Chloride WASP Model Setup...................................................... 111 6.3.2 Chloride WASP Model Calibration ............................................. 115 6.3.3 Chloride WASP Model Calibration Results ................................. 116 6.3.4 Chloride WASP Model Validation Results .................................. 120 6.3.5 Discussion and Further Analysis of the Chloride WASP Model... 123

iii

CHAPTER 7: Phosphorus Water Quality Modeling .................................................... 130 7.1. Introduction .............................................................................................. 130 7.2 Phosphorus WASP Model Setup................................................................ 130 7.3 Phosphorus WASP Model Calibration ....................................................... 132 7.4 Phosphorus WASP Model Calibration Results........................................... 135 7.5 Phosphorus WASP Model Validation ........................................................ 138 7.6 Discussion and Further Analysis of the Phosphorus WASP Model............. 141 CHAPTER 8: Conclusions and Future Developments ................................................. 145 8.1 Water Budget Model Conclusions.............................................................. 145 8.2 Water Budget Future Developments........................................................... 146 8.3 Chloride Model Conclusions...................................................................... 147 8.4 Chloride Model Future Developments........................................................ 148 8.5 Phosphorus Conclusions ............................................................................ 148 8.6 Phosphorus Future Developments.............................................................. 149 Literature Cited .......................................................................................................... 150 APPENDIX A Removed Chloride and Phosphorus Outliers: ...................................... 160 APPENDIX B: Daily Chloride Excel Model Results................................................... 164 APPENDIX C: Daily Chloride WASP Model Results................................................. 169

iv

List of Tables

Table 3.1: Available rainfall data in the Loxahatchee Refuge for the POR (1995 to 2004).................................................................................................................... 33 Table 3.2: Availability of flow data in the Loxahatchee Refuge for the POR (1995 to 2004).................................................................................................................... 43 Table 4.1: Marsh and canal statistics in the Loxahatchee Refuge for the calibration period January 1, 1995 to December 31, 1999 ...................................................... 69 Table 4.2: Marsh and canal statistics in the Loxahatchee Refuge for the validation period January 1, 2000 to December 31, 2004 ...................................................... 71 Table 4.3: Marsh and Canal Statistics for Complete POR ............................................. 72 Table 4.4: Marsh and canal statistics for complete POR (1995 to 2004) using the regulation schedule to predict outflows in the Loxahatchee Refuge....................... 75 Table 4.5: Comparison of the marsh modeled water budget statistics to the ELM v.2.1 model ................................................................................................................... 76 Table 4.6: Comparison of the marsh modeled water budget statistics to the SFWMM model ................................................................................................................... 76 Table 5.1: Total annual chloride loads going in and out of the Refuge through hydraulic structures and the total percent of chloride retained in the Refuge ......... 87 Table 5.2: Total phosphorus loads going in and out of the Refuge through hydraulic structures and the total percent of phosphorus retained in the Refuge.................... 91 Table 5.3: Comparison of the calculated inflow loads against the SFWMD’s loads published in their annual reports for Florida Water Years 2002 to 2004................ 93 Table 5.4: Comparison of the calculated outflow loads against the SFWMD’s loads published in their annual reports for Florida Water Years 2002 to 2004................ 93 Table 5.5: Distance of each cell from the Refuge canal and its area .............................. 97 Table 5.6: Location water quality stations in reference to the canal and interior cells used in calibration of the chloride and phosphorus models.................................... 98 Table 6.1: Initial and long term average concentrations for chloride in each cell......... 102

v

Table 6.2: Chloride Excel model performance measures for the calibration period ...... 107 Table 6.3: Chloride Excel model performance measures for the validation period ....... 110 Table 6.4: Chloride Excel model performance measures for the POR.......................... 110 Table 6.5: Initial volumes for the canal and interior cells............................................ 113 Table 6.6: Fraction of flows used in WASP................................................................ 114 Table 6.7: Areas and distance used to calculate dispersion in the WASP chloride model ................................................................................................................. 115 Table 6.8: Performance measures for the calibration period using the chloride WASP model. ................................................................................................................ 119 Table 6.9: Performance measure for the validation period using the chloride WASP model. ................................................................................................................ 122 Table 6.10: Performance measure for the POR using the chloride WASP model ........ 123 Table 7.1: Initial conditions for phosphorus and the average observed phosphorus concentration for each cell. ................................................................................. 131 Table 7.2: Fraction of flows used in for calculating settling rate for each cell ............. 132 Table 7.3: Performance measures for the calibration period using the phosphorus WASP model...................................................................................................... 137 Table 7.4: Performance measure for the validation period using the phosphorus WASP model...................................................................................................... 140 Table 7.5: Performance measure for the POR using the phosphorus WASP model ..... 141 Table 7.6: Statistics in the canal comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model ....................................................................... 142 Table 7.7: Statistics in the cell 1 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model ....................................................................... 143 Table 7.8: Statistics in the cell 2 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model ....................................................................... 143 Table 7.9: Statistics in the cell 3 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model ....................................................................... 143

vi

Table A.1: Chloride outlier values; and dates and values when there were more than recording ............................................................................................................ 161 Table A.2: Dates and values of days when there were more than one phosphorus reading at a inflow or outflow structure .............................................................. 162 Table A.3: Dates and values of days when there were more than one phosphorus reading at a inflow or outflow structure .............................................................. 163

vii

List of Figures

Figure 1.1: Satellite image and location of the Arthur R. Marshall Loxahatchee National Wildlife Refuge. Inset shows the image location within the State of Florida.................................................................................................................... 1 Figure 1.2: Historic and altered flow patterns for the Everglades system. ....................... 2 Figure 1.3: Boundaries of the Loxahatchee Refuge......................................................... 4 Figure 1.4: Map of Water Conservation Areas (WCAs).................................................. 4 Figure 1.5: Water Regulation Schedule for WCA 1 ........................................................ 8 Figure 1.6: Plant communities located inside the Refuge ................................................. 9 Figure 1.7: Refuge vegetation map. .............................................................................. 10 Figure 1.8: Loxahatchee Refuge 2003 USGS topographic data..................................... 12 Figure 1.9: North to South ground profile of the Loxahatchee Refuge .......................... 13 Figure 1.10: West to East ground profile of the Loxahatchee Refuge............................ 13 Figure 1.11: Location of canals around the perimeter of the marsh ............................... 14 Figure 3.1: Rain gage locations in and around the Loxahatchee Refuge ........................ 33 Figure 3.2: Seasonal variation of average monthly rainfall in the Loxahatchee Refuge for the POR (1995 to 2004) .................................................................................. 35 Figure 3.3: Variation of total annual rainfall in the Loxahatchee Refuge for the POR (1995 to 2004)...................................................................................................... 35 Figure 3.4: Spatial distribution of annual average rainfall in the Loxahatchee Refuge from January 1, 1997, to December 31, 2004........................................................ 36 Figure 3.5: Seasonal variation of average monthly ET at STA-1W for the Loxahatchee Refuge for the POR (1995 to 2004) ...................................................................... 38 Figure 3.6: Annual variation in total ET at STA-1W for the Loxahatchee Refuge for the POR (1995 to 2004)........................................................................................ 38 Figure 3.7: Location of hydraulic structures located in the Loxahatchee Refuge ........... 39

viii

Figure 3.8: Various inflow pump stations located in the Loxahatchee Refuge............... 40 Figure 3.9: Various outflow structures located in the Loxahatchee Refuge ................... 41 Figure 3.10: Various structures with bidirectional flows located in the Loxahatchee Refuge.................................................................................................................. 42 Figure 3.11: Water level sites located in the Loxahatchee Refuge................................. 44 Figure 3.12: XYZ and EVPA water quality monitoring sites located inside the Loxahatchee Refuge ............................................................................................. 46 Figure 3.13: Chloride and TP arithmetic means at Refuge XYZ transect stations with increasing distance from the rim canal.................................................................. 48 Figure 4.1: Sketch of Water Budget double-box model................................................. 51 Figure 4.2: An example of one of the sixteen “Theissen Polygon Method” area distributions used for calculating average daily rainfall in the Loxahatchee Refuge for the POR (1995 to 2004) ...................................................................... 55 Figure 4.3: Canal stages in the Loxahatchee Refuge for the calibration period January 1, 1995, to December 31, 1999 using the water budget model............................... 64 Figure 4.4: Marsh stages in the Loxahatchee Refuge for the calibration period January 1, 1995, to December 31, 1999 using the water budget model............................... 64 Figure 4.5: Canal stages in the Loxahatchee Refuge for the validation period January 1, 2000, to December 31, 2004 using the water budget model............................... 70 Figure 4.6: Marsh stages in the Loxahatchee Refuge for the validation period January 1, 2000, to December 31, 2004 using the water budget model............................... 70 Figure 4.7: Canal stage results using the regulation schedule to predict outflow for the period January 1, 1995, to December 31, 2004 in the Loxahatchee Refuge ........... 74 Figure 4.8: Canal stage results using the regulation schedule to predict outflow for the period January 1, 1995, to December 31, 2004 in the Loxahatchee Refuge...... 74 Figure 4.9: A comparison of the reduction of inflow from STA1-W to the Refuge based on Alternatives 1 and Alternative 2 in respect to Alternative 0.................... 79 Figure 4.10: Comparison of Marsh stages using the water budget model to compare the Alternatives 1 and 2 against Alternative 0 ....................................................... 79 Figure 4.11: Time series of estimated marsh Stages for the three alternatives ............... 80

ix

Figure 4.12: Comparison of Canal stages using the water budget model to compare the Alternatives 1 and 2 against Alternative 0 ....................................................... 80 Figure 4.13: Time series of estimated canal Stages for the three alternatives................. 81 Figure 4.14: The total number of days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives........................... 82 Figure 4.15: The average number of consecutive days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives ........................................................................................................... 83 Figure 4.16: The longest number of consecutive days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives ........................................................................................................... 83 Figure 5.1: Total annual chloride loads going in and out of the Refuge through hydraulic structures .............................................................................................. 87 Figure 5.2: The correlation between the net flow for the POR and the percent chloride retained in the Refuge........................................................................................... 88 Figure 5.3: Schematic explaining how the composite phosphorus samples were filled to make a complete time-series ............................................................................. 90 Figure 5.4: Total annual phosphorus loads going in and out of the Refuge through hydraulic structures .............................................................................................. 91 Figure 5.5: The correlation between the net flow for the POR and the percent of phosphorus retained in the Refuge ........................................................................ 92 Figure 5.6: Location of EVPA and XYZ water quality monitoring sites in relation to the various cells.................................................................................................... 97 Figure 6.1: Schematic of cells used to calculate chloride concentrations...................... 100 Figure 6.2: Canal calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 105 Figure 6.3: Cell 1 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 105

x

Figure 6.4: Cell 2 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 106 Figure 6.5: Cell 3 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 106 Figure 6.6: Canal validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 108 Figure 6.7: Cell 1 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 108 Figure 6.8: Cell 2 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 109 Figure 6.9: Cell 3 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data...................................................................................... 109 Figure 6.10: Canal calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 117 Figure 6.11: Cell 1 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 118 Figure 6.12: Cell 2 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 118 Figure 6.13: Cell 3 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 119 Figure 6.14: Canal validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 120

xi

Figure 6.15: Cell 1 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 121 Figure 6.16: Cell 2 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 121 Figure 6.17: Cell 3 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 122 Figure 6.18: Modeled loads in the canal compared to the observed outflow loads from the canal structures. Solid line is a trendline with forced zero origin generated by Excel .......................................................................................... 124 Figure 6.19: Observed (1/5/1995, to 1/12/1995 plotted without a line) and modeled (1/11/1995 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 125 Figure 6.20: Observed (4/15/1996, to 4/25/1996 plotted without a line) and modeled (4/24/1996 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 125 Figure 6.21: Observed (6/3/1997, to 6/11/1997 plotted without a line) and modeled (6/3/1997 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 126 Figure 6.22: Observed (1/5/1998, to 1/13/1998 plotted without a line) and modeled (1/13/1998 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 126 Figure 6.23: Observed (1/4/1999, to 1/12/1999 plotted without a line) and modeled (1/4/1999 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 127 Figure 6.24: Observed (1/3/2000, to 1/11/2000 plotted without a line) and modeled (1/11/2000 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 127 Figure 6.25: Observed (10/9/2001, to 10/16/2001 plotted without a line) and modeled (10/9/2001 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal ............................................. 128

xii

Figure 6.26: Observed (1/8/2002, to 1/15/2002 plotted without a line) and modeled (1/15/2002 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal............................................................... 128 Figure 6.27: Observed (12/4/2003, to 12/16/2003 plotted without a line) and modeled (12/4/2003 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal ............................................. 129 Figure 6.28: Observed (10/18/2004, to 10/21/2004 plotted without a line) and modeled (10/18/2004 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal ............................................. 129 Figure 7.1: Canal calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 135 Figure 7.2: Cell 1 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 136 Figure 7.3: Cell 2 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 136 Figure 7.4: Cell 3 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 137 Figure 7.5: Canal validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 138 Figure 7.6: Cell 1 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 139 Figure 7.7: Cell 2 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 139 Figure 7.8: Cell 3 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data......................................................................... 140 Figure 7.9: Modeled loads in the canal compared to the observed outflow loads from the canal structures.................................................................................. 142

xiii

Figure B.1: Chloride Excel model results for the canal for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 165 Figure B.2: Chloride Excel model results for the canal for the validation period January 1, 2000, to December 31, 2004 ........................................................... 165 Figure B.3: Chloride Excel model results for the cell 1 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 166 Figure B.4: Chloride Excel model results for the cell 1 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 166 Figure B.5: Chloride Excel model results for the cell 2 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 167 Figure B.6: Chloride Excel model results for the cell 2 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 167 Figure B.7: Chloride Excel model results for the cell 3 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 168 Figure B.8: Chloride Excel model results for the cell 3 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 168 Figure C.1: Chloride WASP model results for the canal for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 170 Figure C.2: Chloride WASP model results for the canal for the validation period January 1, 2000, to December 31, 2004 ........................................................... 170 Figure C.3: Chloride WASP model results for the cell 1 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 171 Figure C.4: Chloride WASP model results for the cell 1 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 171 Figure C.5: Chloride WASP model results for the cell 2 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 172 Figure C.6: Chloride WASP model results for the cell 2 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 172 Figure C.7: Chloride WASP model results for the cell 3 for the calibration period January 1, 1995, to December 31, 1999 ........................................................... 173

xiv

Figure C.8: Chloride WASP model results for the cell 3 for the validation period January 1, 2000, to December 31, 2004 ........................................................... 173

xv

CHAPTER 1: Introduction

1.1 Background

The Arthur R. Marshall Loxahatchee National Wildlife Refuge (hereafter referred to as the Loxahatchee Refuge or simply the Refuge) is a remnant of the Northern Everglades in Palm Beach County, Florida, that once extended to Lake Okeechobee (USFWS, 2000). The Refuge is approximately 143,238 acres (58,000 hectares) and is located seven miles west of Boynton Beach, Florida (Figure 1.1).

0

1.5

3

± 6

9 Miles

Figure 1.1: Satellite image and location of the Arthur R. Marshall Loxahatchee National Wildlife Refuge. Inset shows the image location within the State of Florida. Image adapted from SFWMD (2000a).

1

The Loxahatchee Refuge is part of a large watershed known as the KissimmeeOkeechobee-Everglades System. Historically, the Kissimmee River discharged into Lake Okeechobee, and during wet cycles, the lake would overflow its south bank, providing additional flow to the Everglades (Douglas, 1947; Light and Dineen, 1994). This water would sheet flow across the Everglades, but now, water flows through canals and structures and through a series of water storage areas termed Water Conservation Areas or WCAs (Loucks and McVoy, 2004). Today, the water not used for municipal water supply and irrigation or lost to evapotranspiration is discharged to the Everglades National Park (ENP) and ultimately flows to Florida Bay. Figure 1.2 shows the historic and the current flow condition for the Kissimmee-Okeechobee-Everglades system.

Loxahatchee Refuge

Figure 1.2: Historic and altered flow patterns for the Everglades system. Adapted from the Comprehensive Everglades Restoration Plan website, http://www.evergladesplan.org/index.cfm.

With the 1845 Swampland Act and the 1907 Everglades Drainage Act, excessive drainage occurred in the Everglades to help establish the agricultural industry and

2

encourage urban development in the area. In the late 1940s, the State of Florida in cooperation with the U.S. Army Corps of Engineers (USACE) and other federal agencies planned the construction of three impoundment areas (WCA 1, 2, and 3), bounded by levees and connected by a series of canals, and placed them under the jurisdiction of what is now the South Florida Water Management District (SFWMD) (Johnson, 1974; Light and Dineen, 1994). In the early 1960s, construction of levees and canals circumscribing WCA 1 was completed. In 1951, a license agreement occurred between the SFWMD and the United States Fish and Wildlife Service (USFWS), under the Migratory Bird Conservation Act; the Loxahatchee National Wildlife Refuge was established overlaying Water Conservation Area 1 (WCA 1). The Refuge land is owned by the State of Florida, but it is the responsibility of the USFWS to properly conserve, protect, and manage it (USFWS, 2000).

The Refuge is now hydraulically isolated from the historic Kissimmee-OkeechobeeEverglades Watershed, as it is completely enclosed within a levee system and a borrow canal along the interior of the levee (Richardson et al., 1990). The marsh and interior canal cover the approximately 140,000 acres of WCA 1. The remaining Refuge acreage outside WCA 1 includes land owned by the USFWS, including four management compartments A, B, C, and D (Figure 1.3).

The Refuge is bordered on the northwest by the Everglades Agricultural Area (EAA) and primarily by urban development on the east. WCA 2A is located at the southwest of the Refuge (Figure 1.4).

3

Figure 1.3: Boundaries of the Loxahatchee Refuge. Adapted from USFWS (2000).

Figure 1.4: Map of Water Conservation Areas (WCAs). Adapted from USFWS (2000), courtesy of South Florida Water Management District.

4

1.2 Refuge Water Management

From a historic perspective regarding water control in the Everglades, Light and Dineen (1994) indicated that the WCAs were designed to accomplish eight objectives: 1) receive and store agricultural runoff from the EAA; 2) prevent water accumulated in the system from overflowing into urban and agricultural areas; 3) recharge regional aquifers; 4) prevent salt water intrusion; 5) store and convey water supply for agricultural, municipal and industrial use, and for the ENP requirements; 6) receive controlled releases from Lake Okeechobee; 7) protect wildlife and promote recreation; and 8) dampen the effect of hurricane-induced wind tides by maintaining marsh vegetation in the system.

According to the Comprehensive Conservation Plan for the Loxahatchee Refuge (USFWS, 2000) “the construction of the levees has had significant effects on the hydrology, vegetation and wildlife in the refuge.” The changes in natural timing of water levels affect wading birds’ feeding patterns, apple snail reproductive output, and alligator nesting. Similarly, changes in the spatial distribution of water levels alter the distribution of aquatic vegetation and tree islands. In addition, and particularly during the dry season, lower water levels increase the potential for fire and damage to vegetation, soils, and wildlife. The USFWS (2000) indicated that for consistency with the South Florida Ecosystem Plan, the Refuge should be used to accomplish the following: 1) reduce exotic species; 2) manage water quality and quantity through partnerships; 3) monitor and inventory wildlife and habitats; 4) promote public awareness about the ecosystem; and 5) provide wildlife-compatible recreation.

5

To control the water quantity and timing, the Refuge is managed under a water regulation schedule; the current one was initiated in May 1995 after approximately five years of analysis and negotiation. The Refuge regulation schedule is administered by the USACE (U.S. Army Corps of Engineers, Jacksonville District 1994). The main purpose of the water regulation schedule is to regulate the water level in WCA 1 in order to produce maximum benefits for flood control, water supply, fish and wildlife, and prevention of salt water intrusion. To meet these objectives, water levels in the Refuge are adjusted during the year primarily by releasing water from the Refuge. The Refuge regulation schedule is described in detail in the Comprehensive Conservation Plan for the Refuge (USFWS, 2000) and is summarized below, along with a schematic diagram of the water regulation schedule shown in Figure 1.5. The water regulation schedule is grouped into four zones (Neidrauer, 2004). •

Zone A1 is the flood control zone from January through June. When water levels reach this zone active water releases will be made through the S-10 spillway (and S-39 when agreed between USACE and SFWMD).

•

Zone A2 is the flood control zone from July through December. In this zone, water levels in the Refuge, which are linked with rainfall amounts and the water level at Lake Okeechobee, are permitted to reach a maximum of 17.5 feet (ft) NGVD 29. Excess water is released from the S-10 and S-39 spillways. When additional water is needed for WCA 2A or other areas, it is released from the Refuge depending on relative water level at Lake Okeechobee. If Lake Okeechobee’s stage is above WCA 1’s stage or no

6

more than one foot below, then water supply releases from WCA 1 must be preceded by an equivalent volume of inflow (Neidrauer, 2004). •

Zone B is the water supply zone. Water levels range from a minimum of 14.0 ft NGVD 29 up to a maximum of 17.5 ft NGVD 29. When water levels in the Refuge are within this zone, water releases are allowed, as needed depending on the water level at Lake Okeechobee. If Lake Okeechobee’s stage is above WCA 1’s stage or no more than one foot below, then water supply releases from WCA 1 must be preceded by an equivalent volume of inflow (Neidrauer, 2004). This is the zone considered to be most beneficial to fish and wildlife of the Refuge (USFWS, 2000).

•

Zone C is characterized when water levels drop to 14 ft NGVD 29 or less; when this occurs the Refuge management should not allow any water supply releases. If water supply releases do occur they must be preceded by an equivalent volume of inflow.

According to the USFWS (2000), some of the benefits of the water regulation schedule relative to earlier schedules include: 1) increased hydroperiod of interior marshes to avoid annual dryout; 2) increased water depth during wet years in the northern portion of the Refuge; 3) increased area of interior marsh which serves as nursery areas for aquatic organisms; 4) improvements in timing in winter stage drawdown to benefit wading birds; 5) restoration of deep water habitats suitable for nesting Everglades snail kites; and 6) greater storage within the central and southern Florida project system during wet and normal rainfall years.

7

18.0 A1

A2

B

Water Elevation (feet NGVD29)

17.5 ZONE A1 Active water releases due to flood conditions

17.0

ZONE A2

16.5 Water releases linked to amount of rainfall and water elevation at Lake Okeechobee

16.0 15.5 15.0 14.5

ZONE B Water releases as needed depending on water elevation at Lake Okeechobee

14.0 ZONE C No net water releases due to drought

13.5 13.0 ry ary ua u r n b Ja Fe

h arc M

ril Ap

ay M

ly Ju

e Jun

r r r r st be be be be gu o t u m m m c e A c ve O pte De No Se

Figure 1.5: Water Regulation Schedule for WCA 1. Adapted from USFWS (2000).

There are currently discussions of revising the Refuge regulation schedule to take into account newer data and understanding of hydrological, ecological, and water quality relationships. Analyses supporting management decisions concerning alternative schedule revisions should utilize the models presented in this report.

Along with the changes in water quantity and timing, the changes in water quality are an important threat to the Everglades ecosystem. High concentrations of nutrients (specifically phosphorus) in runoff from agricultural areas cause proliferation of cattails and other undesirable species that negatively affect the ecosystem’s balance. Other negative impacts from increased nutrients include: increased soil phosphorus content; changed periphyton communities; loss of native sawgrass communities; increased

8

organic matter in water; reduced dissolved oxygen; conversion of wet prairie plant communities to cattail; and loss of important habitats for wading birds (Stober et al., 1996).

