Citation
Effects of Water Retention on the Water Dynamics of Wetland-Upland Systems on the Ranchlands in the Lake Okeechobee Watershed, Florida

Material Information

Title:
Effects of Water Retention on the Water Dynamics of Wetland-Upland Systems on the Ranchlands in the Lake Okeechobee Watershed, Florida
Creator:
Wu, Chin-Lung
Publisher:
University of Florida
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Agricultural and Biological Engineering
Committee Chair:
SHUKLA,SANJAY
Committee Co-Chair:
GRAHAM,WENDY DIMBERO
Committee Members:
BROWN,MARK T
KIKER,GREGORY A
ANNABLE,MICHAEL D
Graduation Date:
5/3/2014

Subjects

Subjects / Keywords:
Constructed wetlands ( jstor )
Groundwater ( jstor )
Groundwater flow ( jstor )
Groundwater level ( jstor )
Hydrological modeling ( jstor )
Rain ( jstor )
Rainy seasons ( jstor )
Surface water ( jstor )
Water flow ( jstor )
Wetlands ( jstor )
evapotranspiration
modeling
ranchland
retention
wetland
Miami metropolitan area ( local )

Notes

General Note:
Ranchland water retention (WR) on the wetland-upland systems in the Lake Okeechobee (LO) watershed, for reducing damaging flows to the lake, was evaluated. Water retention was implemented by raising the spillage level at the outlets of a shallow and deep wetland site. Climatic and hydrologic data were collected at two sites to: 1) quantify the evapotranspiration (ET) for two wetlands using Eddy-Covariance (EC) method and develop ET models; 2) construct water budgets to quantify groundwater fluxes; and 3) use a field-verified model, MIKE-SHE/MIKE11, to evaluate different levels of WR with regards to volume of surface and subsurface flows. EC-based ET for the deep and shallow wetlands was 127cm/year and 84cm/year, respectively, and accounted for 93% and 62% of annual rainfall (136cm). Use of commonly used crop coefficient (KC) method with literature KC for estimating ET resulted in 23% error in ET, highlighting the importance of deriving wetland KC for improved ET estimates. Two regression models were developed, one for predicting KC (R2 = 0.58-0.80) and another for daily ET predictions (R2 = 0.80). The multi-site evaluation of MIKE-SHE indicated good to very good performance (Nash-Sutcliffe Efficiency, E = 0.70-0.90) on predicting surface water levels and less than satisfactory to very good (E = 0.38-0.79) for groundwater levels. Different levels of WR can achieve surface flow reductions of 9 to 20cm (42-86% reductions compared to baseline) at the deep wetland site, while the reductions were 1 to 24cm (5-93% reductions) at the shallow wetland site. Although almost all the retained water left as groundwater, WR reduced the flow volume and peak flows. The scale-up analyses for the entire ranchlands within the LO watershed showed a reduction of 2.1cm in surface flow and represents a 22% storage target for the LO watershed. However, these reductions are likely to be much lower if there is a watershed-scale rise in groundwater levels, and it will take the spillage level of 110cm to achieve 2.1cm reduction. When combined with effects on ranch economic and ecology, results can be used to develop water storage strategies under current and changed climate in the greater Everglades.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Wu, Chinlung. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
5/31/2016

Downloads

This item has the following downloads:


Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EJLNYVEJM_BH664A INGEST_TIME 2014-10-03T21:26:11Z PACKAGE UFE0046243_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

1 EFFECTS OF WATER RETENTION ON THE WATER DYNAMICS OF WETLAND UPLAND SYSTEMS O N THE RANCHLANDS IN THE LAKE OKE E CHOBEE WATERSHED FLORIDA By CHIN LUNG WU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 4

PAGE 2

2 201 4 Chin Lung Wu

PAGE 3

3 To my f amily

PAGE 4

4 ACKNOWLEDGMENTS I am very grateful to my supervisor, Dr. Sanjay Shukla, for his motivation, guid ance, and continuous support on my research over the past 7 years. I would like to thank my co advisor, Dr. Wendy D. Graham for her help, time and invaluable advice s I would also like to thank my co mmittee members Dr. Michae l D. Annable, Dr. M ark T. Brown and Dr. Gregory A. Kiker Without their helps this study would not have been possible. Also, I would like to thank Mr. James M. Knowles, Dr. Debashish Goswami, Dr. Alphonce Guzha, Dr. Gregory S. Hendricks, Dr. Niroj Shrestha, and Ms. Asmita Shukla for their friendship s and support throughout the past few years, especially during some tough times. I would like to acknowledg e all the good friends for their friendship s Finally, I would like to thank my parents, siblin gs and my girlfriend, Dr. Ming Wei S. Kao, for their unconditional love, patience and encouragement. Without their understanding it would have been even more difficult for me to complete this research.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 2 EDDY COVARIANCE BASED EVAPOTRANSPIRATION FOR A SUBTROPICAL WETLAND ................................ ................................ .................... 34 Overview ................................ ................................ ................................ ................. 34 Materials and Methods ................................ ................................ ............................ 40 Site Description and Hydrology ................................ ................................ ........ 40 Eddy Covariance Measurement ................................ ................................ ....... 41 Flux Footprint Analysis ................................ ................................ ..................... 45 FAO Penman Monteith Based Evapotranspiration ................................ ........... 46 Eddy Covariance Based Vegetation Coefficient (K CW ) ................................ ..... 47 Statistical Analysis ................................ ................................ ............................ 47 Results and Discussion ................................ ................................ ........................... 48 Climate and Wetland Inundation ................................ ................................ ...... 48 En ergy Fluxes ................................ ................................ ................................ .. 50 Eddy Covariance Based Evapotranspiration ................................ .................... 51 Comparison with FAO Penman Monteith Based Evapotranspiration ............... 53 Wetland Vegetation Coefficient ................................ ................................ ........ 55 Ch apter Summary and Conclusions ................................ ................................ ....... 60 3 MEASUREMENTS AND MODELING OF EVAPOTRANSPIRATION FROM TWO HYDROGEOMORPHICALLY DISTINCT WETLANDS I N SUBTROPICAL FLORIDA ................................ ................................ ................................ ................ 73 Overview ................................ ................................ ................................ ................. 73 Materials and Methods ................................ ................................ ............................ 78 Background ................................ ................................ ................................ ...... 78 Meteorological and Hyd rologic Measurements ................................ ................. 81 Eddy Covariance Measurements ................................ ................................ ..... 82 Flux Footprint Analysis ................................ ................................ ..................... 84 Developm ent of Wetland Vegetation Coefficient ................................ .............. 85 Evapotranspiration from Literature Based Crop Coefficient .............................. 86 Multivariate Evapotranspiration Model ................................ ............................. 86

PAGE 6

6 Evaluation of Evapotranspiration Methods ................................ ....................... 87 Results and Discussion ................................ ................................ ........................... 88 Climate and Hydrology ................................ ................................ ..................... 88 Evapotranspiration ................................ ................................ ........................... 92 Wetland Vegetation Coefficient ................................ ................................ ........ 94 Mu ltivariate Regression Evapotranspiration Model ................................ .......... 96 Comparisons of Evapotranspiration Methods ................................ ................... 98 Chapter Summary and Conclusions ................................ ................................ ....... 98 4 WATER BUDGETS AND GROUNDWATER FLUXES FOR TWO WETLAND UPLAND SYSTEMS IN LAKE OKEECHOBEE WATERSHED ............................. 107 Overview ................................ ................................ ................................ ............... 107 Materials and Methods ................................ ................................ .......................... 111 Site Description and Data Collection ................................ .............................. 111 Water Budget ................................ ................................ ................................ 114 Uncertainty Analysis ................................ ................................ ....................... 116 ................................ ............................ 118 Results and Discussion ................................ ................................ ......................... 120 Rainfall ................................ ................................ ................................ ........... 120 Evapotranspiration and Surface Flow ................................ ............................. 121 Water Storage ................................ ................................ ................................ 122 Water Budget ................................ ................................ ................................ 123 Groundwater Flux ................................ ................................ ........................... 126 Uncertainty Analysis ................................ ................................ ....................... 128 Chapter Summary and Conclusions ................................ ................................ ..... 129 5 SIMULATING THE EFFECTS OF WATER RETENTION ON WATER DYNAMICS OF RANCHLANDS IN THE NORTHERN EVERGLADES WATERSHED ................................ ................................ ................................ ....... 139 Overview ................................ ................................ ................................ ............... 139 Materials and Methods ................................ ................................ .......................... 148 Site Description and Instrumentation ................................ .............................. 148 Model Description ................................ ................................ ........................... 150 Model Development ................................ ................................ ....................... 151 Topography ................................ ................................ ................................ .... 151 Overland Flow and Channel Flow ................................ ................................ .. 152 Evapotranspiration ................................ ................................ ......................... 153 Unsaturated Flow ................................ ................................ ........................... 154 Saturated Zone ................................ ................................ ............................... 155 Model Calibration and Validation ................................ ................................ .... 155 Water Retention Alternatives ................................ ................................ .......... 158 Results and Discussion ................................ ................................ ......................... 159 Calibration ................................ ................................ ................................ ...... 159 Validation ................................ ................................ ................................ ........ 164 Effects of Water Ret ention Alternatives ................................ .......................... 166

PAGE 7

7 Chapter Summary and Conclusions ................................ ................................ ..... 178 6 SUMMARY AND CONCLUSIONS ................................ ................................ ........ 202 LIST OF REFERENCES ................................ ................................ ............................. 209 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 225

PAGE 8

8 LIST OF TABLES Table page 2 1 Components of the eddy covariance system ................................ ...................... 62 2 2 Evapotranspiration and vegetation coefficient values from wetland studies ....... 63 3 1 Components of the eddy covariance system at the deep wetland (DW) and the shallow wetland (SW) ................................ ................................ ................. 101 3 2 Monthly average reference ET (ET 0 ), actual ET (ET C ), and wetland vegetation coefficients (K CW ) for the deep and shallow wetland s ..................... 101 3 3 Comparison between EC based ET and ET estimated using multivariate regression model (ET CR ), average K CW (ET CW ) and K C from literature (ET CL ) .. 102 3 4 Comparison between EC based ET and ET estimated using multivariate regression model (ET CR ), average K CW (ET CW ) and K C from literature (ET CL ) .. 102 4 1 Soil type and associated areas in Site1 and Site4 ................................ ............ 132 4 2 Annual water budget for Site1 for two years (May 2009 April 2011) ............... 132 4 3 Annual water budget for Site4 for two years (May 2009 April 2011) ............... 132 4 4 Seasonal water budget components for Site1 ................................ .................. 132 4 5 Seasonal water budget components for Site4 ................................ .................. 133 5 1 Land use areas and associated MIKE SHE parameters ................................ ... 183 5 2 Monthly leaf area index ................................ ................................ .................... 183 5 3 Soil type, associated areas, and physical properties at the study site .............. 183 5 4 Average elevations for upland and wetland areas for Site1 and Site4, and the spillage elevations at the culvert and riser board structure ............................... 183 5 5 Comparisons of mean, root mean square error (RMSE), coefficient of determination (R 2 ), index of agreement (d), and Nash Sutcliffe coefficient (E) 184 5 6 Comparisons of the simulated net groundwater fluxes and the net groundwater fluxes estimated as residual term in the water budget ................. 184 5 7 Predicted surface flows for baseline and other water retention alternatives ..... 185 5 8 Simulated water budget components for baseline and other water retention alternatives for Site1 ................................ ................................ ......................... 186

PAGE 9

9 5 9 Simulated water budget components for baseline and other water retention alternatives for Site4 ................................ ................................ ......................... 187 5 10 Simulated maximum peak flow and peak flow for major rainfall events for baseline and other water retention alternatives ................................ ................ 188 5 11 Maximum inundated area for Site1 and Site4 for baseline and other water retention alternatives ................................ ................................ ........................ 189 5 12 Comparisons of simulated total surface flows for baseline and other water retention alternatives under the zero groundwater flux boundary condition ...... 190

PAGE 10

10 LIST OF FIGURES Figure page 2 1 Locations of study site and the eddy covariance station ................................ ..... 64 2 2 The distribution of inundation within the flux footprint area for the dominant wind direction (southwest) on December 20, 2010 ................................ ............. 64 2 3 Meteorological data collected during the study period (October 2008 May 2011) ................................ ................................ ................................ .................. 65 2 4 Weekly average soil moisture (February 2009 May 2011) at 3 cm depth at the eddy covariance station ................................ ................................ ................ 66 2 5 Weekly average percent inundation for the flux footprint for the study period .... 66 2 6 Daily average heat fluxes estimated at the study site during the study period ... 67 2 7 Monthly average heat fluxes and flux ratios for the study period ........................ 68 2 8 Monthly average ET 0 (mm/d). ET C EC (mm/d), ET C PM (mm/d) and K CW .............. 69 2 9 Comparisons of EC based ET and reference ET for the study period ................ 69 2 10 Comparisons of EC based ET and PM based ET for the study period ............... 70 2 11 Monthly wetland vegetation coefficient (K CW ) from this study and cattail and open water crop coefficient from Mao et al. (2002) ................................ ............. 70 2 12 Relationships between K CW and climatic and hydrologic variables, and measured and modeled K CW during the dry season ................................ ........... 71 2 13 Relationships between K CW and climatic and hydrologic variables, and measured and modeled K CW during the wet season ................................ ........... 72 3 1 Maps of the study site and topography for two wetlands ................................ .. 103 3 2 Monthly total rainfall and monthly average net radiation for two wetlands during the study period (June 2009 May 2011) ................................ .............. 104 3 3 Surface water depth at the deep and shallow wetlands during the study period (June 2009 May 2011) ................................ ................................ ......... 104 3 4 Monthly average percent inundation for two wetlands during the study period (June 2009 May 2011) ................................ ................................ ................... 104 3 5 Groundwater de pths for both wetlands during the study period (June 2009 May 2011) ................................ ................................ ................................ ......... 105

PAGE 11

11 3 6 Soil moisture at 3 cm depth at the eddy covariance station during the study period (June 2009 May 2011) ................................ ................................ ......... 105 3 7 Daily ET C for the deep and shallow wetlands during the study period (June 2009 May 2011) ................................ ................................ ............................. 105 3 8 Average wetland vegetation coefficients (K CW ) for the deep and shallow wetlands ................................ ................................ ................................ ........... 106 4 1 Aerial photo of the study site and its sub watersheds ................................ ....... 134 4 2 The locations of surface water wells (SU1, SU2), groundwater wells (GW26, GW32, GW33, GW42), and soil moisture stations (SM34, SM13) .................... 135 4 3 Monthly water budget components for Site1 ................................ .................... 136 4 4 Monthly water budget components for Site4 ................................ .................... 136 4 5 Monthly total precipitation and water storages for both sites ............................ 136 4 6 Relationships of stage volume below the top elevation of the culvert and riser board structure at Site1 and Site 4 ................................ ............................ 13 7 4 7 Surface water and groundwater levels at Site1 ................................ ................ 138 4 8 Surface water and groundwater levels at Site4 ................................ ................ 138 5 1 Watershed and sub watershed boundaries and the locations of the hydrologic monitoring systems at the study site ................................ ............... 191 5 2 Wetland water retention im plementation (control structure) at Site1 and Site4 192 5 3 Schematic representation of the components of MIKE SHE (Refsgaard et al. 1999) ................................ ................................ ................................ ................ 192 5 4 The 20 m digital elevation model (DEM) used in the MIKS SHE model. The DEM was derived from 1 m light detection and ranging (LIDAR) data ............. 193 5 5 Soil map for the study site ................................ ................................ ................ 193 5 6 Observed and simulated surface water levels behind the culvert and riser board structure at Site1 and rainfall for October 1, 2008 May 31, 2011 .......... 194 5 7 Observed and simulated surface water levels behind the culvert and riser board structure at Site4 and rainfall for October 1, 2008 May 31, 2011 .......... 194 5 8 Observed and simulated groundwater levels for wells at Site1 (Figure 5 1) for the calibration and validation periods ................................ ............................... 195

PAGE 12

12 5 9 Observed and simulated surface flow at both sites for the calibration and validation periods ................................ ................................ .............................. 196 5 10 Simulated surface water flow for baseline and A5 water retention alternative (Table 5 4) ................................ ................................ ................................ ........ 197 5 11 Simulated surface water flow for baseline and A2 2 water retention alternative (Table 5 4) ................................ ................................ ...................... 197 5 12 Simulated surface flows and groundwater levels at the middle of wetland under baseline and A4 water retention alternative (Table 5 4) for Site1 ........... 198 5 13 Simulated surface flows and groundwater levels at the middle of wetland under baseline and A4 water retention alternative (Table 5 4) for Site4 ........... 199 5 14 Inundated areas corresponding to predicted maximum water level for baseline and other alternatives (Table 5 4) at Site1 ................................ ......... 200 5 15 Inundated areas corresponding to predicted maximum water level for baseline and other alternatives (Table 5 4) at Site4 ................................ ......... 201

PAGE 13

13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EFFECTS OF WATER RETENTION ON THE WATER DYNAMICS OF WETLAND UPLAND SYSTEMS ON THE RANCHLANDS IN THE LAKE OKEECHOBEE WATERSHED, FLORIDA By Chin Lung Wu May 201 4 Chair: Sanjay Shukla Cochair: Wendy D. Graham Major: Agricultural and Biological Engineering W ater retention (WR) on the wetland upland systems in the Lake Okeechobee (LO) watershed, for reducing damaging flows to the lake, was evaluated. Water retention was implemented by raising the spillage level at the outlets of a shallow and a deep wetland site Climatic and hydrologic data were collected at two sites to: 1) quantify the evapotranspiration ( ET ) for two wetlands using Eddy Covariance (EC) method and develop ET models; 2) construct water budgets to quant ify groundwater fluxes; and 3) use a field verified model, MI KE SHE/MIKE11, to evaluate different levels of WR with regards to volume of surface and subsurface flow s EC based ET for the deep and shallow wetlands were 127 cm /year and 84cm /year respectively, and accounted for 93 % and 62% of annual rainfall (136cm). Use of commonly used crop coefficient (K C ) method with literature K C for estimating ET result ed in 23% error in ET highlighting the importance of deriving wetland K C for improved ET estimates. Two regression models were developed, one for predicting K C (R 2 = 0.58 0.80) and another for daily ET predictions (R 2 = 0.80). The multi site evaluation of MIKE SHE indicated good to very good performance (Nash Sutcliffe

PAGE 14

14 Efficiency, E = 0.70 0.90) for predicting surface water level s and less than satisfactory to very good (E = 0.38 0.79) for groundwater levels. D ifferent levels of WR can achieve surface flow reductions of 9 to 20cm (42 86% reductions compared to baseline) at the deep wetland site while the reductions were 1 to 24cm (5 93% reductions) at the shallow wetland site Although almost all the retained w ater left as groundwater WR reduced the flow volume and peak flows. The scale up analyses for the entire ranchlands within the LO watershed showed a reduction of 2.1cm in surface flo w and represents a 22% storage target for the LO watershed. However, thes e reductions are likely to be much lower if the re is a watershed scale rise in groundwater levels and it will take the spillage level of 110cm to achieve 2.1cm reduction. When combined with effects on ranch economic s and ecology, results can be used to de velop water storage strategies under current and changed climate conditions in the greater Everglades.

PAGE 15

15 CHAPTER 1 INTRODUCTION Lake Okeechobee (LO) is a large, multi functional lake located at the center of the Kissimmee Okeechobee Everglades aquatic ecosys tem. The lake provides habitat for a wide variety of wading birds, migratory waterfowl and the federally endangered Everglades Snail Kite. It is a source of drinking water for surrounding municipalities, a backup water supply for the communities of lower east coast Florida, supplies irrigation water to the large Everglades Agricultural Area (EAA), and is a critical supplemental water supply for the Everglades. During the 20th Century, due to increasing demands for improved agricultural production and flood control for an exp a nding population, much of the land around LO was converted to agricultural use s (SFWMD, 2000 ; NRC 2008 ) To the north, dairy farms and beef cattle ranching became the major land uses ; in the south, sugar cane and vegetable farming incr eased rapidly (Harvey and Havens, 1999). About 62 percent of the area north of the lake is devoted to agricultural uses. Agricultural areas with in the LO watershed normally contain extensive networks of drainage ditches to prevent root damage under prolong ed condition of high water table ( Harvey and Havens, 1999 ) This construction of surface drainage networks has decreased the watershed storage and changed the volume and rate of flow of surface water that enters LO. As a result of drainage, a large part of the surface and groundwater storage, especially in the wetland has since been lost Wetlands account for up to 25 percent of the LO watershed ( Flaig and Havens, 1995 ). During the past 45 years, the coverage of wetlands in the watershed has decreased from approximately 25 percent to 15 percent, as drainage networks have been constructed for improved flood control and as more land has been converted to

PAGE 16

16 pasture for beef cattle production (Flaig and Havens, 1995; Flaig and Reddy, 1995; SFWMD, 1997). Wetlands in the cattle ranches served as storage areas for runoff resulting from storms and slowly release d this storage to downstream locations through surface and subsurface pathways. Since these wetlands were drained to increase pasture acreage, the storage capa city of these wetlands to store runoff has been partially lost. The wetland drainage also reduced the storage from the upland areas which contributed flow s to the wetlands. As a result of th e drainage network development large amounts of water collected f rom the uplands now move quickly into the lake during the wet season, which results in rapid water level rises in the lake. A consequence of these large flows during the wet season is a reduction in the amount and distribution of surface and groundwater st orage s in the watershed that extends to the dry season. Due to the artificial connections between LO and the St. Lucie Canal and the Cal oosahatchee River, e xcessively high peak flows especially during the wet season result in large releases from LO to th e east coast via the St. Lucie Canal and to the west via the Caloosahatchee River to prevent flooding (NRC, 200 8 ) Large discharges to the Caloosahatchee River and St. Lucie estuaries are adversely affecting the vitality of those ecosystems. Deterioration of th e se ecosystems results in loss of oyster reefs which provide important and productive habitats. Reduced dry season storage has resulted in frequent water shortages for natural areas (Gary et al., 2007) such as the upland and coastal wetlands and the estuarine systems. Reducing the excessively high levels in the lake and the associated water bodies will require increasing the water storage in the wetland s which also to some extent will increase storage in uplands

PAGE 17

17 Another impact of drainage has been the water quality degradation of the lake. Increased drainage has resulted in increasing the rate of Phosphorus (P) inputs from the uplands to the lake. Excess flows and P loading are the main concern s for ecosystem because it has been identified as t he key nutrient that contributes to the eutrophication of the lake (Davis and Marshall, 1975; Federico et al., 1981). Since the early 1970s, P loads and concentrations enter ing LO have more than doubled (James et al., 1995). The 1997 Lake Okeechobee Surfac e Water Improvement and Management (SWIM) Plan, developed by South Florida Water Management District (SFWMD), reported that the lake faced a serious problem of excessive P loading (Harvey and Havens, 1999). T he current average total P loading to LO is 387 metric tons/year ( 2008 2012, AF 2013 ). Total P concentrations in LO when first measured in the early 1970s were 40 to 50 parts per billion (ppb); however, as a result of the continued influx of elevated P loadings, concentrations have gradually increased to an average of 158 ppb ( 2002 to 2006 ) (Zhang et al., 2007). This rapid rise in P content has accelerated eutrophication of the lake (Havens et al., 1996) and has exacerbated a series of complex and interrelated ecological responses. A numeric goal of 40 ppb total P, as measured in the pelagic zone of the lake, was established by the s tate of Florida achieve this goal, a Total Maximum Daily Load (TMDL) of 140 metric tons P p er year was set. In 20 12 the incoming P loading was 377 metric tons/year ( SFWMD, 2013 ) which was 169 % higher than the TMDL. Different approaches have been proposed to increase water storage capacity and reduce P loadings from agricultural lands to the la ke. Increasing the water storage

PAGE 18

18 capacity within the LO watershed could help achieve reduction in high flows and levels in the wet season that require discharges to estuaries, meet water supply needs during the dry season, and prevent harmfully low levels in the lake during the dry season. Increased storage of runoff on the ranchland in the LO watershed is considered one of the alternatives to mitigate the harmful water levels in the lake. Regulatory agencies and nonprofit conservation organizations such a s The Nature Conservancy, have pursued the goal of increasing storage in the wetlands within the cattle ranches to increase the LO storage and help achieve the TMDL for the LO. Although ranchland water retention is being promoted as one of the B est M anagement P ractice s (BMPs) no data is available to confirm its effectiveness in reducing the flows from ranchland (Steinman et al., 2003). Water retention is generally defined as the prevention of stormwater runoff from being discharged into receiving waters by storing it in a storage area (Boman et al., 2002). Water is retained and stored until it is lost through percolation, removed by evapotranspiration (ET) by plants or through evaporation from the free water surface (Boman et al., 2002). From the standpoint of the LO watershed water retention can be defined as a practice or modification that can reduce the surface flows from cattle ranches. On ranch water retention could be pursued by retaining water in depression areas in a ranch by mod ifying the drainage infra structure. For example, water retention could be implemented at a wetland outlet or at the ranch outlet. While wetland water retention (WWR) may help reduce surface water discharges from a ranch to a certain level, combining it wit h upland water retention could increase the water retention volume. To optimize on ranch water retention, a combination of wetland and upland

PAGE 19

19 water retention may be needed to increase storage in the wetlands, ditches and adjacent upland areas and reduce t he flow volume and peak flows from c attle ranches. A large part of ranchland in the LO watershed has wetland upland connectivity, part of it is due to naturally occurring topographic gradients and the other is due to artificial drainage infrastructure that includes ditches and swales. Although WWR seems to be an attractive alternative, its effects on wetland/watershed water dynamics and their pathways in the flatter landscape of South Florida are not completely understood (Shukla et al., 2007). Due to the c omplexity of the wetland/watershed relationship, there is still uncertainty over the hydrologic budgets and the hydrologic functions of wetlands (Carter, 1986; Owen, 1995). Effect of water retention on wetland and upland hydroperiod and water budget compon ents (e.g. ET and groundwater flow) are not yet fully understood in the LO watershed and elsewhere. storage of surface water) generally leads to change in other components (e.g. groundwater discharge and recharge) in a contiguous area (Winter, 1988). Rehydrating a wetland by increasing the water discharge elevation (spillage level) can change water storage, surface and subsurface flows, and groundwater discharge and recharg e areas (Winter, 1988). Assessments of the cumulative effects of one or more of these changes are affected by the uncertainty in the measurements as well as the assumptions reg arding the hydrologic process. Wetlands in many hydrogeologic settings may appea r to be hydraulically isolated from a surface perspective, but they are not hydraulically isolated from a groundwater perspective (Winter and LaBaugh, 2003). Increasing water storage and its effects on water budget components of a wetland may change the lo cal

PAGE 20

20 hydraulic gradients and result in the groundwater flow to downstream water bodies and adjacent depression al areas through subsurface pathways However, these effects have not been field verified. For example, Nungesser and Chimney (2006) estimated wate r budgets in a constructed wetland in the Everglades in South Florida and found that the groundwater outflow was underestimated because the groundwater flow to adjacent levees through seepage was not included in the water budget. Although it is often assu med that WWR can reduce surface flows, it may at times reduce the wetland storage capacity and increase the total and peak flows after a rainfall event. I f wetlands are already saturated, they may have little capacity to store additional water (Verry and B oelter, 1979) Although several studies have addressed the site specific relationships between the wetland storage, drainage to stream flow, and ET, the se results may not be transferable to other sites especially for regions such as South Florida, which has nearly flat topography, sandy soils, and a shallow water table. One of the unique features of the South Florida hydrology is that overland flow mainly occurs as the saturation excess after the water table has reached the surface. Water budget provide s the framework from which one can investigate the linkages and fluxes between the hydrologic, biochemical, and ecological systems of the wetland and its relationship to the surrounding terrain (Drexler et al. 1999). T here have been few wetland water budget studies (Owen, 1995; Bradley, 1997; Mitsch and Gosselink, 2000; Nungesser and Chimney, 2006). Drexler et al. (1999), in a detailed study of a small peatland, concluded that there was a wide margin of err or in all components of the water budget, with the exception of precipitation. Review of several studies (e.g. Drexler et al. 1999; Bradley, 2002) suggests that the main error in

PAGE 21

21 estimating a water budget comes from two components ET and groundwater flux. Constructing a more precise water budget with improved ET and groundwater estimates could help enhance the understand ing of the effects of water retention on surface and subsurface flows in the LO watershed The importance of ET in the hydrologic cycle of different regions including Florida has long been recognized. In Florida, ET is second only to precipitation in magnitude (Nachabe et al., 2005). Approximately 70 percent of mean annual rainfall has been estimated to return to the atmosphere as ET in Flor ida (Tibbals, 1990; Sumner, 1996). The importance of ET is in its influence over surface water depth, temperature, areal extent of water coverage and inundation duration. Despite the importance of ET in the hydrologic cycle of a wetland, its magnitude and seasonal distribution is still not fully understood (Sumner, 2001). Several studies have addressed the measurement and estimation of wetland ET for specific ecosystems such as reedbeds freshwater marshes (Linacre, 1976) and bogs and fens (Ingram, 1983). Although considerable efforts have been focused on quantifying wetland ET, ET from the wetland upland system which dominates the ranchlands of South Florida remains poorly characterized (Souch et al., 1996). Due to the complexity of surface characterist ics and diversity of wetland types, there exists high uncertainty in ET estimates. E vapotranspiration estimates for wetlands are often highly variable depending on the variety of approaches and the substantial differences in their relative accuracies (Drex ler et al., 2004). Wetlands are challenging environments for estimating ET because of the lack of uniformity in wetland shape,

PAGE 22

22 surface coverage, hydrology and topography, and all these components have a significant effect on ET ( Drexler et al., 2004) Several approaches have been developed to estimate ET: 1) as a residual term in the water budget (Winter, 1981; LaBaugh, 1986); 2) using pan evaporation method (Hayashi et al., 1998); 3) using lysimeters (Abtew and Obeysekera, 1995; Lott and Hunt, 2001); a nd 4) by using any of several empirical equations (Winter, 1988). Large errors in estimating wetland ET are common when using pan evaporation method because the open water evaporation from a pan is not a surrogate for transpiration by plants (Winter, 1988) Studies of ET using lysimeters containing wetland vegetation are not common (Carter, 1986) due to problems such as resource intensive setup and difficulty in capturing plant, topographic, and hydrologic diversity. Empirical equations, such as Penman Mont eith (PM) (Monteith, 1965) equation and Priestley Taylor (PT) (Priestley and Taylor, 1972) equation, have been developed to estimate the potential ET (PET). T he PET can be generally defined as the amount of water that could evaporate and transpire from a v egetated landscape without restrictions other than the atmospheric demand (Penman, 1948). The PM equation, a theoretically based combination approach that incorporates energy and aerodynamic considerations, has been used for plant canopies having adequate fetch and a canopy distribution sufficient s are at the same height and temperature ( Drexler et al., 2004 ). Sensible heat is heat energy transferred between the surface and air when there is a difference in temperature between them. The heat used in the phase change from a liquid to a gas is called latent heat. The inaccuracy in the PM equation has been attri buted to not accounting for

PAGE 23

23 surface resistance of the exposed soil and water surface; therefore, it needs a site specific surface resistance coefficient which is usually not available for wetlands (Stannard, 1993; Kustas et al., 1996). The PT equation is based on the assumption that the effect of turbulence is small compared to the effect of radiation and has been found to lead to good estimates of ET rates for large regions, small regions under vegetative coverage, or regions with heterogeneous land cover at the scale of a few kilometers (Pauwels and Samson, 2006). The PM and PT equations could produce estimates of actual ET (AET) only when the PM canopy resistance or empirical term ( ) for the PT are provided (Sumner and Jacobs, 2005). Sumner and Jacobs (2005) studied a non irrigated pasture in North Florida using eddy covariance (EC) method and found that both PM and PT equations required calibration of site specific derivation of parameters for accurately estimating ET. The recent emergence of EC metho d has significantly increased the accuracy of ET estimates. The EC method is a conceptually simple, one dimensional approach for measuring the turbulent fluxes of vapor and heat above a surface. The EC sensor measures the vertical motions and admixtures (s uch as flux of heat, water vapor) between the surface and the atmosphere. The EC method estimates both latent heat and sensible heat flux es Sensible heat is equal to the mean air density multiplied by the covariance between deviations in instantaneous ver tical wind and temperature and converted to energy unit using the specific heat. Latent heat flux is equal to the latent heat of vaporization multiplied by the covariance between deviations in instantaneous vertical wind and vapor density. The EC technique offers several advantages to alternative water budget approaches (lysimeter or regional water budget) by providing

PAGE 24

24 better areal integration and less site disruption than lysimeters and eliminating the need to quantify other terms of a water budget (precip itation, deep percolation, runoff, and storage) (Sumner, 2001). It avoids soil surface heterogeneity issues by placing the sensors above the crop canopy, which makes it better suited for measuring ET for various types of vegetation including those found i n wetlands (Sumner, 1996, 2001; Gholz and Clark, 2002; Sumner and Jacobs, 2005; Jia et al., 2007). Several recent studies used EC technique to estimate ET in a variety of wetlands such as subtropical fresh marshes (Bidlake et al., 1996), a cypress swamp (B idlake et al., 1996), and temperate marshes (Souch et al., 1996; Souch et al., 1998; Bidlake, 2000). Pauwels and Samson (2006) measured ET rates for a wet sloping grassland in Belgium using EC method and noted that results were in good agreement with those measured using the Bowen ratio method during a 2.5 year study (R 2 = 0.89). Bowen ratio ( B ), the ratio of the sensible heat to the latent heat, is calculated from the differences in air temperature and vapor pressure measured at two heights above the crop canopy. Bowen ratio method is used to estimate ET by using measurements of net radiation, soil heat flux and calculated B The AET has traditionally been estimated by multiplying the reference ET with a crop coefficient, K C he rate of ET from a hypothetical reference crop with an assumed crop height of 0.12 m, a fi xed surface resistance of 70 s/m and an albedo of 0.23, closely resembling the ET from an extensive surface of green grass of uniform height, actively growing, well watered, and completely shading the ground." In the reference ET definition, the grass is specifically defined as the reference vegetation and this vegetation is assumed to be free of water stress and