1.3 Site Description

1.3.1 Vegetation

The Refuge landscape consists of a complex mosaic of wetland communities that grade from wetter areas such as sloughs and wet prairies to sawgrass, brush, and finally tree islands occurring at the dryer end of the scale (Figure 1.6) (USFWS, 2000).

(a) Sloughs

(b) Wet Prairies

(d) Sawgrass

(c) Tree Islands

Figure 1.6: Plant communities located inside the Refuge. Photographs a, b, and c taken by J. Arceneaux; photograph d adapted from USFWS (2007).

9

Sloughs are the deepest natural marsh communities in the Everglades with water depth that may exceed 3 ft in the wet season; slough annual average depth is about 1 foot. In contrast to sloughs, wet prairies have shallow water levels. They are the prevalent vegetative community in the Refuge, with approximately 50 % land coverage (Figure 1.7) (USFWS, 2000).

Figure 1.7: Refuge vegetation map. Adapted from USFWS (2000).

Sawgrass accounts for about 25% of land coverage. It is present on all parts of the Refuge including a vast area on the west side. The tree islands cover approximately 20% of the Refuge interior. They are basically located at the northern portion of the Refuge, ranging in size from less than 1 acre to more than 300 acres (USFWS, 2000).

10

In addition to the aforementioned species, cattails also grow on the Refuge. Cattails are a native species and are naturally found around wading bird colonies, tree islands, and alligator holes. Cattail growth is dependent on nutrients. Excessive cattail growth has occurred along the perimeter of the canal as a response to the anthropogenic load of nutrients in the incoming water (USFWS, 2000). According to Richardson et al. (1990), almost all of the cattails are found within the first 0.621 miles (1000 meters) of the canal, and most remaining cattails are found within the next 0.621 miles (1000 m); Childers et al. (2003) documented additional expansion of cattail in the Refuge. Cattails are more abundant in the west side of the Refuge.

1.3.2 Geology

The Refuge wetland communities occur on top of a bed of peat (Richardson et al., 1990) from seven to nine feet deep (Scheidt et al., 2000; Stober et al., 1996). The peat is lightly colored, fibrous and spongy, and reflective of high organic content (USFWS, 2000; Stober et al. 1996).

1.3.3 Marsh Topography

The Refuge topography is characterized by a fairly flat interior marsh elevation and a varying-section rim canal. The latest marsh elevation data for the Refuge are available from the USGS on a 400 by 400 m grid. The horizontal and vertical data have an accuracy of +/- 15 cm (Desmond, 2003). Figure 1.8 shows the bathymetric contours for

11

the Loxahatchee Refuge based on the USGS’s data. Results of this survey indicate that in the Refuge the bathymetry contours (excluding the rim channel) range from 18.50 to 10.61 ft NGVD 29, with a mean elevation of about 15.17 ft (4.62 m) NGVD 29.

Figure 1.8: Loxahatchee Refuge 2003 USGS topographic data. Adapted from Meselhe et al. (2006).

A North-South profile (Figure 1.9) shows that the Refuge has a very mild north to south slope, which results in a slow southward flow movement of water. Lin (1979) indicates that the flow through the heavily vegetated marsh is slower than the flow in the canals. The North-South slope is estimated to be about an inch per mile. The West-East profile shows mounds and depressions in the terrain, but maintains a relatively horizontal slope (Figure 1.10).

12

18

Ground Elevation (ft NGVD29)

17 16 15 Average Slope = 0.085 ft/mile

14 13

Overland Surface 12 11 10 0

2

4

6

8

10

12

14

16

18

20

22

24

Distance from Point A (miles)

Figure 1.9: North to South ground profile of the Loxahatchee Refuge. Adapted from Meselhe et al. (2005).

18

Ground Elevation (ft NGVD29)

17 16 15 Average Slope = 0.004 ft/mile

Overland Surface

14 13 12 11 10 0

2

4

6

8

10

12

14

Distance from Point C (miles)

Figure 1.10: West to East ground profile of the Loxahatchee Refuge. Adapted from Meselhe et al. (2005).

13

1.3.4 Canals

The Refuge is bordered by the L-7 and L-39 Canals to the west and south and by the L-40

L-7 Ca na l

Canal on the east (Figure 1.11).

al an 0C L-4

oro lsb Hil

Cana l (L-3 9)

0

1.5

3

6

9 Miles

±

Figure 1.11: Location of canals around the perimeter of the marsh.

All the water that is pumped into the Refuge goes into these canals and some of this water moves through the canals around the perimeter and leaves the Refuge through the southwestern and eastern structures. The rim canal bathymetric data were collected by the University of Florida’s Institute of Food and Agricultural Sciences (Daroub et al., 2002). For the western canals, the sediment surface elevations range between 7.0 and -1.5 ft NGVD 29 and between 6.7 and -5.7 ft NGVD 29 for the L-40 Canal. The top width ranges between 120 and 205 ft for the western canals, and between 88 and 173 ft for the L-40 Canal. It is noted that modeling of sheet flow and water surface levels in the wetlands of South Florida is very sensitive to changes in elevation due to the expansive 14

and extremely low relief terrain. Therefore, vertical accuracy on the order of +/- 15 centimeters is required for the elevation data to be used as input to hydrologic models (Desmond, 2003).

1.4 Objective of Study

According to the Comprehensive Conservation Plan for the Refuge (USFWS, 2000), “Water quality, quantity and delivery timing affect the welfare of fish, wildlife and plants… Because the Everglades is no longer a free-flowing system that relies on temporal weather patterns to sustain it, humans must now attempt to provide water when and where the system can most benefit.” The Refuge is impacted by elevated concentrations of nutrients, particularly phosphorus, in pumped stormwater (Harwell et al., 2005; USFWS, 2007b). Such nutrients enhance the growth of non-indigenous and invasive species to the detriment of native species (USFWS, 2000). It is a priority for the Refuge to better understand and minimize the impacts of this excessive nutrients loading. Hence, the goal of this modeling effort is to provide a quantitative framework for management decisions related to Refuge inflow and outflow quantity, timing, and quality. Therefore, this report will present the methodology and results of simple water budget and water quality models for the Refuge, which has the potential of providing the needed management and scientific support related to these concerns. The simplified modeling presented here is part of a larger project that will also develop more complex, 2dimensionsional models of hydrology and constituent fate and transport.

15

When fully calibrated and validated, the water budget and water quality models should assist in answering questions and provide information such as those listed below (Brandt et al., 2004). • What is the impact of different management scenarios on the water distribution inside the Refuge? • Which management scenarios will cause portions of the Refuge to dry out and for how long? In other words what is the impact of the management scenarios on the hydroperiod? • Does the water depth (duration and frequency) satisfy the needs of plants and wildlife? • What are the spatial and temporal distributions of phosphorus levels within the Refuge? • What are the impacts of management decisions and strategies on the water quality? • What are the impacts of alternative regulation schedules on the water quantity and quality in the Refuge? • What are the effects of the surface-groundwater interactions on the Refuge? • How does the surface and ground water interact in the Refuge?

16

CHAPTER 2: Literature Review

2.1 Introduction

There have been various noteworthy efforts devoted to modeling the hydrology and water quality of the Loxahatchee Refuge, alone or as a part of the greater Everglades. A great deal can be learned from these models, but none of them meet the current management needs for the Loxahatchee Refuge. This chapter briefly covers some of these modeling efforts. Some similar modeling techniques that were used in the completion of this report are also discussed.

2.2 Everglades Water Budget Modeling

2.2.1 Lin (1979)

Lin (1979) adapted and modified the Receiving Water Quantity Model to model the WCAs in order to investigate the hydraulic impact of additional inflow under different pumping scenarios. Lin (1979) modeled WCAs 1, 2A, and 3A with 20 link-nodes each. The network system for WCA 1 contained 20 nodes and 57 channels. The calibration of the model was based on a comparison of predicted and observed stages at selected gages. The model was calibrated for the year 1974 and was later applied to the period 1962 to 1973. For WCA 1, modeled and observed values at gages S-6, 1-8, 1-7, and 1-9 were compared. Gages S-6 and 1-8 are located in the existing canal system, while 1-7 and 1-9

17

are located in the central marshland of the Refuge. For the validation period, important deviations were observed between the model results and the measurements. The deviation for interior gages was far less than that seen in the canal system. Lin (1979) recommended that the number of nodes in the network system should be increased in order to provide a better representation of the real water body. Neither groundwater nor water quality were modeled.

2.2.2 MacVicar et al. (1984)

MacVicar et al. (1984) presented the application of the South Florida Water Management Model (SFWMM) to two planning areas, the Lower East Coast (LEC) and the Upper East Coast (UEC). The WCA 1 was included in the LEC model that also included the other WCAs, the Everglades Agricultural Area, and some other nearby areas. A two by two mile node spacing was used to cover the 6,880 square mile area modeled. A time step of one day was used. The model was able to simulate overland, channel, and groundwater flow. McVicar et al. (1984) indicated that simplified mathematical formulations were implemented in order to make the model computationally efficient. For example, the canal routine developed for this model was a mass balance procedure that sums all the inflows and outflows of a canal to determine the water surface position at the end of each day. The canals were defined as continuous channel reaches with flow control structures at the upstream and downstream ends. The overland flow was simplified using a diffusion flow approximation based on Manning’s equation. According to MacVicar et al. (1984), the model did simulate regional flooding in undeveloped areas, and also

18

indicated excessive groundwater drawdowns when they occurred, although it was unable to provide detailed flood routing results for single events or define detailed depression cones around municipal wells.

MacVicar et al. (1984) indicated that the period 1969 to 1971 was chosen as the calibration period, and the period of 1973 to 1975 was selected as validation period. The investigators reported a good agreement between simulated and recorded water levels at two gages in WCA 1 (gages 1-8 and 1-7). They reported that evapotranspiration and overland friction losses were the two major calibration parameters. Water quality and mass transport were not simulated during this study.

The SFWMM continues to be developed and its period-of-record for simulation was extended in order to support water resources management in the South Florida area (SFWMD, 2003). A companion model, the Natural Systems Model (NSM), also continues in development. The NSM is essentially the SFWMM with human alterations of the system (e.g., canals, levees, and water control structures) removed, and topography restored to an estimate of pre-development conditions.

2.2.3 Richardson et al. (1990)

Richardson et al. (1990) studied the distribution of water over space and time and how vegetation was being structured on the Refuge by hydroperiod pattern. A hydrologic model was developed to better understand the hydrologic characteristic of the Refuge.

19

For this task, topographic data and water depths were gathered and the percent covered by each vegetation class was recorded. A flat pool of water in the Refuge was obtained by holding water at the 17 foot level during the time that the grid survey was being conducted. Marsh surface elevations were determined by subtracting measured water depths at each of the grid locations from an assumed horizontal water level.

A hydrologic simulation model was constructed utilizing the Adaptive Environmental Assessment Everglades Simulation Model (AEA Everglades Model) developed by Carl Walters (Walters, 1990; Tait, 1990). Some modifications were made to the AEA Everglades model to make it applicable to the Refuge; some of these modifications included reducing the cell size, adjusting Manning’s roughness coefficient, tagging cells located around the edge as canal cells, and using data from the Refuge. The stage of the rim canal was not modeled, but rather inputted as a boundary. The input and output to the canal were controlled using the historic monthly canal levels (data from SFWMD) by adjusting the water depths in canal cells.

A sixteen-year period, 1970 through 1985, represented the standard base run of the model. The simulations were compared to two stage stations, 1-7 and 1-9. Observed data indicated that, during the 192 month time period, there were 33 and 11 months of recession at the 1-7 and 1-9 gages, respectively. With water depths smaller than 0.075 feet set as dry, it was predicted that there were 30 months of drawdown at the 1-7 gage and 14 months of drawdown at the 1-9 gage. The model slightly underestimated the 16year hydroperiod for gage 1-7 and slightly overestimated the hydroperiod at gage 1-9.

20

The model was later used for approximating spatial hydroperiods for the Loxahatchee Refuge. The 16 year hydroperiod over the entire Refuge ranged from 70% to 98% (wet period over total period) exhibiting an obvious north-south trend of increasing hydroperiod with localized anomalies corresponding to topographic features. Mean water depth for the study period ranged from 0.2 ft in the north to 3.2 ft in the south. It was found that the north end of the Refuge had much greater variance in hydroperiod than the south end. Richardson et al. (1990) stated that during dry years, the north end of the Refuge is much more susceptible to staying dry for long periods, while in the south the dry season is not as likely to completely dewater the marsh for months at a time.

2.2.4 Welter (2002)

Welter (2002) used the Regional Simulation Model (RSM) to simulate the hydrology of the Loxahatchee Refuge. The model used a grid with 16,292 triangular cells with average element size of 650 ft. Overland, canal, and groundwater flows were modeled. Welter (2002) expressed that the groundwater portion of the model was simplified as much as possible, because the overland processes seemed to be more important.

The RSM was calibrated over the period of record of 1988 to 1990, and validated for the four-year period, 1991 to 1994. The model results showed the same trends observed in the field measurements. However, some deviations were observed. Welter indicated that “the most disappointing aspect of these results is that measured data shows a larger slope in the canal’s water level than the model calculates.” He attributed this discrepancy to

21

inaccurate cross section data which, according to Welter, overestimated depths. Welter also stated that “the limiting factor in this modeling effort is the sparse network of stage monitoring stations in the Refuge.”

2.3 Previous Modeling Completed on Similar Wetlands

Many water quality constituents can be well modeled using a very simplified formulation (Kadlec and Knight, 1996). Some constituents undergo no significant transformation over their residence time within the modeled system. These constituents may be modeled as conservative substances. That is, they are affected only by transport processes. Disappearance of other constituents may be acceptably modeled using a first-order disappearance rate analogous to a settling velocity (Bowie et al., 1985). Some substances, including total phosphorus and total nitrogen may in some situations be well modeled using a settling velocity with a minimum limiting concentration. The k-c* model of Kadlec and Knight (1996) is an example of such a model formulation. Some simplified models of wetland water quality constituents are briefly surveyed in this section.

2.3.1 Kadlec and Hammer (1982) and Kadlec and Knight (1996)

Kadlec and Hammer (1982) presented a theoretical paper discussing the transport of pollutant in wetland systems. They indicated that water flow in wetlands ecosystems usually occurs in thin-sheet flows at slow rates, which are controlled by the ground slope,

22

water depths, type of vegetation and by the degree and type of channelization. Kadlec and Hammer (1982) indicated that removal rates in wetland systems are fast in comparison to typical biological processes, and can be represented by a first-order reaction. Kadlec and Knight (1996) also suggested nitrogen and phosphorus removal in wetland systems can be approximated by first-order models. They indicated that corrections need to be made to account for non-ideal flow, infiltration, and atmospheric inputs and outputs.

Kadlec and Knight (1996) introduce the k-c* model (Equation 2.1) which is an area based, first order concentration or bacterial die-off model:

d (QC ) = − k (C − C*) dA

(2.1)

where, Q is the volumetric flow rate in m3/day, C is the concentration in g/m3, k is the removal rate constant in m/yr, C* is the background pollutant concentration in g/m3, and A is area in m2. Assuming depth h is constant Equation 2.1 can be written as Equation 2.2:

dhC = − k (C − C*) = − kC + kC * dt

(2.2)

The left side of Equation 2.2 is the derivative of the areal constituent mass, and the final term on the right side in Equation 2.2 is analogous to a constant areal mass loading rate

23

(g/m2/yr). Kadlec and Knight (1996) conclude that this model is appropriate for wetland treatment systems because surface area for constituent removal or for bacterial inactivation does not increase proportionally to water volume as water covers the vegetated zones. Kadlec and Knight (1996) listed some key assumptions of this model. One such assumption is that there are no adaptation trends, as implied by a stationary state for all active wetland storage; therefore the k-c* model cannot predict certain longterm changes. They also assume that the model will not capture any rapid changes. Therefore the k-c* model is best applied when there are intermediate changes or small changes over a long period of time.

2.3.2 Mitsch (1988) and Mitsch and Reeder (1991)

Mitsch (1988) and Mitsch and Reeder (1991) stressed the importance of developing a proper hydrologic model as the first step in producing a productivity and/or nutrient mass balance simulation. Mitsch and Reeder (1991) developed a hydrologic-nutrient removal model to estimate the fate of phosphorus in a wetland area adjacent to Lake Erie (one of the North American Laurentian Great Lakes). The only state variable in the hydrologic model was the volume of water in the marsh, which was affected by rainfall, inflow, evapotranspiration and outflow. The TP model included incoming phosphorus, macrophyte and plankton uptake, and sedimentation and resuspension of phosphorus. The calibration of the TP model was done by varying a resuspension coefficient until the model predicted phosphorus concentrations similar to field data. They also modeled plankton and macrophyte biomass productivity.

24

2.3.3 Wang and Mitsch (2000)

Wang and Mitsch (2000) used a similar model to the one presented by Mitsch and Reeder (1991) for the evaluation of phosphorus dynamics in a created riparian wetlands. The hydrology module was updated to include seepage, and bank storage in the water volume balance calculation, and periphyton community was included in the productivity model. The authors indicated that simulated TP concentrations did not follow observed data well, especially during times where there was no outflow or in low flow periods. They conjectured that it was due to the fact that the model itself is a steady-state lumped model, unable to capture influences of disturbance and random effects such as wind stirring of sediments. The lack of an atmospheric deposition term may have also introduced errors in the phosphorus budget calculations.

2.4 Everglades Water Quality Modeling

Water quality within the Everglades has been a central issue for management of this ecosystem for decades (Richardson, 1990; USFWS, 2000). Models of water quality constituents in the Everglades have been developed to improve understanding and to support management decisions. This section covers some of the models that have been developed to project aspects of water quality within the Everglades ecosystem.

25

2.4.1 Raghunathan et al. (2001)

Raghunathan et al. (2001) developed the Everglades Water Quality Model (EWQM) to predict phosphorus fate and transport in the Everglades. The WCAs and the Everglades National Park (ENP) were included in the model. The output from the SFWMM was used to transport phosphorus between model cells and canals. As in the SFWMM, the model used two-by-two mile grid-cells. A simplified relationship based on a single apparent net settling rate coefficient was used to represent the combined effect of all biogeochemical processes that control the dynamics of phosphorus in the water column. This simplified relationship indicated a net deposition of phosphorus in the sediments. An apparent net settling rate equal to 6.30 m/year was found for WCA 1 during the calibration period. The model was simulated from 1979 to 1989. Model results indicated that the interior of WCA 1 exhibits much lower concentration than actually found in the areas near the rim canal. However, the rim canal was simulated with a single water quality segment without nutrient concentration gradients (the EWQM assumed a constant canal water depth of 3 m). Model results also suggested that reduction of phosphorus concentrations leaving the EAA will result in lower concentrations entering the Everglades National Park (Raghunathan et al., 2001). It was concluded that this model proves to be a good tool for screening the effects of nutrient reduction options in the regional scenario of the EAA-WCAs-ENP system; however, it lacks the level of detail necessary to accurately model the phosphorus dynamics, and the temporal and spatial distribution of water within the Loxahatchee Refuge.

26

2.4.2 Munson et al. (2002)

Munson et al. (2002) developed the Everglades Phosphorus and Hydrology (EPH) model to simulate water movement and phosphorus dynamics in the water that flows from the EAA through WCAs and into the Everglades National Park. The EAA-WCAs-ENP system was modeled as a series of cells with water flowing from one cell to the next, using a monthly time step. In this application, the Loxahatchee Refuge was modeled with only three cells, cell 1 had a surface area of 250 ha representing the rim canal, cell 2 had a surface area of 46,952 ha representing the north-central portion, and cell 3 with 11,734 ha represented the southern part of the Refuge. The hydrologic processes simulated by the EPH model included precipitation, evapotranspiration, inflow and outflow. Total phosphorus in the water column was the only nutrient modeled in this application.

Evapotranspiration parameters and stage-discharge relationship were adjusted during the calibration process to obtain the best results for flows and water surface elevations. The period of record of 1980 to 1988 was used for this purpose. The phosphorus removal rate in each cell was adjusted in order to match simulated and observed concentrations. During the calibration, the average deviations between simulated and observed values for water depths and phosphorus concentrations were 7 and 6%, respectively. The model was recently applied to simulate the impacts on annual average total phosphorus concentrations in each cell as a result of the implementation of the management plan mandated by the Everglades Forever Act. Model results indicate that reductions in input

27

phosphorus concentrations will have little impact on phosphorus concentrations for 85% of the area of the WCAs and on the water entering the ENP.

2.4.3 Fitz et al (2002a)

Fitz et al. (2002a) presented the calibration of the Everglades Landscape Model (ELM) to match the observed data on water stages and total phosphorus concentration in the water column at about 60 point locations distributed throughout the greater Everglades using a 1 km x 1 km square grid. ELM simulates surface, canal, and groundwater flow, but it only considers advective flow (dispersion is not directly modeled). Surface and groundwater flows are solved using a finite difference, alternating direction explicit technique, providing for propagation of water and water-borne constituents across space. The simulation of phosphorus cycles includes uptake, remineralization, sorption, diffusion, and organic soil loss/gain. Sixty gages were used for the calibration of water stages (during the period from 1979 to 1995), but only three gages were located inside the Loxahatchee Refuge (gages 1-7, 1-9 and 1-8T). The water quality data used in the calibration was total phosphorus (TP) concentration sampled in the surface water column during the period from 1979 to 1995. Of 57 monitoring sites, 21 were located inside the Loxahatchee Refuge. A goodness of fit statistic indicated that for water levels, the ELM v.2.1 simulated values explained 68% of variability in observed values. When each simulated and observed depth weighted-seasonal mean surface water TP concentration (at all stations) were compared, simulated values explained more than 50% of variability in

28

observed values (Fitz et al., 2002a). However, differences close to ten orders of magnitude could be found at specific locations.

2.4.4 Walker (1995)

Walker (1995) presented the development of a mass-balance model for predictions of long-term-average phosphorus removal in WCA 2. The model was driven by inflow volumes, precipitation, evapotranspiration, phosphorus loads in the influent and atmospheric deposition, and by a calibrated first-order settling rate. Walker (1995) concluded that a settling rate of 8.9 to 11.6 m/yr was supported by peat-accretion and water column data. He stated that over a long time period, accumulation of phosphorus in plant biomass approaches zero as the ecosystem matures and approaches dynamic equilibrium.

2.4.5. Walker and Kadlec (2006)

The Dynamic Model for Stormwater Treatment Areas (DMSTA) was developed by Walker and Kadlec as an improvement of the total phosphorus models originally used in Everglades stormwater treatment area (STA) design (Walker, 1995). The DMSTA model has been applied to numerous wetlands and wetland treatment systems including STA1W located at the northwest boundary of the Refuge. Walker and Kadlec state that the main goal of DMSTA is to develop and calibrate the simplest, highly aggregated model that could mimic the major features of events driven behavior of treatment wetlands in

29

the runoff environment. DMSTA simulates daily water and mass balances in a user defined series of wetland treatment cells. The model allows a maximum of six cells to be linked in series or parallel. At the present time, DMSTA does not support bidirectional flows. Water balance terms included in this model include inflow, bypass, rainfall, evapotranspiration, outflow, seepage in, and seepage out. This model is coded in visual basic and uses Microsoft Excel as the user interface. DMSTA is an advance over the kc* equation for modeling phosphorus within the STAs. By dynamically incorporating a phosphorus storage state-variable, DMSTA is capable of providing greatly improved projections of the transient behavior of phosphorus in wetlands. The model may be calibrated using the settling rate, k and the c* value determined in the simpler k-c* model. Based on experience in modeling a diverse set of wetland systems, multiple parameter sets are suggested by the authors depending on wetland vegetation type. The calibrated c* value ranged from 4 to 20 µg/L.

30

CHAPTER 3: Data Collection and Analysis

3.1 Introduction

As part of the Everglades, the Loxahatchee Refuge recently is a highly monitored area and could be termed data-rich. Initial modeling efforts were devoted to data identification, compilation, and processing (Meselhe et al., 2005). Many of the datasets are spatially variable, while others are both temporally and spatially variable such as all meteorological, hydrologic, and water quality parameters.

This chapter includes a brief summary of the data collected and analyzed for use in the water budget and water quality models documented in this report. A detailed description of the data acquisition and processing can be found in Meselhe et al. (2005), which describes the selection of periods of record, the sources of the data, the compilation process, and data quality of assurance. Meselhe et al. (2005) also concludes that some additional data would be useful in improving model performance and credibility, and recommends needed additional monitoring.

A ten-year simulation period from January 1, 1995, to December 31, 2004 was selected for this modeling effort. This selection was based on the quality of the data collected during this period, as well as on analysis of the temporally variable data showing significant variability in precipitation and stage over this time period. It should be noted that unless otherwise specified “year” in this report refers to a calendar year.

31

3.2 Precipitation

Rainfall is the predominant type of precipitation in South Florida. Based on data records of varying lengths from a varying number of historical meteorological monitoring stations, Abtew et al. (2005) concludes that South Florida is a high-rainfall region, with an annual average rainfall of approximately 52.8 inches for a period of record from 1900 to 2000. Frontal, convective, and tropical system-driven rainfall events occur within this region.

Daily rainfall data are available at different locations inside and close to the Refuge. There are five weather stations inside the Refuge: S-5A, S-6, S-39, WCA1ME, and LOXWS. One additional SFWMD station is located in the former Everglades Nutrients Removal Project (ENRP), within what is now termed Storm Water Treatment Area 1 West (STA-1W). STA-1W is located adjacent to the northwestern boundary of the Refuge (Figure 3.1).

These six rainfall measurement stations are operated by the SFWMD, and data are available through their Environmental Database website DBHYDRO2. Table 3.1 shows the availability of the rainfall data for the POR.

2

Available at www.sfwmd.gov/org/ema/dbhydro/

32

Figure 3.1: Rain gage locations in and around the Loxahatchee Refuge.

Available Data Available Data 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Start Date End Date S-5A 1/1/1995 12/31/2004 S-6 1/1/1995 12/31/2004 S-39 1/1/1995 12/31/2004 STA1W 1/1/1995 9/30/2004 WCA1ME 2/12/1996 12/31/2004 LOXWS 12/31/1995 12/31/2004 Gage 1 1/1/1997 12/31/2004 Gage 2 1/1/1997 12/31/2004 Gage 3 1/1/1997 12/31/2004 Gage 4 1/1/1997 12/31/2004 Gage 5 1/1/1997 12/31/2004 Gage 6 1/1/1997 12/31/2004 Gage 7 1/1/1997 12/31/2004 Gage 8 1/1/1997 12/31/2004 Gage 9 1/1/1997 12/31/2004 Gage 10 4/1/2000 12/31/2004 Station

Missing Data Days from Available Period Total Continuous 0 0 0 0 32 7 0 0 640 359 216 85 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Rainfall data are available for this period Structure was not in operation during this period

Table 3.1: Available rainfall data in the Loxahatchee Refuge for the POR (1995 to 2004). Adapted from Meselhe et al. (2005).

33

Stations S-5A, S-6, and S-39 have daily average rainfall measurements since 1956, 1960 and 1963, respectively. The weather station WCA1ME has rainfall measurements since 1994, and weather stations LOXWS and ENRP have measurements since 1996.

There are ten additional rain gages located in and near the Village of Wellington adjacent to the Refuge in the ACME Drainage District’s Northern Basin A and Southern Basin B (Figure 3.1). Daily rainfall measurements from these gages are available since January 1997. Gage 10 was added to this rain gage network in April 2000, and its daily rainfall data are available since then. Due to the location of the gages in reference to the Refuge, only Gages 6 to 10, located in Acme Basin B, were used here for analysis.

The Refuge has two distinct seasons, wet and dry (Figure 3.2). The “wet season” runs five months from June through October, and the “dry season” runs seven months from November through May (USFWS, 2000). The “wet season” accounts for 66% of the annual rainfall (Abtew et al., 2005). Accordingly, Meselhe et al. (2005) found that a monthly rainfall analysis for the studied POR indicates that June is the wettest month averaging 7.7 inches, followed by September with 7.5 inches. The driest months for the POR were found to be January and December with 1.8 inches and 1.9 inches, respectively.

Annual (calendar year) total rainfall (Figure 3.3) for the POR shows a steady distribution for the first five years (1994 to 1999) with an annual value of about 58 inches/year. From 2002 to 2004, the annual rainfall dropped below 50 inches, with an average value of

34

about 46 inches/year. A severe drought occurred in 2000 with an annual total equal to 38.9 inches/year. The wettest year during the POR was in 1999 with an annual rainfall total of 59.1 inches/year.

9 8

Monthly Rainfall (inches)

7 6 5 4 3 2 1 0 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Figure 3.2: Seasonal variation of average monthly rainfall in the Loxahatchee Refuge for the POR (1995 to 2004). Adapted from Meselhe et al. (2005).

70

Annual Rainfall (inches)

60 50 40 30 20 10 0 1994

1996

1998

2000

2002

2004

2006

Figure 3.3: Variation of total annual rainfall in the Loxahatchee Refuge for the POR (1995 to 2004). Adapted from Meselhe et al. (2005). 35

The spatial distribution of annual average rainfall in the Loxahatchee Refuge was estimated for a period of record between January 1, 1997 and December 31, 2004 (Figure 3.4). This figure is based on the information of 8 active rain gages during the aforementioned period (S-5A, WCA1ME, LOXWS, S-39, S-6, STA-1W, Gage 8, and Gage 10). This period was selected because gages 8 and 10 started operating on January 1, 1997. As can be observed in Figure 3.4, the northeastern part of the Refuge received more rainfall compared to the other areas. Conversely, the west and southwest received the least amount of rain. The difference between the zones with the highest and the least amount of rainfall is notable. This difference is about 19 inches of rain per year. It is important to note that Meselhe et al. (2005) conducted a thorough evaluation of the rain gages’ data and did not find reasons to avoid the use of any particular gage.

S-5A

STA1W (ENRP) Gage 8

Gage 10

WCA1ME LOXWS S-6

S-39

Figure 3.4: Spatial distribution of annual average rainfall in the Loxahatchee Refuge from January 1, 1997 to December 31, 2004. Adapted from Meselhe et al. (2006).

36

3.3 Evapotranspiration

Rainfall and evapotranspiration (ET) are the main drivers in the hydrologic balance of the Everglades. The balance between rainfall and ET maintain the hydrology system in both the wet and dry seasons (Abtew et al., 2005). According to Abtew et al. (2005) the average annual ET for the Loxahatchee Refuge was approximately 51.1 inches for the years 2003 and 2004.

ET data for the Refuge are available from the ENRP (STA-1W) site, where a lysimeter is used to measure ET. Pan evaporation and potential ET data are also available from station S-5A and LOXWS respectively, but were not used in modeling efforts. These data are available through SFWMD’s Environmental Database, DBHYDRO. The locations of these ET sites can be seen in Figure 3.1.

The seasonal variation of ET was estimated using site STA-1W for the POR (Figure 3.5). As can be observed, ET is higher during the months of March to August with values ranging from 4.5 inches to 6 inches. The average annual ET for the POR from station STA-1W is approximately 52.1 inches with the range being between 49.3 inches and 56 inches (Figure 3.6).

37

7

Monthly Evapotranspiration (inch)

6 5 4 3 2 1 0 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Figure 3.5: Seasonal variation of average monthly ET at STA-1W for the Loxahatchee Refuge for the POR (1995 to 2004). Adapted from Meselhe et al. (2005).

70

Annual Evapotranspiration (inch)

60 50 40 30 20 10 0 1994

1996

1998

2000

2002

2004

2006

Figure 3.6: Annual variation in total ET at STA-1W for the Loxahatchee Refuge for the POR (1995 to 2004).

38

3.4 Flows

There are 19 hydraulic structures located around the perimeter canal, which play an important roll in water management (Figure 3.7). The inflows and outflows associated with these structures are important components of the water budget of the Refuge. The flow data for the Refuge are available through the SFWMD’s Environmental Database, DBHYDRO.

G-310!( !(

L-7 Ca nal

S-5A S-5AS G-301!( !(!(!( G-300 S-362 ACME-1 !( !( G-94D / ACME-2

!(

l na Ca 40 L-

G-251

G-94C

(!

S-6 !( G-338 !( S-10E

(!

oro llsb Hi

(!

4

G-94A

Ca

S-10D !( n!( al (L-3 9) !( !( S-10C S-10A S-39 0 1 2

G-94B

6 Miles

±

Figure 3.7: Location of hydraulic structures located in the Loxahatchee Refuge. Adapted from Meselhe et al. (2005).

Sources of inflow into the Refuge include pump stations, S-6, S-5A, G-310, G-251, S362, ACME-1, and ACME-2 (via gate G-94D). Some of these pump stations are pictured in Figure 3.8. At times, flows can be discharged from S-5A through bypass gates G-300 and G-301 directly into the Refuge. Similarly bypass of the S-6 discharge directly to the

39

Refuge is possible through bypass gate G-338, but such bypass has not occurred since the S-6 diversion. Pump station S-362 began discharging into the Refuge from STA1E in fall 2004. The S-5A station pumps water from the West Palm Beach Canal, while pump stations G-251 and G-310 pump water from the STA-1W, and the pump station S-6 pumps water to the Hilsboro (L-39 canal) (Meselhe et al., 2005).

(a) Pump Station S-362

(b) Pump Station G-251

(c) Pump Station G-310

(d) Pump Station S-6

Figure 3.8: Various inflow pump stations located in the Loxahatchee Refuge. Photographed by J. Arceneaux.

40

Water is released from the Refuge through gated structures S-10E, S-10D, S-10C, S-10A, S-39, G-94C, G-94A, and G-94B. Some of these structures can be seen in Figure 3.9. The S-10 series consists of three spillways, S-10A, S-10-C, and S-10D (S-10B was proposed but was never constructed), which function as flood control gates operated by the USACE. Also included in the S-10 series is S-10E, which consists of three 6 ft diameter culverts, and is operated as an outlet from the Refuge. The S-39 gate is operated to make water supply releases from the Refuge during dry seasons and to also release excess water to the ocean when water is not needed in WCA 2 and WCA 3 (Meselhe et al., 2005).

a) Spillway S-10D

b) Spillway S-39

c) Culvert G-94C

Figure 3.9: Various outflow structures located in the Loxahatchee Refuge. Photographed by J. Arceneaux.

41

Some structures are bidirectional (Figure 3.10), with both inflows and outflows occurring; these structures include S-5AS, G-338, G-301, and G-300 (Meselhe et al., 2005).

a) Spillway S-5AS

b) Spillway G-301

c) Spillway G-300 Figure 3.10: Various structures with bidirectional flows located in the Loxahatchee Refuge. Photographed by J. Arceneaux.

Table 3.2 shows the availability of the data at the various hydraulic structures. Not all 19 structures were in operation during the complete POR. The S-5A pump station discharged into the Refuge until August 1999, when it was diverted to STA-1W. Structures S-5AS and S-6 were diverted away from the Refuge in June 1999 and May 2001, respectively. Structures G-301 and G-300 started operating in August 1999. Structure G-310 started operating in May 1999. During the POR, only one brief inflow event occurred at the G-94C (Meselhe et al., 2005).

42

Operational Dates

Total Operative Daily Days during Average the POR Flow (cfs)

Net Inflow Volume (Ac-ft)

Net outflow Volume (Ac-ft)

Structure

Type of Flow

Type of Flow

Start

End

S-5A

Pump Station

Inflow

1/1/1995

8/26/1999

1698

391.8

1,319,556

0

S-5AS

Spillway

Bidirectional

1/1/1995

6/7/1999

1618

112.8

0

362,004

2.4

9,302

0

G-300

Spillway

Bidirectional

8/26/1999 12/31/2004

1954

G-301

Spillway

Bidirectional

8/26/1999 12/27/2004

1950

28.4

109,845

0

G-310

Pump Station

Inflow

7/7/2000

12/31/2004

1638

411.0

1,335,308

0

G-251

Pump Station

Inflow

1/1/1995

12/31/2004

3652

118.6

859,095

0

S-6

Pump Station

Inflow

1/1/1995

5/15/2001

2326

398.6

1,838,963

0

S-10E

Culvert

Outflow

1/1/1995

12/31/2004

3652

33.4

0

241,937

1/1/1995

5/15/2001

2326

0.0

0

0

G-338

Culvert

Inflow

S-10D

Spillway

Outflow

1/1/1995

12/31/2004

3652

175.9

0

1,274,156

S-10C

Spillway

Outflow

1/1/1995

12/31/2004

3652

146.3

0

1,059,744

S-10A

Spillway

Outflow

1/1/1995

12/31/2004

3652

141.4

0

1,024,250

S-39

Spillway

Outflow

1/1/1995

12/31/2004

3652

184.7

0

1,337,900

S-362

Pump Station

Inflow

9/21/2004 12/31/2004

101

99.2

19,873

0

ACME # 1

Pump Station

Inflow

1/1/1995

12/31/2004

3652

21.4

155,014

0

ACME # 2

Pump Station

Inflow

1/1/1995

12/31/2004

3652

19.8

143,424

0

G-94C

Culvert

Bidirectional

1/1/1995

12/31/2004

3652

38.7*

0

280,329

G-94B

Culvert

Outflow

1/1/1995

12/31/2004

3652

4.7*

0

34,045

G-94A

Culvert

Outflow

1/1/1995

12/31/2004

3652

20.3*

0

147,046

Total

5,790,380

5,761,411

Table 3.2: Availability of flow data in the Loxahatchee Refuge for the POR (1995 to 2004). Adapted from Meselhe et al. (2005).

For the 10 year POR from 1995 to 2004, the yearly total inflow to the Refuge was 579,038 acre-ft, and the yearly total outflow was 576,141 acre-ft. Pumping stations G310, S-6, and S-5A present the highest mean of daily average inflows, with flows averaging close to 400 cubic feet per second (cfs). The maximum daily average discharge was equal to 4,779 cfs through pump station S-5A. Structures S-39 and S-10D had the highest mean daily average outflow from the Refuge with flows close to 180 cfs. The maximum daily average outflow from the Refuge, approximately 4,921 cfs, was from spillway S-10A (Meselhe et al., 2005).

43

3.5 Water Levels

Precipitation, ET, seepage, and surface water management all affect changes in Refuge water levels. There are five continuous recording stations located in the Refuge interior; 1-7, 1-9, 1-8T, Lox North, and Lox South (Figure 3.11). There is an additional site, 18C, which is located in the perimeter canal (Figure 3.11). These data may be obtained from SFWMD’s Environmental Database, DBHYDRO. These Refuge water level sites are currently maintained by the USGS. Sites 1-7, 1-9, and 1-8C have been in operation since 1954, while site 1-8T did not go into operation until 1979. Lox North and Lox South were recently installed in 2001 (Meselhe et al., 2005).

North ^_ 1-7 ^_

1-8T ^_ 1-8C ^_ 1-9 ^_

South ^_

0 1 2

4

±

6 Miles

Figure 3.11: Water level sites located in the Loxahatchee Refuge. Photograph by J. Arceneaux.

For the POR, the arithmetic means of daily average water levels for the interior stations (1-7, 1-8T, and 1-9) range between 16.55 ft and 16.26 ft NGVD 29, and the maximum 44

and minimum daily average stages are 18.12 ft and 13.94 ft NGVD 29, respectively. For gage 1-8C, located in the perimeter canal, the arithmetic mean of daily average water level is 16.31 ft NGVD 29, and the maximum and minimum daily average stages are 18.19 ft and 12.06 ft NGVD 29, respectively. Lox North has an average stage of 16.73 ft NGVD 29, which is higher than the other stations. While Lox South has an average stage of 16.10 ft NGVD 29, which is lower than the other stations (Meselhe et al., 2005).

Other stage data are available at the SFWMD’s Environmental Database, DBHYDRO website for the inflow and outflow structures. It is important to recognize that these water level observations are at times impacted by local influence of structure flows (Lin and Gregg 1988).

3.6 Water Quality

Water quality data for the Loxahatchee Refuge are available from 5 different monitoring efforts: 1) Everglades Protection Area (EVPA) water quality monitoring sites; 2) Enhanced water quality monitoring sites; 3) District Transect monitoring sites, also known as the XYZ sites; 4) water quality monitoring sites located at the hydraulic structures; and 5) additional independent monitoring sites (Harwell et al., 2005; Meselhe et al., 2005). Meselhe et al. (2005) did a complete data analyses for all 5 sources, and based on the period of record from 1995 to 2004, only the data from the EVPA and XYZ monitoring sites (Figure 3.12), and the hydraulic structures were used for modeling. Also, the only constituents analyzed for modeling by Meselhe et al. (2005) were chloride

45

and total phosphorus (TP). The data from the EVPA monitoring sites and from the hydraulic structures are available through SFWMD's environmental database, DBHYDRO, and the XYZ data are available by request from the SFWMD.

Legend # XYZSites !

EVPA

!

!

! !

!

### #

#

## #

# #

!

!

!

!

# !

! ! !

0 1 2

4

!

6 Miles

±

Figure 3.12: XYZ and EVPA water quality monitoring sites located inside the Loxahatchee Refuge.

3.6.1 EVPA Monitoring Sites

There are fourteen EVPA water quality monitoring sites located in the Refuge interior that were active during the POR (Figure 3.12). These stations were designed to monitor the physical, chemical, and biological quality of the Refuge. Most of the constituents are measured monthly; however, the sampling frequency is irregular (Meselhe et al., 2005).

46

For TP, the sample size for the POR varies between 65 and 122 samples, with the arithmetic average TP concentrations varying between 7.3 and 11.8 micrograms per liter (µg/L) (Meselhe et al., 2005).

The chloride data from the EVPA sites were also analyzed. The sample size varied between 41 and 112 data points per site for the POR, with the arithmetic site means ranging between 13.5 and 67.6 milligrams per liter (mg/L). The arithmetic mean over all EVPA sites of chloride concentration during the POR is equal to 31.8 mg/L (Meselhe et al., 2005).

3.6.2 XYZ Monitoring Sites

There are eleven XYZ water quality monitoring sites located inside the Loxahatchee Refuge, with two stations located in the rim canal and nine stations located inside the marsh (Figure 3.12). According to the SFWMD (2000b), these stations were established along a nutrient gradient in the southwestern corner of the Refuge for biological and chemical sampling. Data from these stations are available beginning April 1996.

For TP, the sample size from these sites varies between 107 and 142 values per site, for the POR. The arithmetic means for the POR ranges between 9.0 and 56.5 µg/L. The highest values are at sites located in the rim canal, with the concentrations declining as the distance from the rim canal increases.

47

Chloride data from the XYZ sites were also analyzed for the POR (Figure 3.13). It was found that the sample size varies between 103 and 121 data points per station, with the arithmetic means ranging between 40.4 and 148.6 mg/L. The arithmetic mean of chloride during the POR is 92.7 mg/L. Chloride follows a pattern similar to that of TP, with the concentrations declining as the distance increases from the rim canal. However, the gradient of TP is steeper with TP concentrations decreasing to a fairly constant value of about 10 µg/L within the first 1.5 km; whereas chloride concentrations decrease less rapidly and seem to drop to a relative constant interior value of about 50 mg/L within the first 3.2 km (Meselhe et al., 2005).

Figure 3.13: Chloride and TP arithmetic means at Refuge XYZ transect stations with increasing distance from the rim canal. Adapted from Meselhe et al. (2005).

3.6.3 Hydraulic Structures

As mentioned in Section 3.4, there are 19 hydraulic structures located around the perimeter canal of the Refuge (Figure 3.7). TP data are available from 16 of these sites

48

for the POR; only sites G-338, S-362, and G-94A do not have water quality monitoring data available. Stations S-5A, G-310, and S-6 have both grab samples and composite (usually flow proportional) TP samples, the rest of the stations only have grab samples available. The composite data are for a 7 day period. Data gathered as grab samples had a range of sample size between 81 and 534 samples per site for the POR, with a mean of 177 samples per station. The TP arithmetic means vary between 35.2 and 127.4 µg/L, with the arithmetic mean for all the sites equal to 80.9 µg/L. For the TP data which were gathered using composite samples the range of samples per site was between 160 and 314 for the POR. The TP arithmetic mean varies between 55.2 and 141.5 µg/L.

There are 14 hydraulic stations with data available for the POR. Those that do not have data include G-300, G-301, G-94A, G-338, and S-362. Chloride data for station G-300 and G-301 were assumed to be equal to the S-5A data due to their close proximity. The ranges of sample size of chloride for the POR are between 81 and 218 samples per site, with a mean equal to 129 samples per station. The chloride arithmetic means vary between 49.7 and 148.7 mg/L, with the arithmetic mean for all the sites equaling 113.2 mg/L.

49

CHAPTER 4: Water Budget Model

4.1 Introduction

It is a top priority for the Loxahatchee Refuge to ensure that appropriate water management will produce maximum benefits for flood control, water supply, and fish and wildlife. As mentioned in chapter 1, the main objective of this project is to develop models that will provide quantitative support for making management decisions. Therefore this chapter will cover the water budget model development, calibration and validation, and results.

This water budget model evolved from a previous modeling effort that modeled the water and constituent masses of the Loxahatchee Refuge; this model was developed by Dr. William Walker (W.W. Walker, personal communication, 2004). Notable modifications were introduced in order to fit the management needs of the Refuge. One particular need of the Refuge is to predict the hydroperiods; therefore the model was derived to predict temporal variations of water levels in the canal and marsh based on observed inflows, outflow, precipitation, and evapotranspiration.

The model was implemented using Microsoft Excel with a daily time step. The calibration period was selected as January 1, 1995, to December 31, 1999, and the validation period from January 1, 2000, to December 31, 2004.

50

4.2 Modeling Assumptions

Initial model assumptions were made to insure that the model remained simple, but could still efficiently predict the marsh and canal stages in the Refuge. An initial assumption was made that the model would be implemented using a double-box (2 compartment) model with canal and marsh compartments (Figure 4.1).

P

P

Marsh

Canal

Qin

AC = 996 acres

QMC

AM = 138,325 acres

Qout Marsh Stage = EM

Canal Stage = EC

ET

GC

ET

GM

Figure 4.1: Sketch of Water Budget double-box model.

This setup, like Walker’s, models these two compartments separately, with the only interaction being an exchange flow between the two compartments. This simple modeling technique is reminiscent of the classical hydrological methods of level pool routing (Chow et al., 1988) or cubature (Rantz, 1982).

Other assumptions include: 1) the water surface for both the canal and the marsh are flat; 2) the marsh is characterized by an average soil elevation of 15.16 ft NGVD 29 (4.62 m 51

NGVD 29), which was obtained from the USGS bathymetry data; and 3) the surface area in the marsh and the canal are constant. Initial water levels were assumed to match the observed water levels for the first day of simulation. Therefore the observed water level in the canal (Gage 1-8C) was 17.19 ft on January 1, 1995. The initial water level in the marsh on January 1, 1995 was 17.15 ft, which is the average water level of gages 1-9 and 1-7 on this day.