PAGE 25

25 diseases (Maidment, 1993). The most widely accepted m easure of reference ET is the American Society of Civil Engineers (ASCE) standard reference ET in which reference ET is estimated using the PM equation for a broad expanse of short vegetation (Walter et al., 2000). The EC method has been used as a measure of AET to develop K C for different land uses and plant species. Sumner and Jacobs (2005) developed K C for non irrigated pasture in central Florida by using the EC method; the K C values ranged from 0.47 in January (dry season) to 0.92 in July (wet season). Jia et al. (2009) developed K C for Bahiagrass from the ratio of the measured AET using EC method and calculated reference ET using on site meteorological data in central Florida. Use of the EC based K C associated with a reference ET could help one to estimate AET when the direct measurements of ET from EC method are not feasible. Using EC method to estimate ET for a wetland can improve the wetland upland water budget analyses. If a model that includes t he climatic and hydrologic variables as well as reference ET to predict AET can be developed, it can further improve the transferability of the EC based data from one to another wetland upland system Improved AET estimates can help quantify the groundwate r component as the residual term which then can result in better evaluation of WWR Groundwater component can be estimated either as the residual term of the equation. R echarg e and discharge of groundwater is one of the most important attributes of a wetland. Groundwater flows from or to wetlands are usually the most difficult to quantify because of the complex nature of groundwater flow through heterogeneous materials such as peat (Drexler et al., 1999; Hunt et al., 1996). The most common

PAGE 26

26 approach of estimating groundwater fl uxes for wetlands is through a water budget method (set as a residual when the other components are known) (Mitch and Gosselink, 2000). However, this metho d is limited by the errors in the other water budget components, which may be larger than the groundwater flux itself (Winter, 1981). Groundwater fl ux could also be estimated by measuring the fluctuation of groundwater levels using pressure transducer Gro undwater wells placed around a wetland, can help determine the direction of groundwater fl ux and the hydraulic gradient which can calculations are often limited by the ability to accurately determine the hydraulic conductivity of the soil s and geologic layers (Winter et al., 1988). In the wetland systems, especially in flatter landscapes such as flatwoods, groundwater flux is difficult to be quantified due to the low to pographic relief and small hydraulic gradient Although direct and indirect methods of estimating groundwater fl ux have associated error, the results obtained from both methods could help one to better evaluate the accuracy of groundwater flux When measur ement errors in water budget components are small and hydraulic conductivity for soils is known in Darcian calculation, groundwater flux estimated from both methods should be close to each other. Although EC based ET combined with more precise groundwater fluxes improve water budget components alone using the measured data for a specific c ontrol elevation at the wetland outlet. For example, consider the case of a study where hydrologic monitoring is conducted for two years, one year each for a low (drained) and high

PAGE 27

27 (increased hydroperiod) control elevation. Even if one has the EC based ET data and groundwater fluxes, the relative difference in ET and groundwater fluxes between the two years cannot be fully attributed to WWR Year to year variability in rainfall and antecedent hydrologic conditions in the watershed can mask the effect of wat er retention on ET and groundwater flux Although the paired wetland design may partly remove the effect of rainfall, it is difficult if not impossible to find two similar wetlands that have similar soil s plant s topographic and hydrologic characteristics. Use of hydrologic models in conjunction with the water budget analys is discussed above is a better alternative to accurately evaluate the effects of WWR. A variety of models have been developed and used for simulating water and nutrient transport of a variety of landscapes throughout the world. These models vary in their accuracy depending on the extent to which the hydrological processes governing the water and nutrient cycling are included and the simplifying assumptions made to repres ent these hydrological processes. Although the building procedures and components of existing hydrologic simulation models are similar, one model may be very different from another in its capability to simulate a land use practice and its applicability to different regions. Several models have been widely used for simulating water quantity and quality, for example, DRAINMOD (Skaggs, 1978, 1999) FHANTM (Campbell et al., 1995) SWAT (Arnold et al., 1998) FLATWOODS (Sun et al., 1998) WAM (Bottcher et al., 2 002) MIKE SHE (Refsgaard and Storm, 1995) and ACRU2000 (Kiker et al., 2006) DRAINMOD (Skaggs, 1978, 1999) was developed to predict the effects of drainage and associated water management practices on water table depths, the soil

PAGE 28

28 water regime and crop y ields. It has been used to analyze the hydrology of certain types of wetlands and to define the minimum, or threshold, drainage intensity that would result in failure of a site to maintain the wetland ecosystem criterion. However, dynamic interactions betw een groundwater and pond areas are not included in the model (Mansell et al., 2000). FHANTM (Field Hydrologic And Nutrient Transport Model), a field scale model, was based on DRAINMOD with modifications to include simulation of water and P movement s on fla twood landscape in Florida (Campbell et al., 1995). SWAT ( S oil and W ater A ssessment T ool) (Arnold et al., 1998), a watershed scale model developed by United State Department of Agriculture (USDA) Agricultural Research Service (ARS), is used to predict th e impact s of land management practices on water, sediment, and agricultural chemical yields in large complex watersheds with varying soils, land use s and management conditions over long periods of time. However, water interactions between unsaturated and saturated zone s are not included in the model FLATWOODS (Sun et al., 1998), a distributed hydrologic model, was developed to study the hydrologic processes of the wetland/upland system for the flatwoods landscape in Florida. WAM (Watershed Assessment Model) is a GIS based hydrologic model which allows engineers and planners to assess the water quality of both surface water and groundwater based on land use s soils, climate, and other factors (Bottcher et al., 2002). MIKE SHE/MIKE 11 (Refsgaard and Stor m, 1995) is an advanced integrated hydrological modeling system. It simulates water flow in the land based phase of the hydrological cycle from rainfall to river flow, via various flow processes such as overland flow, infiltration into soils, ET from veget ation, and groundwater fl ux ACRU2000 ( A gricultural C atchments R esearch U nit), an object

PAGE 29

29 oriented hydrologic model, is a multi purpose, daily time step, physical conceptual model that can simulate stream flow, ET, and land cover/management and the abstract ion impact on water resources at a daily time step (Kiker et al., 2006). Some of the above mentioned models can simulate wetland water dynamics and have been used to evaluate the effects of human alterations on wetland hydrology. Wang et al. (201 1 ) modifie d SWAT to simulate the artificial water input to the designated wetlands in Qingdianwa depression in Tianjin, China and evaluate the effect of recharging wetlands on local hydrological cycle and estuary ecology. FLATWOODS (Sun et al., 1998) was used to ev aluate the hydrologic effects of forest management practices on a pine flatwood landscape of North Florida. Although a variety of models have been developed and used in the different regions of the world, these models have limitations with regards to their applicability in flat landscapes of South Florida. South Florida hydrology is unique especially with regards to the extensive network of ditches, low slope ( less than 6 cm/km Obeysekera et al., 1999 ) highly interactive surface and ground water systems, shallow water table condition (approximately 1.2 to 1.5 m below the surface) and sandy soils (hydraulic conductivity > 1.410 4 m/s Boman and Tucker, 2002 ) One of the unique aspects of South Florida hydrology is that the infiltration capacity of these s oils is rarely exceeded and overland flow occurs due to saturation excess after the water table has reached the surface. Only few models are accurately able to simulate the surface and ground water interactions in South Florida. WAM has been used to simul ate daily flows in several Florida watersheds such as the Suwannee River and Caloosahatchee River (Bottcher et al., 2003). However,

PAGE 30

30 lack of documentation of the WAM model and its simplified approach of cell to stream water and solute delivery, simplified i n stream water quality processes, inability to adequately represent small scale short term storm event impacts (especially in Florida, for simulating cell to cell int eraction s (i.e. the re infiltration of ponded water) and peak flow caused by the short term storm event (Graham et al., 2009). MIKE SHE/MIKE 11 was used to simulate the different hydrologic processes and their interactions in a storm water impoundment in a citrus grove in South Florida to assess the potential use of impoundments as sou rces of water supply (Jaber and Shukla, 2004). Jaber and Shukla (2004) noted that MIKE SHE/MIKE 11 was effective in modeling the complex surface/groundwater interactions in addition to the different hydraulic structures such as pumps and culverts in South Florida. Advantages of the MIKE SHE/MIKE 11 model over other existing models (e.g. DRAINMOD, SWAT) in simulating wetland hydrology are: 1) it offers options to simulate the infiltration processes and water movement in the unsaturated zone (e.g. Richards e q uation gravity flow, and two layer water balance ); and 2) the model can simulate both wetland and upland lateral movement of surface and ground water flows. ACRU2000 has been used to simulate South Florida hydrology. ACRU2000 was shown to adequately predi ct water table depths, soil moisture distributions, and saturation excess runoff in the flatwoods of the LO watershed (Martinez, 2008). However, d ue to lack of representation of ditch drainage network in the model, ACRU2000 cannot explicitly simulate the s torage of the drainage network and hydraulic structures including weirs, gates, bridges and culverts that are commonly found within wetland environments.

PAGE 31

31 Hydrologic models are usually calibrated to measured historical data (e.g. water levels and flows) to identify unknown model parameters such as crop coefficient in simulating ET process and saturated hydraulic conductivity in simulating water movement in unsaturated and saturated zones and assess model performance. However, the parameter uncertainty in m ost of hydrologic models degrades the utility of these models. E vapotranspiration and groundwater flux are the two main components models, accurate estimations of ET and groundwater flux are needed. Shoemaker et al. (2008) studied the sensitivity of wetland saturated hydraulic heads and water budgets to ET in South Florida and concluded that reliable estimates of ET are necessary for estimating a wetland water budget. Sumn er and Jacob (2005) conducted a study of estimation of pasture ET at a commercial farm in central Florida and indicated that ET estimates from the EC method can improve the quality of hydrologic model calibration through reduction in the uncertainty of the AET component of the model. Use of improved estimates of ET and groundwater flux could improve the accuracy of the model in two ways: 1) input as measured values to better calibrate model parameters (difficult but not impossible) ; and 2) comparison of the simulated values of ET and groundwater flux with the measured and estimated data and refine the conceptualization or parameterization of the model. Improved model calibrat ion can performance on simulat ing the hydrologic processes for the specific applications for example, evaluating the effects of WWR for a variety of control elevation alternatives Designing a WWR strategy requires evaluating different water retention alternatives with regards to peak flow reduction and volume of wat er retained.

PAGE 32

32 By using field verified model s different alternatives can be simulated, and the simulated results can be used to design on wetland water storage strategies. This study is aimed at filling the knowledge gaps regarding the hydrology of wetland upland system in the ranchlands in South Florida and the effects of WWR on surface flows, ET, and groundwater fl uxes This research attempts to answer the following questions concerning effects of WWR : a) ET and groundwater fl ux account for how much of th e water inflows and outflows? b) h ow does WWR affect ET, groundwater fl ux and surface water flows? c) w hat strategies can be designed to achie ve different water retention goals? The goals of this study are to use measured water components to quantify the water budget of wetland upland systems located in ranchlands of South Florida and use this to field verify the hydrologic model and use it to evaluate the effects of water retention under different water retention alternatives. Specific objectives include : 1. Quantify ET losses for the two wetlands using EC method, compare it with ET estimates from the commonly used ET method, and develop specific coefficients to improve the predictability of empirical methods 2. Construct a water budget using the EC based ET t o quantify groundwater flux and evaluate its accuracy by comparing it with the groundwater flux estimated from 3. Use the hydrologic model, MIKE SHE/ MIKE 11 in conjunction with soil, plant, topography and 3 year hydrologic data collected at the ranch to quantify the effects of different levels of WWR and select an optimum alternative for volume and/or peak flow reductions

PAGE 33

33 This dissertation has been organized into six chapters including this introduction Chapter 1 Chapter s 2 and 3 address obje ctive 1. Evapotranspiration from a deep wetland and a shallow wetland located in a ranchland in South Florida w ere quantified using EC method M onthly wetland vegetation coefficients were developed for ET estimation when EC based ET data are not available For the deep wetland, m ultivariate regression equations were developed to predict wetland vegetation coefficient s for the dry and wet seasons. For both wetlands, a multivariate regression equation was developed to predict daily ET for two wetlands. Chapte r 2 has been accepted and Hydrological Processes (Wu and Shukla, 2013 ) Chapter 4 is focused on constructing water budget s for two wetland upland systems and understanding the role of ET and groundwater flux es from these systems, and as such is organized to address objective 2. In Chapter 5, an integrated model, MIKE SHE/ MIKE 11 was calibrated and validated using field observations and the field verified model was used to simulate different level s of WWR Based on model predictions of r eductions in flow volume and peak flow and increment s in hydroperiod and inundated area, the optimum alternative is selected In Chapter 6 the main findings from this study are summarized an d recommendation s for future research are made.

PAGE 34

34 CHAPTER 2 EDDY COVARIANCE BASED EVAPOTRANSPIRATION FOR A SUBTROPICAL WETLAND Overview W 2 ), and their characteristics vary widely by ve getative type, climate, hydrology and topography (Mitsch and Gosselink, 2000). The size, development, persistence and ecology of wetlands are primarily controlled by the hydrologic processes (Mitsch and Gosselink, 2000). Understanding and quantifying diffe rent sinks and sources of water are critical for calculating the nutrient and chemical budgets (Lott and Hunt, 2001). Evapotranspiration (ET) as one of the two largest components of wetland water budgets (with precipitation) is critical to precise determin ation of water dynamics that affects ecology of wetlands (Abtew, 1996; Lott and Hunt, 2001). Evapotranspiration can vary depending on a variety of factors that include climatic conditions, vegetation cover, and proportion of open water surface among others Evapotranspiration rates from studies conducted worldwide across wetlands with varying climatic, hydrologic and vegetation characteristics have been shown to vary significantly (Allen, 1995; Souch et al., 1996; Bidlake, 2000; Mao et al., 2002; Peacock an d Hess, 2004; Sanchez Carrillo et al., 2004; Drexler et al., 2008; Zhou and Zhou, 2009). Li et al. (2009) reported a mean ET rate of 2.2 mm/day for a reed wetland in Northeast China, a warm, temperate region of semi arid monsoonal climate. Mao et al. (2002 ) measured ET rates for a freshwater marsh in subtropical Florida with mean ET of 3.2 mm/d. Drexler et al. (2008) estimated ET rates for a marsh in semi arid California, USA and reported a mean ET rate of 3.9 mm/d. Mean ET rate from a freshwater wetland in semi arid central Spain was estimated by Sanchez Carrillo et al. (2004) at 8

PAGE 35

35 mm/d, a rate much higher than reported in similar studies. Given such large variability in ET rates, quantification of ET for specific climatic and hydrologic conditions is cruci al for improved water budgets for wetlands as well as watersheds. Wetland plays an important role in South Florida, a subtropical climatic region and home to one of the largest wetlands, the Everglades (Fig ure 2 1). The watersheds of the Lake Okeechobee (L O) part of the Everglades watershed have shallow water tables and were dominated by wetlands prior to the construction of an extensive drainage network in the 19 0 0s for agricultural development. A large fraction of the water storage capacity of the water shed has been lost resulting in abrupt rise and fall of water levels during the wet (May October) and dry (November April) seasons. Within the LO watershed cattle ranching is the dominant land use (36 % of the area) of which 15 % is occupied by wetlands ( Tweel and Bohlen, 2008). Quantification of ET from these wetlands is crucial for developing state and federal plans to increase the watershed storage capacity in the Northern Everglades region to partially restore the natural flows and levels. Approximatel y 70 % of mean annual rainfall has been estimated to return to the atmosphere as ET in Florida (Sumner, 1996). The importance of ET is in its influence over surface water depth, temperature, areal extent of water coverage and inundation duration. Although several studies have been conducted in Florida to measure ET, most of these were for agricultural crops (Jia et al., 2007; Shukla et al., 2012). Few studies have been conducted in south Florida on quantifying the ET for wetland environments but these studi es have focused on large natural wetlands that have been mostly unaltered. Abtew (1996) used a lysimeter experiment to measure ET rates from cattail

PAGE 36

36 ( Typha domingensis ) and mixed marsh vegetation (Spikerush ( Eleocharis spp. ), Pickerel weed ( Pontederia cord ata ), Arrowhead ( Sagitteria latifolia ), Duckpotato ( Sagitteria lancifoha ), Maidencane ( Panicium hemitomon ), and Sawgrass ( Cladium jamaicense )) in the Fort Drum Marsh Conservation Area in south Florida. German (2000) estimated ET rates using eddy covariance (EC) and Bowen ratio energy balance method for nine natural wetlands in the Everglades region extending from south of the lake to the southern part of Everglades National Park (Fi gure 2 1). These two wetland ecosystems differ considerably from ranchland w etlands that have much smaller area as well as hydroperiods with different vegetation. In addition to the complexity of surface characteristics and diversity of wetland types, ET estimates can also differ significantly due to the relatively accuracy of the method used (Drexler et al., 2004). Compared to cropped areas (e.g. corn) wetlands are challenging environments for estimating ET because of the lack of uniformity in wetland shape, surface coverage, hydrology and topography. These variations have a significant effect on ET estimates (Drexler et al., 2004) Several approaches have been devel oped to estimate ET: 1) as a residual term in the water budget (Winter, 1981); 2) using pan evaporation method (Hayashi et al., 1998); 3) using lysimeters (Abtew and Obeysekera, 1995; Abtew, 1996; Lott and Hunt, 2001; Mao et al., 2002); 4) using EC method (Souch et al., 1996; Jacobs et al., 2002; Sanchez Carrillo et al., 2004); or 5) by using any of several empirical equations (Winter, 1988). Use of pan evaporation method commonly results in large errors in estimating wetland ET (Winter, 1988). Studies of E T using lysimeters containing wetland vegetation are not common due to problems such as resource intensive setup and difficulty in capturing plant, topographic, and hydrologic

PAGE 37

37 diversity. Wetlands often contain more than one type of vegetation with variable open water areas which can introduce uncertainty in estimating ET from lysimeter data. Empirical methods, such as Penman Monteith (PM) (Monteith, 1965) equation and Priestley Taylor (PT) (Priestley and Taylor, 1972) equation, have been used to estimate ET The PM equation, which incorporates energy and aerodynamic considerations, has been used to estimate ET for areas with uniform surface (plant he sources of sensible and latent heat are at the same height and temperature (Seller s et al., 1996; Drexler et al., 2004). Souch et al. (1996, 1998) used PM equation to estimate wetland ET in the Indiana Dunes National Lakeshore, Indiana, USA; Bidlake (20 00) used a fixed surface resistance in PM equation to estimate ET for a wetland in the Upper Klamath National Wildlife Refuge of south central Oregon, USA. However, the inaccuracy in the PM equation has been attributed due to inaccurate accounting of surfa ce resistance of plant canopies; therefore, it needs a site specific surface coefficient which is usually not available for wetlands (Stannard, 1993). The PT equation, which assumes that the effect of turbulence is small compared to the effect of radiation has been found to lead to good estimates of ET rates for regions with uniform vegetation cover or with land cover which is heterogeneous at the scale of a few square km (Shuttleworth, 1989, 1992). The PM and PT equations could produce estimates of actual ET (ET C ) only when the PM surface resistance or empirical term for the PT are provided (Sumner and Jacobs, 2005). Sumner and Jacobs (2005) in an ET study for a pasture in Florida found that both PM and PT equations required calibration of site specific de rivation of parameters. They noted that use of EC based crop coefficient (K C )

PAGE 38

38 can improve the accuracy of ET estimates. Although studies have reported the K C for few wetland vegetation located in specific climatic regions (Abtew and Obeysekera, 1995; Mao e t al., 2002; Drexler et al., 2008; Zhou and Zhou, 2009), these are for specific vegetation (e.g. cattails and sawgrass) which limits their use for wetlands which often have a variety of vegetation. The ability of the EC method to provide relatively accurat e areal ET estimates has made this method popular for wetlands. The EC method is a direct, non intrusive and areal integrated approach to measure and calculate turbulent fluxes within the atmospheric boundary layer. The EC sensor measures the vertical moti ons and admixtures (fluxes of heat, water vapor) between the surface and the atmosphere. The EC method estimates both latent and sensible heat flux es Sensible heat is the mean air density multiplied by the covariance between deviations in instantaneous ve rtical wind and temperature and converted to energy unit using the specific heat (Burba and Anderson, 2010). Latent heat flux is equal to the latent heat of vaporization multiplied by the covariance between deviations in instantaneous vertical wind and vap or density. The EC technique offers several advantages to alternative water budget approaches (lysimeter or water budget) by providing better areal integration and less site disruption than lysimeters and eliminating the need to quantify other terms of a w ater budget (precipitation, deep percolation, runoff, and storage) (Sumner, 2001). It avoids surface coverage heterogeneity issues by placing the sensors above the crop canopy, which makes it better suited for measuring ET for various types of vegetation i ncluding those found in wetlands (Sumner, 1996, 2001; Gholz and Clark, 2002; Sumner and Jacobs, 2005; Jia et al., 2007). The EC method has been used for a variety of wetlands such as

PAGE 39

39 subtropical fresh marshes (Bidlake et al., 1996), cypress swamp (Bidlake et al., 1996), peatlands (Hatala et al., 2012), and temperate marshes (Souch et al., 1996; Souch et al., 1998; Bidlake, 2000). Pauwels and Samson (2006) measured ET rates for 2.5 years for a wet sloping grassland at a pasture in Belgium using EC method and noted that results were in good agreement with those measured using Bowen ratio energy balance method. Bowen ratio energy balance method estimates latent heat using measurements of air temperature and humidity gradients, net radiation, and soil heat flux (Fritschen and Simpson, 1989). Due to the difficulties in measuring wetland ET, it is frequently estimated by the crop coefficient approach which involves multiplying the reference ET (ET 0 ) with specific K C values (ET C =ET 0 K C ) to obtain ET C (Allen et al., 1998). The crop coefficient approach has been the primary method for estimating ET for the last 40 years (Jensen 1973; Allen et al., 1998; Tasumi et al., 2005). It has been used for estimating ET for a variety of purposes (Kang et al., 2003; Vazquez and Fe yen 2003; Mutiibwa and Irmak, 2013) ranging from irrigation scheduling to field, farm, and watershed scale water budgets using hydrologic models such as MIKE SHE/MIKE11 (Jaber and Shukla, 2012). Reference ET is defined as "the rate of evapotranspiration f rom a hypothetical reference crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s/m and an albedo of 0.23, closely resembling the evapotranspiration from an extensive surface of green grass of uniform height, actively growing, wel l watered, and completely shading the ground" (Allen et al., 1998). The Food and Agricultural Organization (FAO) modified PM equation, here onwards termed as FAO PM, has become the most widely used method for determining ET 0 (Allen et al., 1998; Mao et al. 2002; Drexler et al,

PAGE 40

40 2008). The EC method has been used in conjunction with ET 0 to develop K C for different land uses and plant species (Mao et al., 2002; Sumner and Jacobs, 2005; Jia et al., 2007; Drexler et al., 2008; Zhou and Zhou, 2009). Sumner and Ja cobs (2005) developed the K C of pasture in central Florida by using the EC method; the K C values ranged from 0.47 in January to 0.92 in July. Use of the EC method for measuring ET for ranchland wetlands and deriving the K C values could help to improve the regional water budgets. The goal of this study is to quantify ET and develop vegetation coefficient ( K C W ) for a wetland containing mixed vegetation in a ranch in south Florida using the EC method. Specific objectives include: 1) to measure the EC based ET for 30 months; 2) to evaluate the differences between the EC based ET and the commonly used FAO PM method (with literature K C ), and 3) to develop K CW and a multivariate regression model to predict K CW as a function of hydrologic and meteorological variable s for estimating ET for similar wetlands. Materials and Methods Site D escription and H ydrology The wetland is located 13 km cattle ranch (area = 2.75 km 2 ) in subtropical central south Florida (Fig ure 2 1) with average annual rainfall of 1385 mm (NOAA, 2010). The soils in this region are sandy and mostly classified as flatwood soils which dominate the Florida landscape. Flatwood region is typically characterized by nearly level topography, shallow wate r table, and poor drainage. The wetland area is 0.21 km 2 It is a freshwater wetland and vegetation in the wetland includes Soft Rush ( Juncus effuses ), Maidencane, Pickerelweed, Cattail, Duckpotato, and Sawgrass. Leaf area index is 1.9 m 2 /m 2 and does not v ary

PAGE 41

41 considerably during the year. The wetland is surrounded by upland pasture (Fig ure 2 1) that contains Bahiagrass ( Paspalum notatum ). Hydrologic variables including the surface and groundwater levels and rainfall were measured during the study period. Ra infall was measured at the wetland outlet (220 m from the EC station). Rainfall was measured on a 15 minute frequency with a tipping bucket rain gauge (Model H 340a, Design Analysis Associates, Inc, Logan, UT). Groundwater and surface water levels in the w etland were measured using two pressure transducers (Solinst Levelogger, Solinist Canada Ltd., Ontario, Canada). During the study period, wetland water level varied from 8.27 m to 9.50 m (above the mean sea level, A MSL) with an average of 8.86 m. The topog raphy of the site, characterized by a combination of site survey as well as LIDAR, varied from 7.95 m to 10.46 m ( A MSL). The inundation area was determined by combining the water level measurements with the topographic data with in Arc GIS (v.10, ESRI, Redla nds, CA, USA). Eddy C ovariance M easurement The EC tower was installed in the center of the wetland with at least 200 m of fetch in all directions. The EC data were collected for the October 1, 2008 May 31, 2011 period. The EC station consists of: 1) a CSI (Campbell Scientific Inc., Logan, UT, USA) CSAT3 three dimensional sonic anemometer; 2) a CSI KH20 krypton hygrometer; 3) a CSI HMP45C temperature and relative humidity (RH) probe; 4) a CSI NR LITE net radiometer; 5) CSI HPF01 soil heat flux plate, thermoc ouples, and CSI 616 water content reflectometer; 6) a CSI CS300 pyranometer; and 7) a CSI 107 L temperature sensor. The heights of these sensors are presented in Table I. The sampling frequency was 10 Hz with 30 min averaging periods.

PAGE 42

42 The anemometer measur es fluctuations in wind speed and virtual temperature, and the hygrometer measures the fluctuations of water vapor density. The temperature and RH were measured using a CSI HMP45C sensor. Using the measured covariance between vertical wind speed and vapor density, the latent heat flux (LE) was calculated as: (2 1) where LE is latent heat flux (W/m 2 ); is the latent heat of vaporization (J/g); s the fluctuation in the water vapor density from the hygrometer (g/m 3 the vertical wind speed from the anemometer (m/s); overbar and primes indicate mean for the sampling interval (30 minute frequency) and deviations from mean, respectively. Using the covariance between the vertical wind speed and the air temperature, the se nsible heat flux (H) was calculated as: ( 2 2 ) where a is the air density (g/m 3 ); C p is the specific heat of moist air (J/g/ C); is the fluctuation in the air temperature ( C). Any misalignment of the sonic anemometer with the airstream was co rrected by applying 2 D coordinate rotation procedure (Tanner and Thurtell, 1969; Baldocchi et al., 1988). The 30 minute LE was corrected for temperature induced fluctuations in air density (Webb et al., 1980) and for the hygrometer sensitivity to oxygen ( Tanner and Green, 1989). The H w as corrected for differences between the virtual temperature and the actual air temperature (Schotanus et al., 1983). Corrected LE and H are termed as LE C and H C

PAGE 43

43 T he energy budget equation along with measured turbulent fluxes was used to estimate corrected LE and H by closing the energy budget (Twine et al., 2000): ( 2 3) where R n is the net radiation (W/m 2 ); G is the soil heat flux (W/m 2 ); W is water heat sto rage (W/m 2 ), and B is the B owen ratio (B) given by: (2 4) Rearranging ( 2 3) ( 2 5) ( 2 6) The soil heat flux was estimated from the sum of the heat flux measured by a heat flux plate (Table 2 I) and the change in heat storage above the plate as: ( 2 7) where G 8cm is the soil heat flux measured at depth of 8 cm (W/m 2 in energy in the soil above the plate (W/m 2 ). The S was calculated as follows (German, 2000): (2 8) S is the difference in average soil temperature for computation interval ( C); C S is the volumetric heat capacity of the soil (J/g/ C); d is the thickness of the soil layer (m); s : ( 2 9) where BD is the dry soil bulk density (kg/m 2 ); C sd is the specific heat of the dry soil (840;

PAGE 44

44 German, 2000) (J/kg/ C); C W is the specific heat of water (4190; German, 2000) (J/kg/ C); and X W is the mass basis soil water content (kg water/kg dry soil). present on the land surface. The heat stored or l ost by the water mass was estimated using the following equation (German, 2000): (2 10 ) 2 ); d W W is the change in mean water temperature in the time interval ( C); C W is the heat capacity of water (J/g/ water depth and mean water temperature at the beginning and end of each 30 min interval. The mean water temperature was calculated by averaging the water temperature on the surface and bottom (Table 2 I) Although water heat storage is relatively small compared to other energy terms and is negligible in the environment without standi ng water, it cannot be ignored in wetland environments (German, 2000; Sanchez Carrillo et al., 2004). Water heat storage was estimated when the soil surface at the station was submerged. For inundated and non inundat ed condition s total available energy (h ere onwards as R n G W ) was estimated as R n W and R n G, respectively. The 30 minute EC based ET (ET C EC mm) was calculated using: (2 1 1 ) where all the terms are as defined above. During nights, early mornings (with dew formati ons), and after rainfall, the hygrometer measurements are not reliable due to water drops on the KH20 lenses.

PAGE 45

45 While nighttime ET has been shown to occur for woody plants and shrubs (Novick et al ., 2009), nighttime LE at our site was mostly zero or negative and therefore, nighttime ET was assumed to be negligible. Quality assurance (QA) was achieved by the following procedure s The s ite was visit ed every two weeks to : clean the hygrometer sensor windows with a cotton swab to remove dust obstructions and restore the signal strength; inspect and clean the net radiometer, pyranometer, temperature and relatively humidity probe s and conduct routine equipment maintenance. Qua lity control (QC) was achieved by visually inspecting the plots of time series of EC data. Erroneous data as a result of instrument malfunction (low anemometer counts and low hygrometer voltage), unexplained spikes, and heavy rainfall events were removed. For short periods with missing data, LE can be estimated from a modified PT model ( Stannard, 1993; Sumner, 2001 ). Due to malfunction of hygrometer, ET could not be calculated for the July and August 2010. Flux F ootprint A nalysis The source area for flux me asurements is the upwind surface area contributing contribution from each element of the upwind surface area source to the measurement concentration or vertical flux (Schuepp et al., 1990). The footprint is sensitive to measurement height, wind direction, atmospheric stability and surface roughness (Leclerc and Thurtell, 1990, Schmid, 1997). Mathematically, the flux footprint area goes to infinity and thus one must alw ays define the % level for the source area. In most of the cases, 90% source areas contributing to a point flux measurement are considered. The determination of the flux footprint is complex and several theoretical models have been derived over the decades These footprint models can be classified into

PAGE 46

46 analytical models, Lagrangian stochastic particle dispersion models, large eddy simulations, and closure models. The footprint was estimated for the prevailing wind direction for each month. It was calculated using the monthly average wind component outputs from EC system (e.g. standard deviation of vertical velocity fluctuations, surface friction velocity) in conjunction with stochastic Largrangian model developed by Kljun et al. (2004). The footprint area wa s estimated by aggregating 90% of source area. The percent inundation within the flux footprint area was estimated by using the daily wetland water levels in conjunction with the Spatial Analyst extension in ArcGIS. For example, the distribution of inundat ion within the flux footprint area for the dominant wind direction (southwest) for December 20, 2010 was shown in Fig ure 2 2. FAO Penman Monteith B ased E vapotranspiration To evaluate errors in computing ET from empirical equations and literature K C the ET C EC was compared with the ET from the FAO PM method. Climatological data (air temperature, incoming solar radiation (R S ), R n wind speed, and RH) collected at the weather station at the site were used to calculate ET 0 (Allen et al., 1998). The equation to calculate ET 0 : ( 2 12) (kPa/ C); T is mean air temperature at 2 m height ( C); U 2 is the wind speed at 2m height (m/s); e s is saturation vapor pressure (kPa); and e a is the actual vapor pressure (kPa).

PAGE 47

47 The ET C PM was calculated using the K C values for cattail and open water from Mao et al. (2002). Mao et al. (2002) used three (single vegetation) lysimeters to estimate K C for cattail, sawgrass, and open water at a marsh located 60 km northeast from our site. Water table within the lysimeter was maintained to achieve saturation for the entire study period. Our wetland has mixed vegetation and varying water levels. However, due to limited published data on wetland K C using cattail and sawgrass K C approach usually taken for estimating ET for water budgets and modeling studies. The monthly K C values for K C was used to estimate ET for areas with plants. Area weighted K C for the wetland was computed using the vegetated and open water areas. Accurate topographic data are needed to capture w etland inundation dynamics (Guzha and Shukla, 2011). High resolution Light Detection and Ranging (LIDAR) based 1 m digital elevation model data for the site and measured surface water levels were used to determine inundated area. Eddy C ovariance B ased V eg etation C oefficient (K CW ) Monthly K CW for the wetland estimated using the daily values of ET C EC and ET 0 w as expressed as: (2 13 ) These K CW values were compared to literature values. Statistical A nalysis Multivariate regression analyses were conducted between K CW and climatic and other environmental variables to develop predictive equations for K CW that can be used

PAGE 48

48 for estimating K CW for site specific conditions for similar wetlands. Statistical significance of the dif ferences between both ET and K CW from our study and other studies were analyzed using two sample t test ( = 0.05). Following comparisons were made: daily ET 0 and ET C EC daily ET C EC and ET C PM monthly ET C EC and ET from literature, and monthly K CW and m onthly K C from literature. Results and Discussion Climate and W etland I nundation A nnual rainfall for the study period (October 2008 May 2011) varied considerably from 878 mm for 2010 (34% lower than the average) to 1360 mm for 2009 (close to the average) indicating 2010 to be drier than average. The regional average rainfall is 1362 mm/y ea r (1992 2011; NCDC, 2012) of which 960 mm falls during the wet season and 404 mm in the dry season. Rainfall for 2010 wet season (417 mm) was below average (57%) while it was above average (25% higher than average rainfall) for 2009 wet season. Dry period rainfall for 2010 was above average (14%, 461 mm) while it was below average (59%) for 2009. The February March 2010 period was unusually colder and wetter in the southea stern US including Florida (NCDC, 2012). Rainfalls in February (132 mm) and March of 2010 (192 mm) were 116% and 123% higher than historical averages and resulted in relatively higher inundation and therefore, increasing the potential for higher ET rates. The R S varied from 95 W/m 2 (January 2011) to 228 W/m 2 (May 2010) with average of 160 W/m 2 for the study period (Fig ure 2 3 B ). The highest R S occurred during the April June periods and decreased afterwards due to the increased cloud cover during the wet sea son. The maximum temperature and RH values occurred during the

PAGE 49

49 wet season. Vapor pressure deficit (VPD) values, indicator of actual evaporative demand, were higher during the wet season compared to the dry season (Fig ure 2 3 C ). The ET process is affected b y wind speed and air turbulence which facilitates transfer of air over the evaporating surface. Monthly average wind speed was 1.57 m/s, with dry season values (1.70 m/s) being higher than the wet season (1.35 m/s) (Fig ure 2 3 F ). March 2010 had the highest wind speed. Monthly average RH varied from 70% (December 2010) to 84% (August and September 2009) with an annual average of 76%. Monthly average RH values were mostly lower for the January March period compared to other months. Low RH during January March is likely due to lower aerodynamic resistance and increase ET. Seventy percent of the weekly soil moisture values in the top 3 cm were at or near saturation indicating adequate moisture for maximum ET (Fig ure 2 4). Wet season had higher percent inundation than dry season with the exception of 2010 January, and 2010 December to 2011 April (Fig ure 2 5). The February April period of the dry season in 2009 experienced less than optimum conditions for ET due to lower soil moisture and inundation compared to 201 0 and 2011. On annual basis, the monthly average percent inundation within the flux footprint area was 64% (SD = 6.8%). For the dry season, the monthly average percent inundation was 50% (SD = 13.2%); while for the wet season the monthly average percent in undation was 73% (SD = 6.5%). The monthly average percent inundation for the wet season was almost 1.5 times higher than the dry season. Evapotranspiration is affected by inundation area. Increased open water area has been reported to result in increased w etland ET (Sanchez Carrillo et al., 2004).