4.3 Model Predictions

It is important for Refuge management to be able to determine and predict the hydroperiods in the Refuge; therefore, for management purposes it was determined that the best parameter for the water budget model to predict would be the stages in the marsh and in the canal. The following equations were used to determine the canal ( EC ) and marsh ( EM ) stages:

Canal Stage, EC :

dEC (Q − QMC − Qout ) = P − ET − GC + in AC dt

(4.1)

and

Marsh Stage, EM :

dEM Q = P − ET − GM + MC dt AM

52

(4.2)

where EC is the average stage in the perimeter canal in feet, EM is the average stage in the marsh; AC and AM are the perimeter canal and marsh, respectively; P is the precipitation; ET is the evapotranspiration; GC and GM are seepage in the canal and marsh respectively; Qin is the external inflow to the perimeter canal, Qout is the outflow from the perimeter canal; and QMC is the flow from the perimeter canal to the marsh, and vice versa.

The differential equations for canal and marsh stages are simulated using the Euler numerical integration method with a one day time step. This method provides a fast solution and is easily implemented using the available daily average time series data. However one problem with this technique is that when net canal flow is large, stage change over one day is so large that the assumption of “small” change in the integration algorithm is not satisfied. This problem can result in failure of convergence and instability. Here, a heuristic approach is used to stabilize the solution that is otherwise unstable at times. This heuristic approach limits the magnitude of the canal stage, and maintains conservation of water volume by shifting flow directly to the marsh. Such an approach is reasonable because under these conditions flow between the marsh and canal is likely being underestimated by the Euler Method with a daily time step. Denoting the revised stage derivative with an asterisk, this heuristic scheme is as follows:

dEC* dEC = dt dt

when

53

dEC ≤ EC' max dt

(4.3)

and

 dEC  dt = dE dt C dt

dEC*

   E'

C max

when

dEC > EC' max dt

(4.4)

where E 'C max is equal to 0.10 m/day. The additional flow into the marsh, QMC*, is calculated using the following equations:

 dE dE *  Q*MC =  C − C  AC  dt dt  

(4.5)

and

dE*M dt

=

* dEM QMC + dt AM

(4.6)

4.4 Observed Parameters

4.4.1 Precipitation

Observed precipitation (P) data were obtained from the nine gages S-5A, S-6, S-39, STA1W, WCA1ME, LOXWS, Gage 6, Gage 8, and Gage 10 (Figure 4.2). When analyzing

54

the data, it was found that there were a few days during the POR where data was missing; therefore, it was decided that using multiple Thiessen Polygons would provide the most accurate spatial distribution of rainfall over the entire Refuge. In the “Thiessen Polygon Method,” a weight is assigned to each station in proportion to its representative area defined by a polygon (Gupta, 1989); the areas of the polygons were determined using ArcGIS 9. For each day on which data are missing for one or more stations, the areas of the polygons were altered so that the stations with the missing data were not included. There were a total of 16 different scenarios; an example of one of these scenarios can be seen in Figure 4.2. It was found that the average annual rainfall for the POR was approximately 52.1 inches, with the maximum daily and monthly values for the POR are about 6.5 inches and 16.6 inches, respectively.

Figure 4.2: An example of one of the sixteen “Theissen Polygon Method” area distributions used for calculating average daily rainfall in the Loxahatchee Refuge for the POR (1995 to 2004).

55

4.4.2 Evapotranspiration

Evapotranspiration (ET) data was obtained from station STA-1W (ENRP) (Figure 4.2). It has been observed that sites that go dry for even a few weeks out of the year have considerably lower annual ET water losses (German, 1999). Therefore, when the marsh stage approaches the average sediment elevation of 15.16 ft NGVD 29 (4.62 m NGVD 29), the measured potential ET is reduced below the observed value. The observed data were modified using the following equation:

ET = f ET * ETobs





H  H ET

where f ET = Maximum  f ET min , Minimum1, 

(4.7)

   ; f ET min is the minimum reduction of  

ET because of shallow depth = 20%; H is the marsh water depth in feet so that H = Maximum(0, EM − E0 ) ; E0 is the marsh ground elevation = 15.16 ft 29 (4.62 m NGVD 29), the average elevation of the Refuge interior (Desmond 2003; Meselhe et al. 2005); and H ET is the depth below which ET is reduced = 0.82 ft (0.25 m). Using a linear reduction in ET over a small depth range as depth approaches zero is expected to achieve more stable results than simple switching at zero depth. Some other models, including SWAT (Arnold et al., 1998) and MODHMS3 use a similar approach. This approach reduced the average annual ET from 52.1 inches/yr to 46.3 inches/yr for the POR.

3

http://modhms.com

56

4.4.3 Inflows and Outflows

Inflow into the perimeter canal through hydraulic structures S-5A, S-5AS, G-300, G-301, G-310, G-251, G-94C, ACME-1, and ACME-2 (G-94D) were used to create a daily time series for the POR. It was found that the inflow from hydraulic structures accounted for approximately 49.8% of the total inflows into the Refuge, with an annual average of approximately 51.74 inches/yr (830 ft 3/sec or 536 mgd).

Outflows from the rim canal through hydraulic structures S-5AS, G-300, G-301, S-10E, S-10D, S-10C, S-10A, S-39, G-94C, G-94B, and G-94A were used to create a daily time series for the POR. The average annual outflow from the Refuge through the hydraulic structures was found to be 49.4 inches/yr (793 ft 3/sec or 512 mgd).

The water budget model was set up to calculate outflows using the Refuge water regulation schedule as an alternative to using historic values. Stage in the Refuge is controlled through guidance from the current regulation schedule adopted in 1995. The regulation schedule is discussed in detail in Chapter 1 and is summarized by a chart that displays stage-dependent zones, termed Zone A1, A2, B, and C, whose boundaries change throughout the year (Fig. 1.5). In the upper Zone, A1, the S-10 gates may discharge at maximum capacity, and the S-39 may discharge at a rate agreed upon between the Corps and SFWMD. Releases of water out of the Refuge in Zone A1 generally aim at returning the stage at least to the floor of the A1 Zone. In Zone A2, releases are more constrained than in Zone A1, with consideration given to forecasts and

57

stage outside the Refuge boundary. In Zone B, water managers are constrained when providing water supply releases from the Refuge but are given flexibility to otherwise release water as needed for environmental purposes related to the Refuge and downstream ecosystems. In Zone C, the lowest zone, no net water release from the Refuge is allowed (USFWS, 2000).

For calibration of the water budget model, the historic outflow releases were used. However, for modeling alternative scenarios of water management the release of water must also be modeled as a function of modeled Refuge stage. Decisions on water releases from the Refuge depend not only on information that is unavailable within the Refuge model (stages downstream and in Lake Okeechobee, weather forecasts, and water supply needs), but also depend on professional judgment of water managers. Thus, any model of operations under the regulation schedule is challenging and will not precisely reproduce historic values.

Here, water release, Qout , is optionally modeled as a function of position within the regulation schedule zones (Figure 1.5). Regulation schedule zone is determined from canal stage. In Zone C, below 14 ft canal stage, no discharge is assumed. In Zone B, discharge is based on the fraction of Zone B at which the stage is located. Below a threshold position in Zone B, PBO (0.75) , release is zero. Above this threshold, discharge increases linearly with position, PB , to a value representative of the ceiling of the B or A2 Zones and the floor of the A1 Zone, Q A1Floor (1.5 million m3/day). In Zone A1, discharge increases linearly with stage position, PA1 , from the floor of the A1 Zone

58

to a discharge, Q18 (21.5 million m3/day), at a stage of 18 ft (5.4864 m). The equations used to calculate the outflows based on the Refuge water regulation schedule are found below in Equation 4.8.

0 Zone C    Max(0, Q A1Floor ( PB − PB 0 ) /(1 − PB0 ) ) Zone B or A2 QRO =   Q  A1Floor + (Q18 − Q A1Floor ( ET − E A1Floor ) /(5.4864 − E A1Floor )) 

(4.8) Zone A1

4.5 Estimated Parameters

4.5.1 Exchange Flow

The bidirectional flow between the marsh and canal is assumed to be controlled by the stage difference between the two compartments. This was calculated using the “Power Law Model” by Kadlec and Knight (1996). This equation is similar to a weir equation:

QMC = CH 3 ( EC − EM )

where C =

(4.9)

107 BW = 2π 107 B = 5.73 x108 ft −1d −1 ; B is the calibrated transport R

coefficient = 9.14 ft -1d-1; W is the average marsh width = 2.67 x 105 ft; R is the average radius of the marsh (obtained assuming an approximated circular geometry) = 4.27 x 104

59

ft; H = Maximum(0, EM − E0 ) ; EM and EC are the canal and marsh stages, and E0 is the average marsh ground elevation of 15.16 ft NGVD 29 (4.62 m NGVD 29). Although this simple equation is derived for a simpler geometry it appears to adequately describe the bidirectional flows between the marsh and canal.

According to Kadlec and Knight (1996), this equation is applicable for wetlands due to the fact the Manning’s “constant” is not constant for a wetland environment, and using a model such as the “Power Law Model” would also describe the depth variability.

4.5.2 Groundwater Recharge

The rate of groundwater recharge in the canal or marsh is calculated from the head difference between the Refuge and the boundary area (Lin and Gregg 1988). Therefore, the seepage rates were determined using the following equation:

Gi = rseep ( E i − E B )

(4.10)

where i = C or M for canal or marsh, respectively; rseep is the seepage rate constant = 0.042 and 0.0001315 d-1 in the canal and marsh, respectively; and EB is the boundary water surface elevation = 11.48 ft NGVD 29 (3.5 m NGVD 29). The seepage rate constant for the both the canal and the marsh were calibration parameters.

60

It was found that groundwater seepage was of importance to the balance of the Refuge hydrologic system, especially in the canal. Originally in order to maintain simplicity in the model, only one seepage rate was calculated for the entire Refuge, and the stage model calibrated adequately under this simple assumption. Later, it was found that the seepage in the canal was needed to be much larger than that in the marsh in order to explain the annual chloride budget. The significance of canal versus marsh seepage was also adjusted during water quality model calibration.

4.6 Calibration

The model was calibrated using the data for the 5-year period January 1, 1995, to December 31, 1999. Calibration compared the modeled canal and marsh stages to observed stages. Parameters wee adjusted to obtain the best reproduction of the observed data and statistics.

The marsh stage was compared to the average water levels recorded from gages 1-7 and 1-9 located inside the marsh (Figure 3.11). These gages have a long historical record and continuous data over the entire POR therefore they were chosen for comparison against the modeled data. For the marsh area the observed arithmetic mean of daily average water level is 16.45 ft NGVD 29, and the maximum and minimum daily average stages are 18.01 ft and 14.94 ft NGVD 29, respectively.

61

Stage gage 1-8C located inside the canal was used for calibration of the modeled canal stages (Figure 3.11). For gage 1-8C the observed arithmetic mean of daily average water level is 16.33 ft NGVD 29, and the maximum and minimum daily average stages are 18.19 ft and 12.06 ft NGVD 29, respectively. A constraint was set on the canal so that when modeled or observed stage fell below 14 ft, a value of 14 ft was used. This was done because the model is not expected to perform well below14 ft, and the canal stages of interest are at and above 14 ft. This restraint value was set based on the proposed water regulation schedule for the Refuge, which shows the water level in the Refuge should not be allowed to drop below 14 ft NGVD 29. This constraint preserved model simplicity, while permitting model calibration within the stage range of greatest interest.

4.6.1 Calibration Parameters

To calibrate the model certain parameters were adjusted to obtain the best fit and also the best statistics. These parameters include: 1) the transport coefficient (B) in the “Power Law Model,” which was used in calculating the exchange flow; 2) canal and marsh seepage rate constants; and 3) the ET reduction factor.

The major calibration parameter was the transport coefficient (B) used in the “Power Law Model” to calculate the exchange flow between the marsh and canal. It was found that the value equal to 9.14 ft -1d-1 produced the best agreement between observed and predicted values.

62

The seepage rate constants were initially calibrated for the canal and marsh as 0.06 per day and 0.000004 per day, respectively. After the completion of the chloride water quality model it was found that the constituent model was more sensitive to seepage, than the water budget. Therefore, the original calibrated seepage rate constant in the canal was decreased by 30%, to equal 0.042 per day, and the marsh seepage rate constant was proportionally increased to 0.0001315 per day. These values are similar to those found in literature (Linn and Gregg, 1988).

The ET reduction factor was also calibrated to be 20%. The range in which this value was calibrated was based on personal communication with a Refuge employee.

4.6.2 Calibration Results

Figures 4.3 and 4.4 show the graphical comparison between the modeled and the observed canal and marsh stages, respectively, for the 5-year calibration period, January 1, 1995, to December 31, 1999.

63

20 Modeled 19

Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 4.3: Canal stages in the Loxahatchee Refuge for the calibration period January 1, 1995, to December 31, 1999 using the water budget model.

20 19

Modeled Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 4.4: Marsh stages in the Loxahatchee Refuge for the calibration period January 1, 1995, to December 31, 1999 using the water budget model.

64

As seen in Figures 4.3 and 4.4, the observed and modeled values are in good agreement for the calibration period. However, it can also be seen that the marsh results appear to be in better agreement than in the canal. The model was unable to capture some of the low stages seen in the observed data for the canal. One possible reason for this is that there may have been observed outflows that occurred during this period that were not recorded.

4.6.3 Calibration Performance Measures

Statistics are used to evaluate the performance measures that the model is capable of producing. The statistics which were used to evaluate the calibration and validation period of this model and other models presented in this report include: 1) bias; 2) root mean square error (RMSE); 3) standard deviations of the modeled data, observed data, and error between the modeled and observed data; 4) correlation coefficient (R); 5) coefficient of determination, R2; 6) variance reduction; and 7) Nash Sutcliffe Efficiency (Nash Sutcliffe, 1970). These statistics were also used in evaluating the ELM v.2.5 (Fitz et al., 2002a) and SFWMM models (SFMWD, 2003).

1. Bias is the difference between the mean of the model prediction and the mean observed values. Bias is calculated using Equation 4.11 (Montgomery et al., 2001).

Bias = M − O

65

(4.11)

where O is the mean of the observed stages over the entire period of study and M is the mean of the modeled stages over the entire period of study.

2. The standard deviation ( σ ) of the modeled and observed data, as well as the error, also termed the residual, was determined using Equation 4.12 (Montgomery et al., 2001)

σ=

1 N ( xi − x )2 ∑ N i =1

(4.12)

where, x represents either the observed, the modeled, or error between the observed and modeled (error=observed-modeled); x is the mean of the modeled, the observed, or error for the entire period; and N represents the number of values. Standard deviation carries the dimensions of the value being analyzed, in this study the standard deviation of observed, modeled, and the error of the observed and modeled stages is being analyzed, therefore it takes the dimensions of feet.

3. RMSE is a weighted average of the absolute value of the model error; it was calculated using Equation 4.13 (Legates and McCabe, 1999)

 N   ∑ (Oi − M i )  RMSE =  i =1 N

66

2

(4.13)

where Oi represents the observed stage, Mi represents the modeled stages; N is the total number of values. The RMSE value carries the dimension of the parameters being analyzed, in this study they represent the stage therefore RMSE is in ft.

4. Variance reduction is one minus the ratio of the variance of the model residual to the variance of the observed data.

σ  Variance Reduction = 1 −  E  σO 

2

(4.14)

where the σ E is the standard deviation of the error between the modeled and observed; and σ O is the standard deviation of the observed data. Variance reduction is typically represented as a percent. Variance reduction is unaffected by bias, and quantitatively measures how well the model follows variations in observed data.

5. The correlation coefficient (R) measures the linear association between the modeled and observed data. R was calculated using Equation 4.15 and is dimensionless (Legates and McCabe, 1999)

 N  Oi − O M i − M ∑   i =1 R = 0.5 N  N 2   ∑ Oi − O  ∑ M i − M  i=1  i=1

(

(

)

67

)(

)

(

   . 0.5   2    

)

(4.15)

6. Equation 4.16 was used to calculate the coefficient of determination (R2) which represents the square of the correlation coefficient (Legates and McCabe, 1999)

N     Oi − O M i − M ∑   2 i =1 R = 0.5 N 0.5  N 2 2     ∑ Oi − O  ∑ M i − M       i =1   i =1

(

(

)(

)

)

(

2

(4.16)

)

where the parameters are the same as those in Equation 4.15. R2 is dimensionless.

7. The Nash Sutcliffe Efficiency was calculated using Equation 4.17 (Legates and McCabe, 1999)

N

∑ (Oi − M i ) 2

Nash Sutcliffe Efficiency = 1.0 − i =N1

∑ (Oi − O)

.

(4.17)

2

i =1

This value is also dimensionless. Efficiency reflects both model bias and reduction of variance. It therefore has the value of combining these independent criteria into a single goodness-of-fit measure. Efficiency has a maximum value of one, corresponding to a perfect fit. A value of zero indicates that the model predicts no better than simply using the average observed value. Negative efficiency values are often considered to indicate that a model is not useful as a predictive tool. Nash Sutcliffe Efficiency can be problematic when applied to observations with limited variation about their mean value.

68

Marsh Canal Calibration Calibration Statistics Statistics Bias (ft NGVD 29) 0.134 0.026 RMSE (ft NGVD 29) 0.458 0.251 Standard Deviation of Observed (ft NGVD 29) 0.718 0.466 Standard Deviation of Modeled (ft NGVD 29) 0.582 0.536 Standard Deviation of Error (ft NGVD 29) 0.438 0.250 Variance Reduction 62.9 % 71.2 % R (Correlation Coefficient) 0.793 0.885 2 R Value 0.629 0.783 Nash Sutcliffe Efficiency 0.594 0.709 Table 4.1: Marsh and canal statistics in the Loxahatchee Refuge for the calibration period January 1, 1995, to December 31, 1999. Statistical Parameter

The statistics in Table 4.1 show that the observed and predicted stages for the marsh are in better agreement than the observed and predicted values for the canal. Based on the bias results, the model slightly overestimated the observed data in both the canal and marsh.

4.7 Validation

The model was validated for the 5-year period January 1, 2000, to December 31, 2004, using the same calibrated parameters and model setup. The same observation gages were also used to validate the modeled canal and marsh stages.

4.7.1 Validation Results

Figures 4.5 and 4.6 show the validation results for the canal and marsh, respectively. Unlike in the calibration period results, it appears from Figures 4.5 and 4.6 that both the

69

canal and marsh stages are in equally good agreement. The restraint on the observed and predicted canal stage of 14 ft can also be observed in the validation period. 20 Modeled 19

Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 4.5: Canal stages in the Loxahatchee Refuge for the validation period January 1, 2000, to December 31, 2004 using the water budget model. 20 19

Modeled Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 4.6: Marsh stages in the Loxahatchee Refuge for the validation period January 1, 2000, to December 31, 2004 using the water budget model.

70

4.7.2 Validation Performance Measures

The validation period from January 1, 2000, to December 31, was also evaluated using the performance measures discussed in Section 4.6.3. These results can be found in Table 4.2 below.

Marsh Canal Validation Validation Statistics Statistics Bias (ft NGVD 29) -0.165 -0.164 RMSE (ft NGVD29) 0.504 0.270 Standard Deviation of Observed (ft NGVD 29) 0.926 0.490 Standard Deviation of Modeled (ft NGVD 29) 0.836 0.521 Standard Deviation of Error (ft NGVD 29) 0.476 0.215 Variance Reduction 73.5 % 80.7 % R (Correlation Coefficient) 0.859 0.911 2 R Value 0.737 0.830 Nash Sutcliffe Efficiency 0.704 0.695 Table 4.2: Marsh and canal statistics in the Loxahatchee Refuge for the validation period January 1, 2000, to December 31, 2004 Statistical Parameter

As can be seen in Table 4.2, it appears that both the marsh and canal showed good agreement with the observed data. Based on the bias results the model slightly underestimates the observed data in both the canal and the marsh, which is the opposite of the results from the calibration period. Contrary to the calibration results the model appears to have captured the low stages in the canal for this period, confirming the assumption that there were possible observed outflow events that were not recorded.

71

4.8 Results for Period of Record

Performance measures were calculated for the 10-year POR from January 1, 1995, to December 31, 2004 for both the canal and marsh areas (Table 4.3).

Statistical Parameter

Canal Marsh Statistics Statistics Bias (ft NGVD 29) -0.015 -0.069 RMSE (ft NGVD 29) 0.481 0.261 Standard Deviation of Observed (ft NGVD 29) 0.836 0.487 Standard Deviation of Modeled (ft NGVD 29) 0.767 0.562 Standard Deviation of Error (ft NGVD 29) 0.481 0.252 Variance Reduction 66.9 % 73.3 % R (Correlation Coefficient) 0.823 0.895 R2 Value 0.678 0.800 Nash Sutcliffe Efficiency 0.669 0.713 Table 4.3: Marsh and canal statistics for complete POR

For the POR it can be seen from the statistics that the marsh performed slightly better than the canal. Possible reasons for this variation include: 1) the area of the rim canal was assumed constant; 2) the variability of the water levels is stronger in the canal than in the marsh; 3) the emphasis during the calibration was to match the observed marsh stages with the model predictions; and 4) water supply delivery flows through hydraulic structures G-94A, G-94B, and G-94C, prior to 2000, were unavailable and set to zero.

4.9 Regulation Schedule Analysis

The major function of this water budget model is to allow Refuge management to evaluate various scenarios. Therefore, the model was set up with the option of allowing

72

the model to calculate the estimated structure outflow based using the Refuge’s water regulation schedule; this was discussed in detail in section 4.4.3.

Canal and marsh stages were calculated using the regulation schedule to predict outflows. Water supply deliveries through the G-94 and S-39 structures were ignored in this simulation. These results were compared to the observed stages; the results are shown in Figure 4.7 and 4.8, for the canal and marsh, respectively. The simulation was run for the entire POR from January 1, 1995, to December 31, 2004. The performance measures are listed in Table 4.4.

It can be seen from the results that by using the regulation schedule to predict outflows, rather than the historic outflows, the results are in better agreement between the observed and modeled values in both the marsh and the canal. However, it can be seen in Figure 4.7 that the modeled canal stages did not drop like the observed data did, whereas when using the historic data, the modeled stages follow the pattern.

73

20 19

Modeled Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04

Figure 4.7: Canal stage results using the regulation schedule to predict outflow for the period January 1, 1995, to December 31, 2004 in the Loxahatchee Refuge

20 19

Modeled Observed

18

Stage (ft)

17 16 15 14 13 12 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04

Figure 4.8: Canal stage results using the regulation schedule to predict outflow for the period January 1, 1995, to December 31, 2004 in the Loxahatchee Refuge.

74

Statistical Parameter

Canal Marsh Statistics Statistics Bias (ft NGVD 29) 0.005 -0.080 RMSE (ft NGVD 29) 0.458 0.253 Standard Deviation of Observed (ft NGVD 29) 0.836 0.487 Standard Deviation of Modeled (ft NGVD 29) 0.651 0.585 Standard Deviation of Error (ft NGVD 29) 0.458 0.241 Variance Reduction 70.0 % 75.6 % R (Correlation Coefficient) 0.839 0.915 R2 Value 0.704 0.838 Nash Sutcliffe Efficiency 0.700 0.729 Table 4.4: Marsh and canal statistics for complete POR (1995 to 2004) using the regulation schedule to predict outflows in the Loxahatchee Refuge.

4.10 Discussion of Results

The Water Budget Model proved to be credible when analyzed graphically and statistically. The Water Budget Model while remaining simple still computes reasonable canal and marsh stages in the Refuge. Statistically, the modeled marsh stages were in better agreement with observed data than the canal. However, the canal modeling effort also performed well with a Nash Sutcliffe Efficiency of 0.669.

The marsh statistics can be compared to those calculated by Fitz et al. (2002a) using the ELM model (discussed in section 2.4.3). Fitz et al. (2002a) calculated statistics based on the results from the ELM model compared to the observation data at stage gage locations 1-9, 1-8T, and 1-7. Statistics calculated for the SFWWM model (section 2.2.2) were also compared to water budget modeled marsh stages. It should be noted that these models were run for a different POR than the water budget model; however this comparison

75

illustrates a general performance of the results of similar models. The comparison of the water budget model compared to these models can be seen in Tables 4.5 and 4.6.

Water Budget Model Marsh Bias, m R2 RMSE, m Nash Sutcliffe Efficiency

-0.021 0.800 0.079 0.713

ELM v.2.1 Model WL Gage 1-7 0.06 0.73 0.16 0.33

ELM v.2.1 Model WL Gage1-9 0.00 0.72 0.15 0.50

ELM v.2.1 Model WL Gage 1-8T 0.04 0.67 0.23 0.06

Table 4.5: Comparison of the marsh modeled water budget statistics to the ELM v.2.1. model.

Water Budget Model Marsh -0.021 0.800 0.079 0.713

SFWMM Model WL Gage 1-7

SFWMM Model WL Gage 1-9

SFWMM Model WL Gage 1-8T

Bias, m 0.00 0.08 0.11 R2 0.71 0.72 0.73 RMSE, m 0.15 0.17 0.19 Nash Sutcliffe 0.44 0.35 0.35 Efficiency Table 4.6: Comparison of the marsh modeled water budget statistics to the SFWMM model.

4.11 Case Study of Model Application

The water budget model was used to predict stage, and compare Refuge alternatives in the “Everglades Agricultural Area Regional Feasibility Study” (EAARFS), which was initiated by the SFWMD to consider how flows and loads to the Everglades STAs and planned reservoirs might be rerouted to improve treatment performance for removal of total phosphorus (A.D.A. Engineering and SFWMD, 2005). Input flow files were 76

provided by Dr. William W. Walker. The purpose of the analysis presented here is to illustrate the use of the model, and not to provide a definitive analysis of the project alternatives.

The EAARFS considered two major alternatives termed Alternative 1 and Alternative 2, relative to a no project alternative that is termed here Alternative 0. Both Alternatives 1 and 2 reduce the annual volume of inflow to the Refuge relative to Alternative 0. EAARFS modeling used MIKE 11 to model the conveyance canals and Dynamic Model for Stomwater Treatment Areas Version 2 (DMSTA 2) to model the reservoir and STA performance. None of the EAARFS modeling explicitly addressed the effects of alternative inflow volume changes on the “downstream” Everglades marshes that receive the STA discharges, such as the Loxahatchee Refuge (A.D.A. Engineering and SFWMD, 2005). Therefore the simple water budget model discussed in this Chapter was used to determine the effects on the hydroperiods of the Loxahatchee Refuge over the 36 year period from May 1, 1965, to April 30, 2000. The period was analyzed using what is termed as the South Florida Water Management Year, which is from to May 1 to April 30. This period was used to evaluate annual seasonal changes due to the variance in wet and dry seasons.

Alternative 1 diverts inflow from the STA-1W/STA-1E complex primarily by the construction of a pump station and some canal improvements. This alternative diverts a portion of the water now entering STA-1W to other STAs south of the Refuge.

77

Alternative 2 diverts all Refuge inflow from STA-1W by routing the outflow south to other treatment facilities. Thus, both alternatives reduce the volume of inflow to the Refuge, with Alternative 2 having the greater reduction in flow (A.D.A. Engineering and SFWMD, 2005). This reduction in inflow can be seen in Figure 4.9.

By simulating the future effects of the EAARFS alternatives on the Refuge, we have assumed that no water supply deliveries will be provided from the Refuge over the simulation period. Outflows from the Refuge are therefore determined in our simulations based solely on the current Refuge regulation schedule (A.D.A. Engineering, and SFWMD, 2005).

Using the water budget model the stages were calculated for both the marsh and canal areas using the three different alternatives. Figures 4.10 and 4.11 shows a comparison of the resulting marsh stages for Alternatives 1 and 2 compared to Alternative 0; figures 4.12 and 4.13 show a similar comparison for the canal stages.

78

Figure 4.9: A comparison of the reduction of inflow from STA1-W to the Refuge based on Alternative 1 and Alternative 2 in respect to Alternative 0.

Figure 4.10: Comparison of marsh stages using the water budget model to compare the Alternatives 1 and 2 against Alternative 0.

79

Figure 4.11: Time series of estimated marsh stages for the three alternatives.

Figure 4.12: Comparison of Canal stages using the water budget model to compare the Alternatives 1 and 2 against Alternative 0

80

Figure 4.13: Time series of estimated canal stages for the three alternatives. Using the estimated stages in the marsh, the hydroperiods were estimated to determine the number of consecutive days when the Refuge water depth was greater than 0.8 ft. The purpose of determining inundation periods is to provide ecologists and Refuge management a basic understanding of the changes in water level and the affects they have on the wildlife and plants in the Refuge. The average elevation of the Refuge is 15.158 ft (4.62 m) NGVD 29, but the elevation used to calculate inundation periods was 16.0 ft (4.88 m) NGVD 29 due to the fact that when the average Refuge elevation is used the Refuge remains inundated throughout the year.

The inundation periods were analyzed based on the Florida water year from May 1 to April 30 of each year. The total, average, and longest inundation periods were analyzed. The total annual inundation periods refers to the total number of days that the depth of

81

water in the marsh was greater than 0.8 ft (Figure 4.14). The average annual inundation period is the average number of consecutive days that the depth of water in the marsh was greater than 0.8 ft (Figure 4.15). The longest annual inundation period allows the Refuge management to know the longest number of consecutive days were the marsh water depth is greater than 0.8 ft (Figure 4.16).

Figure 4.14: The total number of days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives.

82

Figure 4.15: The average number of consecutive days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives.

Figure 4.16: The longest number of consecutive days when the water depth in the Refuge is greater than 0.8 ft, based on the stage results from the three alternatives.

83

CHAPTER 5: Water Quality Constituents, Model Selection, and Modeling Approach

5.1 Introduction

Along with the changes in water quantity and timing, changes in water quality are introducing negative impacts to the Everglades ecosystem (USFWS, 2000). The Everglades ecosystem is characteristically low in nutrients and is comprised of species that have evolved under these conditions (Childers et al., 2003; USFWS, 2000). Nutrient loading from urban areas and the EAA has significantly increased nutrient concentrations, particularly phosphorus, in the WCAs (USFWS, 2000).

Wetlands respond to nutrient enrichment with characteristic increases in soil nutrients and shifts in plant community compositions (Childers et al., 2003). Among the negative effects from increased nutrients in the Everglades are: loss of native sawgrass communities, conversion of wet prairie plant communities to cattails, invasion of exotic plants, and loss of important habitats for wading birds (USFWS, 2000). Major efforts are being made to reduce the nutrient load entering the Everglades ecosystem, for example the construction of the STAs.

Development of a simple water quality model allows for Everglades’ scientists and managers the opportunity to evaluate the effects of various scenarios and their impacts on the water quality within the Refuge. These individuals can then identify areas of concern and, if necessary, apply a more complex water quality model to gain a more detailed

84

understanding of their impacts. This chapter will present constituents that will be modeled and the corresponding inflow and outflow loads, the model selection process, and the water quality modeling approach.

5.2 Constituents to be Modeled

5.2.1 Chloride

Before the modeling of chloride began a simple mass balance estimating how much chloride was coming into and leaving the Refuge through hydraulic structures was completed. This allowed for a general estimate of how much chloride was apparently being retained in the Refuge. The amount of chloride load retained in the Refuge refers to the amount of chloride that remains in the Refuge, as well as the chloride that may have left through other means of outflow, such as groundwater seepage or transpiration. This simple mass balance was completed for the ten year period January 1, 1995, to December 31, 2004.

Chloride data were downloaded from the SFWMD DBHYDRO database from 14 of the hydraulic structures located around the perimeter of the canal. Chloride data were available at the following structures S-5A, S-5AS, G-310, G-251, S-6, S-10E, S-10D, S10C, S-10A, S-39, ACME-1, ACME-2, G-94C, and G-94B (Figure 3.7). Chloride samples were taken from these locations on a somewhat irregular basis. These data were then filtered and analyzed removing any extreme (outlying) values. When there were

85

dates with more than one recorded concentration, the average of the two was used. A table listing all outlier values that were removed, as well as any dates were multiple concentrations were recorded can be found in Appendix A.

Linear interpolation between known concentrations was used to create a complete daily time series at each structure. The hydraulic structures that had no data (G-300, G-301, and G-94A), or a limited number of recorded concentrations (G-94C), used the data from nearby stations. For example, G-300 and G-301 used the data recorded from S-5A; and structures G-94A and G-94C used the data recorded from structure G-94B.

Once the chloride concentration time-series had been constructed for all stations, the chloride load at each hydraulic station was able to be calculated using Equation 5.1:

Q * C = Load

(5.1)

where Q is in m3/day and C is in kg/m3 ; resulting in a Load in kg/day. The load time series at each station were summed and separated into inflow (positive flows) and outflow (negative flows) load time series. The total annual chloride retained inside the Refuge could be calculated, (see Table 5.1) from the difference between inflow and outflow loads divided by inflow load.

As can been seen in Table 5.1, the annual amount of chloride retained varies from 7.39 % in 1995, to 49.13 % in 2000, with the 10-year average being around 26.66 % and the total

86

chloride retained over the 10-year period being approximately 25 %. This can also be seen in the bar graphs shown in Figure 5.1.

Year

LoadIN LoadOUT Difference Percent Retained kg/year kg/year kg/year % 1995 147,853,910 136,925,206 10,928,704 7.39 1996 107,069,584 87,359,234 19,710,350 18.41 1997 119,601,977 70,856,138 48,745,349 40.76 1998 111,078,190 80,534,338 30,543,852 27.50 1999 109,418,942 94,312,865 15,106,077 13.81 2000 75,346,798 38,331,608 37,015,190 49.13 2001 46,268,615 32,423,886 13,844,729 29.92 2002 85,733,766 61,936,071 23,797,695 27.76 2003 72,656,556 56,942,900 15,713,656 21.63 2004 66,385,821 46,278,615 20,107,206 30.29 Table 5.1: Total annual chloride loads going in and out of the Refuge through hydraulic structures and the total percent of chloride retained in the Refuge.

160,000,000 Total Annual Chloride Load, kg/year

Inflow Chloride Load 140,000,000

Chloride Outflow Load

120,000,000 100,000,000 80,000,000 60,000,000 40,000,000 20,000,000 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Figure 5.1: Total annual chloride loads going in and out of the Refuge through hydraulic structures.

87

It can be assumed that the majority of the percent chloride retained is associated with groundwater seepage. Since we do not have any direct groundwater seepage data or any knowledge of the percent of evapotranspiration that is transpiration, we can only assume that when the net flow (Inflow – Outflow) volume is high then more water is consumed through groundwater seepage or transpiration. Figure 5.2 shows the correlation between net flow and percent chloride retained, which has a R squared value of 0.7327.

60 Percent Chloride Retained in Refuge

2

R = 0.7327 50

2000 1997

40 2001

30

2004 1998

2002

2003 1996

20 1999 10 0 -250

1995

-200

-150

-100

-50

0

50

100

150

200

250

3

Inflow-Outflow (hm )

Figure 5.2: The correlation between the net flow for the POR and the percent chloride retained in the Refuge.

5.2.2 Phosphorus

A similar balance of phosphorus loads was completed, much like the one discussed in Section 5.2.1 on chloride. As in Section 5.2.1 the percent retained refers to the percent of phosphorus that did not leave the Refuge through a hydraulic structure, therefore, it either

88

remained in the Refuge or exited the Refuge through some other means of outflow such as groundwater seepage or transpiration.

Phosphorus data were downloaded from SFWMD’s DBHYDRO database for the hydraulic structures located around the perimeter of the canal. Of the 19 hydraulic structures 16 of them had phosphorus data for the period of record from January 1, 1995, to December 31, 2004. These structures include: S-5A, S-5AS, G-300, G-301, G-310, G251, S-6, S-10E, S-10D, S-10C, S-10A, ACME-1, ACME-2, S-39, G-94C, and G-94B. As with chloride, G-94C only had three days of data for this period and G-94A did not have any data, therefore, the concentrations recorded at G-94B were used to fill these structures. All of the phosphorus samples taken from the structures above were done so by grab samples; although stations S-5A, G-310, G-251, and S-6 also had seven-day composite samples taken. According to SFWMD and Refuge scientists when seven-day composite samples are available it is best to use these data. All of the data were processed and evaluated, removing any outliers and averaging the concentrations when multiple recordings were recorded on a day. On days were the lab was unable to detect a reading the phosphorus concentration was recorded as the negative of the detection level (-0.004 mg/L), this value was divided by two and made positive (0.002 mg/L). A list of these values can be found in the Appendix A.

Similar to chloride, phosphorus data were also recorded periodically; therefore, phosphorus concentration time-series (mg/L) were generated for each hydraulic station.

89

For the stations where grab samples were available the data were filled linearly from January 1, 1995, to December 31, 2004.

The process of applying values from a seven-day composite sample to each day was a little more difficult. SFWMD scientists fill the time-series for such seven-day composite samples as follows: on the date the sample is recorded that concentration is the concentration for that date and the six days prior to that date. To fill the dates between sixth day prior to the composite reading and the next composite samples the average of the two samples is determined and that value is used to fill all the days between the two readings. A simple schematic can be seen in Figure 5.3 to explain this method.

Composite Value

Take the Average of the 2 Composite Values and Fill Missing Days

Fill 6 Days Prior with this Value 6 Days

Composite Value

Figure 5.3: Schematic explaining how the composite phosphorus samples were filled to make a complete time-series.

Once the daily time series were generated the total load could be calculated at each hydraulic structure using Equation 5.1. Then, the total load going in and out of the Refuge could be calculated by summing all of the daily loads from the individual structures together. The results can be found in Table 5.2 and Figure 5.4.

90

Year

LoadIN LoadOUT Difference Percent Retained kg/year kg/year kg/year % 1995 104,473 95,165 9,308 8.91 1996 75,960 45,544 30,416 40.04 1997 115,186 37,675 77,511 67.29 1998 99,616 49,801 49,815 50.01 1999 87,434 74,720 12,714 14.54 2000 58,563 21,495 37,068 63.30 2001 21,331 14,895 6,436 30.17 2002 32,409 19,946 12,463 38.45 2003 33,916 20,000 13,916 41.03 2004 46,363 48,755 -2,392 -5.16 Table 5.2: Total phosphorus loads going in and out of the Refuge through hydraulic structures and the total percent of phosphorus retained in the Refuge.

Total Annual Phosphorus Load, kg/year

140,000 Inflow Phosphorus Load 120,000

Outflow Phosphorus Load

100,000 80,000 60,000 40,000 20,000 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Figure 5.4: Total annual phosphorus loads going in and out of the Refuge through hydraulic structures.

Opposed to the results seen in chloride, there appears to be more phosphorus retained in Refuge. However, once again there is a correlation between net flow (Inflow – Outflow) and the percent of phosphorus retained in the Refuge (Figure 5.5). We can assume that as

91

the net flow increase so does the amount of phosphorus that exits the Refuge through groundwater seepage.

80 2

Percent Phosphorus Retained in the Refuge

R = 0.5555

1997

70

2000

60 50

1998 20031996

40 30

2002

2001

20 1999 10

1995

0 -250

-200

-150

-100

-50

0

2004 50

100

150

200

250

-10 Inflow-Outflow (hm3)

Figure 5.5: The correlation between the net flow for the POR and the percent of phosphorus retained in the Refuge.

The calculated loads going in and out of the Refuge were compared to those recorded in South Florida Environmental Report for the South Florida Water Years 2002, 2003, and 2004, that is produced by the SFWMD. To properly compare the results the total inflow and outflow loadings were calculated based on the Florida Water Years; for example, Florida Water Year 2002 is from May 1, 2001, to April 30, 2002. The comparison between the calculated results and SFWMD’s results can be seen in Table 5.3 for the inflow loads and Table 5.4 for the outflow loads. It should be noted that the outflow loads posted by SFWMD did not include the loads from structures G-94A, G-94B, or G94C for Florida Water Years 2002 and 2003, and they did not include the loads for G94A in Florida Water Year 2004. Therefore, in order to properly compare the loads; the

92

loads from the G-94 structures were appropriately subtracted from the calculated yearly loads.

Florida Water Year Calculated LoadIN SFWMD’s Loads Difference Percent kg/year Accuracy, % kg/year kg/year 2002 19,162 18,814 348 98.18 2003 43,706 43,409 297 99.32 2004 22,750 22,282 468 97.94 Table 5.3: Comparison of the calculated inflow loads against the SFWMD’s loads published in their annual reports for Florida Water Years 2002 to 2004.

Florida Calculated Water LoadOUT Year kg/year

Total Load Corrected SFWMD’s Difference Percent from G-94 Calculated Loads kg/year Accuracy, Structures LoadOUT kg/year % kg/year kg/year 2002 14,982 1,127 13,855 12,801 1,054 92.39 2003 25,964 3,756 22,208 21,628 580 97.39 2004 16,771 818 15,953 15,996 -43 100.27 Table 5.4: Comparison of the calculated outflow loads against the SFWMD’s loads published in their annual reports for Florida Water Years 2002 to 2004.

The calculated loads compare well, with the percent accuracy ranging from 100% to 92%. One possible reason for the minimal discrepancy in phosphorus total load could be attributed to the filling of missing data. Although, emphasis was taken in attempting to follow similar procedures as those followed by the SFWMD.

5.3 Model Selection

When first developing the water quality model it was important to find a modeling program or technique that would allow for simple model setup and operation; but could still be computationally efficient. Therefore, some previously developed modeling

93

programs and techniques were briefly analyzed to determine the one that could be utilized for the water quality modeling in the Refuge.

One possible modeling technique that was evaluated was the use of Artificial Neural Networks (ANN’s) as a way of forecasting water quality parameters. ANN’s are mathematical models that consist of interconnected nodes, that can extract a complex non linear relationship from a set of input and output data (Habib and Meselhe, 2006). It was determined that although ANN’s have been successful in modeling water quality constituents in the past, their predictive capabilities diminish when any changes would be made to the physical settings in the Refuge. Only things like inflow and outflow could be changed, therefore, limiting the scenarios that could be examined. ANN’s would have been a good choice if the water quality model would have not had a lot of data, or if future modeling would only be projecting within the range of calibration

Another modeling program that was looked at was the Dynamic Model for Everglades Stormwater Treatment Areas (DMSTA) (section 2.4.5) by Walker and Kadlec (2006). This model was developed for the U.S. Department of Interior and the U.S. Army Corp of Engineers for use in modeling the water quality in the STAs located just northwest and northeast of the Refuge, where flow is unidirectional. As canal-marsh flow in the Refuge is bidirectional, this model was not pursued further.

The USGS Branched Lagrangian Transport Model (BLTM) was also considered as a possible modeling program for water quality in the Refuge. The BLTM model was

94

developed to simulate the unsteady movement, dispersion, and chemical reactions of various constituents that move through a series of one-dimensional channels (Jobson, 2001). An advantage of this model is that it is open source code, and can easily be modified as needed. A disadvantage of BLTM is that it does not have any user friendly pre or post processors, and one of the major objectives is to develop a model that can be operated and modified easily by Refuge staff.

The modeling program that was chosen as the most suitable for the water quality modeling effort was the U.S. Environmental Protection Agency’s (EPA) Water Quality Analysis Simulation Program, Version 7.1 (WASP 7.1, hereafter referred to as simply WASP). WASP is a dynamic compartmental model that allows users the ability to interpret and predict water quality responses due to natural occurrence and man made pollution. The flexible compartmental approach allows users to investigate one, two, and three dimensional systems. The model includes the following data requirements: water body hydrogeometry, advective and dispersive flows, settling and resuspension rates, boundary concentrations, pollutant loadings, and initial conditions. The area being modeled can be separated into multiple segments or compartments. The segment volumes, connectivity, and type, such as surface water, must be known. Each segment or compartment acts independently, with the water quality constituents modeled as spatially constant within each segment. A possible limitation with this modeling program is that WASP does not allow the cells to go completely dry (US EPA, 2006). Some benefits of selecting this model is that it is free to the public, user friendly (does not require any computer programming experience), has been widely applied, and although it

95

can be used for simple simulations it can also be used for more complex simulation in the future. Another major advantage of using WASP is that it has a data preprocessor that allows for quick development of input datasets, and a postprocessor that enables efficient reviewing of model results.

WASP has a long history of application that even includes use in projects located in Florida, such as the examining of the eutrophication of Tampa Bay, FL, and the phosphorus loading in Lake Okeechobee, FL (US EPA, 2006).

5.4 Water Quality Modeling Approach

After reviewing the constituent data (sections 3.6 and 5.2), past modeling efforts within the Refuge, and consulting with Refuge scientists, it was determined that the water quality model would be best implemented by separating the Refuge into 4 cells (boxes). These cells would consist of the canal, and three inner marsh cells. Based on the distribution of chloride and phosphorus with distance away from the canal (Figure 3.13) the cells were set so that the first marsh cell fell within the first kilometer from the canal, the second marsh cell fell between one and four kilometers from the canal, and the third marsh cell included the remaining interior marsh area. The areas of each cell can be found in Table 5.5, and a sketch of the three interior cells and the XYZ and EVPA water quality monitoring station locations can be seen in Figure 5.6.

96

Cell Number

Distance from Canal Area Miles (km) Acres Canal 0 996 1 0.621 (1) 22,072 2 2.484 (4) 55,353 3 Remaining Interior 60,901 Table 5.5: Distance of each cell from the Refuge canal and its area.

LOX3

LOX4 !

!

LOX5 !

LOX10 LOX9

LOX8

!

!

!

LOX7 !

LOX6 !

X1#*X2 # * X4 X0 #*#* #* X3 Z1 * # *# # * Y4 * Z2 # Z0 # * Z3 #*

LOX11 !

Z4 LOX13

!

!

LOX12

LOX14 !

LOX15 LOX16

!

0

1

2

4

!

6 Miles

±

Figure 5.6: Location of EVPA and XYZ water quality monitoring sites in relation to the various cells.

Table 5.6 shows the relation of the various XYZ and EVPA monitoring stations to the canal and interior cells. These monitoring stations were used for calibration of both chloride and phosphorus. Because there were only two observation stations located inside the canal, X0 and Z0, the outflow structure data from structures S39, S10-E, S10D, S10-C, S10-A, and G94-B were also included as observed canal concentrations. Cell

97

1 has a limited number of observation stations located within it. The stations that are located within cell 1 are also all XYZ gages, that are located on the west side of the Refuge where higher concentrations typically occur. Therefore, data from observation stations LOX4 and LOX6 were used for the calibration of both cells 1 and cell 2.

Canal

Cell 1

X0 Z0 S39 S10-A S10-C S10-D S10-E G-94B

X1 Z1 Z2 LOX4 LOX6

Cell 2

Cell 3

X2 X4 Y4 LOX3 Z3 LOX5 Z4 LOX7 LOX4 LOX8 LOX6 LOX9 LOX10 LOX11 LOX12 LOX13 LOX14 LOX15 LOX16 Table 5.6: Location water quality stations in reference to the canal and interior cells used in calibration of the chloride and phosphorus models.

This model setup was used in the completion of the chloride and phosphorus models that will be discussed in chapters 6 and 7.

98

CHAPTER 6: Chloride Water Quality Modeling

6.1 Introduction

Chloride is modeled here as a conservative tracer with flows determined in the Refuge water budget model (Chapter 4). As a conservative, chloride, is assumed to not undergo any significant chemical or biological transformations or degradations (Kadlec and Knight, 1996), therefore, it was easily modeled here using both a simple spreadsheet model (Microsoft Excel) and WASP 7.1. Both modeling techniques are discussed in this chapter. This chloride model provides a better understanding of the transport of other surface water constituents including nutrients throughout the Refuge, and additionally provides insight supporting a better calibration of the water budget model.