PAGE 50

50 Energy F luxes The LE and total available energy (R n G W) followed similar patterns (Fig ure 2 6 B and 2 6 C ). Monthly average energy fluxes, fractions of R n used for H and LE (H/R n LE/R n ), Bowen ratio (H/LE), and evaporative fraction (LE/(R n G W)) are presented in Fig ure 2 7 B Evaporative fraction is the fraction of available energy partitioned toward LE During the study period, monthly average of daily H was 94 W/m 2 with maximum and minimum values in March 2009 (240 W/m 2 ) and June 2010 (36 W/m 2 ), respectively (Fig ure 2 7 A ). Monthly average H was higher during the dry season than in the wet season indicating there was greater heat energy transfer and temperature diffe rence between surface and the air in dry season. Daily average LE for the study period was 227 W/m 2 (Fig ure 2 7 A ). The highest and lowest monthly average of daily LE occurred in June 2010 (374 W/m 2 ) December 2008 (95 W/m 2 ), respectively (Fig ure 2 7 A ). The LE and air temperature were positively correlated ( r 2 = 0.68). The monthly average of daily LE peaked when the highest monthly average of daily air temperature was observed (Fig ures 2 3 D and 2 6 B ); similar observations were made by Drexler et al. (2008) fo r a marsh in California, USA. The LE was greater than H during April November period, indicating a large fraction of available energy was used for ET (Fig ures 2 6 A and 2 6 B ). Similar observations were made by others such as Jia et al. (2009) who used EC method to measure ET for a pasture in central Florida, USA. Overall, 60% of the R n was used for LE, 14% for G and W, and 26 % for H. in Indiana, USA. They found that LE accounts for approximately 50% of R n while W

PAGE 51

51 accounts for 30% and H for 20%. Percentage of R n used in ET at our study site is (1996) study. This is likely due to the fact that our wetland was not inundated for the entire yea r, therefore, less heat was stored in standing water and more heat was used in ET. The evaporative fraction of available energy used for ET ranged from 39% in February 2009 to 91% in June 2010 with an average of 70% for the study period (Fig ure 2 7 B ). High evaporative fraction values were observed during the wet season indicating that most of available energy was used for ET during these periods (Fig ure 2 7 B ). Low evaporative fraction values were observed during February March 2009 which is likely due to lo wer vegetation cover, soil moisture (Fig ure 2 4) and inundation (Fig ure 2 5). The Bowen ratio ranged from 1.55 in February 2009 to 0.10 in June 2010 with an average of 0.48 for the study period (Fig ure 2 7 B ). The lower Bowen ratio observed during June Octo ber indicates that the wetland plants had sufficient water supply and were actively transpiring. In February and March of 2009 and December of 2010, the Bowen ratio was greater than 1 (Fig ure 2 7 B ) due to lower water availability (Fig ures 2 4 and 2 5) or t emperature (Fig ure 2 3 D ). Eddy C ovariance B ased Evapotranspiration The average annual total ET C EC was 1152 mm. The wet season ET C EC (707 mm) was higher than the dry season ET C EC (506 mm). Annual ET C EC was almost equal to rainfall (103% of rainfall) indicating that there were other water inputs to the wetland from upland pasture. Wet and dry season ET C EC values were 88%, and 185% of rainfall (wet = 806 mm and dry = 274 mm), respectively. Dry season ave rage ET C EC was greater than total rainfall indicating surface and subsurface inputs from upland to the

PAGE 52

52 wetland. During the dry period, deeper parts of the wetland were either saturated or have open water due to shallow water table. The a verage annual dail y ET C EC for the study period was 3.33 mm/d with the wet season (3.98 mm/d, SD = 1.33) being higher than the dry season (2.79 mm/d, SD = 1.46). Monthly average ET C EC w as higher for the March October period compared to the rest of the months, mainly due to higher R S Highest ET C EC (5.33 mm/d) for the study period occurred in May 2011 (Fig ure 2 8) which represents the end of the dry season or the beginning of the wet season depending on the arrival of daily showers and thunderstorms in Florida which occurs between May and June. However, for 2009 and 2010, the highest average monthly ET C EC occurred in June (4.54 and 4.84 mm/d, respectively). The lowest monthly average ET C EC (1.28 mm/d) was consistently observed in December (Fig ure 2 8). The ET C EC values fo r 2011 were unusually higher compared to other years during the latter part of the dry season (March 2011 and May 2011) (Fig ure 2 8), a result of high VPD (50% higher than 2009 and 2010) combined with relatively low RH. The ET C EC was strongly correlated w ith ET 0 ( r 2 = 0.80, Fig ure 2 9 B ). Annual total ET C EC was almost same as (99 %) ET 0 The average daily ET 0 and the annual average total ET 0 were 3.33 mm/d and 1257 mm, respectively. There was no statistical difference between the daily ET C EC and ET 0 ( p = 0.85). Average seasonal total ET 0 were 757 and 510 mm for the wet and dry seasons, respectively. Similar to ET C EC highest monthly average ET 0 (5.05 mm/d) occurred in May 2011 (Fig ure 2 8). There were large differences between average ET C EC and ET 0 for February and March,

PAGE 53

53 mainly due to larger temperature differences between air, water, and soil which resulted in part of the available energy being used to heat the water. Although the monthly average ET estimates from this study were within the range reported elsewhere for wetlands (Fig ure 2 8), the minimum and maximum observed in this study were much lower than those observed for semi arid climate such as Utah and California (Allen, 1995; Drexler et al., 2008). Differences between our and southwest US A values (Drexler et al., 2008) are due to dry conditions, continuous saturation, and measurements limited to the growing period for the latter. Our ET rates were similar to the values observed by German (2000). The vegetation at their sites (sawgrass, cat tail, spike rush and other emergent plants) were similar to our wetland. German (2000) reported that annual total ET rates for nine sites ranged from 1078 mm to 1458 mm with an average of 1183 mm. Results from our study fall within the range shown in Germa n (2000). The difference between the average seasonal ET C values from German (2000) and our study for the dry and wet se asons was less than 4% (Table 2 2 ). Comparison of monthly average ET C EC from our study with the German (2000) study showed similar patt erns with no statistically significant differences between the two estimates ( p = 0.80). Our ET C values were also not statistically different ( p = 0.40) than another lysimeter based wetland ET study (Mao et al. 2002) in central Florida. Comparison with FA O Penman Monteith B ased E vapotranspiration The FAO PM method (with literature K C ) was unable to accurately estimate ET C The annual ET C EC was 2 3 % higher than the ET C PM Weekly ET C EC were higher than ET C PM for the entire study period with the exception of few dry months (Fig ure 2 10 A ). The average daily ET C PM was 2.44 mm/d. The annual ET C PM was 939 mm. Dry and wet season ET C PM were 329 mm (ET C EC = 507 mm) and 616 mm (ET C EC = 707 mm),

PAGE 54

54 respectively. Mean monthly as well as daily ET C EC were significan tly higher than ET C PM ( p = 0.00). The differences in two estimates were larger in dry months compared to wet months. Although statistically significant differences between dry and wet season ET C EC and ET C PM were also found, the evidence for the wet season was moderate ( p = 0.06). Trends in the monthly ET C PM were similar to the ET C EC with the exception of lowest monthly ET C PM occurring in January 2009 (1.04 mm/d) compared to December 2008 (1.28 mm/d) for E T C EC (Fig ure 2 8). The largest difference (44%) between monthly average ET C EC and ET C PM occurred in March 2010. This was mainly due to unusually high rainfall (192 mm, 123% higher than the historical average) which resulted in much higher inundation (Fi g ure 2 5) than the other two years. Such increased wetness is not specifically accounted in the FAO PM method. Assuming that ET C EC represents the actual ET, results show that use of FAO PM method, even with the regional K C values can result in underestim ation of average seasonal total ET by 35% and 13% during dry and wet seasons, respectively. This divergence is likely due to the following reasons; 1) K C values study were estimated from a lysimeter experiment where vegetation in e ach lysimeter was homogeneous as compared to the mixed vegetation present at our study site; 2) average temperature for the wet season at our study site (28 o C) was higher than Mao o C); 3) use of monthly average K C to estimate daily ET; and 4) unlike Mao duration, our study site is a seasonally flooded wetland with water levels naturally varying above and below the surface both temporally and spatially Unlike irrigated agricultural crops where optimum soil moisture is maintained, transferability of

PAGE 55

55 lysimeter based K C for wetlands is likely to cause significant errors, especially during the dry season. Part of these errors can be reduced by considering the local conditions such as soil moisture and climatic variables while developing site specific K C Use of FAO PM method can introduce significant errors in wetland water budget. To put this in perspective, the above underestimation in ET by FAO PM method (1152 mm ET C EC minus 939 ET C PM = 213 mm) represents 101% of the total annual surface flows from our wetland (Shukla et al., 2011) and is 16% of the annual rainfall. Scaling up the above underestimation to all wetlands (648 km 2 ) within the ranchlands of the LO watershed results in 143 million m 3 of water which equals 7 % of the average annual inflow to the LO. These errors are likely to be higher when all the wetlands within the Everglades watershed are considered. Such errors will result in overestimatin g the groundwater recharge or surface flows in modeling efforts and clearly highlight the need for accurate ET estimates for a variety of applications ranging from developing regional water budgets to quantifying the environmental services of storage and f low modulation by hydration of wetlands in the agricultural areas. Wetland V egetation C oefficient Monthly K CW values ranged from a minimum of 0.66 in February 2009 to a maximum of 1.22 in March 2010 (Fig ure 2 8). The annual average K CW estimated from the two years of data (January 2009 December 2010) was 0.98; indicating that on an annual basis, ET C EC values are close to the ET 0 values. The highest monthly K CW value of March 2010 is likely due to the relatively higher rainfall and wind speed, and perhaps more importantly the average inundated area for this month being 154% larger than the average inundated area for the study period. Higher than expected K CW values were also observed in January 2009 and 2010. This is likely due to the re latively high R S and

PAGE 56

56 high R n as well as the active transpiration by vegetation. Water hyacinth ( Eichhornia crassipes ) was observed despite the cold conditions observed in January 2010. Although our K CW values are within the range of K C values observed in o ther studies, there were large differences in minimum and maximum values (Table 2 2 ). Compari son of our K CW values with those from Allen (1995) study for cattail and bulrush for the initial (May June) and death stages (October) showed our values to be high er which is likely due to difference s in plant growth. Our maximum K CW C value which is likely due to the drier and warmer conditions in Utah. The K CW in months of May in California; there was no statistically significant difference between the two K C values ( p = 0.92). Our K CW values were significantly higher than those reported by Peacock and Hess (2004) for a wetland site in the UK for the May August period ( p = 0.00) Comparison of our K CW values with lysimeter K C values from Mao et al. (2002) for cattail and open water showed our values to be consistently higher (Fig ure 2 11). The weighted average monthly K C calculated using the values for open water and cattail fr om Mao et al. (2002), was compared to K CW from our study. Results from t wo sample t test revealed that our K CW were statistically higher than weighted K C (from Mao et al., 2002) ( p = 0.00). Comparison of average dry (K CW = 0.96 and K C = 0.62) and wet (K CW = 0.96 and K C = 0.8) season values showed much larger differences for the dry season. The difference in K C earlier mentioned differences in vegetation and wetness, and to some extent climatic conditions compared to our site. Comparison of K CW values with lysimeter based values K C values for a cattails marsh in the Everglades of Florida for the May October period

PAGE 57

57 showed both values to be close (K CW = 0.96 and K C = 1.0) (Table 2 2 ). Over all, while our K CW value seems to be close to the lysimeter based K C values (within 60 km from our site) for the wet season, there are large differences in dry season values. The K CW values from our study will help improve ET estimates for wetlands in ranc hland and other agricultural areas within the Northern Everglades watershed as well as other areas with similar environmental conditions. It has been known that K C is affected by climate (Allen et al., 1998; Drexler et al., 2004; Allen et al., 2005, Zhou a nd Zhou, 2009). Allen et al. (2005) noted that K C was affected by R n wind speed and RH. Jia et al. (2007), in a study of bahiagrass ( Paspalum notatum ) ET measurements using EC method, observed that the bahiagrass K C was affected by wind speed and vapor pr essure deficit. Rijal et al. (2012) analyzed the effects of subsurface drainage on corn and soybean ET and K C in North Dakota and found that corn and soybean K C values were affected by soil moisture and subsurface water level. Although those studies have m ade observations regarding the effects of climatic variables on K C limited efforts have been made to develop predictive models for adjusting K C as a function of hydrologic variables to allow for its wider applicability beyond the study site. Zhou and Zhou (2009) used EC method to estimate ET C for common reed ( Phragmites australis ) for a wetland in China and developed an equation to estimate K C as a function of three variables namely, R n temperature, and RH. Irmak et al. (2013) estimated ET and K C from a c ommon reed dominated, cottonwood and peach leaf willow mixed riparian plant community and they found that vapor pressure deficit and surface/subsurface water level were the major factors controlling K C Although water level has been included in the regress ion analysis for K C model

PAGE 58

58 development (Irmak et al., 2013), inundation characteristics (percent inundation and inundated area) are likely to be better indicators. We developed a multivariate linear regression model using Best Subset Regression technique (M initab, v.15, State College, PA) and measurements of climatic variables as well as hydrologic variables (e.g. soil moisture and inundation) to enable its use beyond our study site. The ET C fluctuates with time as a function of surface conditions and climat ic variables, which vary considerably with seasons. Due to large variations in climatic and hydrolog ic factors by seasons, a general K C model utilizing all the data yielded poor results ( r 2 = 0.34). Results improved significantly when separate models for t he dry and wet seasons were developed. For the dry season, results from Best Subset Regression showed following variable s to be statistically significant predictors of monthly K CW : soil moisture ( p = 0.03), temperature ( p = 0.02) and RH ( p = 0.09; moderate evidence). Soil moisture, temperature, and RH for the dry season were likely to have varying levels of linear relationships with K CW (Fig ures 2 12 A 2 12 B and 2 12 C ). The inundat ed area reduces considerably during the dry season making t he soil moisture a better hydrologic variable for predicting ET and K CW than inundation. A positive linear relationship shown in Fig ure 2 12 A is consistent with Allen et al. (2005) indicating K C can be considered to have a linear relationship with soil moi sture when it is higher than field capacity. Fig ure 2 12 B shows a positive linear relationship between temperature and K CW Increase in temperature can lead to increase in ET, and consequently leads to increase in K CW ET and K CW were also influenced by RH When RH increases, the air becomes more saturated which reduces the ability of air to absorb the water and subsequently reduces

PAGE 59

59 evaporation. Therefore, ET and K CW decrease when RH increases (Fig ure 2 12 C ). This is consistent with the relationship reporte d in Allen et al. (1998).Although some variables may not exhibit a clear linear relationship with K CW these variables became important when they are combined with others for predicting K CW For the wet season, percent inundation in flux footprint area ( p = 0.01), R n ( p = 0.02), and wind speed ( p = 0.00) were important predictors of K CW The relationships between K CW and percent inundation, R n and wind speed are shown on Fig ures 2 13 A 2 13 B and 2 13 C respectively. It is expected that R n and K CW have p ositive liner relationship (Figure 2 13 A ). The higher the R n the more the available energy that can be used for ET, and therefore higher K CW The relationship between K CW and percent inundation seems to be negative. It is likely due to the fact that durin g the wet season, C is higher than open water K C ; therefore, ET and K CW decrease when percent inundation increases (Fig 2 13 C ). The negative relationship between K CW and wind sp eed in Equation 15 is due to the fact that wind speed is used to adjust the effects of R n and percent inundation. Monthly K CW models for the dry ( Equation 2 14) and wet (Equation 2 15) seasons are: ( R 2 =0.58, RMSE=0.12, p = 0.03) (2 14 ) ( R 2 =0.80, RMSE=0.07, p = 0.00) (2 15 )

PAGE 60

60 where SM is volumetric soil moisture content (m 3 /m 3 ); N u is the percent inundation; and rest of the terms as defined before. Graphica l (Fig ures 2 12 D and 2 13 D ) and quantitative goodness of fit indicators indicated satisfactory model performance. For the dry season, the model was satisfactory ( r 2 = 0.58, RMSE=0.12, p = 0.03), while for the wet season the model performed very well ( r 2 = 0.80, RMSE=0.07, p = 0.00). The 1973), a statistical measure to check whether the model is free from overfitting, was 1.4 approaching the number of predictors (3 for this study for both dry an d wet seasons). The K CW values and regression models developed from this study, first for the wetlands in agricultural areas in the subtropical climate, are likely to provide better estimates of wetland ET in south Florida and other areas with similar clim atic condition. The regression model allows for adjusting K CW based on the hydrologic and climatic conditions as opposed to using single values of K CW from literature. Chapter Summary and Conclusion s A 2.5 year eddy covariance based ET study was conducted at a partially drained wetland located within a cattle ranch in S outh Florida. Average annual total ET C EC for the wetland was 1152 mm. Average seasonal total ET C EC was 506 mm and 707 mm for the dry and we t seasons, respectively. Average ET C EC in the dry season and the wet season were 2.79 mm/d and 3.98 mm/d. Results showed that the use of FAO PM method (with K C from literature ) could underestimate ET by 2 3 %. The differences in monthly ET rates between th e ET C EC and ET C PM were statistically significant for the dry season but not for the wet season. Average monthly K CW estimated from our study ranged from 0.79 in December to 1.06 in November. T he

PAGE 61

61 K CW from this study were statistically higher than the lite rature values. These differences were mainly due to differences in the dry season K C values Use of K C values from literature ( lysimeters studies) where optimum water conditions are maintained for specific vegetation for wetlands with diverse vegetation a s well as natural variations in surface wetness is likely to cause significant errors. Errors in estimating ET from wetlands can result in errors in developing regional water budgets and hydrological models and quantifying the environmental services of s torage and flow regulation by hydrat ing wetlands in the agricultural areas This study provides the first estimate of ET and K C W from partially drained ranchland wetlands in subtropical Florida and will help improve the ET estimates for wetlands with similar environmental conditions. The K C regression models developed from this study along with the climatic and environmental data can be used to develop site specific K CW Considering that wetlands account for a considerable part of the agricultural area s, our study will help improve ET estimates for a variety of applications ranging from farm scale water budgets to ecosystem services of water storage on agricultural landscape in subtropical Florida and elsewhere.

PAGE 62

62 Table 2 1. Components of the eddy covariance system Instrument Measurements Height* (m) CSAT3 3 D Sonic anemometer Fluctuations of horizontal and vertical wind 2 .0 KH20 Krypton hygrometer Fluctuations in atmospheric water vapor 2 .0 HMP45C Temperature and relatively humidity probe Air temperature and relative humidity 2.3 NR LITE Net radiometer Incoming and outgoing short and long wave radiation 2 .0 CS300 Pyranometer Incoming solar radiation 2.8 HPF01 Soil heat flux plate Heat flux through the plate 0.08 Thermocouples (1) Soil temperature 0.02, 0.06 CS616 Soil water reflectometer Soil moisture 0.03 107 L Temperature sensor (2) Water temperature 0 .0 ,** *from the soil surface at the eddy covariance station ** floats with water surface

PAGE 63

63 Table 2 2. Evapotranspiration and vegetation coefficient values from wetland studies Wetland vegetation Location Measurement method ET (mm/day) K C Reference Typha latifolia (cattail) Utah, USA Lysimeters 7.5 13.5* Initial period, 1 May 15 June (0.3); midseason, 15 June 15 September (1.6); Death or dorm, 1 October (0.3) Allen (1995) Scirpus lacustris (bulrush) Utah, USA Lysimeters 7.3 13.2* Initial period, 1 May 15 June (0.3); midseason, 8 July 7 August (1.8); Death or dorm, 1 October (0.3) Allen (1995) Typha domingensis (cattail) Florida, USA Lysimeters 2.6~5.8 Initial period, 15 March (0.6); midseason, 1 May 15 September (1.0); Death or dorm, 15 October (0.6) Abtew and Obeyseke ra (1995) Cladium jamaicense (sawgrass) Florida, USA Bowen ratio energy balance 3.0 4.0 ** N/A German (2000) Cladium jamaicense (sawgrass) Florida, USA Lysimeters 1.5 6.4 May October (1.09); November April (0.73) Mao et al. (2002) Typha domingensis (cattail) Florida, USA Lysimeters 1.4 4.7 January (0.51); February (0.61); March (0.64); April (0.73); May June (0.87); July (0.78); August (0.76); September (0.86); October (0.78); November (0.65); December (0.56) Mao et al. (2002) Phragmites austrailis (common reed) Kent, UK Bowen ratio energy balance 0.5 5.0* 0.53 for dry days and 0.88 for wet days within 24 May 28 August 2001 Peacock and Hess (2004) Typha latifolia Typha domingensis, T ypha angustifolia (cattail) California USA Surface renewal 0.8 12.2 May (0.8); June (0.92); July (1.02); August (1.09); September (1.01); October (0.9) Drexler et al. (2008) Mixed vegetation*** Florida, USA Eddy covariance 1. 3 5.3 January (0.99); February (0.84); March April (1.04); May (0.94); June (1.02); July (0.98); August (0.90); September (0.94); October (0.99); November (1.06); December (0.79) This study *values were estimated from graph s; ** values were estimated from monthly average ET; *** including Juncus effuses( soft rush), Panicium hemitomon (maidencane), Pontederia cordata (pickerelweed), Typha domingensis (cattail), Sagitteria lancifoha (duckpotato), Cladium jamaicense (sawgrass)

PAGE 64

64 Figure 2 1. Locations of study site and the eddy covariance station. Figure 2 2. The distribution of i nundation within the flux footprint area for the dominant wind direction (southwest) on December 20, 2010.

PAGE 65

65 Figure 2 3. Meteorological data collected during the study period (October 2008 May 2011) A ) M onthly total precipitation (mm) B ) M onthly average incoming solar radiat ion (W/m 2 ) C ) M onthly average vapor pressure deficit (kPa) D ) M onthly average temperature ( o C) E ) M onthly average relative humidity (%) F ) M onthly average wind speed (2 m height, m/s)

PAGE 66

66 Figure 2 4. Weekly average soil moisture (February 2009 May 2011) at 3 cm depth at the eddy covariance station. Figure 2 5. Weekly average percent inundation for the flux footprint for the study period.

PAGE 67

67 Figure 2 6. Daily average heat fluxes estimated at the study site during the study period A ) S ensible h eat (H) B ) L atent heat (LE) C ) T otal available energy (R n G W)

PAGE 68

68 Figure 2 7 Monthly average heat fluxes and flux ratios for the study period A ) Heat fluxes B ) F lux ratios

PAGE 69

69 Figure 2 8 Monthly average ET 0 (mm/d). ET C EC (mm/d), ET C PM (mm/d) and K CW Figure 2 9 Comparisons of EC based ET and reference ET for the study period A ) T he trends of weekly average EC based ET and reference ET B ) W eekly average EC based ET and reference ET, and the solid line is 1:1

PAGE 70

70 Figure 2 10 Comparisons of EC based ET and PM based ET for the study period A ) T he trends of weekly average EC based ET and PM based ET B ) W eekly average EC based ET and PM based ET, and the solid line is 1:1 Figure 2 11 Monthly wetland vegetation coefficient ( K CW ) from this study and cattail and open water crop coefficient from Mao et al. (2002)

PAGE 71

71 Figure 2 12 Relationships between K CW and climatic and hydrologic variables, and measured and modeled K CW during the dry season A ) M easured K CW and soil moisture (SM) B ) M easured K CW and temperature (T) C) M easured K CW and relative humidity (RH) D ) M easured K CW and modeled K CW the line is 1:1

PAGE 72

72 Figure 2 13 Relationships between K CW and climatic and hydrologic variables, and measured and modeled K CW during the wet season A ) M easured K CW and net radiation (R n ) B ) M easured K CW and wind speed (U 2 ) C ) M easured K CW and percent inundation (N u ) D ) M easured K CW and modeled K CW the line is 1:1

PAGE 73

73 CHAPTER 3 MEASUREMENTS AND MODELING OF EVAPOTRANSPIRATION FROM TWO HYDROGEOMORPHICALLY DISTINCT WETLANDS IN SUBTROPICAL FLORIDA Overview Wetlands occupy about 8.6 million km 2 which is 6.4 percent of the land surface of the world and they are important regulator s of water cycle (Mitsch and Go sselink, 2000). Among wetlands in the world, 56 percent of wetlands are found in the tropical (2.6 million km 2 ) and subtropical (2.1 million km 2 ) regions. Wetlands are recognized as important features in the landscape that provide beneficial services incl ud ing storing flood waters, improving water quality and providing habitats for wildlife and plant species. However, the world has lost 50 percent of its wetlands since the 1900s (OECD, 1996) Much of this loss was caused by drainage for agricultural and urban development (Mitsch and Gosselink, 2000). At present, efforts have been made worldwide to mitigate wetland losses and reestablish wetlands to reduce acreage losses Wetland s play an important role in S outh Florida, a subtropical climatic region and home to one of the largest wetlands, the Everglades. The watersheds of Lake Okeechobee (LO), part of the Everglades watershed have shallow water tables and were dominated by wetlands prior to the constructio n of an extensive drainage network in the 1900s for agricultural development. A large fraction of the water storage capacity of the watershed has been lost consequently, resulting in abrupt rise and fall of water levels in the LO during the wet (May Octob er) and dry (November April) seasons. Within the LO watershed cattle ranching is the dominant land use accounting for 36 percent of the area and of which 15 percent is occupied by wetlands (Tweel and Bohlen, 2008). Increased water storage capacity on the ranchland in the LO watershed

PAGE 74

74 is considered one of the alternatives to mitigate the harmful water levels in LO Regulatory agencies and nonprofit conservation organization s have aimed to increas e storage in the wetlands within the cattle ranches to increa se the LO storage. Although ranchland water retention seems promising no data is available to confirm its effectiveness (Steinman et al., 2003). and persistence are governed by a large extent by the hydrological processes that interact both within them and with the surrounding areas (Acreman and Miller 2006) Wetlands gain water via surface water inflow, groundwater discharge, and precipitation a nd lose water through runoff, groundwater recharge, and evapotranspiration (ET). Studying wetland water balance provides a basis for understanding hydrologic processes in a wetland. Evapotranspiration, one of the major components in the wetland water budge t, is difficult to be accurately estimated due to the variability and complexity of wetlands (Drexler et al., 2004). Therefore, accurately quantifying ET from ranchland wetlands is crucial for developing state and federal plans to increase the storage capa city in the LO watershed, part of the Northern Everglades region to partially restore the natural flows and levels. Evapotranspiration rates from studies conducted across wetlands worldwide have shown to be varied significantly with climatic, hydrologic a nd vegetation characteristics (Allen, 1995; Souch et al., 1996; Bidlake, 2000; Peacock and Hess, 2004; Sanchez Carrillo et al., 2004; Drexler et al., 2008; Zhou and Zhou, 2009). Li et al. (2009) reported a mean ET rate of 2.2 mm/day for a reed wetland in N ortheast China, a warm, temperate region of semi arid monsoonal climate. Mao et al. (2002) measured ET rates for a freshwater marsh in subtropical Florida with mean ET of 3.2 mm/d.

PAGE 75

75 Drexler et al. (2008) estimated ET rates for a marsh in semi arid Californi a, USA and reported a mean ET rate of 3.9 mm/d. Given such large variability in ET rates, quantification of ET for specific climatic and hydrologic conditions is crucial for improved water budgets for wetlands as well as watersheds. Wetland ET rates can va ry spatially despite wetlands are located in the similar climatic region German et al. (2000) concluded that annual ET can vary spatially by 26% when compared ET estimates from nine freshwater marshes located in the Everglades, USA. Lott and Hunt (2001) e stimated ET from a natural wetland and a constructed wetland in Wisconsin, USA using lysimeter and found 36% difference in ET despite the two wetlands being composed of same types of vegetation and under same climate These differences demonstrate that usi ng ET related variables (e.g. K C surface resistance of plant canopy, etc.) from off site can result in error in ET estimates Error in ET estimates can propagate error into other components of the wetland water budget especially subsurface flow and can le ad to misinterpretation of wetland and watershed hydrology. S patial variation of surface characteristics of wetlands should also be taken into consideration while estimating wetland ET. A n approach which can represent the on site integrated surface charact eristic in ET estimate is needed. Various approaches have been developed to estimate wetland ET: 1) as a residual term in the water balance (Winter, 1981); 2) using pan evaporation method (Hayashi et al., 1998); 3) using lysimeters (Abtew and Obeysekera, 1 995; Abtew, 1996; Lott and Hunt, 2001; Mao et al., 2002); 4) using eddy covariance (EC) method (Souch et al., 1996; Jacobs et al., 2002; Sanchez Carrillo et al., 2004); 5) using empirical equations (Winter, 1988) or 6) using crop coefficient method that ut ilize the generic crop

PAGE 76

76 coefficient (K C ) and combine it with reference ET ( ET 0 Jensen et al., 1990, Allen et al., 1998; Read et al., 2008). Use of pan evaporation method only provides open water evaporation and it commonly results in large error in estimating wetland ET (Winter, 1988). Using lysimeters containing wetland vegetation to quantify ET is not common due to problems such as resource intensive setup and difficulty in capturing plant, topograph y and hydrologic diversity. In addition, u nlike the lysimeter, which is usually composed of homogeneous vegetative cover, wetlands often contain more than one type of plant with variable open water areas. Extrapolating ET estimated from lysimeter can cause significant error in estimating ET from the wet land. Furthermore, soil water condition in the lysimeter is always maintained at saturation to ensure that water is removed at the potential ET rate. However, in a natural wetland soil water content varies not only temporally but also spatially based on th e climate and topography. The most common approach for estimating ET is to use the crop coefficient method where ET is estimated by multiplying K C with the reference ET (ET 0 ). This approach has been widely used for irrigation scheduling and research of agr icultural and engineering scopes (Jensen 1973; Allen et al., 1998; Tasumi et al., 2005). Crop coefficient method only provides good performance on ET estimation for uniform or low diversity vegetation (Allen et al., 1998; Drexler et al., 2004) where the sp ecific K C is applied. Application of crop coefficient method for wetland ET can result in high uncertainty depending on site specific climate, vegetation, and hydrology. Crop coefficient developed for several wetland plants, such as common rush ( Juncus eff usus ), duck potato ( Sagittaria lancifolia ), Pickerelweed ( Pontedaria cordata ), unlike monoculture crops (e.g. corn, citrus), are unavailable due to the diversity in vegetation

PAGE 77

77 species and spatial variation of vegetation coverage in the natural wetland syst ems. Furthermore, K C is not constant and changes with plant growth stages, which makes it more difficult to determine for wetlands where vegetation may have different growing seasons. Crop coefficient can also vary widely due to differences in study site c haracteristics, such as climate, topography, and hydrology (Peacock and Hess, 2004; Drexler et al., 2008), and that limits its use outside the region where it was developed. C rop coefficient method has also been applied in hydrologic modeling studies. Kien zle and Schmidt (2008) simulated both the natural hydrology of the catchment and a number of irrigation scenarios by using a physically based model, Agricultural Catchments Research Unit, which uses crop coefficient method to simulate ET. Zhao et al. (2012 ) used crop coefficient method to estimate ET in MIKE SHE/MIKE 11 to simulate the dynamics exchange between overland flow plain, groundwater system, and river system in Florida, USA. However, uncertainty associated with ET estimation by using off site K C v alues can lead to poor calibration and validation of the physical based hydrological models and therefore error in simulating the watershed hydrology. Agricultural development (e.g. construction of drainage ditches and swales) can gic processes, such as water tables dropdown, soil saturation reduction, and surface water flow alteration; therefore, all wetland water components can be affected (Blann et al., 2009) Drainage, one of hydrological modifications for developing suitable la nd for agricultural uses in South Florida may reduce inundated area and frequency and lower the groundwater table in the wetland. Drainage driven t emporal variations in water levels, combined with uneven topography of wetland, can result in chang e of open water area (inundated area) and wetland plant species

PAGE 78

7 8 composition (Kadlec and Knight, 1996) Flood intolerant upland plants may shift in to the wetland footprint area which lead s to widely varying ET (Kadlec and Knight, 1996). Ranchlands in LO watershe d contain different types of wetlands from the standpoint of drainage intensity. Quantifying ET for differently drained wetlands can help us to better understand not only ET but also other water budget components such as runoff and groundwater fluxes. Howe ver, ET from drained wetlands located at ranch land s has not been quantified. Evapotranspiration from two wetlands which contain drainage ditches and located at the same cattle ranch in S outh Florida were estimated using EC technique for a 2 year period ( June 1, 2009 to May 31, 2011). Specific objectives are: 1) quantify ET for two hydrogeomorphically distinct wetlands; 2) determine wetland vegetation coefficient that can be used for similar environments; 3) identify climatic and hydrologic factors that h ave significant influence on ET and develop a multivariate regression model for estimating daily ET; and 4) compare ET estimated using the developed regression model with ET estimated using K C from literature and average K CW developed in Chapter 2. Materials and Methods Background The study site is a cattle ranch, with an area of 275 ha, located 13 km northwest of LO (27 o o classified as flatwood soils which dominate approxima tely one third of Florida landscape. The study site is a typical flatwood region dominated by herbaceous and shrubby vegetation with scattered pine trees and is characterized by nearly flat topography, shallow water table, and poor drainage. The ranch is d ominated by

PAGE 79

79 improved pastures laced with wetlands and shallow drainage ditches. The surface water (drainage and runoff) from the ranch flows southeasterly and discharge s to the Kissimmee River which eventually empties into the L O ( Fig ure 3 1 C ). In this stu dy, t wo wetlands located in the ranch were instrumented with EC systems to collect E C data for ET estimation B oth wetlands can be categorized as herbaceous marsh depressional wetland based on the hydrogeomorphic approach (Brinson, 1993). One of them, larg er in size (21 ha), is considered as a deep herbaceous marsh depressional wetland, hereafter termed as DW ( Fig ure 3 1 A ). Water inputs for the wetland include rainfall groundwater discharge, overland flow and shallow subsurface seepage from adjacent upland pasture area Hydraulic gradient and field slope are primarily from the upland surrounding the wetland toward the deepest part of the wetland (Figure 3 1C) The average elevation in upland pasture is 10.21 m above mean sea level (AMSL) ; while t he average elevation found in the DW footprint is 8.85 m AMSL with a minimum elevation of 7.95 m AMSL Difference in average elevation between upland pasture and DW is 1.36 m. The DW has an average slope of 3.8%. Water is lost from the wetland through ET, groundwate r recharge, or through perennial or intermittent surface water flow through a drainage ditch Standing water is present most of the time and can be found during extended dry periods. The DW contains two drainage ditches ( Fig ure 3 1 A ). The two ditch segment s are 669 m and 332 m in length, and they converge and flow toward east for 147 m before exiting the wetland ( Fig ure 3 1 C ). The small depressional area at the northwestern part of DW was separated by the road but it is still connected to the DW by the culv ert buried under the road ( Fig ure 3 1 C ).

PAGE 80

80 The second wetland, with 8 ha in area is a shallow herbaceous marsh depressional wetland, hereafter term ed as SW ( Fig ure 3 1 B ). The dominant source of inflow to SW is rainfall and groundwater Water loss es through ET, s urface flow, and groundwater recharge Standing water appears at the beginning of the wet season but declines rapidly once dry season sets in. Standing water is rarely observed in the dry season except on days with high rainfall. The SW contains a dr ainage ditch which flows southeasterly for about 282 m before exiting the ranch boundary (Figure 3 1C) A depressional area at the northwestern of SW is considered as part of SW due to the linkage to SW through the drainage ditch. The average elevation in surrounding upland pasture is 10.53 m AMSL and t he average elevation of SW footprint is 9. 81 m AMSL with the lowest elevation being 8.97 m AMSL D ifference in elevation between upland pasture and SW is 0.72 m. The average slope in SW is 2. 4 % which is smal ler than DW (3.8%) In addition to the differences in elevation two wetlands also differ in the way they interact with the regional groundwater system which is mainly due to differences in topography. The DW is located at a lower elevation compared to SW (average elevation in DW is 0.96 m lower than SW) therefore, DW tends to receive more regional groundwater discharges and has longer hydroperiod. Since SW is located at a higher elevation, the stored water is likely to be lost to the surrounding low lying areas through subsurface pathways. As a result, when dry season begins surface water levels decline faster in SW than in DW. Based on the plant species survey, the two wetlands were also different f rom plant species found at the inner and outer zones. At DW, the outer zone of footprint

PAGE 81

81 area was mostly covered with leafy bladderwort ( Utricularia foliosa ), knotgrass ( Paspalum distichum ), southern watergrass ( Luziola fluitans ), common rush ( Juncus effusus ); while water spangles ( Salvinia minima ), duck potato, brook crowngrass ( Paspalum acuminatum ), common duckweed ( Lemna minor ), common rush, and water hyacinth ( Eichhornia crassipes ) were found in the inner zone. At SW, the outer zone of footprint area was dominated by bahiagrass ( Paspalum notatum ) and limpograss ( Hemarthria altissima ), while dotted smartweed ( Polygonum punctatum ), bahiagrass, limpograss, and common rush were found in the inner zone. Due to the drainage wetland footprint of SW has reduced to area along the ditch. As a result, most of th e footprint area was covered by bahiagrass due to the cattle ranching operation Bahiagrass is predominantly used in cattle ranching operation in South Florida (Chambliss et al., 2001). It can be established from seed and is widely adapted, very dependable persistent, and easy to manage (Chambliss et al., 2001). Bahiagrass has good quality during spring, but quality drops in July and August due to the high temperatures and rainfall. With minimum fertilization bahiagrass is moderately productive during spri ng and summer, but it shows little fall growth (Chambliss et al., 2001). Meteorological and H ydrologic M easurements A weather station was installed in the ranch to collect rainfall and other meteorological data ( incoming solar radiation, air temperature, r elative humidity, etc. ) Surface water level was monitored at the lowest point in the wetland using the pressure transducers during the study period. Groundwater levels were monitored at the groundwater wells located in the wetland areas (Figure 3 1C)

PAGE 82

82 Acc urate topographic data is needed to capture the spatial variation of wetland inundation (Guzha and Shukla, 2011). Guzha and Shukla (2011) estimated water storage for a wetland in South Florida using 10m USGS and 10m Light Detection and Ranging (LIDAR) digital elevation model ( DEM ) data They indicated that using USGS DEM data can result in an average 23% higher annual water storage than LIDAR DEM data The topographic data for the entire ranch was collected in May 2008 using LIDAR technique to develop a n accurate 1m DEM data for both wetlands (Figures 3 1B and 3 1C) The LIDAR based high resolution DEM data (1m) for the study site and measured surface water levels were used to determine the inundated area by using Spatial Analyst extension in ArcGIS (v.1 0, ESRI, Redlands, CA, USA). H ydroperiod, the length of time that there is standing water at a location (Gaff et al., 2000), was also calculated for two wetlands when surface water depth equals or greater than 20 cm Eddy C ovariance M easurements One EC tow er was installed in each of the monitored wetlands to estimate ET for the study period from June 1, 2009 to May 31, 2011. The EC tower was installed in the center section of the wetland with fetch of at least 100 times of the instrument height in all dire ctions ( Fig ure 3 1 C ). Each EC station consisted of: 1) a CSI (Campbell Scientific Inc., Logan, UT, USA) CSAT3 three dimensional sonic anemometer; 2) a CSI KH20 krypton hygrometer; 3) a CSI HMP45C temperature and relative humidity (RH) probe; 4) a Kipp & Zonen NR LITE (Campbell Scientific Inc., Logan, UT, USA) net radiometer; 5) a Hukseflux HFP01 (Campbell Scientific Inc., Logan, UT, USA) soil heat flux plate, thermocouples, and CSI 616 water content reflectometer; 6) a CSI CS300 pyranometer; and 7) a CSI 107 L temperature sensor. The heights of sensors and their monitored

PAGE 83

83 components are presented in Table 3 1. The sampling frequency was 10 Hz with 30 min averaging period The latent heat (LE) was estimated using the measured covariance between vertical wi nd speed and vapor density. The sensible heat (H) was estimated using the covariance between the vertical wind speed and the air temperature. The soil heat flux (G) was estimated from the sum of the heat flux measured by a heat flux plate (Table 3 1) and t he change in heat storage above the plate. The water heat storage (W) was calculated when standing water was observed on the land surface. Several corrections were applied to the raw EC data: 1) any misalignment of the sonic anemometer with the airstream w as corrected by applying 2 D coordinate rotation procedure (Tanner and Thurtell, 1969; Baldocchi et al., 1988); 2) the 30 minute LE was corrected for temperature induced fluctuations in air density (Webb et al., 1980) and for the hygrometer sensitivity to oxygen (Tanner and Green, 1989); 3) the H was corrected for differences between the virtual temperature and the actual air temperature (Schotanus et al., 1983). During certain periods (e.g., night times, early mornings with dew formations), and after rainf all, the hygrometer measurements were not available due to water drops formed on the hygrometer sensor windows. While nighttime ET has been shown to occur for woody plants and shrubs (Novick et al., 2009), nighttime LE at this site was mostly zero or negat ive and therefore, nighttime ET was assumed to be negligible. Quality assurance was achieved by the following procedure. The site was visited biweekly to clean the hygrometer sensor windows with a cotton swab and distilled water to remove dust obstruction s and restore the signal strength inspect and clean the net

PAGE 84

84 radiometer, pyranometer, temperature and relatively humidity probes and conduct routine equipment maintenance (Sumner, 2001). Quality control was achieved by visually inspecting the plots of tim e series of EC data. Erroneous data as a result of instrument malfunction (low anemometer counts and low hygrometer voltage), unexplained spikes, and heavy rainfall events were removed. For short periods with missing data, LE can be estimated from a modifi ed Priestley Taylor model (Stannard, 1993; Sumner, 2001). T he only period with long term data missing, due to malfunction of hygrometer, was July 2010 to August 2010 for DW This period was not included in this study and it impacted month average ET estima tes Flux F ootprint A nalysis The source area for flux measurements is the upwind surface area that as the surface area contributing water vapor and heat fluxes from e ach element of the upwind surface area source to the measurement concentration or vertical flux (Schuepp et al., 1990). The flux footprint is sensitive to wind direction, atmospheric stability, surface roughness, and measurement height (Leclerc and Thurtel l, 1990, Schmid, 1997). Theoretically the flux footprint area extends to infinity and thus one must always define the ratio of cumulative footprint ( e.g. 95%) ( Kljun et al., 2004 ) In most cases, 90% source areas contributing to a point flux measurement a re considered (Kljun et al., 2004). Determining the flux footprint is a complex task, yet several theoretical models have been developed over the decades. The footprint models can be categorized into analytical models, Lagrangian stochastic particle dispersion models, large eddy simulations, and closure models. The flux footprint in this study was estimated by

PAGE 85

85 aggregating 90% of source area using stochastic Largrangian model developed by Kljun et al. (2004). The percent inundation within the flux footprint area was estimated by using the wetl and water levels in conjunction with the Spatial Analyst extension in ArcGIS. Development of W etland V egetation C oefficient Evapotranspiration from a single vegetative surface is estimated as a product of the ET 0 and K C (Allen et al., 1998). Natural wetlands always have mixed vegetation similar to wetlands at the study site. To differentiate from the single vegetation crop coefficient, the vegetation coefficient for a wetland with mixed vegetation determined in this study is denoted as K CW Meteorolo gical data (e.g., air temperature, incoming solar radiation, net radiation wind speed, and relative humidity ) collected at the weather station in the site were used to calculate ET 0 The equation to calculate ET 0 is, (3 1 ) pressure curve (kPa/ o C); R n is the net radiation (W/m 2 ); G is the soil heat flux (W/m 2 ); T is air temperature ( o C); constant (kPa/ o C); U 2 is the wind speed at 2 m height (m/s); e s is saturation vapor pressure (kPa); and e a is the ac tual vapor pressure (kPa). Monthly K CW for two wetlands were estimated by using averag e daily values of actual ET ( ET C ) and ET 0 as: (3 2 )

PAGE 86

86 Evapotranspiration from Literature Based C rop C oefficient To e valuate errors in ET C estimates, ET C estimated using K C from literature and EC method were compared. The crop coefficient based ET was calculated using the K C values for cattail and open water from Mao et al. (2002), and bahiagrass from Jia et al. (2009). Both studies were con ducted in Florida. Mao et al. (2002) developed cattail, sawgrass, and open water K C values at a marsh located 60 km northeast from the study site ; while Jia et al. (2009) estimated bahiagrass K C from a well established grassed area in central Florida. The DW has mixed wetland vegetation, while SW has areas of both mixed wetland vegetation and bahiagrass. Therefore, K C values for cattail and open water from Mao et al. (2000) were used to estimate ET C for DW. At SW, cattail open water, and bahiagrass K C values were needed to estimate ET. Multivariate Evapotranspiration Model The Minitab ( Minitab, Inc., State College, Pennsylvania, USA ) a statistic al software, was chosen to perform regression analy sis. The best subset regression analysis in Minitab helps in identifying the important variables that could be used in developing daily ET model. Minitab displays statistic al results with coefficient of determination ( R 2 or different set of variables to help identify important variables to be considered in model development. The R 2 represents the percentage of variation in daily ET C accounted for by the independent variables Cp is a technique for model selection in regression (Mallows, 1973). The Cp statistic is defined as a criterion to assess fits when models with different numbers of expla natory variables are being compared. Acceptable models in the sense of minimizing the total bias of the predicted values are those models for which Cp is close to or smaller than

PAGE 87

87 the number of independent variables. Using the b est s ubset r egression techniq ue, probable models with different combinations of variables and their goodness of fit measures (e.g. R 2 were obtained Among those models, the model with largest R 2 smallest standard error than number of model predictors was selected. The most explanatory variables affecting daily ET C were identified as predictors in the selected model Evaluation of E vapotranspiration M ethods To better understand the difference in wetland ET estimates, ET C estimated using th r ee different methods were compared These methods are: 1) using the multivariate regression model and the estimated ET was denoted as ET CR ; 2) using crop coefficient method in conjunction with average K CW ( site specific K CW are known bu t hydrologic conditions are unknown ) and the estimated ET was denoted as ET CW ; 3) using crop coefficient method in conjunction with K C developed from Mao et al. (2002) and Jia et al. (2009) ( lack of site specific K C W ) and the estimated ET was denoted as ET CL Evapotranspiration estimated using three methods were evaluated by comparing the estimated ET with EC based ET C The EC based ET C was divided into two sets; one set was for multivariate regression ET model validation (randomly selected 5 values from ea ch month) and the rest of ET C values were used for model development. To evaluate the performance of each method we compared EC based ET C with ET CR ET CW and ET CL values, and used statistical criteria to evaluate the accuracy of each method Three statistical criteria were used to evaluate the ET model performance: root mean square error (RMSE), R 2 and Nash Sutcliffe model efficiency coefficient (E). The root mean square error value is used to measure the discrepancy between simulated

PAGE 88

88 and obs erved values and indicates the overall predictive accuracy of the two models. the model to the observed data. The R 2 lies between 0 and 1. A perfect fit of the model to explain the variation is represented by a value of 1, and a value of 0 indicates that the model does not explain the variation at all. The closer R 2 is to 1, the better the regression explains the relationship between simulated results and observed data Nash Sutcliffe coefficient value is an indicator of goodness of fit. The E can range from the observed data. An efficiency of 0 ( E = 0) indicates that the model pr edictions are as accurate as the mean of the observed data, whereas an efficiency less than 0 ( E < 0) indicates the observed mean is a better predictor than the model. Results and Discussion Climate and H ydrology Meteorological ( rainfall incoming solar rad iation, air temperature, relative humidity, wind speed etc. ) and hydrologic data ( i.e. surface water depth soil moisture ), were collected and analyzed at two wetlands for the 2 year period from June 2009 to May 2011 Weather conditions were similar at th e two wetlands due to their proximity (1.6 km apart from each other ). Average temperature during the study period was 21C and it averaged of 18C and 26 C for the dry (November April) and wet (May October) seasons, respectively. In South Florida, p eriods of near or below average temperatures and sporadic rainfalls are usually associated with cold fronts in the dry season (primarily from November to March Chen and Gerber 1990 ; Obeysekera et al., 1999 ) Due to cold fronts, the February March 2010 period was unusually colder in the southeastern US including Florida (NCDC, 2012). During the study period, several freezes

PAGE 89

89 (temperature drops below 0 o C) were recorded at the study site. Freezes were observed in 7, 2, and 7 days in January, February, and December 2010 respectively. In 2011, freezes occurred on January 13 and January 23. Freeze s can result in frost injury to pasture plants such as bahiagrass and therefore result in decrease in transpiration Co mpar ing to seasonal temperature variations, rainfall was more variable throughout the study period. Average annual rainfall for the 2 year period was 1050 mm and that was 23% less than the long term average rainfall in South Central Florida region ( 1362 mm/y ear, 1992 2011; NCDC, 2012) The monthly rainfall during the study period varied from a maximum of 2 20 mm ( March 2009 ) to a minimum of 0 mm ( October 2010 ) ( Fig ure 3 2). Rainfall in March 2010 ( 220 mm) and April 2010 ( 1 33 mm) were 1 57 % and 91 % respecti vely, higher than long term averages ( 1992 2011; NCDC, 2012 ) and resulted in unseasonably higher inundation and therefore, increased the potential for higher ET rates. At DW, the monthly average wind speed was 1.51 m/s with the dry season (1.69 m/s) being more windy than the wet season (1.29 m/s). Annual average wind speed (1.5 6 m/s) at SW was slightly higher than DW which was mainly due to higher wind speed (1.82 m/s compared to 1.69 m/s for DW) observed during the dry season. L ower wind speed observed at DW in the dry season is likely due to forest type vegetation (e.g. pines and palmettos) on the border that reduces wind speed from west (prevailing wind direction between November and March) of DW. Monthly average relative humid ity ( RH ) varied from 70% (December 2010) to 84% (August and September 2009) with an average of 77% at DW. At SW, monthly average RH varied from 72% (December 2010) to 84% (August and September 2009) with an average of 79% during the study period.

PAGE 90

90 Higher mo nthly average RH at SW relatively saturated air compared to DW, is likely to result in low er ET due to low vapor pressure gradient. M onthly average RH values for the dry season (76%) were lower than those in the wet season (81%) especially during January March period Temporal variations of surface water depth observed at both wetlands closely followed rainfall and have a seasonal similarity ( Fig ure 3 3) Average surface water depth at DW was 0.32 m and 0.68 m for the dry and wet seasons, respectively; and the respective values for SW were 0.11 m and 0.29 m. Average surface water depth at DW was higher than SW for both dry and wet seasons. D ifference in surface water depth between two wetlands were statistically significant ( p = 0.00). Surface water depth at SW declined faster than DW when there was no rainfall especially during the 2010 wet season. Since DW is located at the lower elevation and tends to receive groundwater discharge from surrounding upland areas ET from DW may not be s tressed because of sufficient water Comparisons of percent inundation for both wetlands during the dry and wet seasons are shown in Fig ure 3 4. The lowest surface water level for both wetlands occurred in May 2011 which is normally the driest part of the year depending on beginning of the wet period which can vary from mid May to the first week of June (Obeysekera et al., 1999) During the study period, maximum surface water level measured at the lowest elevation at DW was 1.16 m above land surface, and mi nimum water level was 0.53 m below the ground For SW, the maximum water level measured at the lowest elevation was 0.77 m above land surface and the minimum water level was 0.58 m below the ground On annual basis, the hydroperiods ranged from 7.7 to 12

PAGE 91

91 m onths for DW and from 2.8 to 7.6 months at SW. Given the longer hydroperiod observed at DW, DW is likely to have higher evaporation than SW. Monthly average percent inundation within the two wetlands indicating the flux footprint was higher at DW than SW ( Figure 3 4) Average percent inundation within the flux footprint was 66% at DW, 10 times higher than SW (6%) Surface water was present within the flux footprint area at DW for the entire study period while at SW the surface water was limited mainly to the wet season. During the dry season, standing water in SW was occasionally observed after rainfall. Average percent inundation for SW was at zero or near zero in the dry season ( Figure 3 4). Large r inundated area in DW enhanced the ET especially the eva poration component. This is consistent with the finding of Sanchez Carrillo et al. (2004) who indicated that larger areas of open water increase water consumption by ET. Net radiation (R n ) is the most significant energy term to determine available energy f or ET (Samani, 2000). The R n at DW was greater than SW for most of the monitoring period ( Fig ure 3 2). The average R n was 368 W/m 2 at DW with seasonal variations of 332 W/m 2 for the dry season s and 412 W/m 2 for the wet season s. At SW, the average R n was 34 0 W/m 2 for the study period with dry season values of 308 W/m 2 and wet season values of 378 W/m 2 Differences in R n between the two wetlands did not vary significantly by the season. Difference in m onthly average R n values for two wetlands w as nearly constant for most of the study period except during April and May 2010 when the difference became significantly large ( Fig ure 3 2). This is likely due to the relatively higher percent inundation that was found in DW ( Fig ure 3 4). Linacre (1968) found that R n is strongly related to albedo and since open water has lower

PAGE 92

92 albedo than wetland vegetation this supports the observed relationship between higher R n and inundation. The l ower the albedo of the surface, the higher the R n (Allen et al., 1998). Jacobs et al. (2007) found that among six types of land uses, including urban, agricultural, ranch land, forest, wetland and open water, open water has the highest R n Given higher R n at DW, it can be assumed that DW has more heat energy to expend towards ET than SW Measurements of groundwater depth and soil moisture for both wetlands are shown in Figure s 3 5 and 3 6. Temporal variation of groundwater depth for two wetlands was consistent. However, during the dry season the groundwater level was lower at SW which m ay resulted in water stress to plants in SW. For the period s when soil moisture data for both wetlands were measured DW soil moisture DW was higher than SW most of time Evapotranspiration Daily ET C at both wetlands showed similar seasonal trends ; higher during the wet season and lower during the dry season ( Fig ure 3 7 ). The annual average ET C for SW was 836 mm, 34% less than ET C from DW (1271 mm). Average daily ET C at DW in the dry season was 2.96 mm, while in the wet season the average daily ET C was 4.20 mm. Average daily ET C at SW was 1.83 mm (38% less than DW) and 2.87 mm (32% less than DW) in the dry and wet seasons, respectively. D aily and monthly average ET C for DW were statistically higher than SW ( p = 0.00) (Table 3 2). Differences in ET C between two wetlands could be explained by the topography and hydrologic controls in the drainage areas surrounding the wetlands and the type and distribution of vegetative coverage within the flux footprint areas. DW has a higher slope and a lower elevati on compared to SW P art of the lower elevation at both

PAGE 93

93 wetlands includes d rainage ditches. As a result, DW tends to retain more surface water and receives more groundwater discharge during the wet season which results in longer hydroperiod at DW. Higher p ercent inundation was observed at DW than SW indicating larger open water area at DW resulting in higher R n Large difference in ET C between DW and SW during the wet season is likely due to the combination of higher R n and percent inundation within the flux footprint area at DW ( Fig ure 3 4). Furthermore, difference in ET C between DW and SW may also be due to the composition of plant community within the flux footprint area in SW. The dominant vegetation within the flux footprint area at SW was bahiagrass. For the wet season, the ET rate for bahiagrass was lower than that for the wetland Difference in monthly ET C between DW and SW was high in March 2010, April 2010, October 2010, March 2011, and May 2011 (Table 3 2) In March and April 2010, the study site received higher than long term average rainfall which provided sufficient water for wetland plants growing in DW so that plant transpiration remained in potential rate and did not decline due to the water stress However, since the most of SW was covered by bahiagrass which remain ed in dormant phase, instead of increasing ET excessive rainfall resulted in increase of soil moisture and rise of groundwater level (Figures 3 5 and 3 6). Difference in monthly ET C betwee n two wetlands in October 2010, March and May 2011 is likely due to the combination s of high percent inundation and soil moisture observed at DW (Figures 3 4 and 3 6). Furthermore, transpiration from bahiagrass in SW may decline due to water stress when gr oundwater level s at SW w ere lower than root zone (approximately 1.5 m below the surface ) from October 2010 to May 2011 (Figure 3 5).

PAGE 94

94 The d ifference in ET C between two wetlands was lowest in December 2010 and January 2011 (Table 3 2). This is likely due to the fact that freezes were observed at both wetlands within those months which caused frost injury to wetland plants and bahiagrass As a consequence the ET C from both wetlands decreased Wetland V egetation C oefficient Although crop coefficient method is most widely used approach for ET estimation, site specific K C is needed for accurate estimation of ET C for agriculture (Allen et al., 1998, Kang et al., 2003) as well as wetlands (Towler et al., 2004; Drexler et al., 200 8; Zhou and Zhou, 2009). Average monthly K CW for two wetlands are shown in Fig ure 3 8 As expected, monthly average K CW for DW were statistically higher than SW K CW ( p = 0.00). Monthly average K CW values for DW varied from a minimum of 0.73 in December 201 0 to a maximum of 1.22 in March 2010, while it varied from 0.45 (January 2010) to 0.85 (June 2010) for SW (Table 3 2). A nnual average K CW for DW and SW were 1.01 and 0.65, respectively. The highest monthly DW K CW value of March 2010 is likely due to the relatively high rainfall and percent inundation (99.7%) observed in that month ( Fig ure 3 4). A higher than expected DW K CW (1.06) was also observed in January 2010 (Table 3 2). This is likely due to the relatively high soil moisture as well as the active transpiration by emergent floating wetland plants. Water hyacinth was observed despite the cold condition that existed in January 2010. In contrast, the lowest K CW for SW was observed in January 2010. This s likely due to low temperature (12 o C, 43% lower than annual average) and the dry condition ( zero percent inundation) and the fact that there was no water hyacinth or other floating plants observed in SW. At SW, higher monthly K CW

PAGE 95

95 values were observed during the mo nths of April June which is likely due to combination of higher R n and percent inundation ( Fig ures 3 2 and 3 4). Difference in K CW between two wetlands was high in January October, and November 2010 (Table 3 2). Several factors have contributed to the di fference in those months: 1) with DW being at the lower elevation, DW stored more water ( higher percent inundation) which provide d sufficient water for ET (Figure s 3 3 and 3 4 ); 2) difference in R n between two wetlands in those months was relatively higher than other months (slightly lower than April and May 2010); 3) plants in SW may experience water stress due to groundwater level dropped below root zone (Figure 3 5) ; 4) bahiagrass in SW remains dormant in January and is with little growth in October and November (Chambliss et al., 2001) The DW K CW from this study was not in good agreement with published K C for wetland vegetation (Allen et al., 1998; Mao et al., 2002; Zhou and Zhou, 2009) and bahiagrass ( Jia et al., 2009), especially for the dry season. This is likely due to the presence of standing water and prolonged hydroperiods during the beginning of the dry season which provided sufficient water for ET. When wetland plants are not water limited, the rate of ET is not reduced until the plants enter into their physiological senescence (Lott and Hunt, 2001). Comparing the shape of K CW curve from this study with the shape of K C curve for single type of vegetation (Mao et al., 2002; Zhou and Zhou, 2009), we found that wetland plants grow in a natural wet land system may have different phase of growth throughout the year, unlike the upland plants. Senescence for most plants at DW does not occur at the same time, and that is different from SW because SW is dominant by bahiagrass for which senescence occurs

PAGE 96

96 For the entire study period, t he shape of the K CW curve for SW did not to follow the one for DW ( Fig ure 3 6). Instead, SW K CW has a similar trend as bahiagrass (Jia et al., 2009). The SW K CW from this study was not statistically different from the bahiagra ss K C values reported by Jia et al. (2009) ( p =0.56). The annual average K CW for SW from this study and bahiagrass K C from Jia et al. (2009) are 0.66 and 0.63, respectively. This is likely due to the fact that the flux footprint area of SW was covered by b ahiagrass and mixed wetland vegetation. Although the SW K CW values were similar to the bahiagrass K C values from Jia et al. (2009), K CW values were higher than bahiagrass K C during the months of June July. This could be attributed to the fact that wetland plants, which have higher K C values than bahiagrass, were observed at SW Furthermore, there was increased wetness and inundat ed area during these months caused by the high rainfall that occurred in the wet season. Past studies report that climate has a major influence on K CW (Drexler et al., 2008; Zhou and Zhou, 2009); therefore, efforts have also been made to develop a regression model to predict daily K CW as well as ET C However, results showed that the developed model (K CW = 0.44 0.101ET 0 + 0.00204 R n + 0.0021N u ) has a low R 2 ( 0.51) and was not accurate enough to be suitable for predicting ET C Since K C is often used on monthly basis in modeling and water budget constructions studies, monthly K CW values developed from this study can be used to esti mate ET for similar types of wetlands. Multivariate R egression E vapotranspiration M odel There is no universally accepted ET model or measurement technique that can be applied in areas across the globe. In a comprehensive review of methods for estimating we tland ET C (empirical models and micrometeorological methods), Drexler et

PAGE 97

97 al. (2004) concluded that due to the fact that wetlands varied largely in plant communities, hydrology, and spatial characteristics of topography, the best approach to improve accurac y on wetland ET C estimation is to account for surface characteristics including R n and inundation. At present, only a few studies have shown success on modeling daily wetland ET C and most of them focus on specific wetland plants (Drexler et al., 2008; Jia et al., 2009). Results showed that ET C was highly affected by R n r = 0.85; p r = 0.39; p = 0.00). Given the fact that the two wetlands differed significantly in R n and percent inundation, these variables can be used to differentiate them. A generic multivariable regression model was developed to predict daily ET C for both wetlands despite significant difference in topography, hy drology, and plant communities. The regression model developed has the form: ET C = 1.14 + 0.477 ET 0 + 0.00652 R n + 0.00796 N u R 2 = 0.80, RMSE = 0.66, (3 3) where ET 0 is the reference ET, R n is net radiation and N u is the perce nt inundation. The daily ET C regression model developed from this study, the first for wetlands with contrasting hydrology and topography, is likely to improve accuracy of ET C estimates in S outh Florida and elsewhere with similar environmental conditions Although EC method provides better estimation of ET C expensive instrumentation and labor intensive maintenance and data processing ha ve limited its application. This model allows estimating ET C based on the relatively easily measured hydrologic and climatic variables when the EC method is not feasible. This daily ET C regression model has important implications to hydrologic modeling. Considering its ability to predict the

PAGE 98

98 daily ET C for a desire wetland, the predicted ET C can be used as direct input or use d to calculate K CW then can be used as input to models such as MIKE SHE (Jaber and Shukla, 2012). Furthermore, once the above model is verified for other regions, it can be further integrated in the physical based models. For example, MIKE SHE, one of the integrated (surface and groundwater) models used world wide, needs specific K C to simulate ET (Jaber and Shukla, 2012) which can be derived from Equ 3 3 and ET 0 Comparisons of E vapotranspirati on M ethods Performances of each ET method in the stages of development and validation for two wetlands were evaluated using statistical criteria (Table s 3 3 and 3 4). For DW, ET CR has the smallest RMSE (0.83) and the highest E (0.69). For SW, the ET CR has the smallest RMSE of 0.57 and E of 0.73 (Table 3 3, Table 3 4). Results showed that the regression method provides the be st daily ET C estimation than both average K CW and K C from literature. It is shown that without a site specific K C large errors cou ld be introduced while using K C from literature even though they were developed under similar climatic conditions. Results from this study also show that u sing the average K CW developed from this study to estimate ET C for wetlands that are different in top ography and hydrology can lead to significant errors. The regression model ( Equ. 3 3) which incorporates site specific hydrolog y and R n is likely to improve the accuracy of ET C estimation on daily basis compared to the K C based approaches Chapter Summar y and Conclusion s In this study, EC method was used to estimate wetland ET C and d evelop K CW for two wetlands in a ranchland. Annual average total ET C were 1271 mm and 836 mm for DW and SW, respectively. Evapotranspiration from SW w as 34% less than DW, and the

PAGE 99

99 differences in daily and monthly average ET C between two wetlands were statistically significant. Evapotranspiration estimates varied significantly even though two wetlands were only 1 6 k m apart and have similar climat es Differ ences in ET C are likely due to higher R n percent inundation in the flux footprint area, and relatively diverse plant communities at DW. This study provides monthly K CW for two hydrogeomorphically distinct, freshwater wetlands, a deep depressional wetland and a shallow depressional wetland, which are dominant in the LO watershed For wetlands with yearlong hydroperiods, DW K CW could be used to improve ET C estimates while for the seasonal ly inundated wetlands similar to SW, SW K CW could help improve ET C est imates. Results from regression analysis indicated that daily ET for both wetlands was largely controlled by R n and percent inundation. Therefore, a multivariate regression model was developed to predict daily wetland ET C clim atic and hydrologic conditions. Using this model, one can acquire accurate ET C estimates with less expensive monitoring equipment compared to the EC system. Given the fact that inundation has an impact on daily ET C the daily ET C regression model can help better quantify ET for wetland studies which relate to evaluating effects of climate change and wetland water retention scenarios since they may result in change of inundation in the wetland Evapotranspiration estimated using average K CW and K C from literature were compared to ET C estimated using the regression method. Results s howed that climatic and hydrologic conditions should be considered when selecting literature values for K C including K CW for DW and SW, to estimate ET C Using the site specific climatic and

PAGE 100

100 hydrologic measurements the regression m odel developed from this study will improve accuracy of daily ET C estimates for wetlands in ranchland and other agricultural areas. It can be integrated with hydrologic model s for improving ET estimates.