6.2 Chloride Excel Model

Chloride was initially modeled using Microsoft Excel to calculate chloride concentrations on a one day time-step for the canal and the three interior marsh cells for the calibration period January 1, 1995, to December 31, 2004 and validation period January 1, 2000, to December 31, 2004. This Excel model parallels the approach used in the Refuge water budget model (Chapter 4).

99

6.2.1 Excel Model Setup

The water budget model discussed in chapter 4 was coupled with the simple spreadsheet water quality model using the observed inflows, outflows, and precipitations; along with adjusted evapotranspiration, and estimated canal and marsh seepage estimates. It was assumed that mass entered the Refuge through inflows ( Qin ), precipitation ( PM ), and M

dry deposition ( DDM ); and left the Refuge by means of outflows ( Qout ), M

groundwater seepage in the canal and marsh ( GS M ), and transpiration ( TM ) (Figure 6.1). Mass was exchanged between cells through the advection of flow between cells ( Qi _ M where i represents the downstream cell). As in the water budget model the exchange of flows was based on the corresponding inflows and outflows from each cell; however unlike the water budget model when the net canal flow was large there was no restriction that limits the magnitude of the canal stage. Downstream

Upstream

PM

PM DDM

MC

Qin

M

M1

Q1

CC

M

Canal

E EM

C1

M2

Q2

M

Cell 1

GS M E

Qout

PM DD M

PM DD M

DDM

C2 Cell 2

TM GS M E

M2

Q3

M

C2 Cell 3

TM GS M TM GS M E

M

Figure 6.1: Schematic of cells used to calculate chloride concentrations.

100

The spreadsheet model calculates concentrations by accumulating the daily change in chloride mass (g) (Equations 6.1a and 6.1b)

M i ,t = PM + GM + TM + Qi _ M + M i ,t −1

(6.1a)

and

M c ,t = PM + GM + TM + Qi _ M + Qin _ M − Qout _ M + M c ,t .

(6.1b)

Concentration is then calculated by dividing mass within the cell by the cell volume (Equations 6.2a and 6.2b)

Ci ,t =

M i ,t Ai * ( Ei − Ei _ 0 )

C c ,t =

M c ,t

(6.2a)

and

Ac * ( Ec − Ec _ 0 )

.

(6.2b)

Equations 6.1a and 6.2a apply to the ith marsh cell, while Equations 6.1b and 6.2b apply to the canal. The subscript t is the day index, A represents the areas of the canal (subscript c) and interior cells (subscript i) as listed in Table 5.4, Ei is the canal or marsh

101

stage calculated from the water budget, and Ei _ 0 is the marsh and canal elevations of 15.158 ft (4.62 m) and 1.641 ft (0.5 m), respectively.

The initial conditions for chloride concentrations were set using the average monthly observed concentration from the XYZ and EVPA sites located in each interior cell for January 1995, the starting month of the simulation, . The initial conditions for the canal and three interior cells can be seen in Table 6.1. Also included in Table 6.1 is the average chloride concentration for each cell based on the average monthly observed values.

Canal Cell 1 Cell 2 Cell 3 Chloride Initial Condition, mg/L 89.6 71.5 30.00 12.19 Average Observed Chloride Concentration for the POR 112.57 94.82 54.64 27.34 Table 6.1: Initial and long term average concentrations for chloride in each cell.

The inflow chloride concentrations through the perimeter canal inflow structures were obtained from the daily time-series which were calculated and discussed in section 5.2.1.

6.2.2 Calibration

The Excel model was initially calibrated on its own, but was later recalibrated using the values that were found to result in the best calibration of the WASP model in order to facilitate a comparison between the two model setups. Therefore, the values found in calibrating the WASP model are the values that will be presented here for both the simple Excel model and WASP model.

102

The calibration parameters for the chloride models include wet deposition, dry deposition, and the percent of evapotranspiration that is transpiration. It is assumed that transpiration transports chloride and other constituents into the root zone while evaporation does not transport any constituents. The major calibration parameter in modeling chloride was found to be the percent of transpiration fraction of evapotranspiration. Through calibration it was estimated that approximately 35% of evapotranspiration is transpiration. This value is relative to the range of 30% to 60% suggested by Dr. Robert H. Kadlec (R.H. Kadlec, personal communication, 2006). It is important to note that the percent of transpiration was calibrated over the entire Refuge, although transpiration does vary considerably based on water depth and vegetation, and it is reasonable to assume that percent of transpiration varies depending on vegetation type and percent cover (German, 1999). The model is relatively sensitive to transpiration.

When calibrating wet and dry deposition in the Refuge it was important to remember that the Refuge is unique in a part of the high nutrient water received from the control structures remains in the rim canals without actually flowing through the interior of the Refuge. Some high nutrient water moves into the Refuge, but evidence indicates that it moves slowly and most acutely impacts only a limited habitat near the canals (USFWS, 2000). There are no known published references for dry deposition of chloride in the Refuge. Therefore this parameter was simply calibrated based on recommended values from experienced wetland modelers (W.W. Walker, personal communication, 2005). Cells 2 and 3, the more interior cells, were more sensitive to the calibration value of dry deposition of chloride than the canal and cell 1.

103

There was some difficulty in calibrating in order to get an overall agreement between modeled and observed data between the various cells; therefore, the models were calibrated by trying to achieve the best overall results in the canal and marsh, while minimizing the biases in the canal and in cell 3.

6.2.3 Calibration Results

Figures 6.2 to 6.5 represent the graphical results of the chloride Excel model based on the calibration parameter from the WASP model. Shown are the modeled and observed monthly averaged values. Also shown are the standard deviations for the average monthly observed values. The daily results can be seen in Appendix B. It should be noted that it was observed that with the one day time step there was a lot of instability and numerical dispersion especially in the canal. There were also months where there were no observed data available within certain cells.

The model was also analyzed using the same performance measures used in the water budget model (section 4.6.3). Performance measures were calculated for the canal, cell 1, cell 2, cell 3, and also for the total marsh area which included cells 1, 2, and 3 together. The results from these calculations can be seen in Table 6.2.

104

Chloride Concentration, mg/L

300 250

Monthly Average Observed Monthly Average Modeled

200 150 100 50 0 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.2: Canal calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.3: Cell 1 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

105

200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.4: Cell 2 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Chloride Concentrations, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.5: Cell 3 calibration results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 106

Statistical Parameter

Cell 2 Marsh Canal Cell 1 Cell 3 Calibrations Calibration Calibration Calibration Calibration Statistics Statistics Statistics Statistics Statistics -5.458 -9.127 12.900 14.287 6.004 22.482 22.123 22.182 18.824 21.139 26.470 26.756 19.189 8.625 33.777

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation 24.022 20.912 16.270 13.666 24.146 of Modeled, mg/L Standard Deviation 21.993 20.331 18.200 12.368 20.328 of Error, mg/L Variance Reduction 31% 42% 10% -106% 64% R (Correlation 0.624 0.664 0.485 0.376 0.803 Coefficient) 0.390 0.440 0.235 0.141 0.644 R2 Value Nash Sutcliffe 0.266 0.304 -0.359 -3.851 0.606 Efficiency Table 6.2: Chloride Excel model performance measures for the calibration period.

6.2.4 Validation Results

The chloride Excel model was validated from January 1, 2000, to December 31, 2004 using the same model setup and parameters that were used for the calibration period. These results can be seen in Figures 6.6 to 6.9.

The performance measures for the validation period and POR can also be seen in Tables 6.3 and 6.4, respectively.

107

Chloride Concentration, mg/L

300 250

Monthly Average Observed Monthly Average Modeled

200 150 100 50 0 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.6: Canal validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Observed Monthly Average Modeled Monthly Average

Chloride Concentration, mg/L

175 150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.7: Cell 1 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 108

200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.8: Cell 2 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Chloride Concentrations, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.9: Cell 3 validation results for the chloride Excel model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 109

Statistical Parameter

Canal Validation Statistics -10.973 26.169 26.418

Cell 1 Validation Statistics -15.548 37.241 31.823

Cell 2 Validation Statistics 2.592 17.161 24.151

Marsh Cell 3 Validation Validation Statistics Statistics 14.404 0.519 25.244 27.652 18.519 37.665

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 22.121 19.192 20.788 21.322 25.534 Modeled, mg/L Standard Deviation of 23.958 34.141 17.108 20.915 27.727 Error, mg/L Variance Reduction 18% -15% 50% -28% 46% R (Correlation 0.525 0.171 0.720 0.396 0.677 Coefficient) R2 Value 0.275 0.029 0.518 0.157 0.458 Nash Sutcliffe 0.002 -0.394 0.487 -0.891 0.458 Efficiency Table 6.3: Chloride Excel model performance measures for the validation period.

Statistical Parameter

Canal Statistics for POR -8.215 24.395 26.412

Cell 1 Statistics for POR -12.337 30.629 30.035

Cell 2 Statistics for POR 7.703 19.810 24.258

Cell 3 Statistics for POR 14.347 22.323 15.475

Marsh Statistics for POR 3.238 24.640 36.519

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 23.005 20.321 19.437 18.694 25.308 Modeled, mg/L Standard Deviation of 23.066 28.159 18.329 17.179 24.462 Error, mg/L Variance Reduction 24% 12% 43% -23% 55% R (Correlation 0.572 0.429 0.669 0.457 0.744 Coefficient) R2 Value 0.327 0.184 0.447 0.209 0.554 Nash Sutcliffe Efficiency 0.140 -0.49 0.327 -1.099 0.543 Table 6.4: Chloride Excel model performance measures for the POR.

110

6.2.5. Discussion of the Chloride Excel Model

In the chloride Excel model it is likely that there was significant numerical dispersion. For the entire POR the model performed relatively well in the marsh as a whole (cells 1, 2, and 3); although, the model shows a negative Nash Sutcliffe Efficiency value. The models all showed qualitatively a good repetition of the observed data catching the overall trend of the data. The large biases are likely due to the fact that the chloride Excel model was calibrated using the calibration values determined in the chloride WASP model.

6.3 Chloride WASP Model

The chloride WASP model was setup using a 0.1 day time step for the calibration period from January 1, 1995, to December 31, 1999 and the validation period January 1, 2000, to December 31, 2004. WASP operates completely on the metric system, therefore all values and input parameters mentioned in this section will use SI units. The equations used by WASP are based on the basic principals of the conservation of mass. WASP operates on a mass balance principle in each cell.

6.3.1 Chloride WASP Model Setup

WASP requires the input of the model segmentation (cells) geometry and their initial conditions, system to be simulated, boundary conditions, source loads, exchanges

111

(dispersion), and flows. This input data along with the general WASP mass balance equations and general kinetics equations then defines a special set of water quality equations. These equations are numerically integrated by WASP.

The mass balance equations for a 1-dimensional stream used by WASP are shown in Equation 6.3

∂ ∂  ∂C  ( AC ) =  − U x AC + E x A  + A(S L + S B ) + AS K ∂t ∂x  ∂x 

(6.3)

where A is the cross-sectional area, m2; C is the concentration of the water quality constituent, mg/L; t is time in days; U x is the longitudinal, advective velocities in m/day; E x is the longitudinal diffusion coefficients, m3/day; S L is the total of direct loading rates in g/m3-day; S B is the boundary loading rates in g/m3-day; and S K is the total kinetic transformation rate in g/m3-day. It should be noted that S K only applies to modeling of phosphorus in this report.

Chloride was modeled using the eutrophication module in WASP. The eutrophication module, rather than the toxics module, was selected so that later the more complex phosphorus model in the eutrophication module could be implemented in the futures. WASP models salinity as a conservative constituent. In the present study, the WASP salinity state variable was chosen as the system in which chloride was modeled.

112

WASP requires initial volumes for each cell to be designated; this was done by assuming, consistent with the water budget model, an initial water depth in the canal of 2 m and a depth in the interior cells of 0.61 m. The water depth in the interior cells was calculated by taking the observed water level in the marsh on January 1, 1995 of 5.23 m and subtracting the average marsh elevation of 4.62 m. The assigned volumes are shown in Table 6.5. The initial chloride concentrations were the same as those shown in Table 6.1. The WASP parameter “fraction dissolved” for chloride was set at 100%.

Canal Cell 1 Cell 2 Cell 3 Volume, m 8,066,971 54,509,080 136,701,113 150,402,747 Table 6.5: Initial volumes for the canal and interior cells. 3

Aerial loads were input into WASP based on calibrated wet and dry deposition. Wet deposition was calibrated in mg/L and multiplied by the daily rainfall rate and area in order to get a load in kg/day. Dry deposition was calibrated in g/m2-yr and multiplied times the cell areas accordingly in order to get a load in kg/day; this dry deposition load was assumed to be constant for each day during the modeled POR. A daily aerial load time series was created by adding the daily wet and dry deposition rates.

Flows used in the modeling chloride were also taken from the water budget model including inflow from canal structures ( Qin ), outflow from canal structures ( Qout ), estimated canal and marsh seepages ( GS ), and estimated exchange flow from the Power Law Model ( QMC ). WASP considers precipitation and evaporation as flows that do not transport mass. As in the simple Excel model the percent of transpiration was also calibrated and is modeled as a flow in WASP. All flows were input into WASP in

113

m3/sec. The flows were input according and distributed to and from the various cells according to the fraction of flow going into or out of each cell. The fractions used for each flow are expressed in Table 6.6.

The boundary inflow chloride concentration time series was input into WASP on a daily time step in mg/L. This time series was obtained from that calculated in section 5.2.1.

Inflow to Canal Boundary to Canal Canal to Boundary Canal to Cell 1 Cell 1 to Cell 2 Cell 2 to Cell 3 Cell 1 to Boundary Cell 2 to Boundary Cell 3 to Boundary

Outflow from Canal

Exchange Flow

Marsh Seepage

Canal Seepage

Transpiration Precipitation & Evaporation

1

0.00715

1 1 1 0.840439 0.443854 0.1559564

0.15842

0.3965825

0.39375

0.4438535

0.44068

Table 6.6: Fraction of flows used in WASP.

Dispersion was also used as a calibration parameter. WASP models dispersion as an exchange function in m2/sec. In order to implement dispersion in WASP the user must assign gross-cross sectional areas representative of the areas through which mixing occurs; and mixing lengths which reflect the length over which mixing occurs. The cross sectional areas were calculated using the perimeter of each cell and an estimated typical depth of 0.5 m for the interior cells and a depth of 2 m for the canal. The lengths are

114

calculated using the center point of adjoining segments (cells). These values are shown in Table 6.7. Dispersion was considered to be constant for the entire POR. Area, m2 Distance, m Canal – Cell 1 46,521.035 522 Cell 1 – Cell 2 42,949.0465 2,000 Cell 2 – Cell 3 31,268.145 4,500 Table 6.7: Areas and distance used to calculate dispersion in the WASP chloride model.

6.3.2 Chloride WASP Model Calibration

The parameters which were calibrated in the WASP chloride model and also used in the chloride Excel model were percent transpiration, wet deposition concentration, dry deposition rate, and dispersion. Dispersion was not implemented in the Excel model. As mentioned earlier there was some difficulty in calibrating chloride based on when improving the statistics in certain cells other cells statistics decreased. Therefore, calibration was based on achieving the best overall statistics in the canal and marsh as a whole (cells 1, 2, and 3 combined). It was also attempted to calibrate to minimize the biases in the canal and cell 3.

The primary calibration parameter was found to be the percent of transpiration. It was calibrated that the percent transpiration was approximately 35% of the total evapotranspiration estimate.

115

Wet and dry depositions were also determined by calibration. It was found that the model calibrated with a wet deposition concentration of 2 mg/L and a dry deposition of 0.5 g/m2-yr. Cell 3 was the most sensitive to the calibration of these parameters.

Longitudinal dispersion was also estimated by calibration. A range of 0.37 to 22 m2/hr was typical in this area based on (Meselhe et al., 2005). Longitudinal dispersion was calibrated to be equal to 22 m2/hr, although when calibrating it was found that dispersion had very little effect in the canal and cell 1, and no effect in cells 2 and 3.

The observed concentrations used in calibration statistics were based on the XYZ, EVPA, and outflow structure concentrations were aggregated to monthly averages (Meselhe et al., 2005). Variability of samples was characterized by the monthly standard deviations of values observed within the cell boundaries. There were some months where there were no observed values within a cell; these months were eliminated from calibration statistics.

6.3.3 Chloride WASP Model Calibration Results

The results from calibration can be seen in Figures 6.10 to 6.13; these plots represent the modeled and observed data along with the standard deviations of the observed data. Performance measures can also be found in Table 6.8. The performance measures can be slightly misleading due to the gaps in data, and the minimal reading in some cells. For example cell 1 has predominantly XYZ stations used for observation data, which tend to

116

have higher concentrations. Also due to the small range in data in cell 3 the statistics do not represent the pattern that was achieved, that can be seen in the daily results which can be found in Appendix C.

It was found when calibrating the chloride model that the modeled chloride concentrations in the canal were low, therefore, it was determined that the water budget model needed further calibration of the canal seepage rate. By lowering the canal seepage rate by 30%, and proportionally adjusting the marsh seepage rate the bias in the canal began to approach zero.

Chloride Concentration, mg/L

300 250

Monthly Average Observed Monthly Average Modeled

200 150 100 50 0 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.10: Canal calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

117

200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.11: Cell 1 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.12: Cell 2 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 118

200

Chloride Concentrations, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-95 -25

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 6.13: Cell 3 calibration results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

Cell 3 Marsh Canal Cell 1 Cell 2 Calibration Calibration Calibration Calibration Calibration Statistics Statistics Statistics Statistics Statistics Bias, mg/L -2.214 -7.192 8.947 5.610 2.494 RMSE, mg/L 17.533 22.033 21.209 12.191 19.100 Standard Deviation of 26.470 26.756 19.189 8.625 33.777 Observed, mg/L Standard Deviation of 23.961 23.930 16.876 10.901 27.906 Modeled, mg/L Standard Deviation of 17.539 21.011 19.395 10.923 18.992 Error, mg/L Variance Reduction 56% 38% -2% -60% 68% R (Correlation 0.782 0.667 0.430 0.318201 0.827 Coefficient) R2 Value 0.581 0.445 0.185 0.101 0.684 Nash Sutcliffe 0.554 0.310 -0.243 -1.035 0.678 Efficiency Table 6.8: Performance measures for the calibration period using the chloride WASP model. Statistical Parameter

119

The results show that both canal and marsh as an entirety perform well in modeling chloride. The graphs show that the general pattern of concentration was obtained.

6.3.4 Chloride WASP Model Validation Results

The chloride WASP model validation results can be seen in Figures 6.14 to 6.17. Daily validation graphs can be found in Appendix C. The performance measures for the validation period and entire POR can be seen in Table 6.9 and 6.10.

Chloride Concentration, mg/L

300 250

Monthly Average Observed Monthly Average Modeled

200 150 100 50 0 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.14: Canal validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

120

200

Observed Monthly Average Modeled Monthly Average

Chloride Concentration, mg/L

175 150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.15: Cell 1 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 200

Chloride Concentration, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.16: Cell 2 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 121

200

Chloride Concentrations, mg/L

175

Observed Monthly Average Modeled Monthly Average

150 125 100 75 50 25 0 Jan-00 -25

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 6.17: Cell 3 validation results for the chloride WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data

Statistical Parameter

Canal Validation Statistics -8.402 23.037 26.418

Cell 1 Validation Statistics -18.420 40.357 31.823

Cell 2 Validation Statistics -8.121 20.311 24.151

Cell 3 Marsh Validation Validation Statistics Statistics -1.918 -9.463 20.672 28.562 18.519 37.665

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 23.983 27.067 25.233 17.789 31.607 Modeled, mg/L Standard Deviation of 21.631 36.228 18.775 20.766 27.027 Error, mg/L Variance Reduction 33% -30% 40% -26% 49% R (Correlation 0.635 0.256 0.712 0.340 0.709 Coefficient) R2 Value 0.404 0.065 0.507 0.116 0.503 Nash Sutcliffe 0.227 -0.637 0.281 -0.268 0.422 Efficiency Table 6.9: Performance measure for the validation period using the chloride WASP model.

122

Statistical Parameter

Canal Statistics for POR -5.308 20.471 26.412

Cell 1 Statistics for POR -12.806 32.513 30.035

Cell 2 Statistics for POR 0.341 20.761 24.258

Cell 3 Statistics for POR 1.779 17.043 15.475

Marsh Statistics for POR -3.536 24.337 36.519

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 23.895 25.457 21.496 14.750 29.811 Modeled, mg/L Standard Deviation of 19.853 30.017 20.846 17.026 24.113 Error, mg/L Variance Reduction 43% 0.13% 26% -21% 56% R (Correlation 0.693 0.431 0.591 0.353 0.754 Coefficient) 0.480 0.182 0.350 0.125 0.569 R2 Value Nash Sutcliffe Efficiency 0.394 -0.182 0.261 -0.224 0.555 Table 6.10: Performance measure for the POR using the chloride WASP model.

6.3.5 Discussion and Further Analysis of the Chloride WASP Model

The chloride WASP model showed better performance measures overall than the chloride Excel model. The canal showed a vast improvement, likely related to more limited numerical dispersion and improved stabilities. The canal, cell 2, and cell 3 all had moderately low biases. Cell 3’s performance measures do not adequately represent the successfulness of the model, however the graphical representation of modeled and observed daily values are viewed in Appendix C; and it can be seen that both the overall pattern and values were achieved.

The chloride WASP model was also analyzed by comparing the calculated outflow loads discussed in section 5.2.1 to the canal modeled loads (this was calculated by multiplying the daily canal modeled concentrations by the daily observed outflows and converting to

123

kg/day). The results showed a good correlation between the two (Figure 6.18) with a R2 value of 0.9285 and a Nash Sutcliffe Efficiency of 0.9254.

3,500,000

R2 = 0.9285 Efficiency = 0.925

3,000,000

Modeled (kg/day)

2,500,000

2,000,000

1,500,000

1,000,000

500,000

0 0

500,000

1,000,000

1,500,000

2,000,000

2,500,000

3,000,000

3,500,000

Observed (kg/day)

Figure 6.18: Modeled loads in the canal compared to the observed outflow loads from the canal structures. Solid line is a trendline with forced zero origin generated by Excel.

The modeled chloride concentrations were also analyzed by plotting the modeled and observed concentrations versus distance from perimeter canal. The distance from the perimeter canal for each of the observed stations was determined, and the modeled concentrations were plotted based on distance of the center of each cell in reference to the canal. Therefore, the concentrations in cell 1 were set at 0.5 km, cell 2 at 2.5 km, and cell 3 at 10 km. Due to the observed chloride concentrations being recorded on an irregular basis the concentrations from the observed stations were taken within a short period either before or after the date of the analyzed model concentration date. Evaluations

124

were completed for one day of each year for the POR. These results can be seen in Figures 6.19 to 6.28. From these figures it can be seen that the WASP chloride model catches the overall trend in reduction of concentration with respect to distance.