PAGE 101

101 Table 3 1 Components of the eddy covariance system at the deep wetland ( DW ) and the shallow wetland ( SW ) Instrument Measurements DW SW CSAT3 3 D Sonic anemometer Fluctuations of horizontal and vertical wind 2 .0 m 1.5 m KH20 Krypton hygrometer Fluctuations in atmospheric water vapor 2 .0 m 1.5 m HMP45C Temperature and relatively humidity probe Air temperature and relative humidity 2.3 m 1.8 m NR LITE Net radiometer Incoming and outgoing short and long wave radiation 2 .0 m 2.3 m CS300 Pyranometer Incoming solar radiation 2.8 m 2.5 HPF01 Soil heat flux plate Heat flux through the plate 8.0 c m 8.0 c m Thermocouples (1) Soil temperature 2.0 c m, 6.0 c m 2.0 c m, 6 .0 m CS616 Soil water reflectometer Soil moisture 3 .0 c m 3 .0 c m 107 L Temperature sensor (2) Water temperature 0 .0 m,* 0 .0 m,* All the measurements are from the soil surface at the eddy covariance station floats with water surface Table 3 2. Monthly average reference ET (ET 0 ), actual ET (ET C ), and wetland vegetation coefficients (K CW ) for the deep wetland (DW) and shallow wetland (SW) Month ET 0 (mm /d ) DW ET C (mm /d ) SW ET C (mm /d ) DW K CW SW K CW Jun 09 4.41 4.54 3.48 1.01 0.78 Jul 09 4.32 4.29 3.35 0.98 0.77 Aug 09 4.07 3.71 2.49 0.90 0.60 Sep 09 3.68 3.23 1.74 0.88 0.47 Oct 09 3.20 3.55 2.15 1.10 0.67 Nov 09 2.34 2.86 1.67 1.20 0.71 Dec 09 1.82 1.77 1.16 0.97 0.62 Jan 10 2.03 2.19 0.95 1.06 0.45 Feb 10 2.46 2.39 1.45 0.94 0.58 Mar 10 3.28 4.10 2.58 1.22 0.76 Apr 10 4.14 4.52 2.98 1.06 0.70 May 10 4.75 4.57 3.43 0.95 0.72 Jun 10 4.68 4.84 3.95 1.03 0.85 Sep 10 3.94 3.97 2.63 0.99 0.67 Oct 10 3.29 3.94 2.11 1.20 0.64 Nov 10 2.42 2.77 1.42 1.13 0.59 Dec 10 1.99 1.46 0.92 0.73 0.50 Jan 11 2.08 1.70 1.22 0.81 0.59 Feb 11 2.78 2.59 1.68 0.93 0.60 Mar 11 3.70 4.10 2.25 1.08 0.61 Apr 11 4.64 5.15 3.73 1.11 0.80 May 11 5.05 5.33 3.42 1.05 0.67 Average 3.41 3.52 2.31 1.02 0.65

PAGE 102

102 Table 3 3. Comparison between EC based ET and ET estimated using multivariate regression model (ET CR ), average K CW (ET CW ) and K C from literature (ET C L ) for the model development and validation periods for the deep wetland Development Validation ET CR ET C W ET C L ET CR ET C W ET C L RMSE 0.78 1.0 1 1.33 0.83 0.99 1.3 6 R 2 0.75 0. 80 0.63 0.71 0. 83 0.66 E 0.7 3 0.78 0.20 0.69 0.57 0.18 Table 3 4. Comparison between EC based ET and ET estimated using multivariate regression model (ET CR ), average K CW (ET CW ) and K C from literature (ET C L ) for the model development and validation periods for the shallow wetland Development Validation ET CR ET C W ET C L ET CR ET C W ET C L RMSE 0.53 0.81 0. 70 0.57 0.8 3 0.6 9 R 2 0.81 0.77 0.64 0.7 7 0.6 4 0.6 4 E 0.7 8 0.48 0.62 0.73 0.4 4 0.61

PAGE 103

103 Figure 3 1. Maps of the study site and topography for two wetlands A ) T he topographic map of the deep wetland (DW) B ) T he topographic map of the shallow wetland (SW) C ) M ap of ranch with drainage ditches, wetland footprints, and locations of eddy covariance station s and gr oundwater well s

PAGE 104

104 Figure 3 2. Monthly total rainfall and monthly average net radiation for two wetlands during the study period (June 2009 May 2011) Figure 3 3. Surface water d epth at the deep and shallow wetlands during the study period (June 20 09 May 2011) Figure 3 4. Monthly average percent inundation for two wetlands during the study period (June 2009 May 2011)

PAGE 105

105 Figure 3 5. Groundwater depths for both wetlands during the study period (June 2009 May 2011) Figure 3 6. S oil moisture at 3 cm depth at the eddy covariance station during the study period (June 2009 May 2011) Figure 3 7 Daily ET C for the deep and shallow wetlands during the study period (June 2009 May 2011).

PAGE 106

106 Figure 3 8 Average wetland vegetation coe fficients (K CW ) for the deep and shallow wetlands

PAGE 107

107 CHAPTER 4 WATER BUDGETS AND GROUNDWATER FLUXES FOR TWO WET LAND UPLAND SYSTEM S IN LAKE OKEECHOBEE WATERSHED Overview Wetlands play a unique role in terms of their functions such as flood protection, water purification, and fish and wildlife habitat. Since the 1900s, natural wetlands have been lost dramatically due to drainage for urbanization and agricultural land uses. Efforts have been made to preserve the existing wetland s as well as restore the lost wetlands. Wetland conservation needs in depth knowledge of wetland hydrology (Bedford, 1999; Hill, 2000; Hayashi and Rosenberry, 2001) Water availability and dynamics are the principal determinants of the structure and the fun ction of wetland ecosystems (Mitsch and Gosselink, 2000). A ccurate quantification of water balance components is important for understanding wetland environments and restoration efforts Water balance is total inflows equal total outflows plus or minus cha nge in storage in a wetland. The water balance approach has been broadly used to characterize the behavior of wetland systems (Carter, 1986; Gilvear, et al., 1993; Boudreau and Rouse, 1995; Guardo, 1999; Raisin et al., 1999; Lee et al., 2006; Wilcox et al. 2006). The mathematical expression of the water balance is termed as water budget. A complete water budget should have an independent and reliable estimate of each of water budget component (Dooge, 1975). Wetland water budget comprises of surface and gro undwater inflow and outflow storage, rainfall, and evapotranspiration (ET) Difficult ies in determining the water budget for a wetland lies with how well the individual inputs, outputs, and storage can be measured or estimated, and the magnitude of the as sociated errors (Carter et al., 1979; Winter, 1981). For wetland

PAGE 108

108 studies with intensive monitoring of hydrologic processes each of the water budget components is measured or estimated individually to calculate an overall water budget (Novitzki, 1982; Hyat t and Brook, 1984); however, in most studies, it is common to estimate one of the components as the residual of the water budget due to the difficulties in accurately measuring the specific water budget component (e.g. groundwater flows) Review of various wetland water balance studies (Owen, 1995; Bradley, 1997; Drexler et al., 1999; Mitsch and Gosselink, 2000; Bradley, 2002; Zhang and Mitsch, 2005; Favero et al., 2007; Nungesser and Chimney, 2006) suggests that among wetland water budget component s ET an d groundwater flux are the least understood and have the highest error Using traditional lysimeter approach to estimate ET from wetland vegetation is limited by its resource intensive setup and difficulty in capturing plant, topographic, and hydrologic diversity. The r ecent emergence of eddy covariance (EC) method, presented in Chapters 2 and 3, has potential to significantly improve the accuracy of wetland ET. Once EC based ET values are available they can be combined with the other measured components ( rainfall storage, and surface flow) to calculate the next most challenging component, groundwater flux as the residual term of the water budget. This estimate of groundwater flux can be compared to the groundwater flux from hydrologic models, such as M IKE SHE, to improve the model performance. Groundwater recharge and discharge are one of the most important attributes of a wetland. Groundwater recharge ( inflow ) is the addition of water to the groundwater systems, and groundwater discharge ( outflow ) re pr esents the loss Groundwater flux is usually the most difficult component to quantify due to its complex nature, especially in

PAGE 109

109 heterogeneous materials such as peat (Drexler et al., 1999; Hunt et al., 1996). The most common approach for estimating groundwat er flux es for a wetland is through the water budget method (as a residual) (Mitch and Gosselink, 2000). However, this is limited by the associated errors in other water b udget components, which may be larger than the groundwater flux itself (Winter, 1981). Groundwater flux can also be estimated by which requires surface and ground water levels and hydraulic conductivity of the soil or sediment. Groundwater monitoring wells placed around the wetland can help determine the direction and the hydraulic gradient which can be combined with equation to calculate the flow Applications of Darc Law are limited by our ability to accurately determine the hydraulic conductivity of the soil (Winter et al., 1988). Furthermore, for the wet lands in relatively flat landscapes such as flatwoods region of Florida, groundwater movements are often limited by low hydraulic gradients, which are difficult to measure with high accuracy Although direct and indirect approaches of estimating groundwater flux have inherent drawbacks, their uses can still provide information of interactions between groundwater and other water budget components and are essential to understand the effects of water retention on ranchland s Since the 1900s, the primary canal and levee system s have been constructed i n the Lake Okeechobee (LO) watershed to prevent flooding. Th e construction included extensive ditching of isolated wetlands and construction of drain ed fields for agricultur e and urban development (Flaig and Reddy, 1995) C attle ranching is the dominant land use (36%) in the LO watershed (Tweel and Bohlen, 2008) Due to the drainage, hydrologic conditions of drained wetland s within the ranchland have a ltered and are

PAGE 110

110 different than natural wetlands. The state and federal agencies are developing plans to partially restore the watershed storage of th e LO watershed One of the strategies to achiev e the desired storage is to store water in the ranchland wetl and s by applying wetland water retention ( W W R ) To evalu at e the effects of W W R, the water budget components of the wetland upland system on the ranchlands need to be understood since the hydrologic conditions have been altered due to drainage. It is assumed that W W R has potential to increase groundwater storage ; however, its extent has not been quantified This study aims at quantifying the groundwater fluxes as a residual term of the budget. The groundwater flux obtained from the water budget m ethod is dependent on the accuracy of other components of which ET is the most important since it represents the largest term in the water budget. The ET rates for the two wetlands w ere determined from the EC method (C ha pters 2 and 3) which is the most a ccurate method for wetland systems. The water b udget method w ere applied in this study to quantify groundwater fluxes to or from two partially drained wetland upland systems located in a commercial cow calf ranchl and in the LO watershed whi ch is part of the Everglades watershed in Florida. Since the EC based ET was estimated for the period from May 1, 2009 to May 31, 2011, the groundwater fluxes were quantified for the same period. Other components, rainfall, surface flow, and soil moisture in the unsaturated zone were also measured for the same period Furthermore, surface and ground water levels were measured for the same period to quantify surface and ground water storage for constructing water budget. Water budgets were constructed at di fferent time steps

PAGE 111

111 (daily, monthly and seasonal) to investigate the short and long term interactions between the wetland upland system and surrounding area Materials and Methods Site D escription and D ata C ollection The study area is a commercial cow calf ranch located approximately 13 km northwest of L O (27 o o The long term average annual rainfall in the South Central Florida region is 1362 mm/year (1992 2011; NCDC, 2012) with 404 mm and 960 mm falling during the dry (November April) and we t seasons (May October), respectively. The ranch has an area of 2.75 km 2 The land uses within the ranch are mainly improved pastures, hardwood forests, and herbaceous marsh depressional wetlands. T he ranch can be divided into four sub watersheds including Site1 (0.8 0 km 2 ), Site2 (0.91 km 2 ), Site3 (0.80 km 2 ), and Site4 (0.24 km 2 ), and surface water outlets for those sites are denoted as Q1, Q2, Q3 and Q4, respectively (Figure 4 1). The main outlet for the ranch is Q5 (Figure 4 1). The wetland upland systems in this stud y are Site1 and Site4 and t he wetland upland system mentioned here includes the actual wetland as well as the associated pasture areas that are drained through the ditch network The wetland in Site1 is a deep depressional wetland (DW) and the wetland in Site4 is a shallow depressional wetland (SW ) (Figure 4 1). Site1 contains two main drainage ditches that drain the wetland and upland area of Site1 (Figure 4 1). The drainage ditches at the ranch are typical of ranchlands of South Florida that were dug in the early 1900s to drain wetlands as part of developing improved pasture for cow calf production systems. Drainage from two ditch segments (669 and 332 m in length) converges and flows together for 147 m before exiting the wetland (Figure 4 1). Site4 contains a single drainage ditch which flows south easterly

PAGE 112

112 for 245 m before reaching the outlet (Figure 4 1). The drainage ditches route both surface flows and groundwater fluxes from the wetlands as well as the upland pasture. Wetland water retention was implemented at both wetland upland systems by installing a control s installed at Site1 (40 m west of Q1) includes a culvert (0.61 m in diameter ) and a riser board (CRB) structure which is 0.92 m in height and 1.07 m in width. The CRB structure i nstalled at Site4 (30 m west of Q4) is 0.49 m in height and 0.76 m in width with a 0.61 m diameter culvert. The CRB structure allows surface water to flow only when the water level exceeds the top elevation of the riser board. The elevation of the top of t he CRB structures at DW (9.12 m above mean sea level, ASML) is lower than that at SW (9.47 m, AMSL). Elevations presented here and throughout the study are reference d to the North American Vertical Datum of 1988 (NAVD 88). The highest and lowest elevations at Site1 were 11.72 and 7.95 m ASML, respectively, with an average elevation of 9.91 m AMSL. Site4 is situated higher than Site1 with an average elevation of 10.29 m AMSL and the highest and lowest elevations are 11.03 m and 8.97 m ASML, respectively. Th e lowest elevations of both Site1 and Site4 were found in the ditch located in the two wetland s Soils in the study site are typically poorly drained and highly sandy (NRCS, 2003) Upland areas are mainly comprised of Basinger, Immokalee Myakka, and Valkaria fine sand s. Soils in DW are comprised of Basinger and Floridana fine sands; while only Basinger fine sand is found in SW. For both sites a fiberglass trapezoidal flume was installed at the ditch outlet (Q1 and Q 4 ) to measure surface f lows (Figure 4 1). Flow rate was determined by two methods. P ressure transducers (KPSI, Pressure System, Hampton, VA), installed at the

PAGE 113

113 upstream and downstream end of the flumes, were used in conjunction with the flume equation to calculate the flow rates. Due to errors associated with measuring submergence and bi directional flows for flat landscape of South Florida, an Acoustic Doppler Velocity meter (SonTek/YSI, San Diego, CA) was also installed in the throat section for accurate flow measurement especia lly when under high submergence W h en the flume is submerged, the flow estimation has significant errors. Groundwater levels were monitored using pressure transducers at 27 locations ( Site1 and Site4 ) out of which 15 wells were installed in the wetlands an d 12 in the upland area s(Figure 4 1). Soil moisture throughout the soil profile was monitored on a 15 minute frequency using the Frequency Domain Reflectometry (FDR Sentek Technologies, Australia ) soil moisture probes at different depths (0 0.10m, 0. 1 0 0. 20m, 0. 2 0 0.30m, 0. 3 0 0.40m, 0. 4 0 0.50m, 0. 5 0 0.60m, 0. 6 0 0.70m, 0. 7 0 0.80m, 0. 8 0 1.00m, and 1. 0 0 1.20m). The meteorological variables (rainfall, wind speed, temperature, solar radiation, net radiation, and relative humidity) were monitored at a weather station located near Q2. Rainfall was measured at weather station and near Q1 on a 15 minute frequency with H 340 tipping bucket rain gauges (Design Analysis Associates, Inc, Logan, UT) located at the weather station and near Q1. Evapotranspiration for upland pasture was estimated using crop coefficient (K C ) method (Allen et al., 1998) with K C obtained from lite rature. The crop coefficient method has been the primary method for estimating ET for the last 40 years (Jensen 1973; Allen et al., 1998; Tasumi et al., 2005). It estimates ET by multiplying the reference ET ( ET 0 ) with crop specific K C values (ET = ET 0 K C ) (Allen et al., 1998). The meteorological data (air temperature, solar radiation, net radiation, wind speed and humidity) collected at the weather station were used to

PAGE 114

114 calculate ET 0 T he ET for the wetland area was quantified using the EC method. The two E C systems were located in the middle section of the wetland s to ensure adequate fetch (Figure 4 1). Each EC system consists of a CSI (Campbell Scientific Inc., Logan, UT) CSAT3 three dimensional sonic anemometer and a KH20 krypton hygrometer. The anemometer measures fluctuations in wind speed and virtual temperature using three pairs of non orthogonal sonic transducers, and the hygrometer measures the fluctuations of water vapor density. The temperature and relative humidity were measured using a C SI HMP45C sensor. Details of the EC measurements are discussed in Chapter 2 and 3. A Light Detection and Ranging (LIDAR) survey of the entire watershed was conducted in May 2008 to develop a 1 m resolution digital elevation model (DEM) data with vertical a ccuracy of 5 30 cm The LIDAR based topographic data provides highly accurate and dense points to capture spatial variations for the surface / terrain. Water B udget The water budget equation (Equ. 4 1) is based on the conservation of mass and shows that a c hange in water volume for a given time period is equal to the difference between water inflows and outflows. In this study, the water budget s for Site1 and Site4 were constructed and the groundwater flux was estimated as a residual component. For each site the horizontal boundary of the control volume of the water budget was defined as boundary of the sub watershed which represents the wetland area and the upland area connected to the wetland through surface flows and the vertical domain was defined as t he region above the deepest point at the two wetlands The w ater budget equation can be written as:

PAGE 115

115 ( 4 1) where S is the change in storage in vadose and saturated zones ( mm/day ) ; P is the precipitation ( m m /day ) ; RI and RO are the surface inflow and outflow ( m m /day ) respectively; GWI and GWO are the groundwater inflow and outflow (mm/day) respectively; and ET is evapotranspiration (mm/day) In this study, groundwater flux was estimated as a net flux ( GW ). Assuming the horizontal boundary of the wetland upland system is well defined, the surface inflow into the watershed was considered negligible. Th us the water budget equation for this study can be simplified as: ( 4 2) Evapotranspiration for the wetland upland system was estimated by combining EC based ET for the wetland area and the ET estimated using crop coefficient method and K C from literature, for the upland pasture The upland areas of both sites were dominated by bahiagrass Since th e wetland vegetation coefficient developed for SW m ost ly represented bahiagrass (Chapter 3), the wetland vegetation coefficient (K CW ) of SW (Chapter 3) and bahiagrass K C reported by Jia et al. (2009) w ere averaged to obtain K C for the bahiagrass (K CB ) For the area covered by pine trees on the west side of DW in Site1 (Figure 4 1) K C for pine tree from the Kissimmee Basin Modeling and Operations Study (KBMOS) (Earth Tech and DHI, 2007 ) was used W ater stor ed at the two sites were estimated as sum of subsur face (unsaturated and saturated) and surface water storages on a daily basis Fifteen minute s oil moisture measurements taken at different depths (0 1.2m) at both sites ( Site1: SM34 ; Site4: SM13) were used to estimate soil water storage (Figure 4 2). S urface water level

PAGE 116

116 measurements and topographic data were used to determine the r elationships of stage area and stage volume for two wetland upland system s via spatial analyst module in ArcGIS (v.10, ESRI, Redlands, California). Uncertainty Analysis A gene ral water budget equation is simple and straightforward However, in reality, errors in other water components may affect the reliability of the one estimated as residual. The net g roundwater flux was estimated as residual from the water budget of two wetl and upland systems in this study As a consequence error s in other water budget measurements are lumped with the net groundwater flux and this could result in unrealistic estimates of the net groundwater flux ( Sutula et al. 2001 ). Water budget analysis without estimate of error can be misleading. Analysis of error in measurements for each component is very important while estimating water budgets. Assuming terms on the right hand side of Equ. 4 2 are not correlated error in the net groundwater flux is t he propagation of error from terms on the right hand side E rror in the net groundwater flux can be presented as ( Winter, 1981 ; Sutula et al., 2001) : (4 3) where e(term) indicates error in different components in the water budget Considering the spatial variability of rainfall error associated with areal averaging of point rainfall data can vary depending on storm type, duration and gage density (Winter, 1981). Winter (1981) evaluated errors associated with the rainfall data in the water balance studies of lakes and noted that using common methods (i.e. arithmetic mean, Thiessen method, and isohyetal method) of interpreting rainfall data could result

PAGE 117

117 in errors in estimating areal daily average rainfall. Error in estimatin g rainfall for a ga u ge density of 2.6 km 2 is 4% (Winter, 1981) Winter (1981) and Carter (1986) also stated that range of error in rainfall would be reduced when the recording ga u ge is located at the study site. In this study, rainfall data was collected from a rain ga u ge located at the study site ( i.e. the ga u ge density is 2.75 km 2 / gauge ). Therefore, range of e rror in rainfall for two wetland upland system was assumed to be 4 %. T he energy budget technique is the most accurate method for estimating ET (Ca rter, 1986). Studies have reported that e rror in estimate of ET using energy budget is 10 % of annual estimates (Winter, 1981 ; Sutula et al., 2001 ). In this study, e rror in wetland ET (estimated using EC method) was assumed to be 10%. Wu and Shukla (2013) used EC method to quantify ET from a ranchland wetland in South Florida, and they found that annual wetland ET can be underestimated by 23% when using crop coefficient method and K C from literature. Since the upland area at both sites has less variety of vegetation and is mainly dominated by bahiagrass error in upland ET estimation was assumed to be 2 0 %. U ncertainties in flow velocity and quantity measured using ADV are 5 % (Meile et al. 2008) However, considering the ADV sensors may be surrounded by fl oating vegetation from wetlands and the flume equation has high error under submerged condition, which occurs most of the times during the wet season, the surface flow measurement error was assumed to be 10% similar to Winter ( 1981 ) and Harmel et al ( 200 6). Estimation of s torage includes errors in soil water and surface water storage s Error in measurement of soil water storage is attributed to error in soil moisture

PAGE 118

118 measurement s Error in the soil moisture was evaluated by c omparing the soil moisture measurements from data set where measured soil moisture were compared to observed values at an agricultural farm near Immokalee, Florida with soils similar to those at Sites 1 and 4 ( unpublished study Shukla and Pandey ) The difference betw een sensor and gravimetric values was 10% and therefore this value was used for Sites 1 and 4. Measurement error in s urface water storage is related to how well the water volume is estimated Determination of water volume depends on the accuracy of topo graphic data which was 1m LIDAR based DEM In this study, m easurement of surface water storage, al though seemingly accurate by using higher resolution LIDAR based DEM data, can still introduce error due to the vertical accuracy of the LIDAR sensors Based on the specification s of the LIDAR sensor (Gemini Airborne Laser Terrain Mapper, Optech Inc, Canada) the elevation accuracy ranges from 5 to 30 cm. To investigate error in elevation measurement in the wetland area two points (within DW) determined by fi eld survey were compared to LIDAR based DEM data. Results showed that there was 5 cm difference in the elevation between the field survey and LIDAR based DEM data. When the 5 cm error was combined with the observed range of surface water fluctuations at the two sites, this transferred to 35 % and 65% errors in surface water storages at DW and SW, respectively. Groundwater Flux Using Law Flow in the groundwater system is generally determined by using an equation ation of continuity. fundamental law of groundwater movement. It relates groundwater flux to a material's ability to transmit water and the gradient that drives it. The Darcy equation can be written as :

PAGE 119

119 (4 3) where Q i s the volumetric flow rate K is the hydraulic conductivity, h is hydraulic head L is the distance nt In this study, a simpler approach is used to calculate steady state groundwater flow by using hydraulic gradients between groundwater levels on the boundary and surface levels in the wetland. To differentiate between the traditional terminology for groundwater movement, the terms of groundwater discharge and recharge used in this study are defined as follows. When groundwater levels at the boundary are higher than the surface water or groundwater levels in the wetland, groundwater flows from the reg ional groundwater system into the wetland upland system this is termed as groundwater inflow (groundwater discharge). When groundwater levels at the boundary are lower than the surface water level in the wetland, groundwater flows from the wetland upland system to the regional groundwater system (outside of the wetland upland system) which is denoted as groundwater outflow (groundwater recharge). Groundwater levels measured at GW26, GW33, GW32, and surface water stage at SU1 were used in conjunction with Equ. 4 3 to estimate groundwater fluxes from or to Site1 (Figure 4 2). At Site4, measurements of groundwater levels at GW42 and surface water levels at S U 2 were used to estimate groundwater fluxes (Figure 4 2). Main soil types at Site1 and Site4 are Basinger fine sand and Immokalee fine sand (Table 4 1). The hydraulic conductivity values for all soil types at both sites were obtained from the Florida Soil Characterization Data Retrieval System (FSCDRS, http:// flsoils.ifas.ufl.edu ). The FSCDRS provides access to a comprehensive soil dataset (1,300 profiles and over

PAGE 120

120 8,000 horizons distributed across 58 counties in Florida) including soil profile descriptions, soil taxonomic information, and physical and chemical soil properties. Depth weighted average hydraulic conductivity (equivalent hydraulic conductivity ) values were calculated using hydraulic conductivity and soil depth data (Das, 2009). With the assumptions of a homogeneous and isotropic medium, groundwater fluxes for both sites were estimated using measured groundwater levels and weighted hydraulic conductivity The e rror in groundwater flux from the Darcy equation was assumed to be entirely due to soil hydraulic conductivity Maximum and minimum hydraulic c onductivity values obtained from FSCDRS ( http://flsoils.ifas.ufl.edu ) for the soils found at the two site s were used to estimate the range of groundwater fluxes Results and Discussion Rainfall Annual ( May April ) rainfall during the study period varied considerably from 1 407 mm for 2009 ( 3 % higher than long term average of 1362 mm ) to 893 mm for 2010 ( 34 % lower than the long term average) with a two year average of 1 1 5 0 mm. Rainfall for 2009 and 2010 wet season s w a s 13% (831 mm) and 35 % (621 mm) below the long term average of 960 mm respectively. Dry season rainfall for 2009 ( 576 mm ) was 4 3 % above long term average of 404 mm while it was approximately 33 % lower than average for 2010 ( 272 mm ). In March 2010, the study site received unusually high rainfall ( 157 % higher than average). October 2010 was the driest period of the two wet seasons Above variability in rainfall volume and distribution had a direct influence on variability in water budget components includ ing the ET and surface flows both of which affect the groundwater fluxes Furthermore, considering the drier condition in 2010 wet

PAGE 121

121 season, the effectiveness of WWR may be masked due to lack of water to be stored in the wetland. Evapotranspiration and Surf ace Flow Both sites display ed a significant seasonal pattern in ET (Figures 4 3 and 4 4) with Site1 being higher than Site4 for the study period ( May 2009 May 2011 ) T otal ET estimated at Site1 (20 62 mm) was 13% higher than at Site4 (18 22 mm) for the perio d of May 2009 May 2011. Although S W ET was 34 % lower than it was in DW (Chapter 3), the difference in total ET was not less when including ET of upland area for two sites. It is due to the fact that upland areas within both sites account large portion of watersheds. For Site1, the ratio of wetland area to upland area is 1:4, while it is 1:3 for Site4. M onthly ET for both sites peak ed during May and was lowest during January (dorman cy of bahiagrass). It is likely due to the wet and warm con dition in May and the dry and cold condition in January. In annual basis, ET was accounted for 89% and 79% of rainfall in Site1 and Site4, respectively. For both sites, surface flows were observed from the middle of May through October (Figures 4 3 and 4 4) In 2009 wet season when rainfall was close to long term average, most of surface flow (89%) in Site1 was drained during August and September 2009. At Site4 33% of surface out flow occurred in June and July 2009 and 67% of surface outflow occurred in A ugust and September 2009. Temporal variation in surface flows between the two sites is due to the fact that Site1 has higher storage capacity behind the CRB structure than Site4 The only period when surface flow was observed during the dry season was in M arch 2010 and April 2010 (Figures 4 3 and 4 4) U nusually high rainfall observed in March 2010 and April 2010 resulted in the highest

PAGE 122

122 surface flows from Site1 in April 2010 ( 80 m m) while it was observed in March 2010 for Site4 ( 84 mm ). Water S torage Change in storage includes change s in surface (mainly in the wetland ) and subsurface storage s (mainly in the upland ) Total storages (sum of surface and subsurface water storage) at both sites are shown in Figure 4 5 Total storage was at the maximum in th e March April 2010 period due to the combination of high rainfall and low ET losses as opposed to wet season when rainfall and ET were both high. S torage in Site4 slightly dropped in August 2009. It is due to high ET losses and temporal variation of rainf all. M ost of the rainfall in August 2009 was observed during August 21 26 Storage at Site4 was lowest in November 2009 It is likely due to the antecedent dryness of Site4 and with Site4 being located at a higher location compared to surrounding area, water is likely to be lost to the surrounding low lying area through subsurface pathways. Difference in total storage between Site1 and Site4 was larger during the period of October 2009 January 2010. It is likely due to the fact that the Site1 is located at lower elevation and has higher storage capacity than Site4; therefore, Site1 tends to store more surface water and receive groundwater for the surrounding area towards to the dry season. Relationships of surface water stage volume for Site1 and Site4 sh ow n in Figure 4 6 show that Site1 has larger storage capacity behind the CRB structure at the top of the board (9.12 m, AMSL) compared to Site4 (9.47 m, AMSL) (Figure 4 6 ). With the implementation of the WWR, Site1 has 67 fold higher storage capacity than Site4. S tage volume relationships show that storage increases gradually at Site4 compared to Site1. At Site4, when water level reaches the top of CRB structure, only the drainage ditch will be full At Site1, when water level reaches the top of CRB structu re not only

PAGE 123

123 the drainage ditch will be full but also part of the wetland will be submerged ; therefore the abrupt increase in storage volume is observed when water level rises higher than 8.85 m, AMSL T he ratio of the storage capacity of the watershed to the area of its watershed is gross measure of the potential magnitude of storage of the watershed (Graf, 1999) Based on the relationship of stage volume for both sites, Site1 (0.03 m 3 m 2 ) has higher storage per unit area than Site4 (0.002 m 3 m 2 ) behind the CRB structure at the top of the board which indicates Site1 has potential to store more surface water than Site4. Water Budget In the first year (May 2009 April 2010), ET accounted for 74 % and 69 % of total water losses for Site1 and Site4, respectively At Site1, the re was a net groundwater outflow of 116 mm for the first year (Table 4 2) and accounted for 9 % of total outflow There was also a net groundwater outflow (79 mm) at Site4 (Table 4 3) and it accounted for 7 % of total outflow. Surf ace flows were 1 6 % and 2 4 % of total outflow for Site1 and Site4 for the first year, respectively. In the second year (May 2010 to April 2011), ET accounted for 95% and 84 % of total outflow for Site1 and Site4, respectively S urface flow w as 5% and 9% of to tal outflow for Site1 and Site4, respectively, for the second year. B oth sites received 34% less than long term annual average rainfall and the groundwater fluxes were mainly driven by ET In the second year, Site1 ET was higher than rainfall which result ed in decrease in storage ( 154 mm) and almost negligible groundwater inflow (1 mm) (Table 4 2) W hile at Site4 rainfall was slightly higher than Site4 ET and exceeded water left the site through groundwater (75 mm, 7% of total outflow).

PAGE 124

124 Annual average su rface flow for Site1 (1 30 mm) and Site4 (191 mm) was 1 1 % and 1 7 % of total outflow, respectively. The annual change in storage at Site1 varied from surplus to deficit, from a net gain of 152 mm for May 2009 April 2010 to a net loss of 154 mm for May 2010 April 2011. At Site4 the change in storage varied from a net gain of 214 mm during May 2009 April 2010 to a net loss of 153 mm during May 2010 April 2011. During May 2009 April 2010, change in storage at Site1 was less than Site4, which was likely due to the fact that more water left Site1 through ET (13% higher than Si te4) During May 2010 April 2011, change in storage at both sites was similar. Annual average net groundwater fluxes, calculated as residual of the water budget, were net groundwater out flow for Site1 ( 58 mm/year) and Site4 ( 77 mm/year) Compared to Site4, there was more water left Site1 through ET and less water left Site1 through subsurface flow During the wet season of 2009 (May 2009 October 2009), rainfall was 831 mm and total surface flows for Site1 and Site4 were 77 mm and 16 1 mm, respectively (Table s 4 4 and 4 5). For Site1, total ET and change in storage were 5 76 mm and 59 mm, respectively (Table 4 4). For Site1 there was a net groundwater outflow of 119 mm to regional groundwater system. At Site1, ET, surface flow and groundwater fl ux accounted for 75 %, 10 % and 15 % of total outflow respectively. During the 2009 wet season, ET and change in storage were 5 23 mm and 4 mm at Site4, respectively (Table 4 5). The re was a net outflow of 150 mm from Site4 (Table 4 5). At Site4, ET surface flow, and the groundwater flux were accounted for 63 % 19%, and 18 % of the total outflow, respectively. Based on percentage of outflow accounted by ET, surface

PAGE 125

125 flow and groundwater flux for both sites, Site1, which has higher ET, lower surface flow and groundwa ter outflow compared to Site4, has potential on retaining more water. In the dry season of 2009, from November 2009 to April 2010, total rainfall was 576 mm and the total surface flow for Site1 and Site4 were same (Tables 4 4 and 4 5). For Site1, ET was th e main outflow ( 7 4 % of total outflow ), and the groundwater flux was a net inflow by 3 mm (1% of total inflow) (Table 4 4). For Site4, ET accounted for 70 % of total outflow, and the groundwater flux was a net inflow by 71 mm (11% of total inflow ) (Table 4 5 ). High s urface flow s occurred in March 2010 and April 2010 for both sites It is primarily due to that both sites received higher than long term average rainfall during the 2009 dry season, especially in March 2010 and April 2010 Both sites received 79 mm rainfall in February 2010 which partial ly filled the watershed storage Additional rainfall in March 2010 and April 2010 exceeded the total storage below the top of the CRB structure resulting in high surface flows. During the 20 1 0 wet season, both s ites followed the similar temporal pattern for each of water budget components (Figures 4 3 4 4 Tables 4 4 and 4 5). T he groundwater flux was a net outflow of 85 mm (11% of total outflow) and 113 mm ( 15 % of total outflow) for Site1 and Site4, respectivel y. During the wet seasons, both sites acted as groundwater recharge sites. During November 2010 to April 2011, the ET was higher than rainfall for both sites. The groundwater flux was driven by ET. Soil water was extracted by ET which caused drop in ground water level in Site1 and resulted in groundwater inflow from the regional groundwater system. T he groundwater flux was a net in flow of 86 mm, and it accounted for 24% of total inflow at Site1. At Site4, the groundwater flux was also a net

PAGE 126

126 inflow of 39 mm w hich was 12% of total inflow. Compared to Site1, Site4 received less groundwater inflow which is likely due to that Site4 ET was 14% less than Site1. During the dry season, both sites functioned as groundwater discharge sites. Groundwater F l ux Surface water levels in the wetlands and groundwater levels on the boundaries throughout the study period for both sites are shown in Figures 4 7 and 4 8. Measured surface and groundwater levels measured at both sites showed that both sites received net gr oundwater inflows during the dry seasons (November 2009 April 2010 and November 2010 April 2011) (Tables 4 4 and 4 5). During dry season, ET depleted soil moisture at both sites and resulted in soil moisture deficit. This moisture deficit was replenished b y capillary rise from the shallow groundwater that dropped the groundwater levels and resulted in positive hydraulic gradient between regional groundwater and the site. To reach the new equilibrium, groundwater from outside started flowing into both sites. Daily groundwater levels monitored at wells GW26, GW33 and GW32 were higher than water level s in the wetland indicating that during the dry seasons groundwater flowed into Site1 (Figure 4 7). Similarly, groundwater flowed into the Site4 during the dry sea sons (Figure 4 8). During the 2009 wet season, the was a net inflow of 9 mm for Site1 T he sign and magnitude of groundwater flux was not consistent with the groundwater flux estimate d from water budget (119 mm) (Table 4 4). Although groundwater levels measured at Site1 boundary were higher than the surface water levels measured at Site1 during the 2009 wet season, it is possible that groundwater lost to surrounding low elevation areas instead of Site1

PAGE 127

127 During the 2010 wet season, there was a net groundwater outflow of 16 mm at Site1 which was not consistent with the groundwater flux (net out flow of 85 mm) estimated from water budget (Table 4 4). At Site1, the net groundwater flux was estimated using groundwater levels measured at three locations. Since the GW32 was the nearest well to SW1 (Figure 4 2), groundwater levels measured at GW32 have larger impact on the net groundwater flux estimation compared to GW33 and GW26. Furthermore, a values of field hydraulic conductivity of soils for the entire systems (Winter et al., 1988). The hydraulic conductivity values obtained from the soil survey is an average value of seve ral measured values at the nearest locations which are far away from the site and are unlikely to represent the actual value. In the wetland systems, especially in flatter landscapes such as flatwoods, groundwater fluxes may also be limited by low hydrauli c gradient, which are difficult to be measured accurately. For Site4, average groundwater levels (9. 23 m, AMSL) in the upland area were higher than average surface water levels (9. 00 m, AMSL) in the wetland area for May 2009 to April 2011 indicating positi ve hydraulic gradient from the regional groundwater system to Site4 (Figure 4 8). However, the sign of groundwater fluxes estimated from water budget for the period of two wet seasons was not consistent with groundwater fluxes estimated from water budget ( Table 4 5) It is likely due to the fact that on ly one was used Groundwater flux is influenced by hydraulic gradient which is related to spatial variation of groundwater level in the surrounding areas Since there was no other site where groundwater level was measur e d during the study period, the spatial distribution of

PAGE 128

128 groundwater levels could not be characterized at Site4, therefore, the groundwater flux calculated using Darc equation was likely associated with errors. The estimated groundwater flux may not represent the actual groundwater flow condition at Site4. Uncertainty Analysis The order of magnitude of water budget estimates is representative of commonly found natur al wetland upland systems within ranchlands in LO watershed In this study, the major components of water budgets for two wetland upland system s are rainfall and ET. E rror in rainfall data was at a minimum because rainfall was measur ed at the study site; t herefore, the largest source of error was ET for both sites (Tables 4 4 and 4 5). Error in surface flow is mostly like to be higher during the wet periods such as wet seasons of 2009 and 2010 and the dry season of 2009. Based on results of uncertainty ana lysis, the groundwater fluxes were net outflows by 11982 mm and 8583 mm for the 2009 and 2010 wet seasons, respectively, for Site1 (Table 4 4) This indicates that Site1 act s as a groundwater recharge site in the wet season when uncertainties in other water budget components are considered For the dry season of 2009, the groundwater was a net inflow of 3 50 mm for Site1; while it was a net inflow of 8647 mm for the dry season of 2010 (Table 4 4). Given t he fact that error in the net groundwater flux was one order of magnitude larger than the net groundwater flux itself for the dry season of 2009, groundwater flux was the least significant component during that period. Considering the range of error of groundwater flux esti mates, the magnitude of groundwater flux es estimated from water budget at Site1 w ere c omparable to those equation for the study period Results showed that the groundwater fluxes were net outflows by 15082 mm and 11383 mm for the 2009 and 2010 wet seasons, respectively, for Site4 (Table 4 5).

PAGE 129

129 For the dry season of 2009, the groundwater was a net inflow of 71 53 mm for Site4; while it wa s a net inflow of 3946 mm for the dry season of 2010 (Table 4 5). Errors in the net groundwater flux and ET were both larger than estimated net groundwater flux for the dry season of 2010 which in dicates the importance of accuracy on ET quantification. Fo r Site4, the sign and magnitude of groundwater flux estimated from It is primarily due to the spatial variation of groundwater levels was not well represented when estimating g roundwater flux using equation Chapter Summary and Conclusion s Water budgets were constructed for two wetland upland sites using EC based ET and other data such as rainfall, surface flow s and levels and soil moisture. Groundwater fluxes for two wetland upland system s were estimated as the residual equation Evapotranspiration was the dominant outflow component at both sites. On an annual basis, ET accounted for 84 % and 7 6 % of total water losses for the deep and shallow wetland sites respectively. For the period of May 2009 April 2010, the groundwater flux accounted for approximately 9 % and 7 % of the total outflow for the deep and shallow wetland sites respectively. For the dr ier year ( May 2010 to April 2011 ) the groundwater flux accounted for only 0.1 % of total inflow at the deep wetland site while it accounted for 7 % of total out flow at the shallow wetland site Results from water budget showed that average groundwater flux from the wetland upland sy stem is a net out flow of 58 (deep wetland) to 77 (shallow wetland) mm/yr. For the wet season, the groundwater flux was a net outflow for both sites and represented average of 13% and 17% of total outflow for the deep and shallow wetland

PAGE 130

130 sites, respectively For the dry season, there were net groundwater inflows for both sites, and they represented 12% of total inflow. Groundwater outflow at the deep wetland site was more than the shallow wetland site for the wet season. Groundwater flux was equal or more th an surface flow in the wet season for both sites which indicates that groundwater flux is as or more important than surface flow. Given that groundwater outflow increases with surface flow in the wetland (2009 wet season), retained water is likely to be lo st to groundwater with part of it may be lost through ET. Higher groundwater outflow at the shallow wetland site is mostly because of its higher elevation compared to the deep wetland site and regional groundwater; therefore it is not a preferable site for water storage compared to the deep wetland site. For both wet seasons, except the 2010 wet season for the deep wetland site, the were not consistent with the groundwat er fluxes estimated using water budget. These differences were attributed to lack of site specific hydraulic conductivity values and inability to capture the spatial variation of groundwater levels within and outside of wetland upland drainage area. Theref ore, a flownet analysis utilizing depth measurements from multiple wells around the site boundary is needed to improve the groundwater flux estimates. In this study, groundwater fluxes for both sites were closer to the fluxes obtained from MIKE SHE model (Chapter 5) Groundwater is an important component of the water budget for two wetland upland systems. Results show that average groundwater fluxes account for 15% of total outflow and 12% of total inflow for the wet and dry seasons, respectively. Both sit es

PAGE 131

131 functioned as groundwater recharging site during the wet seasons and discharging site during the dry season. Surface and ground water outflows from the deep wetland site were lower than the shallow wetland site. Since most of rainfall comes during the w et season and the purpose of WWR is retaining water in the LO watershed to prevent LO from receiving high surface flow s implementing WWR at the deep wetland site can help retain more water than the shallow wetland site. Results from this study will help i n understanding the fluxes from and to the wetlands for a variety of applications ranging from hydrologic modeling to wetland restoration.

PAGE 132

132 Table 4 1. S oil type and associated areas in Site1 and Site4. Site1 Site4 Type Area (m 2 ) Percentage Area (m 2 ) Percentage Immokalee fine sand 442209 55 110182 45 Basinger fine sand 166768 21 133467 55 Myakka fine sand 91161 11 Floridana fine sand 59652 8 Valkaria fine sand 40533 5 Total 800323 100 243649 100 Table 4 2. Annual water budget for Site1 for two years (May 2009 April 2011) Rainfall (mm) ET (mm) Surface flow (mm) (mm) ("GWI GWO") (mm) May 2009 Apr il 2010 1 407 93 5 20 4 15 2 116 May 2010 Apr il 2011 893 99 2 5 6 15 4 1 p ositive value for the net groundwater flux indicates a net groundwater inflow and negative value indicates opposite. Table 4 3. Annual water budget for Site4 for two years (May 2009 April 2011) Rainfall (mm) ET (mm) Surface flow (mm) (mm) ("GWI GWO") (mm) May 2009 April 2010 1 407 82 5 28 9 214 7 9 May 2010 April 2011 893 87 8 94 153 75 p ositive value for the net groundwater flux indicates a net groundwater inflow and negative value indicates opposite Table 4 4 Seasonal water budget components for Site1 Rainfall (mm) ET (mm) Surface flow (mm) Change in storage Net groundwater "GWI GWO", mm) Net groundwater Law, mm) Wet Season 2009 (May. 2009 Oct. 2009) 831 ( 33.2 ) 5 76 ( 74 0 ) 77 ( 7. 7 ) 5 9 (5.8) 119 ( 81. 7 ) 9 { 5 to 21 } ** Dry Season 2009 (Nov. 2009 Apr. 2010) 576 ( 23. 1 ) 3 58 ( 41. 8 ) 127 ( 12.7 ) 93 (9.2) 3 ( 50.2 ) 79 { 44 to 155 } Wet Season 2010 (May. 2010 Oct. 2010) 621 ( 24. 9 ) 6 1 4 ( 77.7 ) 5 6 ( 5.6 ) 133 (13. 3 ) 8 5 ( 82.8 ) 16 { 2 to 16 } Dry Season 2010 (Nov. 2010 Apr. 2011) 272 ( 10.9 ) 37 9 ( 45.3 ) 0 (0 .0 ) 2 1 (2.0) 8 6 ( 46.6 ) 132 { 69 to 242 } Positive value for the net groundwater flux indicates a net groundwater inflow and negative value indicates opposite. values in the parentheses indicate absolute error ** values in the brackets indicate range of groundwater flux.

PAGE 133

133 Table 4 5 Seasonal water b udget components for Site4 Rainfall (mm) ET (mm) Surface flow (mm) Change in storage Net groundwater "GWI GWO", mm) Net groundwater Law, mm) Wet Season 2009 (May 2009 Oct. 2009) 831 ( 33.2 ) 5 23 ( 73.7 ) 161 ( 16.1 ) 4 (0. 4 ) 150 ( 82. 5 ) 195 { 1 24 to 348 } ** Dry Season 2009 (Nov. 2009 Apr. 2010) 576 ( 23. 1 ) 302 ( 40.7 ) 127 ( 12.7 ) 21 8 (21. 7 ) 71 ( 53.1 ) 243 { 155 to 434 } Wet Season 2010 (May. 2010 Oct. 2010) 621 ( 24. 9 ) 5 52 ( 77. 3 ) 94 ( 9.4 ) 138 (13.7) 113 ( 82. 9 ) 467 { 298 to 836 } Dry Season 2010 (Nov. 2010 Apr. 2011) 272 ( 10. 9 ) 32 6 ( 44.5 ) 0 (0 .0 ) 1 5 (1. 5 ) 3 9 ( 45.8 ) 289 { 184 to 516 } Positive value for the net groundwater flux indicates a net groundwater inflow and negative value indicates opposite. values in the parentheses indicate absolute error ** values in the brackets indicate range of groundwater flux.

PAGE 134

134 Figure 4 1. Aerial photo of the study site and its sub watersheds. Surface flows from Site1 Site2, Site3 and Site4 were measured a t Q1, Q2, Q3, and Q4, respectively, and the main outlet of the study site wa s measured at Q5.

PAGE 135

135 Figure 4 2. The locations of surface water wells (S U 1, S U 2), groundwater wells (GW26, GW32, GW33, GW42), and soil moisture stations (SM34, SM13).

PAGE 136

136 Figure 4 3 Monthly water budget components for Site1. Figure 4 4 Monthly water budget components for Site4. Figure 4 5 Monthly total p recipitation and water storage s for both sites.

PAGE 137

137 Figure 4 6 Relationships of stage volume below the top elevation of the culvert and riser board structure at Site1 and Site4. Dash line indicates top elevation of the culvert and riser board structure (Si te1: 9.12 m, Site4: 9.47 m). A) Site1. B) Site4.

PAGE 138

138 Figure 4 7 Surface water and groundwater level s at Site1. Figure 4 8 Surface water and groundwater level s at Site4.

PAGE 139

139 CHAPTER 5 SIMULATING THE EFFECTS OF WATER RETENTION ON WATER DYNAMICS OF RANCHLANDS IN THE NORTHERN EVERGLADES WATERSHED Overview Excessive flow s and unnatural peak flows to Lake Okeechobee (LO), consequences of the extensive drainage of the land s through ditches and canals have endangered (SFWMD, 2013) Lake Okeechobee a multi functional lake located at the center of Kissimmee Okeechobee Everglades aquatic eco system provides a number of values and benefits to the st economy and environment, including environmental, public and agricultural water suppl ies flood protection, and natural habitat for a variety of endangered and threatened animal s and plant species. Since the 1900s due to the demands of improved agricultural production and flood control for expanding population, large areas around LO were drained for agricultural land use s and urban development (SFWMD, 2000; NR C 2008) Agricultural land uses with in the LO watershed contain extensive net work of drainage ditches to prevent root damage under prolonged condition of high water table ( Harvey and Havens, 1999 ) This construction of drainage network has not only decreased the watershed storage but also resulted in high flows and nutrient loads t o LO (SFWMD, 2008) Different approaches such as distributing storage on ranchlands or in agricultural impoundments, have been proposed to increase water storage in LO watershed to reduce flow s (TMDL) Cattle ranching is the dominant land use (36% by area) in the LO watershed, and wetlands on ranches have been considered for their potential to store water (Tweel and

PAGE 140

140 we tlands. Most of the ranches in the LO watershed still contain functioning wetlands, though most of them have been partially or completely drained. These wetlands provide valuable habitat, and water and nutrient storage services (Dunne et al. 2007). Wetland s Re taining water in these ranch land wetlands is expected to help restore the natural flows to the lake by increasing storage (Bohlen et al., 2009) Wetland w ater retention (WWR) can be defined as the prevention of surface water from being discharged into receiving waters by storing it in a wetland and adjacent upland areas. Water is retained and stored until it is lost through percolation, through transpiration from plants or through evaporation from bare soil and the free water surface. Although WWR is being promoted as one of the b est m anagement p ractice s (BMPs) no data is available to confirm its effectiveness in reducing the flows from ranchland s (Steinman et al., 200 3). Although WWR seems to be an attractive alternative, its effects on wetland/watershed water dynamics and their pathways in the flatter landscape of South Florida are not completely understood (Shukla et al., 2007). Due to the complexity of the wetland/w atershed relationship, there is still uncertainty over the hydrologic budgets and the hydrologic functions of wetlands (Carter, 1986; Owen, 1995). Effect of water retention on wetland and upland hydroperiod s and water budget components (e.g. evapotranspira tion and groundwater flow) are not yet fully understood in the LO watershed hydrologic processes are interrelated C hange in one component (e.g. storage of surface water) generally leads to change for oth er components (e.g. groundwater discharge and recharge) in a contiguous area (Winter, 1988). Rehydrating a wetland by increasing the water

PAGE 141

141 discharge elevation can change water storage, surface and subsurface flows, and groundwater discharge and recharge ar eas (Winter, 1988). Assessments of the cumulative effects of one or more of these changes are affected by the uncertainty in the measurements as well as the assumptions regulating the hydrologic process. Wetlands in many hydrogeologic settings may appear t o be hydraulically isolated from a surface perspective, but they are not hydraulically isolated from a groundwater perspective (Winter and LaBaugh, 2003). Increasing water storage and its effects on water budget components of a wetland may change the local hydraulic gradients and result in the groundwater flow to downstream water bodies and adjacent depression al areas through subsurface flows However, these changes are not known Although it is often assumed that WWR can reduce surface flows, it may at ti mes reduce the wetland storage capacity and increase the total and peak flows after a rainfall event. I f wetlands are already saturated, they may have little capacity to store additional water (Verry and Boelter, 1979) Although several studies have addres sed the site specific relationships between the wetland storage, drainage to stream flow, and evapotranspiration ( ET ) the se results may not be transferable to other sites especially for regions such as South Florida, which has nearly flat topography, sand y soils, and a shallow water table. Water budgets provide the framework from which one can investigate the linkages and fluxes between the hydrologic of the wetland and its relationship to the surrounding terrain (Drexler et al. 1999). T here have been few wetland water budget studies (Owen, 1995; Bradley, 1997; Mitsch and Gosselink, 2000; Nungesser and Chimney, 2006). Drexler et al. (1999), in a detailed study of a small peatland, concluded

PAGE 142

142 that there was a wide margin of error in all components of the wate r budget, with the exception of precipitation. Review of several studies (e.g. Drexler et al. 1999; Bradley, 2002) suggests that the main error in estimating water budget comes from two components: ET and groundwater flux. Constructing a water budget with precisely quantified ET and groundwater fluxes could help enhance the understand ing of the effects of water retention on surface and subsurface flows in the LO watershed can be improved by constructing water budget components alone using the measured data for a specific control elevation at the wet land outlet. For example, consider the case of a study where hydrologic monitoring is conducted for two years, one year each for a low (drained) and high ( retention ) control elevation. Even if one has accurate ET data and groundwater fluxes, the relative d ifference in ET and groundwater fluxes between the two years cannot be fully attributed to WWR Year to year variability in rainfall and antecedent hydrologic conditions in the watershed can mask the effect of water retention on ET and groundwater flux Al though the paired wetland design may partly remove the effect of rainfall, it is difficult if not impossible to find two similar wetlands that have similar soil s plant s topographic and hydrologic characteristics. Use of hydrologic models in conjunction w ith the water budget analys is, discussed above is another alternative to evaluate the effects of WWR. A variety of models have been developed and used for simulating water and nutrient transport of a variety of landscapes throughout the world. These model s vary in

PAGE 143

143 their accuracy depending on the extent to which the hydrological processes governing the water and nutrient cycling are included and what simplifying assumptions are made to represent these hydrological processes (Demissie et al., 1997) Although the building procedures and components of existing hydrologic simulation models are similar, one model may be very different from another in its capability to simulate a land use practice and its applicability to different regions (Demissie et al., 1997) South Florida hydrology is unique especially with regards to the extensive network of ditches, low slope (less than 6 cm/km, Obeysekera et al., 1999) highly interactive surface and ground water systems, shallow water table condition (approximately 1.2 to 1.5 m below the surface) and sandy soils (hydraulic conductivity > 1.410 4 m/s, Boman and Tucker, 2002) One of the unique aspects of South Florida hydrology is that the infiltration capacity of these soils is rarely exceeded and overland flow occurs du e to saturation excess after the water table has reached the surface ( Hernandez et al., 2003; Jaber and Shukla, 2004) Only few models are accurately able to simulate the surface and groundwater interactions in South Florida (Sun et al., 1998) for example SWAT ( S oil and W ater A ssessment T ool Arnold et al., 1998 ), WAM ( Watershed Assessment Model Bottcher et al., 2002) ACRU2000 ( A gricultural C atchments R esearch U nit Kiker et al., 2006) and MIKE SHE/MIKE 11 (Refsgaard and Storm, 1995). SWAT a watershed scale model developed by United State Department of Agriculture (USDA) Agricultural Research Service (ARS), is used to predict the impact s of land management practices on water, sediment, and agricultural chemical yields in large complex waters heds with varying soils, land use s and management

PAGE 144

144 conditions over long periods of time. For example, Wang et al. (2011) modified SWAT to simulate the artificial water input to the designated wetlands in Qingdianwa depression in Tianjin, China and evaluate the effect of recharging wetlands on local hydrological cycle and estuary ecology. Although SWAT can represent runoff generation areas at a watershed scale. However, it does not allow runoff routing across the sub watersheds, runoff was assumed to reach t he stream or watershed outlet. Such assumption may not be valid at regions like Florida where surface runoff can infiltrate before reaching the streams (Srinivasan et al., 2005) Furthermore, since SWAT has inability to simulate water exchange between sub watersheds, it cannot represent the spatial variation of groundwater within the watershed. WAM is a GIS based hydrologic model which allows engineers and planners to assess the water quality of both surface water and groundwater based on land use s soils, climate, and other factors (Bottcher et al., 2002). WAM has been used to simulate daily flows in several Florida watersheds such as Suwannee River and Caloosahatchee River (Bottcher et al., 2003). However, lack of documentation of WAM model and its simpli fied approach of cell to stream water and solute delivery, simplified in stream water quality processes, inability to adequately represent small scale short term storm event impacts (especially in Florida, afternoon thunderstorms are frequent during June t s (i.e. the re infiltration of ponded water) and peak flow s occurred after the short term storm event ( Graham, 2009; Graham et al., 2009). Therefore, WAM is not applicable for evaluating the effects of WWR in ranchland s

PAGE 145

145 ACRU2000 an object oriented hydrologic model, is a multi purpose, daily time step, physical conceptual model that can simulate stream flow, ET, and land cover/management and the abstraction impact on water resources at a daily time step (Kiker et al., 2006). ACRU2000 was developed and has been tested for use in South Florida hydrology (Martinez, 2008). ACRU2000 was shown to adequately predict water table depths, soil moisture distributions, and saturation e xcess runoff in the flatwoods of the LO watershed (Martinez, 2008). However, d ue to lack of representation of surface water convey ing feature (e.g. channel, drainage ditches) in the model, ACRU2000 cannot explicitly simulate the channel routing and hydraul ic structures including weirs, gates, bridges and culverts that are commonly found within wetland environments (Graham, 2009) Therefore, ACRU is not suitable for evaluating the effects of WWR. MIKE SHE/MIKE 11 (Refsgaard and Storm, 1995) is one of few a dvanced integrated hydrological modeling system s that has the ability to represent South Florida hydrology (Jaber and Shukla, 2012 ) It simulates water flow in the land based phase of the hydrological cycle from rainfall to river flow, via various flow processes such as overland flow, infiltration into soils, ET from soils and vegetation, and groundwater fl ux MIKE SHE/MIKE 11 has bee n used to simulate the different hydrologic processes and their interactions in an impoundment containing wetlands in citrus grove s in South Florida to assess several water storage alternatives (Jaber and Shukla, 2004). Jaber and Shukla (2004) noted that M IKE SHE/MIKE 11 was effective in modeling the complex surface/groundwater interactions in addition to the different hydraulic structures such as pumps and culverts used in agricultural areas of South Florida. Advantages of the MIKE SHE/MIKE 11 model over o ther existing models (e.g. SWAT) in simulating

PAGE 146

146 wetland hydrology are: 1) it is designed for application in low relief terrains (Graham and Butt, 2005 ; Jaber and Shukla, 20 12 ) ; 2) it offers options to simulate the infiltration processes and water movement i n the unsaturated zone (e.g. Richards e quation gravity flow, and two layer water balance ); and 3 ) the model can simulate both wetland and upland lateral movement of surface and ground water flows. It is important to ensure the selected hydrologic model ha s an internal consistency of results while calibrating and validating the model to represent site specific hydrologic processes Previous studies suggested that hydrologic model s calibrated only against the discharge measurements at the watershed outlet ma y not perform well in simulating other hydrologic components (such as the groundwater discharge/recharge, groundwater level, and unsaturated water content) (e.g. Ambroise et al., 1995; Refsgaard, 1997) (Wang et al., 2012). It has been long recognized that multi objective framework (e.g. multi site, multi variable) is essential for model calibration (Ambroise et al., 1995; Andersen et al., 2001; Khu et al., 2008). Bergstr om et al. (2002) suggested that the model calibrated against more measured internal vari ables rather than stream flow can greatly increase confidence in the physical relevance of the model. Hydrology in South Florida is known for its intense surface and ground water interactions; therefore, it is crucial that the calibrated model can represen t surface and groundwater level s correctly. With better representation of spatial variability in surface and groundwater levels, the effects of WWR on water storage and ecological enhancements can be better evaluated. Hydrologic models are usually calibrated to measured historical data (e.g. surface water flow) to identify unknown model parameters such as crop coefficient in

PAGE 147

147 simulating ET process and saturated hydraulic conductivity in simulating water movement in saturated zones, to assess model p erformance. However, the parameter uncertainty in most of hydrologic models degrades the utility of these models. E vapotranspiration and groundwater flux are the two main components that affect a Commonly used ET method can res ult in almost 23% error in wetland ET (Chapter 2 and 3) which is more than surface flow (Chapter 4) and sometimes groundwater flux (Chapter 4) Error in ET may also result in under or over estimation in groundwater flux. To improve the accuracy and perform ance of models, accurate estimations of ET and groundwater flux are needed. Shoemaker et al. (2008) studied the sensitivity of wetland saturated hydraulic heads and water budgets to ET in South Florida and concluded that reliable estimates of ET are necess ary for simulating wetland water budget. Sumner and Jacob (2005) conducted a study of estimation of pasture ET at a commercial farm in central Florida and indicated that ET estimates from the EC method can improve the quality of hydrologic model calibratio n through reduction in the uncertainty of the AET component of the model. Use of improved estimates of ET and groundwater flux could improve the accuracy of the model in two ways: 1) input as measured values to better calibrate model parameters (difficult but not impossible) ; and 2) comparison of the simulated values of ET and groundwater flux with the measured and estimated data and refine the conceptualization or parameterization of the model. Improved model calibrat ion performance on simulat ing the hydrologic processes for the specific applications, for example, evaluating the effects of WWR for a variety of control elevation alternatives Designing a WWR strategy requires

PAGE 148

148 evaluating different water retention alternatives with rega rds to peak flow reduction and volume of water retained. By using field verified model s different alternatives can be simulated, and the simulated results can be used to design on wetland water storage strategies. The objectives of this study are to: 1) e valuate the MIKE SHE/MIKE 11 coupled model for simulating water dynamics of two wetland upland systems in ranchlands of South Florida; 2) simulate the effectiveness of different WWR alternative s with regards to surface water level and flow and water storage and identify an optimum alternative Materials and Method s Site D escription and I nstrumentation The study area is a commercial cow calf ranch located 13 km northwest of LO (27 o o 2 The land uses within the study site can be classified into three types: improved pastures, upland hardwood forests, and depressional wetlands. Based on the topographic data and field survey, the study site can be divided into four sub watersheds (Figure 5 1). They are Site1, Site2, Sit3, and Site4. Wetlands located in Site1 and Site4 are deep (DW) and shallow depressional wetlands (SW), respectively. Outlets for Site1, Site2, Site3 and Site4 are Q1, Q2, Q3, and Q4, respectively (Figure 5 1). Site1 receives flows from the upland areas through the drainage ditches and sheet flow s The collected flows exit Site1 and combine with flows from Site3. Flows from Site3 then combines with flows from Site2 and eventually exit the study site at Q5. Site4 has the separat e drainage system that drains the site at Q4 (Figure 5 1). Soils in the area are typically poorly drained and highly sandy (NRCS, 2010) and are mainly comprised of Immokalee fine sand and Basinger fine sand in upland areas.

PAGE 149

149 For the wetland area in Site1 F loridana fine sand is dominant. The WWR evaluation project was started in June 2005 by a research team of Institute of Food and Agricultural Sciences (UF/IFAS), University of Florida. The WWR was implemented at two wetland upland sites (Site1 and Site4) by installing a riser board structure with a corrugated metal culvert, at the outlets to regulate drainage (Figure s 5 1 and 5 2 ). Topography of the study site was characterized using airborne Light Detection and Ranging (LIDAR) technique in May 2008 to devel op a high resolution digital elevation model (DEM) data (1m 1m) for the study site. A fiberglass trapezoidal flume was installed at Q1 and Q4 to measure surface flows. F low rates w ere determined by two methods: 1) two pressure transducer s (KPSI, Pressure System, Hampton, VA), installed at the upstream and downstream end of the flumes, were used in conjunction with the flume equation to estimate the flow; 2) an Acoustic Doppler V elocity meter (SonTek/YSI, San Diego, CA) installed in the throat section of e ach flume for accurately measuring the flow velocity Flow was estimated by multiplying flow velocity and cross section area ( according to instantaneous water level measured at the flume) Rainfall was recorded on 15 minute interval with two H 340 tipping bucket rain gauges (Design Analysis Associates, Inc, Logan, UT) located near Q1 and at the weather station near Q2. Groundwater levels were monitored at multiple locations at Site1 and Site4 by using pressure transducers (Figure 5 1). Two eddy covariance ( EC) towers were installed at the middle section of wetland areas for both sites to quantify ET (Figure 5 1) and has already been discussed in Chapters 2 and 3, respectively.

PAGE 150

150 Model D escription MIKE SHE (Syste`me Hydrologique Europeen) is a spatially distri buted, physically based and grid based hydrologic model that can simulate integrate d surface water and groundwater systems It can simulate all the major hydrologic processes (e.g. ET, overland and channel flow, and groundwater flow) and is comprised of several independent modules that represent each hydrologic process. The governing partial differential equations for these hyd rologic processes are solved numerically by finite difference methods. Model development began in 1977 as a collaborative research project by the Institute of Hydrology in the United Kingdom, SOGREAH in France, and the Danish Hydraulic Institute in Denmark (Graham and Butts, 2005). The Danish Hydraulic Institute (now called DHI Water and Environment) is the developer of the commercial version of MIKE SHE. The model was first proposed as a blueprint for distributed hydrological modeling by Freeze and Harlan (1969) using the physics based representation of the underlying catchment processes. The overall model structure is illustrated in Figure 5 3. Water inputs into the system in the form of precipitation or snow. After the canopy interception has been account ed for, the net precipitation either flows as overland flow or is infiltrated into the unsaturated zone as a one dimensional (vertical) flow for each grid element. Once water reaches the groundwater table, a three dimensional saturated flow groundwater mod el with rectangular grid elements simulates the spatial distribution of groundwater and exchange between boundaries including surface water bodies. MIKE SHE is also integrated with a flow routing model which is commonly referred as MIKE 11. MIKE 11 include s comprehensive functions for simulating complex channel networks, lakes and reservoirs, and river structures, such

PAGE 151

151 as gates, sluices, and weirs. MIKE 11 is useful in representing highly managed river systems with structures with operations rules (DHI, 200 6). Model D evelopment The model development for the ranch where the WWR was implemented is described in the following sections. T he process follows the recommended steps to build a MIKE SHE/MIKE 11 model (DHI, 2006). Topography MIKE SHE allows users to define the size of the grid cell. In general, increasing the level of discretization increases the accuracy of the simulation, but there is a level beyond which the model performance cannot be improved (Mamillapalli et al., 1996). Reduction of grid size usually means increase in computational time and requires considerable resources in collection and processing of data Vazquez et al. (2002) found slight change s in model performance measures on surface flow, Nash Sutcliffe coefficie nt varies from 0.76 to 0.70 with a range of grid cell sizes (300 1200 m) for a catchment (586 km 2 ). They also found that small flows were better simulated by using grid size of 1200 m and peak flows were better simulated by using grid size of 600 m. McMich ael et al. (2006) used MIKE SHE/MIKE 11 grid cell with 30 m and 270 m resolution to predict stream flow in a catchment with an area of 34 km 2 They found that grid cell of 30 m and 270 m result ed in similar values of Nash Sutcliffe coefficient 0.96 and 0. 93, respectively Site specific results are not transferable; however, the grid size can affect the accuracy of model prediction on specific components. For this study, using 20 m as the grid size can divide Site1 (0.8 km 2 ) into 2000 grid cells which is more than the number of grids (1600) used by Vazquez et al. (2002). Considering the spatial resolution and the running time, the appropriate grid size for this

PAGE 152

152 study is 20 m. Original LIDAR based DEM map (1m resolution) was converted to 20 m resolution usi ng geographic information system (GIS) software, ArcGIS (v.10, ERSI, Redlands, California) (Figure 5 4). Overland F low and Channel Flow MIKE SHE simulates the o verland flow using a two dimensional diffusive wave approximation. MIKE 11 simulates the channel flow (e.g. water levels and flow in rivers and estuaries) using an implicit, 1D, finite difference formulation. MIKE SHE exchanges water (overland and groundwater flow) with MIKE 11 river feature which intersects the MIKE SHE grids. efficients, detention storage values, and the separated flow areas are important parameters Initial values of Manning roughness coefficients and detention values were adapted from the Kissimmee Basin Modeling and Operations Study (KBMOS) (Earth Tech and D HI, 2007 ). Surface water is routed using the overland flow component of MIKE SHE and channel flow component of MIKE 11. Inputs for the MIKE 11 consist of the ditch network path, ditch cross sections, boundary conditions of the downstream ends of the draina ge ditches (Q4 and Q5) and bed resistance ( Mannin roughness coefficient ) The ditch network in the study site was delineated and digitized using ArcGIS software based on the aerial photos and field survey conducted as par t of this study The aerial photo and LIDAR based DEM data combined with field survey were used to identify the cross sections necessary to provide adequate spatial representat ion of ditches. The control structure, culvert with riser board (CRB) structure s at the outlets of Sites 1 and 4 (Figure 5 2) w ere implemented in MIKE 11 to represent the drainage management by adjusting the boards for achieving the desired discharge levels Before

PAGE 153

153 the CRB structures were installed at both sites, the drainage levels at outlets of Sites 1 and 4 were 8.21 m and 8.98 m AMSL, respectively. The CRB structure implemented at Site1 has dimensions of 0.92 m in height and 1.07 m in width; while at Site4 it was 0.49 m height 0.76 m width (Figure 5 2) The CRB structure only al lows water flow when the water level exceeds the top elevation of the CRB structure. From October 1, 2008 to May 31, 2011, the top elevations of the CRB structure at Site1 and Site4 were maintained at 9.1 3 m and 9.47 m AMSL respectively. Evapotranspiratio n MIKE SHE uses the Kristensen and Jensen (1975) model, which uses leaf area index (LAI), vertical root distribution characteristics, interception parameters and crop coefficient (K C ) to estimate actual ET (AET) Actual ET represents combination of evaporation from canopy, evaporation from soil or ponded water, and transpiration from the plants which is sourced from unsaturated and saturated zone. The spatial distribution of vegetation, incorporated in land use data in MIKE SHE, w as b ased on the management record provided by the rancher and the field survey s MIKE SHE also includes a simplified water balance method for both ET and the unsaturated zone storage. The t wo l ayer w ater b alance method divides the unsaturated zone into a root zone, from which ET can be extracted, and a zone below the root zone, where plant cannot extract water for transpiration (Yan and Smith, 1994). Evapotranspiration is extracted first from intercepted water (based on LAI ), then ponded water and finally via t ranspiration from the root zone, based on an average water content in the root zone. The two layer water balance method is preferred for modeling the hydrologic system in the environments with shallow unsaturated zone (Graham and Butts, 2005).