Chloride Concentrations 1/11/1995 ( 1/5/1995 - 1/12/1995 )

Chloride Concentration, mg/L

200 Observed Modeled Canal

180 160 140 120 100 80 60 40 20 0 0

1

2

3

4

5

6

7

8

9

10

Distance from Canal, km

Figure 6.19: Observed (1/5/1995, to 1/12/1995 plotted without a line) and modeled (1/11/1995 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

Chloride Concentration, mg/L

Chloride Concentrations 4/24/1996 ( 4/15/1996 - 4/25/1996 )

200 180 160 140 120 100 80 60 40 20 0

Observed Modeled Canal

0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.20: Observed (4/15/1996, to 4/25/1996 plotted without a line) and modeled (4/24/1996 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal. 125

Chloride Concentrations 6/3/1997 ( 6/3/1997 - 6/11/1997 )

Chloride Concentration, mg/L

200

Observed Modeled Canal

180 160 140 120 100 80 60 40 20 0 0

1

2

3

4

5

6

7

8

9

10

Distance from Canal, km

Figure 6.21: Observed (6/3/1997, to 6/11/1997 plotted without a line) and modeled (6/3/1997 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

Chloride Concentration mg/L

Chloride Concentrations 1/13/1998 ( 1/5/1998 - 1/13/1998 )

200 180 160 140 120 100 80 60 40 20 0

Observed Modeled Canal

0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.22: Observed (1/5/1998, to 1/13/1998 plotted without a line) and modeled (1/13/1998 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

126

Chloride Concentrations 1/4/1999 ( 1/4/1999 - 1/12/1999 ) 200 Observed Modeled Canal

Chloride Concentration mg/L

180 160 140 120 100 80 60 40 20 0 0

1

2

3

4

5

6

7

8

9

10

Distance from Canal, km

Figure 6.23: Observed (1/4/1999, to 1/12/1999 plotted without a line) and modeled (1/4/1999 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

Chloride Concentration, mg/l

Chloride Concentrations 1/11/2000 ( 1/3/2000 - 1/11/2000 )

200 180 160 140 120 100 80 60 40 20 0

Observed Modeled Canal

0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.24: Observed (1/3/2000, to 1/11/2000 plotted without a line) and modeled (1/11/2000 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

127

Chloride Concentrations 10/9/2001 ( 10/9/2001 - 10/16/2001 )

Chloride Concentration, mg/L

200 Observed Modeled Canal

180 160 140 120 100 80 60 40 20 0 0

1

2

3

4

5

6

7

8

9

10

Distance from Canal, km

Figure 6.25: Observed (10/9/2001, to 10/16/2001 plotted without a line) and modeled (10/9/2001 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

Chloride Concentration, mg/l

Chloride Concentrations 1/15/2002 ( 1/8/2002 - 1/15/2002 )

200 180 160

Observed Modeled Canal

140 120 100 80 60 40 20 0 0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.26: Observed (1/8/2002, to 1/15/2002 plotted without a line) and modeled (1/15/2002 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

128

Chloride Concentration mg/L

Chloride Concentrations 12/4/2003 ( 12/4/2003 - 12/16/2003 )

260 240 220 200 180 160 140 120 100 80 60 40 20 0

Observed Modeled Canal

0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.27: Observed (12/4/2003, to 12/16/2003 plotted without a line) and modeled (12/4/2003 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

Chloride Concentration, mg/l

Chloride Concentrations 10/18/2004 ( 10/18/2004 - 10/21/2004 )

200 180 160 140 120 100 80 60 40 20 0

Observed Modeled Canal

0

1

2

3

4 5 6 7 Distance from Canal, km

8

9

10

Figure 6.28: Observed (10/18/2004, to 10/21/2004 plotted without a line) and modeled (10/18/2004 plotted with a solid line) chloride concentrations using the WASP model versus distance from the canal.

129

CHAPTER 7: Phosphorus Water Quality Modeling

7.1 Introduction

High concentrations of nutrients, particularly phosphorus, in runoff from agricultural areas cause proliferation of cattails and other undesirable species that negatively affect the ecosystem’s balance in the Refuge. Therefore, the monitoring of phosphorus levels in the Refuge has become a priority to the Refuge staff. Developing a model that efficiently predicts the phosphorus concentrations in the Refuge gives the Refuge the ability to make proper management decision. In order to meet these objectives, a simple phosphorus model was implemented using WASP 7.1 and the k-c* model by Kadlec and Knight (1996) (see section 2.3.1 for a description of this modeling technique).

7.2 Phosphorus WASP Model Setup

The phosphorus WASP model was setup using a 0.1 day time step for the calibration period January 1, 2000, to December 31, 2004 and the validation period January 1, 1995, to December 31, 1999. These calibration and validation periods were chosen based on that there were more data for this period due to the increase in monitoring of phosphorus. The Refuge was modeled using the modeling approach discussed in section 5.4.

Like chloride, phosphorus was modeled using the eutrophication module in WASP. The same initial volumes were used assuming an initial depth in the canal and marsh of 2.0 m

130

and 0.61 m, respectively (Table 6.5). Phosphorus was modeled as carbonaceous biological oxygen demand (CBOD), using the k-c* model by Kadlec and Knight (1996) (Equation 2.2).

The initial phosphorus concentrations used in modeling can be found in Table 7.1, along with the average observed monthly average phosphorus concentration for the POR.

Canal 0.0341

Cell 1 0.0065

Cell 2 0.0144

Cell 3 0.0133

Phosphorus Initial Condition, mg/L Average Observed Phosphorus 0.0608 0.0241 0.0106 0.0111 Concentrations for the POR Table 7.1: Initial conditions for phosphorus and the average observed phosphorus concentration for each cell.

Aerial loads were input into WASP based on calibrated wet and dry deposition and the areal mass loading rate that was calculated from the calibration of the k-c* model..

Wet deposition was calibrated in mg/L and multiplied by the daily rainfall rate and area in order to get a load in kg/day. Dry deposition was calibrated in mg/m2-yr and multiplied times the cell areas accordingly in order to get a load in kg/day; this load was assumed to be constant for each day during the modeled POR. The areal mass loading rate was calculated in mg/m2-day and was also multiplied times the cell areas in order to obtain a loading rate in kg/day. The areal mass loading rate was assumed to be a constant value for each day during the POR. A daily aerial load time series was created by adding the daily wet and dry deposition rates and the areal mass loading rates.

131

The same flow values used in modeling chloride were also used in the phosphorus WASP model. The calibrated settling rate (m/yr) from the k-c* model is entered into WASP as a flow by multiplying the rate times the total area of the Refuge. It should be noted that all flows in WASP are inputted in m3/sec.

The flows were input accordingly and distributed to and from the various cells based on the fraction of flow going into or out of each cell. The fractions used for each flow are expressed in Table 6.6; the settling rate fraction of flows can be found in Table 7.2.

Settling Rate Fraction of Flows Canal to Boundary 0.00715 Cell 1 to Boundary 0.15842 Cell 2 to Boundary 0.39375 Cell 3 to Boundary 0.44068 Table 7.2: Fraction of flows used in for calculating settling rate for each cell.

The boundary inflow phosphorus concentration time series was input into WASP on a daily time step in mg/L. This time series was obtained from that calculated in section 5.2.2. Dispersion was set to the calibrated values found in the chloride model.

7.3. Phosphorus WASP Model Calibration

In the phosphorus WASP model the following parameters were used for calibration: 1) wet deposition; 2) dry deposition; 3) settling rate (k); and 4) the c* concentration value.

132

As in the chloride models, calibration was difficult in achieving the best results in each cell uniformly. Therefore, for phosphorus calibration it was aimed at achieving a low bias in the canal and cell 3.

The model was primarily calibrated using the k-c* model. The recommended range for settling rate based on Walker and Kadlec’s (2005) DMSTA2 model is between 16.8 and 52.5 m/yr. Through calibration it was found that the model responded best overall, especially in the canal and cell 3, when the settling rate was set to 16.8 m/yr. Through the use of the k-c* model in the DMSTA model, Walker and Kadlec (2005) found that the c* value within the STAs ranged between 4 and 20 µg/L. The c* value was also calibrated; it was initially assigned as 3 µg/L but through calibration it became aware that the value would need to vary between the canal and interior cells. Therefore, the c* value for the canal was calibrated to be 80 µg/L and the interior cells were calibrated to have a value of 8 µg/L. The canal was not very sensitive to the c* value, ranges between 10 and 90 µg/L were tested and it was found that the model showed a minimal bias when 80 µg/L was used. Based on the calibrated settling rate and c* values an areal mass loading rates in the canal and marsh were calculated to be 3.68 and 0.368 mg/m2-day, respectively.

Wet and dry depositions were also calibrated for in the model. These parameters are an important source of nutrients coming into the Refuge. According to Richardson et al. (1990) the atmospheric deposition reported from 1979 through 1988 accounted for 25% of the phosphorus entering the Refuge compared to 75% of the phosphorus entering via

133

S-5 and S-6 structures combined. Analysis of wet and dry deposition data is statistically challenging (Ahn 1999a; Ahn 1999b; Walker and Jewell 1997). Measurements of atmospheric deposition rates are complicated by numerous sources of contamination such as ash, vegetation, insects, spider webs, and bird droppings that can cause positive bias. Estimates of atmospheric phosphorus deposition have ranged from 17 to 96 mg/m2-yr for different locations at South Florida (Walker, 1995). Most modeling approaches for the Everglades have used a constant value for the atmospheric phosphorus deposition. Walker (1995) assumed a constant value of 43 mg/m2-yr for an area adjacent to the Refuge. Raghunathan et al. (2001) used a temporally and spatially constant value of 43 mg/m2-yr. For the phosphorus WASP model it was found that the model calibrated best with a wet deposition concentration of 0.010 mg/L and a dry deposition of 40 mg/m2-yr.

The observed concentrations used in calibration were based on the XYZ, EVPA, and outflow structure concentrations were aggregated to monthly averages (Meselhe et al., 2005). The variability of samples was characterized by the monthly standard deviations of values observed within the boundaries. As with chloride, there were some months were there were no observed values within a cell, these months were eliminated from calibration statistics. There were also quite a few months were there was only one value recorded. These values were not thrown out due to the fact that they were consistent throughout the cells, however, they did have an effect on the overall statistics in model.

134

7.4 Phosphorus WASP Model Calibration Results

The results from calibration can be seen in Figures 7.1 to 7.4; these plots represent the modeled and observed data along with the standard deviations of the observed data. Performance measures can be found in Table 7.3. As with chloride the performance measures can be slightly misleading due to the gaps in data and the minimal number of values in some cells.

0.30

Phosphorus Concentration, mg/L

0.25

Monthly Average Observed Monthly Average Modeled

0.20 0.15 0.10 0.05 0.00 Jan-00 -0.05

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 7.1: Canal calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

135

0.20 Phosphorus Concentration, mg/L

0.18

Observed Monthly Average Modeled Monthly Average

0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 7.2: Cell 1 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 0.08

Observed Monthly Average

Phosphorus Concentration, mg/L

0.07

Modeled Monthly Average

0.06 0.05 0.04 0.03 0.02 0.01 0.00 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 7.3: Cell 2 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

136

0.05 Phosphorus Concentrations, mg/L

Observed Monthly Average Modeled Monthly Average

0.04

0.03

0.02

0.01

0.00 Jan-00

Jan-01

Jan-02

Jan-03

Jan-04

Jan-05

Figure 7.4: Cell 3 calibration results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

Cell 3 Marsh Canal Cell 1 Cell 2 Calibration Calibration Calibration Calibration Calibration Statistics Statistics Statistics Statistics Statistics Bias, mg/L -0.0046 -0.0085 0.0009 -0.0010 -0.0028 RMSE, mg/L 0.0182 0.0270 0.0061 0.0058 0.0161 Standard Deviation of 0.0229 0.0245 0.0056 0.0056 0.0160 Observed, mg/L Standard Deviation of 0.0194 0.0083 0.0019 0.0002 0.0054 Modeled, mg/L Standard Deviation of 0.0177 0.0258 0.0060 0.0057 0.0159 Error, mg/L Variance Reduction 40% -11% -16% -3% 1% R (Correlation 0.6594 0.0085 -0.0606 -0.3233 0.1911 Coefficient) R2 Value 0.4348 0.00007 0.0037 0.1045 0.0365 Nash Sutcliffe 0.3580 -0.2366 -0.1884 -0.0633 -0.0190 Efficiency Table 7.3: Performance measures for the calibration period using the phosphorus WASP model. Statistical Parameter

137

7.5 Phosphorus WASP Model Validation

The phosphorus WASP model validation results can be seen in Figures 7.5 to 7.8. The performance measures for the validation period and entire POR can be seen in Table 7.4 and 7.5.

0.30

Phosphorus Concentration, mg/L

0.25

Monthly Average Observed Monthly Average Modeled

0.20 0.15 0.10 0.05 0.00 Jan-95 -0.05

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 7.5: Canal validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

138

0.20 Phosphorus Concentration, mg/L

0.18

Observed Monthly Average Modeled Monthly Average

0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 7.6: Cell 1 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data. 0.08

Observed Monthly Average

Phosphorus Concentration, mg/L

0.07

Modeled Monthly Average

0.06 0.05 0.04 0.03 0.02 0.01 0.00 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 7.7: Cell 2 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

139

0.08

Observed Monthly Average

Phosphorus Concentrations, mg/L

0.07

Modeled Monthly Average

0.06 0.05 0.04 0.03 0.02 0.01 0.00 Jan-95

Jan-96

Jan-97

Jan-98

Jan-99

Jan-00

Figure 7.8: Cell 3 validation results for the phosphorus WASP model, representing the average monthly observed data and its standard deviations; and the average monthly modeled data.

Statistical Parameter

Canal Validation Statistics 0.0056 0.0313 0.0236

Cell 1 Validation Statistics 0.0009 0.0171 0.0165

Cell 2 Validation Statistics 0.0023 0.0052 0.0042

Cell 3 Marsh Validation Validation Statistics Statistics 0.0006 0.0009 0.0063 0.0171 0.0063 0.0120

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 0.0303 0.0113 0.0021 0.0002 Modeled, mg/L Standard Deviation of 0.0311 0.0172 0.0047 0.0063 Error, mg/L Variance Reduction -74% -8% -25% 0% R (Correlation 0.3547 0.2840 -0.0119 0.0369 Coefficient) R2 Value 0.1258 0.0807 0.0001 0.0014 Nash Sutcliffe -0.7946 -0.0856 -0.5462 -0.0069 Efficiency Table 7.4: Performance measure for the validation period using the phosphorus WASP model.

140

0.0087 0.0108 18% 0.4904 0.2405 0.0856

Statistical Parameter

Canal Statistics for POR 0.0005 0.0256 0.0237

Cell 1 Statistics for POR -0.0037 0.0225 0.0208

Cell 2 Statistics for POR 0.0016 0.0056 0.0049

Cell 3 Statistics for POR 0.00003 0.00618 0.00599

Marsh Statistics for POR -0.0008 0.0137 0.0141

Bias, mg/L RMSE, mg/L Standard Deviation of Observed, mg/L Standard Deviation of 0.0272 0.0105 0.0021 0.0023 0.0074 Modeled, mg/L Standard Deviation of 0.0257 0.0223 0.0054 0.00620 0.0137 Error, mg/L Variance Reduction -18% -15% -21% -7% 5% R (Correlation 0.469 0.1062 -0.0467 -0.1690 0.3156 Coefficient) 0.2469 0.0113 0.0022 0.0286 0.0996 R2 Value Nash Sutcliffe Efficiency -0.1817 -0.1851 -0.3153 -0.0109 0.0512 Table 7.5: Performance measure for the POR using the phosphorus WASP model.

7.6 Discussion and Further Analysis of the Phosphorus WASP Model

The phosphorus WASP model followed the relative trend in the canal and cell 1; however, in cells 2 and 3 the transient changes in concentrations were not captured. The bias was able to be reduced relatively close to zero in cell 3 for the POR. The canal calibrated well, however, the validation results did not show the same conclusions. For the entire POR the marsh as a whole performed better than the canal.

The phosphorus WASP model was also analyzed by comparing the calculated outflow loads discussed in section 5.2.2 to the canal modeled loads (this was calculated by multiplying the daily canal modeled concentrations by the daily observed outflows and converting to kg/day). The results showed a good correlation between the two (Figure

141

6.18) with a R2 value of 0.6872 and a Nash Sutcliffe Efficiency of 0.6862. However, the results were not as good as those seen in the chloride model.

2,500 2

R = 0.6872 Efficiency = 0.6862

Modeled (kg/day)

2,000

1,500

1,000

500

0 0

500

1,000

1,500

2,000

2,500

Observed (kg/day)

Figure 7.9: Modeled loads in the canal compared to the observed outflow loads from the canal structures.

The statistics from the phosphorus water quality model were compared to results from the ELM v.2.1 model (Fitz et al., 2002a). These comparisons can be seen in Tables 7.6 to 7.9.

Phosphorus ELM ELM v.2.1 ELM v.2.1 Model Model v.2.1. Model Statistics Statistics for Model Statistics Statistics L-7 the POR Statistics L40-2 Canal L40-1 Bias, mg/L 0.0005 -0.009 -0.027 0.012 2 R 0.2469 0.13 0.17 0.00 RMSE 0.0256 0.058 0.057 0.097 Nash Sutcliffe Efficiency -0.1817 0.00 -0.23 -0.57 Table 7.6: Statistics in the canal comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model 142

ELM v.2.1. Phosphorus Model Statistics Phosphorus Model Statistics Statistics LOX LOX For the POR 4 6 Cell 1 Bias, mg/L -0.0042 0.024 0.005 2 R 0.0096 0.01 0.03 RMSE 0.0226 0.027 0.009 Nash Sutcliffe Efficiency -0.1867 -105.73 -2.69 Table 7.7: Statistics in the cell 1 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model.

Statistics

Phosphorus Model Statistics For the POR Cell 2

ELM v.2.1. Phosphorus Model Statistics LOX 4

LOX 6

LOX 10

LOX 12

LOX 14

LOX 15

LOX 16

Bias, 0.0015 0.024 0.005 0.002 0.013 0.014 0.018 0.016 mg/L 2 R -0.0577 0.01 0.03 0.08 0.12 0.08 0.00 0.00 RMSE 0.0056 0.027 0.009 0.011 0.014 0.016 0.023 0.019 Nash Sutcliffe -0.2936 -2.69 -0.28 -19.95 -18.06 -18.50 -10.14 105.73 Efficiency Table 7.8: Statistics in the cell 2 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model.

Statistics

Bias, mg/L R2 RMSE

Phosphorus Model Statistics For the POR Cell 3 0.00002 0.0295 0.00618

ELM v.2.1. Phosphorus Model Statistics LOX 3 0.00 1 0.11 0.01 5

LOX 5

LOX 7

LOX 8

LOX 9

LOX 11

LOX 13

-0.002

0.002

-0.002

-0.003

-0.004

-0.003

0.19

0.09

0.06

0.04

0.03

0.00

0.006

0.005

0.005

0.008

0.006

0.006

Nash Sutcliffe -0.0106 -0.33 -0.71 -2.41 -0.73 -0.15 -1.17 -0.54 Efficiency Table 7.9: Statistics in the cell 3 comparing the phosphorus water quality model and the ELM v.2.1 phosphorus model.

143

When comparing the two models it can be seen that although the phosphorus WASP model is cell oriented, it does show better statistical results than the ELM v.2.1 model. In the canal, the phosphorus WASP model results show a lower bias than the ELM v.2.1 model and a relatively similar Nash Sutcliffe Efficiency.

144

CHAPTER 8: Conclusion and Future Developments

8.1 Water Budget Model Conclusions

The double-box water budget model has proven to be computationally efficient in performing multi-decadal simulations within minutes. Also the simplicity of the water budget model allows the Refuge managers to assess strategies (at least on a preliminary basis) and make management decisions quickly and efficiently. The model allows for rapid testing of the model sensitivity to parameters and supports quick tests of a broader suite of management scenarios than can feasibly be examined using a more complex model. Selected scenarios can later be verified using a more complex model.

The simple water budget model is capable of predicting temporal variations of water levels in the canal and marsh. It can also help to quantify the different components of the Refuge’s water budget, particularly the importance of seepage. There are no measurements of overall seepage rate in the Refuge, therefore the simplified model can be used to estimate seepage rates based on water balance.

Some limitations of the model are that no spatial variability within the Refuge is modeled. For example, elevation differences between the northern and southern portions of the interior marsh are not modeled. Another finding was that the model was particularly sensitive to the area-average rainfall estimates and the seepage estimates. An

145

interesting finding was that the water budget model was relatively sensitive to the assumed average marsh elevation.

The simple water budget model was also a valuable tool in filling time series of water quality constituents such as chloride and phosphorus, particularly in the canal. The water budget model also proved that by using “The Power Law Model” and simple geometry the bidirectional flow between the marsh and canal could be appropriately estimated.

8.2 Water Budget Future Developments

Overall the simple water budget model is computationally efficient while maintaining its simplicity, which was a major objective in the production of this model. However, as with any model there are some minor improvements which can be made in future developments that would allow the model to possibly perform more efficiently and maintain the desired simplicity.

One possibility would be to allow the marsh elevation to be spatially variable, as mentioned earlier the model showed some sensitivity to marsh elevation and by varying it spatially it might allow for the model to have better efficiency. To do this the model would have to be slightly more advanced, for example, the marsh areas would need to be divided into multiple cells (boxes) such as a North, South, and midsection.

146

Additional modeling endeavors should include the completion of an uncertainty analysis on the model parameters such as seepage coefficient, aerial average precipitation, and the ET reduction coefficient. This would allow a better understanding of the parameters which make up and drive the simple water budget model.

The model can also be used to continue to assess a variety of management scenarios and alternatives. By continuing to assess various scenarios the model can be further improved to meet certain needs that may arise.

8.3 Chloride Model Conclusions

Chloride was modeled as a tracer which allowed for a better understanding of the transport of all constituents including nutrients in the Refuge. The chloride WASP model can be used to rapidly test the affects of changes in flow on water quality within the Refuge. The Excel chloride model also provides users with the ability to quickly test calibration parameters and determine their relative sensitivity to the model.

This model is also helpful in testing and finalizing the calibration of the simple water budget model. It allowed for better calibration in identifying the canal and marsh seepage rates. The chloride model also proved to be essential in apportioning the percent of evapotranspiration that was transpiration. The chloride model was rather sensitive to this parameter especially in cell 3.

147

Richardson et al. (1990) indicated that there is a large central core area of water in the interior of the Refuge whose nutrient composition is typical of rain water atmospheric deposition, surrounded by an area with a higher nutrient composition affected by the pumped inflows to the perimeter canal. However, this was not the case in this model. It was found that although the interior cells were the most sensitive to the dry and wet depositions, they were not predominately driven by these parameters. For example, much of the chloride that came into cell 3 may have originated in the canal.

8.4 Chloride Model Future Developments

The modeled chloride results could be improved in the future by dividing the marsh into more cells. This would provide the ability to adjust parameters that the model is sensitive to such as the percent of evapotranspiration that is transpiration.

In the future the model should be run for the years 2005 to 2007 using data from the additional 39 enhanced water quality stations which were recently installed (Meselhe et al., 2005). The chloride model could also be extended to model other conservative or semi-conservative constituents such as sulphate, total nitrogen, and calcium.

8.5 Phosphorus Conclusions

The phosphorus WASP model overall proved to be a helpful tool in better understanding the mass transport of phosphorus within the Refuge. The model produced canal results

148

that were comparable to those found in the ELM v.2.1. model. All of the interior cells showed better statistics than the ELM v.2.1 model.

From the phosphorus WASP model it was determined that the k-c* may be too simple of a model. The model was able to capture the transients in the canal and cell 1, but it was unable to do so in cells 2 and 3.

The model was also simulated using different c* and k values for each cell. The results showed that the statistics were slightly improved, however the model was still unable to capture the transients in the interior cells.

8.6 Phosphorus Future Developments

By completing the phosphorus model it was found that the k-c* model may be too simple, therefore, it is suggested that future attempts in modeling phosphorus in the Refuge be completed using a more complex model. WASP offers an eutrophication model that uses a phosphorus cycle to directly model the constituent. This module was not used in these modeling attempts because it was more complex and the main objective of this report was to keep the models simple.