PAGE 154

154 Management r ecords of the study site showed that the upland areas were pre dominant ly covered by b ahiagrass ( Paspalum notatum ) and pinewoods Wetland areas comprise of mixed wetland plants including leafy bladderwort ( Utricularia foliosa ), knotgrass ( Paspalum distichum ), southern watergrass ( Luziola fluitans ), common rush ( Juncus effusus ), duck potato ( Sagittaria lancifolia ), brook crowngrass ( Paspalum acuminatum ), and common duckweed ( Lemna minor ). The land use types and the corresponding areas have been summarized in Table 5 1. Four land use parameters affecting ET process are LAI, root depths, detention storage, and K C for the upland and wetland adapted from Earth Tech and DHI (2007) are presented in Tables 5 1 and 5 2 respectively The K C for wetland areas were est imated using EC method (Chapter s 2 and 3). Since K C for the wetland area in Site4 was close to the reported K C values for bahiagrass in Florida ( Jia et al., 2009; Chapter 3), K C of b ahiagrass for this study was calculated by taking the average of K C derived from EC measurements for Site4 and the values from Jia et al. (2009). The K C for pinewood w as adapted from the Kissimmee Basin Modeling and Operations Study (KBMOS) (Earth Tech and DHI, 2007 ). Unsaturated F low In this study, the t wo l ayer w ater b al ance method was used to simulate unsaturated zone flow. It assumes a uniform soil profile for the entire. The four principle parameters related to each soil type were saturated water content, field capacity, wilting point and infiltration rate. The soil ma p for the study area was obtained from Natural Resources Conservation Service (NRCS) SSURGO soil coverage maps (Figure 5 5). Soils in the area are typically poorly drained and highly sandy (NRCS, 2010) and are comprised of Immokalee fine sand, Basinger fin e sand, Myakka, Valkaria, and Floridana fine sand. Soil characteristics (Table 5 3) were derived from a soil database in the

PAGE 155

155 Florida Soil Characterization Data Retrieval System (FSCDRS) ( http://flsoils.ifas.ufl.edu ). Saturated Z one The groundwater component in MIKE SHE requires a three dimensional geological model describing the extent, thickness, and hydrogeological parameters of aquifers. One of three groundwater layers in the region was represented in the model: the Surficial Aquifer System (SAS) The Intermediate Confining Unit and the Upper Floridan Aquifer were not considered because of their limited effects on the surface hydrological processes in the area and lack of site specific hydraulic property data The thickness of the SAS was obtained from a U.S. Geological Survey (USGS) study for the LO watershed (Reese, 2004). The thickness of the SAS was 45 m. The hydrogeological parameters required for the model are: horizontal (K H ) and vertical (K V ) hydraulic cond uctivities and storage coefficients for SAS In MIKE SHE, the spatial and temporal variation of the groundwater level in the saturated zone is described mathematically by the three dimensional Darcy equation and solved numerically by an iterative implicit finite difference technique The model uses groundwater levels as the boundary conditions. G roundwater water levels measured at GW 11, GW 23, GW 33 and GW10 (Figure 5 1 ) were used as the boundary conditions in this study For periods when the groundwater lev el data at the above mentioned wells was missing due to malfunction of the levelogger s groundwater levels from the nearest well were used. Model C alibration and V alidation Evapotranspiration is the largest component of the water balance. Since EC based ET measurements were only available for the periods of October 2008 May 2011

PAGE 156

156 and June 2009 May 2011 for DW and SW, respectively; therefore, model calibration and validation periods were limited to October 2008 May 2011 The traditional split data approach (R efsgaard, 1997; Jaber and Shukla, 2004) was used for model calibrat ion and validat ion The calibration period was October 1, 2008 to October 31, 2009 which included one dry season (November to April) and one wet season (May to October). Model validation pe riod spans from November 1, 2009 to May 31, 2011 and includes two dry seasons and one wet season A combination of graphical and statistical techniques was used to evaluate the model performance in simulat ing surface water levels during the calibration and validation periods. Measured surface water levels behind the CRB structure at Site1 and Site4 were compared with the simulated surface water levels within the MIKE 11. Surface water levels have been used in other MIKE SHE/MIKE 11 studies such as Jaber and Shukla (2007), in place of measured flows for evaluating model performance. T he main reasons for selecting levels over flows is difficulty in measuring flows in flat topography of S outh Florida which often l eads to non ideal conditions (high submergence) for measuring flows. Another reason is the wetland hydroperiod is controlled by the seasonal pattern of surface water level. It is necessary to ensure that calibrated model can represent the temporal variatio n of surface water level in the wetland for evaluating enhancements in ecosystem influencing by hydroperiod. Therefore, it is essential to calibrate model against surface water level. The groundwater levels observed at both sites were also compared with th e simulated groundwater levels. Wells with least missing data which represent ed spatial variation of

PAGE 157

157 groundwater levels in upland and wetland areas were selected for evaluating the model performance in predicting groundwater level Statistical criteria for model performance evaluation included root mean square error (RMSE), coefficient of determination (R 2 ), index of agreement (d) ( Willmott, 1981 ) and Nash Sutcliffe coefficient (E) (Nash and Sutcliffe, 1970) were expressed as follows: ( 5 1) ( 5 2) ( 5 3) ( 5 4) where O and S are the observed and simulated value s respectively ; t he overbar indicates the mean value ; and n is the total number of observations. The RMSE or standard deviation provides a direct measure of the error between the model and the observed data (Thomann, 1982). The RMSE is used to measure the discrepancy between simulated and observed values and indicates the overall predictive accuracy of the model at the two sites The R 2 value provides insight into the

PAGE 158

158 2 lies between 0 and 1. A perfect fit of the model to explain the variation is repres ented by a value of 1, and a value of 0 indicates that the model does not explain the variation at all. The closer the R 2 is to 1, the better the regression explains the relationship between simulated results and observed data. The d was developed by Willm ott (1981) as a standardized measure of the degree of model prediction error and varies between 0 and 1. A value of 1 indicates a perfect agreement between the measured and predicted values, and a value of 0 indicates no agreement at all (Willmott, 1981). The E is an indicator of goodness of fit (Nash and Sutcliffe, 1970) and is most widely used measure for model performance. The E can range from perfect match of model predictions to the observed data. An efficiency of 0 ( E = 0) indicates that the model predictions are as accurate as the mean of the observed data, whereas an efficiency less than 0 ( E < 0) occurs when the observed mean is a better predictor than the model. W ater Retention A lternatives The W WR alternatives simulated at Sites1 and 4 included the baseline ( pre WWR ) and the implemented WWR Baseline represents the post drainage (prior to the start of the project) condition with the flow invert being the bottom of the CRB structure. Additional alternatives included implemented WWR and six levels of WWR with incremental placement of boards (e ach board has a height of18.4 cm) are described in Table 5 4. They were termed as alternative1 (A1), alternative2 1 (A2 1), alternative2 2 (A2 2), alternative3 (A3), alternative4 (A4), alternative5 (A5), and alternative6 (A6). Among those alternatives, the WWR alternative implemented at Site1 was A5 (five boards, 0.92 m above the bottom of the CRB structure), while for Site4, it was A2 2 (2.6

PAGE 159

159 boards, 0.49 above the bottom of the CRB structure). For better understanding of the effects of WWR, simulations of WWR alternatives were conducted from May 1, 2008 to May 31, 2011 which includes three wet seasons. Results and Discussion Calibration The main MIKE SHE/MIKE 11 parameters calibrated in this study were fficient), the leakage coefficients (L C ) for the ditches and subsurface drainage level (D V ) and drainage time constant (D T ). The L C represents the exchange between the ditch and groundwater. The D V and D T constant are parameters that are used to control th e discharge rate of water routed using the drainage module in MIKE SHE. natural stream channel and floodplain range between 5 m 1/3 /s and 35 m 1/3 /s depending on the density of the vegetation in the flow paths. However, the wetland areas in this study have the value lower than those values because of dense vegetation. The respectively. The L C was 1 10 5 1/s for the drainage ditches in Site1, while it was 1 10 3 1/s for the drainage ditch in Site4. The calibrated value for D V and D T w ere 0.8 m and 1 10 7 1/s respectively The latter three parameters are MIKE SHE specific, and their values are within the range reported in other studies in S ou th Florida (Jaber and Shukla, 2004; Earth Tech, 2007). O bserved and simulated surface water levels at Site1 and Site4 for the calibration period are shown in Figures 5 6 and 5 7. The RMSE, R 2 d, and E values for the calibration period are shown in Table 5 5. The mean water level simulated by model was 0.05 m higher than the mean observed water level at Site1 (Table 5 5) While at

PAGE 160

160 Site4, the mean difference between simulated and observed surface water level was 0.04 m (Table 5 5) The R 2 was 0.9 3 for both s ites (Table 5 5). The d values for Site1 and Site4 were 0.9 7 and 0.9 5, respectively The E values for Site1 and Site4 0. 90 and 0.8 6 respectively (Table 5 5) Using Very Good Good Satisfactory Unsatisfactory ), the model performance was Very Good for both sites Model performance criteria presented above indica wetland, the main outcome of WWR. There was a good agreement between observed and simulated surface water levels for Site1 during the calibration period. There were periods that model under predicted and over predicted the surface water levels at Site1. During the period of January 200 9 April 2009, model over predicted the surface water levels at Site1 and the largest error was observed during the middle of the dry season (Fe bruary 2009) (Figure 5 6). Since the benefit of water retention is mainly observed during the wet season, over prediction of surface water level during the period of January 2009 April 2009 is acceptable. To improve model performance in predicting surface water levels, L C for the drainage ditch w as decreased and the flood code was added to the wetland area. The flood code is a special procedure that enables the simulation of inundation from the ditches defined in MIKE 11 onto MIKE SHE grid cells. This proce dure is employed which compares the water level simulated from MIKE 11 with the surface topography of each of the MIKE SHE grid cells that are specified as potentially flooded. When the simulated water level in the ditch is higher than the ditch banks, wat er can overspill and

PAGE 161

161 flood the grid cells specified with flood code. However, these changes did not improve the simulation results and the calibration statistics. For Site1, the trend of surface water levels between observed and simulated surface water lev els was similar during the 2009 wet season (Figure 5 6). The peaks of water levels after rainfall events were over predicted during the 2009 wet season However, adjusting parameters to solve the problem of this over prediction deteriorated the predictions of surface water level in the 2010 wet season. One of the reasons for this discrepancy was higher rainfall during the wet season in the 2009 compared to 2010, which was drier than long term average Rainfall for the 2010 wet season ( 621 mm) was 35 % below the long term average (960 mm). Another reason would cause t his over prediction is likely due to the effect of aggregating grid cell size from 1 m to 20 m. Using grid size of 20 m can underestimate the surface water storage in wetland areas in Site1 and Si te4 by 45 and 44%, respectively. Kenward et al. (2000) concluded that q uality and resolution of DEM data can affect the accuracy of any extracted hydrological features including wetland water storage and hydroperiod Guzha and Shukla (2012) refer to the la rge errors caused due to quality of spatial representation for South Florida ranches. They found that for a ranch containing wetlands and drainage ditches the USGS DEM resulted in 43% higher average water storage than the LIDAR DEM in the wet season. In th e dry seasons, the LIDAR DEM resulted in 28% higher storage than the USGS DEM. Shallow water table environment and highly conductive soils of South Florida require that models be evaluated for its predictions of surface and ground water levels as they affe ct surface as well as subsurface stores and exchanges within and across the boundaries. Furthermore, despite ET being the largest water balance component

PAGE 162

162 (Chapter 4), models are rarely evaluated for ET. This limitation is due to lack of site specific ET me asurements. MIKE SHE/MIKE11 was evaluated by comparing the simulated and measured wetland ET (EC based ET) and groundwater levels. During the calibration period, the simulated total ET ( 990 mm) for the wetland area at Site1 was 19% less than the observed ET (1217 mm) This under prediction is likely due to that the groundwater level was under predicted and it was lower than the root zone. Consequently, plants were experienced water stress and the transpiration was reduced. Evapotranspiration from DW was al so estimated using daily ET regression model developed in Chapter 3 with percent inundation estimated from the MIKE SHE predicted surface water levels and 20 m DEM data. Results showed that daily ET model combined with MIKE SHE predicted surface water leve ls was able to accurately predict ET within 2%. These results indicate need for improving ET module within MIKE SHE. The groundwater levels measured at three locations in Site1 were compared with the simulated groundwater levels (Figure 5 8). During the ca libration period, model performance in predicting the groundwater levels at GW21, GW 31 and GW 37 were Satisfactory (E = 0.63), Good (E = 0. 68 ) and Very Good (E = 0.7 9 ), respectively. Although the model performance in predicting ground water level at GW 34 was less than satisfactory (consistently under predicted ) (Figure 5 8) there was a high correlation (R 2 = 0.95) between measured and simulated groundwater level. Two main factors contributed to this discrepancy : 1) inability to represent the spatial variabil ity in soil hydraulic properties (e.g. hydraulic conductivity); 2) error in ET estimation in upland area which affected the soil water balance and thus groundwater level s The footprint of

PAGE 163

163 EC measurements was limited to the wetland. Given that upland area at Site1 accounts for 75% of the watershed area and drains to the wetland, ET in upland area is likely over predicted which could result in under prediction in groundwater level considering that groundwater levels interact with unsaturated zone through cap illary action. At Site4, the timing and the magnitude of the simulated surface water levels were close to the observed surface water levels (Figure 5 7). Surface w ater level at Site4 during the dry period was over predicted. It is likely due to an artifact in the model that would not allow for no flow during the prolonged dry condition. Therefore, a minimum depth of water was maintained in the ditch by MIKE 11 to maintain the stability of the model simulation This limitation of the model could result in er rors in groundwater flux simulation This problem has also been reported by Dai et al. (2010) in simulating surface flow and groundwater depth s for a forested watershed in South Carolina. Since the EC based ET measurements started in June 2009 at SW evaluation of June 2009 October 2009. Within th is period, the simulated total ET (413 mm) for the wetland area was almost same as observed ET ( 40 4 mm ) Since the available groundwater level measurements w ere used as boundary condition, model performance in predicting groundwater levels at Site4 was not evaluated. During the calibration, model performance in predicting wetland ET and groundwater level are good while it is more than s atisfactory in predictin g surface water levels. Discrepancies between simulated and observed surface water levels were observed at Site1 for the February 2009 April 2009 and May 2009 October 2009 periods. Several factors may have contributed to these discrepancies: 1) inability t o

PAGE 164

164 represent the spatial variation of soil hydraulic properties; 2) errors in predicting ET for the upland and wetland areas; 3) the resolution of grid cell used in MIKE SHE could affect the surface water storage capacity and consequently affect the surface flow routing. Use of 20 m grid size resulted in under estimating the surface water storage in wetland areas at Site1 and Site4 by 45 and 44%, respectively. The calibrated model adequately represented the va riations of surface water level, ET in the wetlan d and groundwater level at both sites. However, the model p redictions on surface flow w ere less than satisfactory (based on Moriasi et al., 2007) for both sites (Figure 5 9) The E and d values for Site1 were 0.44 and 0.71. At Site4, E was 0.41 and d was 0 .68. Less than satisfactory model performance on surface flow was primarily due to: 1) flat topography at the study site (average slope < 4%) combined with dispersed swales results in errors in delineation of surface flow boundaries which adds uncertainty in capturing sheet flow exchanges across the boundary; 2) errors associated with flow measurements; 3) disproportionally high effect of under or over predictions of surface water level due to the nature of weir equation (Q h 1.5 ). Validation Time series of simulated and observed surface water levels behind the CRB structure at Site1 and Site4 for the validation period are shown in Figures 5 6 and 5 7, respectively M odel performance statistics (RMSE, R 2 d, and E) for surface water level were provided in Table 5 5. Mean simulated was equal to the mean observed surface water level (mean level = 8.60 m above AMSL). At Site4 too the mean difference between observed (9.13 m) and simulated (9.2 m) surface water levels were small ( 0.07 m ; Table 5 5). Although the R 2 values for model validation f or both sites were slightly

PAGE 165

165 lower than the calibration period, they were still equal to or higher than 0.89 The d values for Site1 and Site4 were 0.97 and 0.91, respectively, indicating good model performance. Although the E values for both sites reduced slightly, they still represented Very Good (Site1; E = 0. 88 ) to Good (Site4; E= 0.7 0 ) model performance Although the overall E value was good for Site1, simulated water levels were lower than the observed values durin g the wet season of 2010, which was drier than normal (35 % lower rainfall than the long term average ) This under prediction is likely due to the fact that model was not able to represent the recession of the surface and ground water levels during the dry period. Specific reasons for this error are likely due to effects of soil moisture stress on transpiration especially for the bahiagrass in upland pasture and errors in accurately capturing the recession of surface and ground water levels due to aggregating the topography from 1m (LIDAR) to 20 m (model grid size) which degraded the simulations During the validation period, the simulated total ET for the wetland area in Site1 was 1768 mm which is 19% less than the observed ET ( 2016 mm). The simulated ET (1 133 mm) for wetland area in Site4 was 1 3 % less than the observed ET (1305 mm). Under prediction of ET may likely be due to inaccurate representation of drought stress for wetland plants when the groundwater level falls below the root zone. Evapotranspiration estimated using daily ET model (Chapter 3) and predicted surface water levels from MIKE SHE for DW was 20% less than EC based ET, while it was 6% higher than the o bserved ET for SW. This under prediction of DW ET is likely due to the surface water level during the 2010 wet season in MIKE SHE was under predicted.

PAGE 166

166 During the validation period, model performance in predicting the groundwater levels at Site1was Good at GW 31 (E = 0.71) and GW 37 (E = 0.68), while it was less than satisfactory at GW21 (E= 0.38). Simulated groundwater fluxes were compared with the groundwater fluxes estimated from the water budget from Chapter 4 (Table 5 6). From May 2009 to April 2011, signs of the predict ed groundwater fluxes were consistent with those from the water budget method for both sites except the period of November 2010 April 2011 for Site4 (Table 5 6). The primarily reason is that the total ET for Site4 was under predicted b y 95 mm during November 2010 to April 2011. Although signs of predicted net groundwater flux were same as those from water budget, magnitudes of the net groundwater flux were different when compared with those estimated from water budget. It is likely due to the stage volume relationship defined for the wetlands, especially in the ditch which yields a smaller storage capacity than actual for a given water level due to effects of grid size used in the model Lower storage capacity led to increas e of surface flow from the wetland area. As a result, most of the water left both sites through surface flow instead of subsurface flow during May 2009 to October 2010. Differences in magnitude of groundwater fl ux e s in other periods for both sites may also due to that the model could not adequately represent the ET processes and the spatial variation of groundwater levels due to lack of field measured soil hydraulic properties being used in the model Effects of W ater Retention A lternatives Simulation results showed t hat adding boards to the CRB structure to implement WWR resulted in reduc ed flows by retaining water at both site s. Among WWR alternatives (A1 to A6), the alternatives implemented by the rancher cooperator were A5

PAGE 167

167 (for Site1) and A2 2 (for Site4). The decision to implement specific alternatives was WWR for both sites were discussed by comparing flows and levels for baseline and imp lemented WWR and baseline and other alternatives. From May 1, 2008 to May 31, 2011, predicted reduction in surface flow at Site1 was 64% for A5 compared to baseline ( baseline = 97.8 cm and A 5 = 35.6 cm ) Reduction in surface flow due to implemented alter native for Site4 was 22 % (baseline = 59.9 cm and A2 2 = 47.0 cm). W et season receives almost 70% of the annual rainfall (Shukla et al., 2010) and m ost ecological damages to the S outh Florida ecosystem including the LO, east and west coast al estuaries, and the Everglades arise from damaging high flows and associated nutrient loads (Haven et al., 1996) during wet season. Therefore, the discussion of effects of WWR on surface flows peak flows ET, and groundwater fluxes is focused on the wet seasons. Simulati on results showed that during the 200 8 wet season, the average water levels behind the CRB structure at Site1 increased by 0.1 6 m (baseline = 8. 69 m and A 5 = 8.85 m). Although the difference in average surface water levels was only 0.1 6 m, the associated i ncreases in inundated area and surface water storage under A5 alternative were much higher. Under A5 alternative, 2.9 ha of additional area was flooded (baseline = 1.2 ha) which increased the surface stores by 3693 m 3 (baseline = 1716 m 3 ) at Site1 Surface flow volume was reduced by 45% compared to baseline (Table 5 7) while reduction in peak flow was 50% (Figure 5 10) at Site1. For Site4 in the 200 8 wet season, the difference in average surface water levels between A2 2 and baseline conditions was 0.2 0 m ( baseline = 9. 10 m and A 2 2 = 9. 30 m), and associated

PAGE 168

168 increases in inundated area and stored volume under A2 2 condition were 0.09 ha (baseline = 0.01 ha) and 104 m 3 ( baseline = 5 m 3 ) respectively. For Site4, implementation of A2 2 reduced surface flow by 10% (Table 5 7) with the associated reduction in peak flow being 6% (Figure 5 11). For 2009 wet season the average simulated water levels behind the CRB structure at Site1 increased by 0.1 1 m (bas eline = 8. 6 7 m and A 5 = 8. 7 8 m). Inundated area was increased by 1.3 ha (baseline = 1.0 ha) and the surface water storage was increased by 1703 m 3 (baseline = 1500 m 3 ) at Site1 for A5 For Site1, surface water flow was reduced by 82% compared to baseline ( Table 5 7), and peak flow was also reduced by 87% (Figure 5 10). For Site4 in the 2009 wet season, the difference in average surface water levels between A2 2 and baseline conditions was 0.2 4 m (baseline = 9. 10 m and A 2 2 = 9. 34 m), and associated increase s in inundated area and stored volume under A2 2 condition were 0.1 ha (baseline = 0.01 ha) and 151 m 3 ( baseline = 5 m 3 ) respectively. For Site4, surface flow was reduced by 20% (Table 5 7), and the associated reduces in peak flow was 8% in average (Figur e 5 11). For the 2010 wet season, the difference in average surface water levels in Site1 between baseline and A5 was 0. 1 1 m. Inundated area was increased by 2.5 ha (baseline = 1.8 ha) and the surface water storage was increased by 3247 m 3 (baseline = 2579 m 3 ) at Site1 for A5 For Site1, surface water flow was reduced by 93% compared to baseline (Table 5 7), and peak flow was also reduced by 98% in average (Figure 5 10). At Site4, the average surface water level for A2 2 was 0.2 2 m higher than baseline cond ition during the 2010 wet season The inundated area and surface storage capacity under A2 2 condition were increased by 0.1 ha (baseline = 0.01 ha) and 120

PAGE 169

169 m 3 (baseline = 4 m 3 ) at Site4, respectively For Site4, the surface water flow was reduced by 29% (Table 5 7); however, average reduction on peak flow was small (5%) (Figure 5 11). Comparing reductions in surface flow for three wet seasons, results showed that the lowest percent reduction was found in the 2008 wet season for both sites under the curren t implemented WWR. However, considering the volume of water retained and the year in which it was retained, the retention in wet years (e.g. 2008) is more important than in dry years (e.g. 2010). Higher the retention volume more the reduction in flows and nutrient loads. Given the fact that the 2010 wet season was drier than the normal wet season, retaining water in the 2010 wet season may cause water shortage to downstream water bodies. Water budget components for Site1 and Site4 for the 2008, 2009 and 201 0 wet seasons are shown in Tables 5 8 and 5 9. Results showed that A5 ha d minor impact on ET at Site1 in harmful excess surface flows to the LO. At Site1, there was no difference in ET between baseline and A5, and the increase in ET for the 2009 and 2010 wet seasons were 2 and 3 mm which translated to only 0. 4 % and 0. 5 % increase in total wet season ET compared to the baseline (Table 5 8). At Site4, the corresponding increases in E T were 0 mm for both the 2008 and 2009 wet seasons, while it was 2 mm for the 2010 wet season (Table 5 9). In contrast to ET, the corresponding increases in groundwater storage and flux were much higher. For the 2008 wet season, groundwater flux at Site1 w as a net outflow of 159 mm under baseline, and it was increased by 120% under A5 (350 mm)

PAGE 170

170 (Table 5 8). For Site4, the groundwater flux was a net outflow under baseline condition during the 2008 wet season, and it was increased by 11% under A2 2 (Table 5 9) During the 2009 wet season, respective implementation of A5 and A2 2 on Site1 and Site4, not only changed the extent of groundwater flux it also reversed the direction of groundwater flow from a net inflow to a net outflow (Tables 5 8 and 5 9) Similar to the 2008 wet season, the net groundwater flux was a net outflow for both sites and it increased by 60% and 20% for Site1 and Site4, respectively, for the 2010 wet season. Overall, the average reduction in total surface flow was 16 c m at Site1 under A5 ( Table 5 8). Most of this reduced flow moved out of Site1 as groundwater flow across the boundary At Site4, total surface flow was reduced by 4 c m at Site 4 under A 2 2 (Table 5 9 ) when averaging reduction for three wet seasons All of this reduced flow was also moved out of Site4 as groundwater flow across the boundary. Although the rancher cooperator agreed to A5 and A2 2 for Site1 and Site4, respectively, his comfort levels were based on the height of board. Model results could help convince the rancher ag ree to higher level of water retention depending on the impact on pasture areas and risk to ranch flooding. Results of reduced surface flow for a variety of alternatives for both sites are presented in Table 5 7. Results showed that for Site1 instead of re ducing surface flow, implementations of A1, A2 1 and A3 may actually increase the surface flows (Table 5 7). It is likely due to that these three alternatives may at times reduce the w etland storage capacity and increase the total and peak flows after consecutive rainfall events Increases in surface water retained for A4 from averaging three wet seasons (2008, 2009, 2010) were 42 % and 6 2 % at Site1 and Site4, respectively (Table 5 7). For A6, there was a n ave rage of 86 % increased

PAGE 171

171 water retention at Site1 while it was an average of 93% increased water retention for Site4 (Table 5 7). Comparing different alternatives with regards to surface flow reductions reveals that t he best performance (the highest surface f low reduction per board added) was achieved by A4 for both Site1 and Site4. Differences in percent reduction of surface flow between two sites were mainly due to the difference in stage volume relationship and the board height which resulted in higher surf ace water storage capacity behind the CRB structure at Site1 compared to Site4. Water budget components for Site1 and Site4 for the 2008, 2009 and 2010 wet seasons are shown in Tables 5 8 and 5 9. Results showed that alternative A5 ha d no to small effect o n ET at Site1 for the three years increased ET may not come back to the place of retention. At Site1, there was no difference in ET between baseline and A5 for 2008 while the increases in ET for the 2009 and 2010 wet seasons were 2 to 3 mm which translated to only 0. 4 % to 0. 5 % increase in total wet season ET compared to the baseline (Table 5 8). At Site4, the corresponding increase s in ET were 0 mm for both the 2008 and 2009 wet seasons, while it was 2 mm for the 2010 wet season (Table 5 9). During 2008, higher and frequent rainfall resulted in water availability throughout the season for optimum evaporation and transpiration. Durin g the drier years of 2009 and 2010 increased storage slightly increased the soil water availability which resulted in a small increase in ET.

PAGE 172

172 In contrast to ET, the corresponding increases in groundwater storage and flux were much higher. For the 2008 wet season, groundwater flux at Site1 was a net outflow of 159 mm under baseline which increased by 120% under A5 (350 mm) (Table 5 8). For Site4, the groundwater flux was a net outflow under baseline condition during the 2008 wet season, and it was increased by 11% under A2 2 (Table 5 9). During the 2009 wet season, the A5 (Site1) and A2 2 (Site4) not only changed the extent of groundwater flux it also reversed the direction of groundwater flow from net inflows of 2mm (Site1) and 13 mm (Site4) for the baselin e to net outflow s ( 142 mm, Site 1 and 37 mm Site 4) (Tables 5 8 and 5 9) Similar to the 2008 wet season, the net groundwater flux was a net outflow for both sites and it increased by 60% and 20% for Site1 and Site4, respectively, for the 2010 wet season Overall, the average reduction in total surface flow was 16 c m at Site1 under A5 (Table 5 8). Most of this reduced did not remain within the watershed but flowed out of Site1 as groundwater flow s across the boundary At Site4, average surface flow reduce d by 4 c m under A 2 2 (Table 5 9 ). All of this reduced flow moved out of Site4 as groundwater flow did not contribute to long term storage but moved out of the retention area. Although the rancher cooperator agreed to A5 and A2 2 for Site1 and Site4, respectively, his comfort levels were based on the height of board. Model results could help convince the rancher agree to higher level of water retention depending on the impact on pasture are as and cattle due to increased risk of flooding. Results of reduced surface flow for a variety of alternatives, lower as well as higher than what was implemented at both sites are presented in Table 5 7. Results showed that raising the

PAGE 173

173 spillage levels at t he CRB structures may not always reduce surface flow. At Site1, instead of reducing surface flow, implementations of A1, A2 1 and A3 increased the surface flows (Table 5 7). It is likely due to that these three alternatives may at times reduce the w etland storage capacity and increase the total and peak flows after consecutive rainfall events Increases in surface water retained for A4 from averaging three wet seasons (2008, 2009, 2010) were 42 % and 6 2 % at Site1 and Site4, respectively (Table 5 7). For A6, there was a n average of 86 % increased water retention at Site1 while it was an average of 93% increased water retention for Site4 (Table 5 7). Comparing different alternatives with regards to surface flow reductions reveals that the best performance (the highest surface flow reduction per unit of height increase in spillage levels) was achieved by A4 for both Site1 and Site4. Differences in percent reduction of surface flow between two sites were mainly due to differences in stage volume relationships and the board height which resulted in higher surface water storage capacity behind the CRB structure at Site1 compared to Site4. Water budget components for baseline and WWR alternatives, for Sites 1 and 4 for the 2008, 2009 and 2010 wet seasons (Tables 5 8 a nd 5 9) showed that while water retention resulted in negligible change in ET it increased the groundwater outflow Consider for example, surface flow and groundwater levels under baseline and A4 conditions for both sites during the wet seasons. Reductions in surface flow volume as well as peak flows for rainfall events are accompanied by rise in groundwater levels at both sites (Fig ures 5 12 and 5 13 ). Increases in groundwater levels are especially evident during July September period when majority of dama ging flows to the LO and

PAGE 174

174 the coasts are experienced ( Figures 5 12 and 5 13 ) At Site1, an average of 96% of the reduced surface flow left the watershed through groundwater flow; while at Site4, all of the reduced surface flow left the boundary through subs urface for all the alternatives except baseline Although reductions in surface flows were evident the surface and/or groundwater storages did not extend beyond the wet season to have a beneficial impact on water availability during the dry season. Althoug h retained water leaves the sites before the beginning of the dry season return flows to the local and regional surface drainage systems from groundwater outflows from the water retention sites still delay the peak flows to the LO which is still a desirable outcome This delay in peak flows reaching the LO has the potential to reduce large releases to the east and west coast estuarine systems through Caloosahatchee and St Lucie rivers. Peak flow reduction is important in LO watershed. High peak flow s especially during the wet season can result in increasing sediment losses and total phosphorus loads to the lake and large releases from the L O to the east and west coast estuaries Increasing board heights have shown reduction in peak flow (average of a ll peak flows of events) and maximum peak flow The largest surface flow observed in each wet season was denoted as maximum peak flow. For Site1, the largest percent reduction in peak flow and maximum peak flow per unit of height increase in spillage level s were achieved by A6 and A5, respectively (Table 5 10). At Site4, A5 can achieve the largest percent reduction in maximum peak flow per unit of height increase in spillage levels while A4 gives the highest reduction in peak flows (Table 5 10). From purely peak flow perspective, A6 provides the highest percent reduction in peak flow and maximum peak flow for both sites (Table 5 10)

PAGE 175

175 For most of wetland ecosystem, increasing hydroperiod would benefit the wetland vegetation, wildlifes and microbes (Mitsch and Gosselink, 2000); therefore, effect of different alternatives on hydroperiod was evaluated in this study. For uniformity in the analyses and comparison, 15 cm was used as the minimum depth of water in the deepest part of the wetland for hydroperiod calcul ations for both sites Results showed that for Site1 annual average hydroperiod for baseline, A1, and A2 1 was 282 days, 283 days and 275 days, respectively; while it was consistently higher than the three alternative (286 days) for A3, A4, A5 and A6. For Site4, annual average hydroperiod for baseline and A1 were 17 and 24 day, respectively. The hydroperiods for alternatives A2 1 to A5 were 145 days The longest annual average hydroperiod for Site4 was predicted for A6 (146 days). Results indicated that the hydroperiod would only increase by 4 days even under the highest spillage levels of A6 at Site1. In contrast to Site1, the increase in hydroperiod at SIte4 was much larger with a minimum of 128 days of hydroperiod (compared to 17 days for baseline) even w ith a lower spillage height of A2 1. In general, increasing board height can increase the water retention storage in the drainage basin. However, potential economic losses arising from flooding of pastures will have to be considered to enable acceptability of specific alternatives unless the ranchers are compensated for the losses. The highest predicted water level and associated inundated area for each alternative were compared for both sites in this study (Figures 5 14 and 5 1 5 ). Results showed that for S ite1, the maximum inundated area did not change significantly for A1, A2 1 and A3; while it was increased by 57% to 140% from A4 to A6 (Table 5 11). For Site4, significant increase in inundated area was

PAGE 176

176 found between A3 and A4 (Table 5 11). For Site1, even though A6 created the largest inundated area under maximum water level, most of the inundated area was still within the wetland area and did not flood the upland pastures (Figure 5 1 4 ). However, A5 and A6 would flood most of the wetland area at Site4 incl uding the productive pasture within the footprint of the wetland (Figure 5 15 ). The rancher cooperator indicated that this pasture is one of the most productive pastures. Therefore, he is not likely to accept significant increase in flooding of this pastur e. Such increases in flooding of productive pasture will require analyses of tradeoffs between water retention and economic losses to the rancher and are likely to require policy intervention in the form of payment to rancher for the economic losses. Consi dering the environmental factor and potential economic impacts (surface flow and peak flow reduction, hydroperiod and maximum inundated area), the best alternative, seems to be A6 and A4 for Site1 and Site4, respectively. Given the flooding of productive pasture at S i te4, the best alternative for all the ranchlands within the LO watershed seem to be A4. Results from W WR alternatives were used to conduct a scale up analysis for the entire LO watershed This scale up was conducted by extrapolating the result s from the two wetland upland systems. The LO watershed covers an area of 1 4 244 km 2 (SFWMD, 2008 ) of which ranchland accounts for 36% ( 5,128 km 2 ). Wetlands like those present at the ranch in this study account for 15% of ranchlands ( 769 km 2 ) The wetlands at the two sites account for 27% and 33% of the drainage area (Fig ure 5 1 ), higher than the average wetland percent (15%) in ranchlands of the LO watershed For a conservative scale up, reductions in surface flow at Sites 1 and 4 were adjusted for