The phosphorus model may also be divided into multiple cells in the future.. The phosphorus model can also be run for the period 2005 to 2007 using the observation data from the 39 additional stations which were installed in 2004 (Meselhe et al., 2005).

149

Literature Cited

Abtew, W., Scott, R., Ciuca, V., 2005. Hydrology of the South Florida environment. Chapter 5 in 2005 South Florida Environmental Report. South Florida Water Management District and Florida Department of Environmental Protection, West Palm Beach, FL.

A.D.A. Engineering, SFWMD, 2005. Everglades Agricultural Area regional feasibility study for period 2010 – 2014. South Florida Water Management District, West Palm Beach, FL. Available online: http://www.sfwmd.gov/org/erd/longtermplan/eaapdf /EAA%20RFS%20Final%20Report.pdf.

Ahn, H., 1999a. Outlier detection in total phosphorus concentration data from south Florida rainfall. Journal of the American Water Resources Association, 35 (2), 301-310.

Ahn, H., 1999b. Statistical modeling of total phosphorus concentrations measured in South Florida rainfall. Ecological Modelling, 116 (1), 33-44.

Arceneaux, J. C, 2007. The Arthur R. Marshall Loxahatchee National Wildlife Refuge Water Budget and Water Quality Models. MS Thesis. University of Louisiana at Lafayette, Lafayette, LA. USA.

150

Arnold, J. G., Srinivasan, R., Muttiah, R. S., Williams, J. R., 1998. Large area hydrologic modeling and assessment - Part 1: Model development. Journal - American Water Resources Association, 34 (1), 73-89.

Bowie, G.L, Mills W.B., Porcella D.B., Campbell C.L., Pagenkopf J.R., Rupp G.L, Johnson K.M., Chan W.H., Gherini S.A., Tetra Tech Incorportaed, Chamberlin, C.E. 1985. Rates, Constants, and Kinetics Formulations in Surface Water Quality Modeling – Second Edition. U.S. Environmental Protection Agency, Environmental Research Laboratory Office of Reasearch and Development, Athen, GA.

Brandt, L. A., Harwell, M. C., Waldon, M. G., 2004. Work Plan: Water quality monitoring and modeling for the A.R.M. Loxahatchee National Wildlife Refuge. Arthur R. Marshall Loxahatchee National Wildlife Refuge, U.S. Fish and Wildlife Service, Boynton Beach, FL. Available online: http://sofia.usgs.gov/lox_monitor_model/ workplans/2004-2006_workplan.html#pdf.

Childers, D. L., Jones, R., Trexler, J. C., Buzzelli, C. P., Dailey, S., Edwards, A. L., Gaiser, E. E., Jayachandran, K., Kenne, A., Lee, D., Meeder, J. F., Pechmann, J. H., Renshaw, A., Richards, J., Rugge, M., Scinto, L. J., Sterling, P., Gelder, W. V., 2002. Quantifying the effects of low-level phosphorus additions on unenriched Everglades wetlands with in situ flumes and phosphorus dosing. The Everglades, Florida Bay, and Coral Reefs of the Florida Keys - An Ecosystem Sourcebook. CRC Press, Boca Raton, FL.

151

Childers, D. L., Doren, R.F., Jones, R., Noe, G.B., Rugge, M., Scinto, L.J. (2003). "Decadal change in vegetation and soil phosphorus pattern across the Everglades landscape." Journal of Environmental Quality 32: 344-362.

Chow, V. T., Maidment, D. R., Mays, L. W., 1988. Applied Hydrology. McGraw-Hill, New York.

Daroub, S., Stuck, J. D., Rice, R. W., Lang, T. A., Diaz, O. A., 2002. Implementation and verification of BMPs for reducing loading in the EAA and Everglades Agricultural Area BMPs for reducing particulate phosphorus transport. Phase 10 Annual Report, WM 754. Everglades Research and Education Center, Institute of Food and Agricultural Sciences, University of Florida, Belle Glade.

Desmond, G., 2003. South Florida high-accuracy elevation data collection project. FS162-96. U.S. Department of the Interior, U.S. Geological Survey, Reston, VA. Available online: http://sofia.usgs.gov/publications/fs/162-96/.

Douglas, M. S. 1947. The Everglades: River of Grass. Rinehart, New York.

Fitz, H. C., Wang, N., Godin, J., Sklar, F. H., Trimble, B., Rutchey, K., 2002a. Calibration Performance of ELM v2.1a: 1979-1995 Water Quality and Hydrology. Report to the RECOVER Model Refinement Team, South Florida Water Management

152

District, West Palm Beach, FL. Available online: http://www.sfwmd.gov/org/wrp/elm/ results/cal_ver/elm2.1/ELMcalibAnalysis_draft.pdf.

Fitz, H. C., Wang, N., Godin, J., Sklar, F. H., Trimble, B., Rutchey, K., 2002b. Everglades Landscape Model, agency/public review of ELM v. 2.1a: ELM developers’ response to reviews. Report to the RECOVER Model Refinement Team, South Florida Water Management District, West Palm Beach, FL. Available online: http://www.sfwmd.gov/org/wrp/elm/news/graphics/ELMreviewResponse _final.pdf.

German, E. R., 1999. Regional evaluation of evapotranspiration in the Everglades. 3rd International Symposium on Ecohydraulics, Salt Lake City, UT.

Gupta, R.S., 1989. Hydrology and Hydraulic Systems. Waveland Press, Inc., Prospect Heights, IL.

Habib, E.H., Meselhe, E.A., 2006. Stage-discharge relations for low-gradient tidal streams using data driven models. Jornal of Hydraulic Engineering, 132 (5), 482-492.

Harwell, M., Surratt, D., Waldon, M., Walker, B., Brandt, L. (2005) A.R.M. Loxahatchee National Wildlife Refuge Enhanced Water Quality Monitoring and Modeling – Interim Report. April, 2005. 106 pp.

153

Johnson, L., 1974. Beyond the Fourth Generation. University Press of Florida, Gainesville, FL.

Kadlec, R.H., Hammer, D.E., 1982. Pollutant Transport in Wetlands. Environmental Progress, 1 (3), 206-211.

Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. CRC Press, Inc, Boca Raton, FL.

Legates, D.R., McCabe Jr., G.J., 1999. Evaluating the use of “goodness-of-fit” measures in hydrologic and hydroclimatic model validation. Water Resources Research 35 (1), 233–242.

Light, S.S., Dineen, J.W., 1994. Water control in the Everglades: A historical perspective. Everglades: The Ecosystem and Its Restoration. St. Lucie Press, Delray Beach, FL.

Lin, S., 1979. The application of the Receiving Water Quantity Model to the Conservation Areas of South Florida. DRE-91, South Florida Water Management District, West Palm Beach, FL.

Lin, S., Gregg, R, 1988. Water budget analysis Water Conservation Area 1. Water Resources Division. South Florida Water Management District, FL.

154

Loucks, D. P., McVoy, C. W., 2004. Chapter 1: Introduction. Habitat Suitability Indices for Evaluating Water Management Alternatives. South Florida Water Management District, Office of Modeling, West Palm Beach, FL, 1-10.

MacVicar, T. K., Van Lent, T., Castro, A., 1984. South Florida Water Management Model Documentation Report. Technical Publication 84-3, South Florida Water Management District, West Palm Beach, FL.

Meselhe, E. A., Griborio, A., Arceneaux, J., 2006. Model Selection Report. LOXA05001, University of Louisiana at Lafayette, prepared for the Arthur R. Marshall Loxahatchee National Wildlife Refuge, USFWS, Lafayette, LA., Available online: http://sofia.usgs.gov/lox_monitor_model/advisorypanel/ Model_Selection_June_2006_ LOXA05-001.pdf.

Meselhe, E. A., Griborio, A. G., Gautam, S , Arceneaux, J.C., Chunfang, C.X., 2005. Hydrodynamic And Water Quality Modeling For The A.R.M. Loxahatchee National Wildlife Refuge, Phase 1: Preparation Of Data, Task 1: Data Acquisition and Processing. Report #LOXA05-014, University of Louisiana at Lafayette, prepared for the Arthur R. Marshall Loxahatchee National Wildlife Refuge, USFWS, Lafayette, LA. Available online: http://sofia.usgs.gov/lox_monitor_model/advisorypanel/ data_acq_report.html.

Mitsch, W.J, 1988. Productivity-hydrology-nutrient models of forested wetlands. Wetland Modeling. Elsevier, Amsterdam, 115-132.

155

Mitsch, W.J, Reeder, B.C., 1991. Modeling nutrient retention of a freshwater coastal wetland: estimating the roles of primary productivity, sedimentation, resuspension and hydrology. Ecological Modeling, 54, 151-187.

Montgomery, D.C., Runger, G.C., Hubele, N.F, 2001. Engineering Statistics - Second Edition. John Wiley and Sons, Inc., New York.

Munson, R., Roy, S., Gherini, S., McNeill, A., Hudson, R., Blette, V., 2002. Model Predication of the Effects of Changing Phosphorus Loads on the Everglades Protection Area. Water, Air, and Soil Pollution, 134 (1/4), 255-273.

Nash, J. E., Sutcliffe, J. V., 1970. River flow forecasting through conceptual models part I - A discussion of principles. Journal of Hydrology, 10, 282-290.

Neidrauer, C.J., 2004. Water Conservation Area Regulation Schedules. Available online: http://www.sfwmd.gov/org/ema/toc/archives/2004_08_26 wca_schedules _082604.pdf.

Raghunathan, R., Slawecki, T., Fontaine, T. D., Chen, Z., Dilks, D. W., Bierman, V. J., Jr, Wade, S., 2001. Exploring the dynamics and fate of total phosphorus in the Florida Everglades using a calibrated mass balance model. Ecological Modeling, 142 (3), 247259.

156

Rantz, S. E., 1982. Measurement and Computation of Streamflow: Volume 2. Computation of Discharge, Water-Supply Paper 2175, Available online: http://water.usgs.gov/pubs/wsp/wsp2175/.

Richardson, J. R., Bryant, W. L., Kitchens, W. M., Mattson, J. E., Pope, K. R., 1990. An evaluation of refuge habitats and relationships to water quality, quantity, and hydroperiod: A synreport report. Florida Cooperative Fish and Wildlife Research Unit. University of Florida, Gainesville, FL.

SFWMD, 2000a. Florida Coastal Everglades LTER Mapserver project. Available online: http://fcelter.fiu.edu/gis/everglades-map/.

SFWMD, 2000b. 2000 Everglades Consolidated Report. January 2000. South Florida Water Management District, West Palm Beach, FL.

SFWMD, 2003. SFWMM v5.0 Calibration (1984-1995) and Verification (19811983,1996-2000) Statistics for Stage Locations. Available online: http://www.sfwmd.gov /org/pld/hsm/models/sfwmm/v5.0/sfwmm_calib_verif_stat_v5.0_rc.pdf.

Scheidt, D., Stober, J., Jones, R., Thornton, K., 2000. South Florida Ecosystem Assessment: Everglades water management, soil loss, eutrophication and habitat. EPA 904-R-00-003, Available online: http://www.epa.gov/region4/sesd/sesdpub _completed.html, EPA.

157

Stober, J. D., Scheidt, R. J., Thornton, K., Ambrose, R., France, D., 1996. South Florida Ecosystem Interim Report. monitoring for adaptive management: implications for ecosystem restoration. EPA-904-R-96-008. U.S. EPA Science and Ecosystem Support Division, Atlanta, GA.

Tait, D., 1990. The University of Florida Adaptive Environmental Assessment Everglades Simulation Model user’s guide. Arthur R. Marshall, Jr. Laboratory, Department of Zoology, University of Florida, Gainesville, FL.

US Army Corps of Engineers Jacksonville District, 1994. Environmental assessment: modification of the regulation schedule Water Conservation Area No. 1. US Army Corps of Engineers, Jacksonville, FL.

USEPA, 2006. Water Quality Analysis Simulation Program (WASP). Available online: http://www.epa.gov/athens/wwqtsc/html/wasp.html.

USFWS, 2000. Arthur R. Marshall Loxahatchee National Wildlife Refuge Comprehensive Conservation Plan. Available online: http://loxahatchee.fws.gov, US Fish and Wildlife Service, Boynton Beach, FL.

USFWS, 2007. Arthur R. Marshall Loxahatchee National Wildlife Refuge: Overview of Northern Everglades Ecosystem. Available online: http://www.fws.gov/loxahatchee /Refuge/overview-ecosystem.asp. US Fish and Wildlife Service, Boynton Beach, FL.

158

USFWS. (2007b). A.R.M. Loxahatchee National Wildlife Refuge - Enhanced Monitoring and Modeling Program – 2nd Annual Report – February 2007. LOXA06-008, U.S. Fish and Wildlife Service, Boynton Beach, FL. 183 pp.

Walker, W.W., 1995. Design for Everglades Stormwater Treatment Areas. Water Res. Bull. 31 (4), 671-685.

Walker, W. W., Jewell, S. D., 1997. Atmospheric deposition of phosphorus in Loxahatchee National Wildlife Refuge. Atmospheric Deposition in South Florida: Measuring Net Atmospheric Inputs of Nutrients.

Walker, W.W. Kadlec, R.H., 2006, Dynamic Model for Stormwater Treatments Areas Version 2. Available online: http://wwwalker.net/dmsta/index.htm.

Walters, C., 1990. The University of Florida Adaptive Environmental Assessment Everglades Simulation Model. Arthur R. Marshall, Jr. Laboratory, Department of Zoology, University of Florida, Gainesville, FL.

Wang, N., Mitsch, W.J., 2000. A detailed ecosystem model of phosphorus dynamics in created Riparian Wetlands. Ecological Modeling. 126, 101-130

Welter, D., 2002. Loxahatchee National Wildlife Refuge HSE model. South Florida Water Management District, West Palm Beach, FL.

159

APPENDIX A: Removed Chloride and Phosphorus Outliers

160

Station

Date

Chloride Concentration mg/L

Reading Used, mg/L or Eliminated 89.189

Comments

S-5AS 28-May-98 28-May-98

88.599 89.779

S-39 10-Apr-95 10-Apr-95

120.967 101.829

111.398

G-251 12-Jul-99

830.67

Eliminated

This value was found to be an outlier the next highest value recorded was found to be 264.19

Two Readings for this date - the average of the two was taken Two Readings for this date - the average of the two was taken

S-6

8-Jul-97

755.96

Eliminated

This value was found to be an outlier the next highest value recorded was 275 mg/L - Conductivity* was 1247 Siemens (giving an approximate Cl value of 157.9 mg/L) confirming the elimination.

S-10C

8-Jul-97

633.929

Eliminated

This value was found to be an outlier, the next highest value was recorded to be 167.494 mg/L. Conductivity* was 1105 Siemens (giving an approximate CL of 139.9 mg/L) confirming the elimination.

S-6

27-Mar-00

131.307 141.437

136.372

S-6

11-Apr-00

148.274 144.096

146.185

Two Readings for this date - the average of the two was taken

Two Readings for this date - the average of the two was taken

*

A conductivity constant of 0.1266 was determined by averaging the daily chloride concentrations divided by daily conductivy values over for all of the stations over the POR. This constant times a daily conductivity value gives an estimated Chloride value. Table A.1: Chloride outlier values; and dates and values when there were more than recording.

161

Station

Date

S-5AS

28-May-98 28-May-98

Phosphorus Concentration mg/L 0.039 0.038

G-310

1-Jun-00 1-Jun-00 1-Jun-00 1-Jun-00

Reading Used, mg/L or Eliminated

Comments

0.0385

Two Readings for this date - the average of the two was taken

0.035 0.023 0.022 0.021

0.02525

Four Readings for this date - the average of the two was taken

G-310

8-Jun-00 8-Jun-00 8-Jun-00 8-Jun-00 8-Jun-00

0.016 0.016 0.017 0.016 0.017

0.0164

Four Readings for this date - the average of the two was taken

S-6

11-Apr-00 11-Apr-00

0.017 0.019

0.018

Two Readings for this date - the average of the two was taken

27-Mar-00 0.187 0.1595 Two Readings for this date - the average of the two was taken 27-Mar-00 0.132 Table A.2: Dates and values of days when there were more than one phosphorus reading at an inflow or outflow structure.

S-6

162

Reading Used, mg/L or Eliminated

Comments

0.027

Two Readings for this date - Used .027 mg/L because .006 mg/L was considered to be an extreme value - The previous data reading was .041 mg/L and the following data reading was .022 mg/L therefore it was appropriate to use .027 mg/L

Date

Phosphorus Concentration mg/L

10-Apr-95

0.006

10-Apr-95

0.027

G-251(G)

1-Jun-99

0.017 0.015

0.016

Two Readings for this date - the average of the two was taken

G-251(G)

6-Jul-99

0.012 0.013

0.0125

Two Readings for this date - the average of the two was taken

G-251(G)

7-Sep-99

0.02 0.021

0.0205

Two Readings for this date - the average of the two was taken

G-251(G)

2-Nov-99

0.01 0.15 0.015

0.0125

Three Readings for this date - the .15 was thrown out because of its extreme value - the average of the other two value was used

Station

S-39

0.025 Two Readings for this date - the average of the two was taken 0.0235 0.022 Table A.3: Dates and values of days when there were more than one phosphorus reading at an inflow or outflow structure.

G-251(G)

18-Jan-00

163

APPENDIX B: Daily Chloride Excel Model Results

164

300 275

Chloride Concentration, mg/L

250 225 200 175 150 125 100 75 50 25 0 Jan-95 X0

Jan-96 Z0

S39

Jan-97 S10E

Jan-98 S10D

S10C

Jan-99 S10A

G94B

Jan-00 Modeld

Figure B.1: Chloride Excel model results for the canal for the calibration period January 1, 1995, to December 31, 1999. 300 275

Chloride Concentration, mg/L

250 225 200 175 150 125 100 75 50 25 0 Jan-00 X0

Jan-01 Z0

S39

Jan-02 S10E

Jan-03 S10D

S10C

Jan-04 S10A

Jan-05 G94B

Modeld

Figure B.2: Chloride Excel model results for the canal for the validation period January 1, 2000, to December 31, 2004. 165

200 180

Chloride Concentrations, mg/L

160 140 120 100 80 60 40 20 0 Jan-95 X1

Jan-96 Z1

Jan-97 Z2

Jan-98 LOX6

Jan-99 LOX 4

Jan-00 Modeled

Figure B.3: Chloride Excel model results for the cell 1 for the calibration period January 1, 1995, to December 31, 1999. 200 180

Chloride Concentrations, mg/L

160 140 120 100 80 60 40 20 0 Jan-00 X1

Jan-01 Z1

Jan-02 Z2

Jan-03 LOX6

Jan-04 LOX 4

Jan-05 Modeled

Figure B.4: Chloride Excel model results for the cell 1 for the validation period January 1, 2000, to December 31, 2004. 166

200 180 Chloride Concentration mg/L

160 140 120 100 80 60 40 20 0 Jan-95 LOX4

Jan-96 LOX10

LOX12

Jan-97 LOX14

LOX15

Jan-98 LOX16

X2

Jan-99 Y4

Z3

Z4

Jan-00 LOX 6

Modeled

Figure B.5: Chloride Excel model results for the cell 2 for the calibration period January 1, 1995, to December 31, 1999. 200 180 Chloride Concentration mg/L

160 140 120 100 80 60 40 20 0 Jan-00 LOX4

Jan-01 LOX10

LOX12

Jan-02 LOX14

LOX15

Jan-03 LOX16

X2

Jan-04 Y4

Z3

Z4

Jan-05 LOX 6

Modeled

Figure B.6: Chloride Excel model results for the cell 2 for the validation period January 1, 2000, to December 31, 2004. 167

200 180 Chloride Concentration, mg/L

160 140 120 100 80 60 40 20 0 Jan-95 LOX3

Jan-96 LOX5

LOX7

Jan-97 LOX8

Jan-98 LOX9

LOX11

Jan-99 LOX13

Jan-00 X4

Modeled

Figure B.7: Chloride Excel model results for the cell 3 for the calibration period January 1, 1995, to December 31, 1999. 200 180 Chloride Concentration, mg/L

160 140 120 100 80 60 40 20 0 Jan-00 LOX3

Jan-01 LOX5

LOX7

Jan-02 LOX8

Jan-03 LOX9

LOX11

Jan-04 LOX13

Jan-05 X4

Modeled

Figure B.8: Chloride Excel model results for the cell 3 for the validation period January 1, 2000, to December 31, 2004. 168

APPENDIX C: Daily Chloride WASP Model Results

169

300

Chloride Concentration, mg/L

275 250 225 200 175 150 125 100 75 50 25 0 Jan-95 X0

Jan-96 Z0

S39

Jan-97 S10E

Jan-98

S10D

S10C

Jan-99 S10A

G94B

Jan-00 Modeld

Figure C.1: Chloride WASP model results for the canal for the calibration period January 1, 1995, to December 31, 1999. 300

Chloride Concentration, mg/L

275 250 225 200 175 150 125 100 75 50 25 0 Jan-00 X0

Jan-01 Z0

S39

Jan-02 S10E

Jan-03

S10D

S10C

Jan-04 S10A

G94B

Jan-05 Modeld

Figure C.2: Chloride WASP model results for the canal for the validation period January 1, 2000, to December 31, 2004. 170

200

Chloride Concentrations, mg/L

180 160 140 120 100 80 60 40 20 0 Jan-95 X1

Jan-96 Z1

Jan-97 Z2

Jan-98 LOX6

Jan-99 LOX 4

Jan-00 Modeled

Figure C.3: Chloride WASP model results for the cell 1 for the calibration period January 1, 1995, to December 31, 1999. 200

Chloride Concentrations, mg/L

180 160 140 120 100 80 60 40 20 0 Jan-00 X1

Jan-01 Z1

Jan-02 Z2

Jan-03 LOX6

Jan-04 LOX 4

Jan-05 Modeled

Figure C.4: Chloride WASP model results for the cell 1 for the validation period January 1, 2000, to December 31, 2004.

171

200

Chloride Concentration mg/L

180 160 140 120 100 80 60 40 20 0 Jan-95

Jan-96

LOX4 X2

LOX10 Y4

Jan-97

Jan-98

LOX12 Z3

LOX14 Z4

Jan-99 LOX15 LOX 6

Jan-00 LOX16 Modeled

Figure C.5: Chloride WASP model results for the cell 2 for the calibration period January 1, 1995, to December 31, 1999. 200

Chloride Concentration mg/L

180 160 140 120 100 80 60 40 20 0 Jan-00

Jan-01

LOX4 X2

LOX10 Y4

Jan-02

Jan-03

LOX12 Z3

LOX14 Z4

Jan-04 LOX15 LOX 6

Jan-05 LOX16 Modeled

Figure C.6: Chloride WASP model results for the cell 2 for the validation period January 1, 2000, to December 31, 2004.

172

200

Chloride Concentration, mg/L

180 160 140 120 100 80 60 40 20 0 Jan-95 LOX3 LOX13

Jan-96 LOX5 X4

Jan-97 LOX7 Modeled

Jan-98 LOX8

Jan-99 LOX9

Jan-00 LOX11

Figure C.7: Chloride WASP model results for the cell 3 for the calibration period January 1, 1995, to December 31, 1999. 200

Chloride Concentration, mg/L

180 160 140 120 100 80 60 40 20 0 Jan-00 LOX3 LOX13

Jan-01 LOX5 X4

Jan-02 LOX7 Modeled

Jan-03 LOX8

Jan-04 LOX9

Jan-05 LOX11

Figure C.8: Chloride WASP model results for the cell 3 for the validation period January 1, 2000, to December 31, 2004.

173