PAGE 177

177 wetland /upland ratios at the two sites by dividing it by 1.8 (27%/15%) and 2.2 (33%/15%) respectively for each WWR alternative. The above adjustments assume that water is retained within the wetland footprint. Part of the water retained at both sites is stored in groundwater beneath the uplands; therefore the above scale up is conservative estimate of water retention. For each alte rnative, adjusted reductions in surface flow volume for both sites were averaged to estimate water retention for the scale up analyses The average adjusted reductions in surface flows for A1, A2 1, A3, A4, A5, and A6 were 0. 2 c m, 0. 3 c m, 1 1 c m, 5 .8 c m, 9 6 c m, and 11.2 c m, respectively. To attain its total maximum daily loads goals and reduce damaging flows to the lake and estuaries, an additional storage of 1,356,830,041 m 3 (9.5 cm) (SFWMD, 2008) is needed within the LO watershed. Using the above scale up and assuming 100% implementation of A6 on ranchlands within the LO watershed will achieve 42% of the desired storage within the wet season. Implementation of other WWR alternatives A1, A2 1, A3, A4, A5 can help achieve 0.6%, 1.3%, 4.3%, 22.0%, 36.4% of the needed, respectively. Given relatively higher flow reductions and lesser flooded pasture area compared to other alternatives, implementation of A4 can achieve 2.1 cm of watershed wide storage which is 22% of the desired storage. The above watershed sc ale analyses is derived from predicted storage from MIKE SHE with observed groundwater boundary conditions for the A5 and A2 2 alternatives implemented at the two sites. Considering that most of temporary storage leaves the retention areas as groundwater, it is likely that the boundary conditions will change and result in a raised water level at the boundary. A logical question is that what happens to the groundwater when water retention is implemented in the entire LO

PAGE 178

178 watershed? If the water table is raise d throughout the watershed as a result of water retention, then it is likely to reduce the groundwater losses because of lower gradients at the boundary of retention areas. Considering that it is nearly impossible to determine the spatial distribution of t he changed boundary, MIKE SHE was run with zero groundwater flux boundary condition to obtain an insight into the effect of watershed wide implementation on site specific retention. The analysis of zero groundwater flux was limited to A4, A5, and A6 altern atives. Results showed that no groundwater flux boundary at Site1 will actually result in increasing the wet season surface flows for A4 and A5 alternatives (Table 5 12). For Site4, although the surface flows were not increased but the reductions were less than those predicted under observed boundary condition (groundwater exchange is allowed to occur) (Table 5 12). Implementation of A5 and A6 (under zero flux) can achieve 8.7% (0.8 cm) and 22.0% (2.1 cm) of the required storage, respectively, which is much less than the corresponding storage under the observed boundary condition. Considering that the groundwater exchange is likely to be between the observed and the zero, the estimated storage is likely to be between 8.7 to 22 .0 % of the goal with 100% areas under water retention. The corresponding storage will be 4.4 to 11 .0 % for a more realistic goal of 50% of the ranchland under water retention. Chapter Summary and Conclusions MIKE SHE/MIKE 11 was used to simulate the effectiveness of different levels of WWR for two wetland upland systems on a cow calf ranch in S outh Florida. The model was calibrated and validated with the measured surface and ground water levels M odel performance was evaluated by a combined approach of graphical and statistical techniques. For calibration period, model performance i n predicting surface water level s

PAGE 179

179 was Very Good for Site1 (E = 0.90) and Site4 (E = 0.86) Model performance in predicting groundwater level s ranges from Satisfactory (0.63) to Very Good (0.79). For the validation period, model performance in simulating surface water level s was Very Good and Good for Site1 ( E = 0. 8 8) and Site4 (E = 0.70), respectively; while it varied from les s than satisfactory (0.38) to Good (0.71) in predicting groundwater level s The A5 (Site1) and A2 2 (Site4) were implemented because of its acceptability to the rancher cooperator. R esults showed that average reduction in surface flow for the wet season wa s 73 % (increase in retention by 16 cm) under A5 at Site1, while it was 20 % (increase in retention by 4 cm) under A2 2 at Site4 Although WWR increase d the ET, the relative increase was less than 3 mm at both sites for the wet season under current implemented WWR condition Of the water that did not leave the CRB structure through surface flow, 99% and 100% left through groundwater flow at Site1 and Site4, respectively. Although most of the retained water left both sites through subsu rface pathways it is still a desirable outcome because it helps reduce the peak flows. The WWR had significant impacts on surface water flow, peak flow, hydroperiod, and inundated area. Surface water flow can be reduced by 42 to 86 % at Site1 while it can be reduced by 4 to 93% at Site4 with different levels of WWR Implementing WWR can also reduce the peak flow s from both sites. Results showed that peak flow was reduced by 2 to 90% at Site1 and it was reduced by 1 to 96% at Site4. For Site1, the annual av erage hydroperiod varied from 282 days (baseline) to 286 days (A6) between baseline and other alternatives. At Site4 the annual average hydroperiod varied from 17 days (baseline) to 146 days (A6). This shows that implementing WWR at Site4 may help re estab Change in

PAGE 180

180 maximum inundated area varied from a decrease of 3% to (A1and A2 1) to an increase by 140% (A6) at Site1; while it is increased from 49% (A1) to 9333% (A6) at Site4. Increases in flooding of p roductive pasture may damage forage adversely impact forage availability Future work should focus on determining the threshold inundation time for forage plants to better evaluate economic losses to the rancher. Considering results from different alternatives with potential economic impacts the alternatives A6 and A4 can be considered for sites similar to Site s 1 and 4 respectively For implementation at the LO watershed level A4 could be considered loads goal and reduce damaging flows to the lake and estuaries, an additional storage of 1,356,830,041 m 3 (9.5 cm) (SFWMD, 2008) is needed within the LO watershed. Considering the results from scale up analys e s and assuming 100% implementation of A4 on al l the ranchlands in LO watershed can achieve 2.1 cm storage which is 22% of the desired storage. The above watershed wide scale up analysis wa s derived with field measured groundwater boundary condition s observed during the monitoring period with (2008 201 1) under specific alternatives ( A5 at Site1 and A2 2 at Site4) Given that most of temporary storage leaves the retention areas through subsurface pathways, it is likely that the boundary conditions will change and result in a raised ground water level at t he boundary. Since it is nearly impossible to determine the spatial distribution of the changed boundary, MIKE SHE was run with zero groundwater flux boundary condition to obtain an insight into the effect of watershed wide implementation on site specific retention. Results showed that no groundwater flux boundary at Site1 will actually result

PAGE 181

181 in increasing surface flows in the wet season under A4 and A5. Although the surface flows were not increased at Site4, but the reductions were less than those predict ed under observed boundary condition (groundwater exchange is allowed to occur). Implementation of A5 and A6 (under zero groundwater flux) can achieve 8.7% (0.8 cm) and 22.0% (2.1 cm) of the desired storage, respectively, which is much less than the corres ponding storage under the observed current boundary condition. Considering that the groundwater level is likely to be between the observed and the zero groundwater boundary conditions, the estimated storage is likely to be between 8.7 to 22 .0 % of the goal with 100% areas under water retention. The corresponding storage will be 4.4 to 11 .0 % of the desired storage for a more realistic goal of 50% of the ranchlands under water retention. The calibrated model adequately represented the variations of surfa ce water level at both sites. Under prediction in groundwater level in upland areas was primarily due to: 1) inability to represent the spatial variability in soil hydraulic properties (e.g. hydraulic conductivity); and 2) error in ET estimation in upland area which may affect the soil water balance and thus groundwater level s Several improvements can be identified for f uture studies: 1) conduct site specific measurements of soil characteristic s (e.g. saturated hydraulic conductivity, field capacity, wilting point, etc.) to parameterize the model ; 2) improve model performance i n predicting ET prediction for the upland areas by collecting LAI and root depth s for bahiagrass; 3) use smaller grid siz e to better represent the surface topography and water storage capacity, especially for the wetland area (45% difference in wetland storage capacity between 1 m and 20 m grid size); 4)

PAGE 182

182 improve the identification of flow contributing areas ; and 5) evaluate the effects of flooding on forage losses to evaluate the economic losses from water retention

PAGE 183

183 Table 5 1. L and use areas and associated MIKE SHE parameters Land use Area (m 2 ) Percentage (%) Detention storage (mm) Root depth (mm) Upland (pasture) 2 294800 83.7 25.4 1000 Wetland 332800 12.2 31.8 500 Pinewood 113600 4.1 25.4 1500 Table 5 2. Monthly leaf area index Land use types Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Pasture 3 .0 3.5 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 3.5 3 .0 Wetland 2 .0 3 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 3 .0 2 .0 Pinewood 2.5 3. 3 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 4 .0 3. 3 2.5 Table 5 3. Soil type, associated areas, and physical properties at the study site. Soil Class Area (m 2 ) Percentage (%) Water content Saturated condition Field capacity Wilting point Basinger 758032 27.6 0.355 0.112 0.024 Floridana 59652 2.2 0.373 0.266 0.156 Immokalee 1551298 56.4 0.420 0.214 0.058 Myakka 270040 9.8 0.388 0.196 0.044 Valkaria 109955 4.0 0.377 0.140 0.015 Table 5 4. Average elevations for upland and wetland areas for Site1 and Site4, and the spillage elevations at the culvert and riser board structure for different water retention alternatives. Site1 Site4 Location/Alternatives Elevation (m AMSL a ) Elevation (m AMSL ) Upland 10.21 10.53 Wetland 9.09 9.81 Baseline 8. 21 8.98 A1 (1 board) 8.40 9.16 A2 1 (2 boards) 8.58 9.35 A2 2 (2.6 boards) 9.47 A3 (3 boards) 8.7 6 9.53 A4 (4 boards) 8.9 5 9.72 A5 (5 boards) 9.1 3 9.90 A6 (6 boards) 9.3 2 10.08 a: above mean sea level

PAGE 184

184 Table 5 5 Comparisons of m ean, root mean square error ( RMSE ) coefficient of determination ( R 2 ), index of agreement ( d ) and Nash Sutcliffe coefficient ( E ) for the surface water level s (m, above mean sea level) at the culvert and riser board structures located at Site1 and Site4 for the calibration and validation periods Mean (m) Calibration Validation RMSE R 2 d E Obs. a Sim. b Obs. Sim. Cali. c Vali. d Cali. Vali. Cali. Vali. Cali. Vali. Site1 8. 59 8.6 4 8.60 8. 60 0. 09 0.1 1 0.9 3 0. 89 0.9 7 0.9 7 0. 90 0. 8 8 Site4 9.18 9.22 9.13 9.20 0.08 0.1 1 0.9 3 0.9 1 0.9 5 0.9 1 0.8 6 0.7 0 a: observed, b: simulated, c: calibration, d: validation. Table 5 6. Comparisons of the simulated net groundwater fluxes and the net groundwater fluxes estimated as residual term in the water budget. Site1 Site4 Period Net groundwater flux (sim. a )(mm) Net g roundwater flux (WB b ) (mm) Net groundwater flux (sim.)(mm) Net g roundwater flux (WB) (mm) May 2009 Oct 2009 142 c 11 9 37 150 Nov 2009 Apr 2010 92 3 9 71 May 2010 Oct 2010 289 8 5 287 113 Nov 2010 Apr 2011 10 8 6 22 39 a: simulated; b: estimated as residual term in the water budget; c: p ositive groundwater fl ux value indicates a net inflow, while the negative value indicates a net outflow.

PAGE 185

185 Table 5 7. Predicted surface flows for baseline and other water retention alternatives Site1 Site4 Period Alternative S urface flow, m 3 (cm) % Change S urface flow, m 3 (cm) % Change 2008 wet season Baseline 374120 (47) 90811 (37) A1 374585 (47) 0 89061 (37) 2 A2 1 376182 (47) 1 94295 (39) 4 A2 2 81672 (34) 10 A3 372948 (47) 0 79840 (33) 12 A4 295959 (37) 21 49052 (20) 46 A5 206575 (26) 45 15384 (6) 83 A6 129382 (16) 65 4285 (2) 95 2009 wet season Baseline 149490 (19) 60867 (25) A1 150020 (19) 0 58652 (24) 4 A2 1 151161 (19) 1 52400 (22) 14 A2 2 48629 (20) 20 A3 150131 (19) 0 46194 (19) 24 A4 83717 (10) 44 18843 (8) 69 A5 26177 (3) 82 4172 (2) 93 A6 5044 (1) 97 4156 (2) 93 2010 wet season Baseline 89014 (11) 38222 (16) A1 91293 (11) 3 35737 (15) 7 A2 1 93652 (12) 5 29535 (12) 23 A2 2 27029 (11) 29 A3 94244 (12) 6 25763 (11) 33 A4 35542 (4) 60 10647 (4) 72 A5 6309 (1) 93 3422 (1) 91 A6 3703 (0.5) 96 3127 (1) 92 Average Baseline 204208 (26) 63300 (26) A1 205299 (26) 1 61150 (25) 4 A2 1 206998 (26) 2 58743 (24) 11 A2 2 52443 (22) 20 A3 205774 (26) 2 50599 (21) 23 A4 138406 (17) 42 26181 (11) 62 A5 79687 (10) 73 7659 (3) 89 A6 46043 (6) 86 3856 (2) 93 v alues in the parentheses denote total surface flow (depth unit, cm).

PAGE 186

186 Table 5 8. S imulated water budget components for baseline and other water retention alternative s for Site1. Period Alternative Rainfall (mm) ET (mm) Surface flow (mm) Storage (mm) Net GW a (mm) 2008 wet season Baseline 1109 533 458 39 159 A1 533 459 39 157 A2 1 533 459 39 157 A2 2 A3 533 455 39 161 A4 533 365 33 244 A5 533 257 30 350 A6 533 163 28 443 2009 wet season Baseline 831 514 183 136 2 A1 514 184 136 3 A2 1 514 186 136 5 A2 2 A3 514 185 136 3 A4 515 102 138 75 A5 516 33 141 142 A6 516 7 142 166 2010 wet season Baseline 621 578 102 239 181 A1 578 106 239 177 A2 1 578 108 239 175 A2 2 A3 578 108 238 176 A4 579 17 241 266 A5 581 0 249 289 A6 581 0 254 296 a: net groundwater fl ux Positive value indicates a net inflow, while the negative value indicates a net outflow.

PAGE 187

187 Table 5 9. Simulated water budget components for baseline and other water retention alternatives for Site4 Period Alternative Rainfall (mm) ET (mm) Surface flow (mm) Storage (mm) Net GW a (mm) 2008 wet season Baseline 1109 507 375 105 335 A1 507 367 105 341 A2 1 507 348 105 360 A2 2 507 338 105 371 A3 507 330 105 378 A4 507 204 105 504 A5 507 67 105 641 A6 507 22 105 686 2009 wet season Baseline 831 467 255 121 13 A1 467 246 121 3 A2 1 467 220 122 23 A2 2 467 205 122 37 A3 467 196 122 47 A4 467 83 122 159 A5 467 24 122 218 A6 467 24 122 218 2010 wet season Baseline 621 504 160 283 240 A1 504 150 283 251 A2 1 504 125 284 276 A2 2 504 114 283 287 A3 504 108 284 293 A4 505 46 299 370 A5 506 17 314 415 A6 506 16 314 416 a: net groundwater fl ux Positive value indicates a net inflow, while the negative value indicates a net outflow.

PAGE 188

188 Table 5 10. Simulated maximum peak flow and peak flow for major rainfall events for baseline and other water retention alternatives. Site1 Site4 % Change % Change Alternative Maximum p eak flow Peak flow Maximum p eak flow Peak flow Baseline 0 0 0 0 A1 0 0 1 1 A2 1 2 2 1 3 A2 2 4 6 A3 8 6 4 7 A4 32 4 2 38 5 6 A5 56 7 8 88 92 A6 82 9 0 97 9 6

PAGE 189

189 Table 5 1 1 Maximum inundated area for Site1 and Site4 for baseline and other water retention alternatives. Site1 Site4 Period Alternative Inundated area (m 2 ) % Change Inundated area ( m 2 ) % Change 2008 wet season Baseline 144961 (18) 1269 (1) A1 144961 (18) 0 1616 (1) 27 A2 1 147641 (19) 2 4215 (2) 232 A2 2 9402 (4) 641 A3 152755 (19) 5 23553 (10) 1756 A4 174007 (22) 20 52228 (22) 4016 A5 208002 (2 6 ) 43 61086 (25) 4714 A6 226077 ( 28 ) 56 70607 (29) 5464 2009 wet season Baseline 40575 (5) 321 (0.1) A1 37988 (5) 6 642 (0.3) 100 A2 1 37988 (5) 6 1872 (1) 484 A2 2 4913 (2) 1432 A3 50634 (6) 25 7280 (3) 2170 A4 97380 (12) 140 37860 (16) 11703 A5 144961 (18) 257 55165 (23) 17098 A6 157657 (20) 289 55165 (23) 17098 2010 wet season Baseline 40575 (5) 361 (0.1) A1 35324 (4) 13 642 (0.3) 78 A2 1 35324 (4) 13 1973 (1) 447 A2 2 4913 (2) 1261 A3 43097 (5) 6 7280 (3) 1917 A4 84671 (11) 109 37860 (16) 10391 A5 136595 (17) 237 56214 (23) 15477 A6 160064 (20) 294 58234 (24) 16037 Average Baseline 75370 (9) 650 (0.3) A1 72758 (9) 3 967 (0.4) 49 A2 1 73651 (9) 2 2687 (1) 313 A2 2 6409 (3) 886 A3 82162 (10) 9 12705 (5) 1854 A4 118686 (15) 57 42649 (18) 6459 A5 163186 (20) 117 57488 (24) 8741 A6 181266 (23) 140 61335 (25) 9333 v alues in the parentheses indicate the percentage of watershed

PAGE 190

190 Table 5 12. Comparisons of simulated total surface flows for baseline and other water retention alternatives (A4 A6) under the zero groundwater flux boundary condition. Site1 Site4 Period Alternative Surface flow, m 3 (cm) % Change Surface flow, m 3 (cm) % Change 2008 wet season Baseline 374120 ( 47 ) 9 0811 ( 3 7 ) A4 529370 (66) 41 75868 ( 31 ) 1 6 A5 433843 (54) 1 6 58199 ( 24 ) 3 6 A6 323424 (41) 1 4 41277 ( 17 ) 5 5 2009 wet season Baseline 149490 ( 19 ) 60867 ( 25 ) A4 238164 (30) 5 9 6371 ( 3 ) 90 A5 164692 (21) 1 0 3462 ( 1 ) 94 A6 89796 (11) 40 3751 ( 2 ) 94 2010 wet season Baseline 89014 ( 11 ) 38222 ( 1 6 ) A4 211635 (27) 138 6898 ( 3 ) 82 A5 147430 (18) 66 5239 ( 2 ) 86 A6 83328 (10) 6 5680 ( 2 ) 85 Average Baseline 204208 ( 2 6 ) 63300 ( 26 ) A4 326390 (41) 80 29712 ( 12 ) 63 A5 248655 (31) 31 22300 ( 9 ) 72 A6 165516 (21) 20 16903 ( 7 ) 78 v alues in the parentheses denote total surface flow (depth unit, cm).

PAGE 191

191 Figure 5 1. Watershed and sub watershed b oundaries and the locations of the hydrologic monitoring systems at the study site. Q1, Q2, Q3, and Q4 represent flumes that measure the surface flow s from Site1, Site2, Site3, and Site4, respectively. Combined flows from Q2, and Q3 are measured at Q5, which is t he main outlet for the ranc h Number s next to the groundwater well represent the name of the groundwater well.

PAGE 192

192 Figure 5 2. W etland water retention implementation (control structure) at Site1 and Site4. A) Site1. B) Site4. Figure 5 3. Schematic representation of the compo nents of MIKE SHE (Refsgaard et al., 1999).

PAGE 193

193 Figure 5 4. The 20 m digital elevation model (DEM) used in the MIKS SHE model The DEM was derived from 1 m l ight d etection and r anging (LIDAR) data Figure 5 5. S oil map for the study site.

PAGE 194

194 Figure 5 6. O bserved and simulated surface water levels behind the culvert and riser board structure at Site1 and rainfall for the October 1, 2008 May 31, 2011 period Figure 5 7. Observed and simulated surface water levels behind the culvert and riser board structure at Site4 and rainfall for the October 1, 2008 May 31, 2011 period.

PAGE 195

195 Figure 5 8. Observed and simulated groundwater levels for wells at Site1 (Figure 5 1) for the calibration and validation periods. A) GW21. B) GW 31. C ) GW 34. D ) GW37

PAGE 196

196 Figure 5 9. Observed and simulated surface flow at both sites for the calibration and validation periods. A) Site1. B) Site4.

PAGE 197

197 Figure 5 10 Simulated surface water flow for baseline and A5 water retention alternative (Table 5 4) Figure 5 11 Simulated surface water flow for baseline and A2 2 water retention alternative (Table 5 4)

PAGE 198

198 Figure 5 12. Simulated surface flows and groundwater levels at the middle of wetland under baseline and A4 water retention alternative (Table 5 4 ) for Site1 A) Surface flow. B) Groundwater level.

PAGE 199

199 Figure 5 13. Simulated surface flows and groundwater levels at the middle of wetland under baseline and A4 water retention alternative (Table 5 4) for Site4. A) Surface flow. B) Groundwater level.

PAGE 200

200 Fi gure 5 1 4 I nundated area s corresponding to predicted maximum water level for baseline and other alternatives (Table 5 4) at Site1. A) Baseline. B) A1. C) A2 1. D) A3. E) A4. F) A5. G) A6.

PAGE 201

201 Figure 5 1 5 I nundated area s corresponding to predicted maximum water level for baseline and other alternatives (Table 5 4) at Site 4 A) Baseline. B) A1. C) A2 1. D) A 2 2 E) A 3 F) A 4 G) A 5. H) A6

PAGE 202

202 CHAPTER 6 SUMMARY AND CONCLUSION S Water retention on ranchlands by raising the spillage levels at the drainage ditches, proposed as a potential solution to reduce the damaging high flows to Lake Okeechobee (LO) and the estuarine systems, has not been evaluated with regards to its effects. Effects of wetland w ater retention (WWR) on ranchlands in So uth Florida on surface and groundwater stor ages and flows were evaluated in this study. Effects of W WR w ere evaluated at two pastures each containing a wetland and associated upland area at a commercial beef cattle ranch in the LO watershed. Although both wetlands were herbaceous marsh depressional wetland s, they differed considerably in area, topography, hydrology, and plant community. One wetland is deep (DW) while the other is shallow (SW). The DW is mainly comprised of wetland vegetation ( water spangles duck potato, brook crowngrass, common duckweed, common rush, and water hyacinth ), while SW is dominated by pasture grass (bahiagrass). Climatic, hydrologic, and management data collected at the two sites were combined with an integrated hydrologic model, MIKE SHE/MIKE11, to evaluate the effects of W WR with regards to surface flow volume, peak flow, evapotranspiration ( ET ) groundwater flow and storage. Evapotranspiration from wetlands, a major component of water budget and source of uncertainty in hydrologic model predictions, was quantified by using t he eddy covariance (EC) method for the DW (Oct 2008 May 2011) and SW (Jun 2009 May 2011). Annual average ET was 1271 mm (93% of average rainfall) and 836 mm (62% of average rainfall) for DW and SW, res pectively. Although DW and SW were located in the s ame climatic region, difference in hydrologic conditions (e.g. hydroperiod and inundation) and composition of plant community lead to significant difference s in ET

PAGE 203

203 between the two wetlands The DW has a hi gher average slope and sits lower in elevation than SW Therefore, DW retains more surface water and receives more groundwater discharge during the wet season which results in longer hydroperiod at DW. The annual hydroperiod at DW (9.9 months) was longer t han at SW (5.2 months ), and the a verage percent inundation was 66% at DW, more than 10 times higher than SW (6%). Higher inundation at DW result ed in higher net radiation at DW The d ifference in ET between DW and SW was 3 4 % despite the fact that they are only 1.6 km apart. Difference in ET between two wetlands w as mainly due to inundation which was controlled by topographic and hydrologic conditions of wetlands and surrounding areas. Differences in plant community also resulted in different levels of trans piration losses. I nstead of having mixed type wetland vegetation growing yearlong at DW, SW i s dominated by bahiagrass which has little growth in fall and remain s dormant for most of the dry season. The fact that DW ET is 34% higher than SW indicates that estimation of wetland ET using the crop coefficient (K C ) method, the most commonly used ET method for the last 40 years which relies on generic K C values is likely to be erroneous. These errors are likely to affect the accuracy of hydrologic models such as MIKE SHE/MIKE 11 which use s K C for estimating ET. To evaluate such potential errors, literature K C values were used to estimate ET for the two wetlands. Results showed that using K C from literature lead to 2 3 % underestimation of ET at DW Although EC meth od provides accurate ET estimation, its high cost and intensive resource needs limit its use in common hydrological investigations To enable the wider use of EC based ET data, m onthly wetland vegetation coefficient ( K CW ) were derived from DW EC based ET d ata for their use s in similar wetlands in the subtropical

PAGE 204

204 climates. Two regression models, one for the dry season (R 2 = 0.58) and one for the wet season (R 2 = 0.80), were developed for estimating K CW using site specific hydro climatic variables. Data from both wetlands were pooled to develop a model for direct estimation of daily ET as a function of net radiation and inundation (R 2 = 0.80). Th e daily ET model utilizes relatively easier to measure parameters and can be useful in differentiating ET from two we tlands based on inundation changes, a main outcome of W WR. Current hydrologic models rarely represent the relationship between ET and inundation accurately. The daily ET model can be integrated with hydrologic models such as MIKE SHE for accurate representation of not only ET but also other water budget components. To understand the relative importance of hydrologic processes within the wetland upland systems in relation to the W WR, water budget analyses were con ducted for the two wetland upland sy stems (watersheds), Site1 (with DW ) and Site4 (with SW). Groundwater fluxes for the two sites were computed as residual component of the water budget. On an annual basis, ET accounted for 85 % and 7 7 % of total outflow s for Site1 and Site4, respectively. Gro undwater flux varied throughout the study period. For the wet year (May 2009 April 2010), the annual groundwater flux es at the two sites were net outflows of 116 mm (Site1) and 79 mm (Site4) During the dry year (2010 2011) groundwater fluxes at the two s ites varied from almost negligible (1 mm, Site1) to a net outflow of 75 mm (Site4) Annual average groundwater flux from the two sites was a net out flow of 58 mm/year (Site1) to 77 mm/year (Site4). Water budget analyses indicated a net groundwater outflow from both sites under average rainfall condition and these outflows may increase due to water retention. When the groundwater fluxes estimated

PAGE 205

205 from water budget method were compared with the estimates from the Darcy method along with the errors in both est imates, the two estimates differed significantly in both sign and magnitude, especially at SW site. These errors are likely due to errors caused by lack of field verified saturated hydraulic conductivity and more importantly due to use of only two wells (o ne at boundary and one in the wetland) to define the gradients across the entire site. Overall, water budget analyses indicated an average groundwater outflow of 135 mm during the wet season with average rainfall. It is likely that part of the increased st orage from W WR is likely to move out of the wetland upland system through subsurface pathways. Using measurements on EC based ET, surface and ground water levels, a multi site and multi variable evaluation of an integrated hydrologic model, MIKE SHE/MIKE 1 1 was conducted for its ability to predict effects of different levels of W WR. During the calibration period (October 2008 October 2009) model performance in predicting the surface water level s at both sites evaluated using the Nash Sutcliffe efficienc y (E) was Very Good ( SDW site, E = 0.90, SW site, E = 0.86) For the validation (November 2009 May 2011) the model performance were Good ( E = 0.7, SW site) to Very Good (E = 0. 88, DW site) Model performance in predicting groundwater levels varied from less than satisfactory (1 out of 4 wells) to Good (1 out of 4 well s ). The less than satisfactory performance was primarily due to the lack of field verified soil hydraulic properties (e.g. soil water characteristic s curve, hydraulic conductivity). T he model under predicted wetland ET for both sites by 13 % (Site4) to 19% (Site1) This under prediction of ET, despite the use of site specific K CW derived from EC measurements, is likely due to: 1) limitation of the ET m odule of MIKE SHE/MIKE 11

PAGE 206

206 in capturing the variability in ET, especially transpiration, as a function of soil water availability; 2) under prediction in groundwater levels due to errors in simulating water budget components resulting in reduced estimates o f both evaporation and transpiration; 3) spatial variability in plant community and therefore, K C values; and 4) errors from the use of 20 m grid for the model. When the MIKE SHE/MIKE 11 predicted surface water levels at the two sites were used in conjunct ion with the daily ET model, errors in ET predictions reduced significantly. Field verified MIKE SHE/MIKE11 was used to evaluate the effects of the W WR implemented at the two sites during the period of measurements as well as other W WR alternatives The cu rrent level of W WR was predicted to r educe the total surface flow volume by 64 % at Site1 and 22 % at Site4 for the period of May 2008 May 2011 Although W WR increase d the ET during the wet season it was negligible (2 3 mm) compared to the increased storage On the other hand W WR increased the magnitude of the net groundwater outflow from both sites This increase is similar to that indicated by the water budget analyses. Almost all the retained water at the two sites left the wetland upland drainage area s through subsurface Although this may not result in surface volume reductions during the wet season, temporary retention of surface water is still desirable as it is likely to change the phase of the flows and reduce the peak flows reaching the lake. Based on the maximum flooding during the three year period s, which may cause potential loss of forage, the feasible alternatives seem to be increasing the spillage level by 1.10 m and 0.74 m for Site1 and Site4, respectively. The average reductions in surface flow in the wet season for these two alternatives were

PAGE 207

207 8 6% and 62% for Sites1 and 4, respectively. The predicted reductions in peak flows for these two alternatives were 82% to 88% for Site1 and Site4, respectively The model predictions for the 0.74 m spillage level alternative were used for scale up analyses of W WR effects at the LO watershed, assuming that the alternative was implemented at all ranchlands within the watershed Results from scale up analyses showed that increased storage represents 22% of desired water storage for the LO watershed to reduce t he harmful impacts of excessive flows to the lake as well as estuarine systems on the east and we s t coasts. This storage is based on predictions from MIKE SHE /MIKE 11 under the observed groundwater boundary condition which may not represent the actual grou ndwater levels for watershed scale implementation of W WR. Watershed scale groundwater levels may rise when the W WR is implemented at 100% of the ranchlands within the LO watershed To simulate the effects of watershed scale implementation of W WR, the model was run under zero flux groundwater boundary condition. Under zero flux boundary condition, reduction in flows became much smaller. To achieve the similar storage (22% of target storage) obtained with the 0.74 m alternative under the open boundary conditi on, the spillage levels will need to be raised by 1 10 m from the ditch bottom which may not be acceptable to ranchers given its potential adverse economic impacts on ranchers unless they are compensated for providing the water storage environmental servic e Given that zero flux represents an extreme case, from the standpoints of 100% implementation of W WR on all ranchlands highly conductive sandy soils in the region, and presence of extensive network of ditches and public canals, the actual storage is lik ely to be between the predicted storage under observed and zero flux boundary.

PAGE 208

208 Future research needs and recommendations. For the water budget analysis a flownet analysis is needed for future study in order to improve the accuracy of groundwater flux est he calibrated model could simulate the surface and ground water levels but model performance in predicting volumes of surface flow was not as good as the water level Future research needs would inclu de: 1) collect site specific measurements of soil hydraulic properties ; 2) collect leaf area index data and measure the root zone depths of upland ( bahiagrass ) and wetland vegetation ; 3) use smaller grid size (e.g. 5 or 10 m as compared to 20 m used for th is study) to better represent the surface topography and water storage capacity for the study site, especially for the wetland area s; 4) improving the ET module in MIKE SHE especially in predicting ET losses from wetland s ; 5) conduct economic analyses including the impact of flooding on loss of forage and other factors; 6) evaluate tradeoffs between water retention, economic losses, carbon sequestration, and ecological impacts; and 7) conduct long term simulation study to evaluate effe cts of W WR under current and changed climate scenarios.

PAGE 209

209 LIST OF REFERENCES Abtew W. 1996. Evapotranspiration measurements and modeling for three wetland systems in south Florida. Journal of American Water Resources Association 32 : 465 473. Abtew W, Obeysekera J. 1995. Lysimeter study of evapotranspiration of cattails and comparison of three methods. Transactions of the American Society Agricultural Engineers 38 : 121 129. Acreman MC, Miller F. 2006. Hydrological impact assessment of wetlands. In: Pro ceedings of the International Symposium on Groundwater Sustainability (ISGWAS). Alicante, Spain. Allen RG. 1995. Evapotranspiration from wetlands. In American Consulting Engineer Council of Colorado Meeting: Evapotranspiration and Irrigation Efficiency Ar vada, CO, USA : 1 13. Allen RG, Pereira LS, Raes D, Smith M. 1998. Crop Evapotranspiration: Guidelines for Computing Crop Requirements. In FAO Irrigation and Drainage, Paper NO. 56 Food and Agricultural Organization of the United Nation: Rome. Allen RG, Pereira LS, Smith M, Raes D., Wright JL. 2005. The FAO 56 dual crop coefficient method for predicting evaporation from soil and application extensions. Journal of Irrigation and Drainage Engineering 131 : 2 13. Ambroise B, Perrin JL, Reutenauer D. 1995. Multicriterion validation of a semi distributed conceptual model of the water cycle in the Fecht ca tchment (Vosges Massif, France). Water Resources Reserch 31 : 1467 1481. Andersen J, Refsgaard JC, Jensen KH. 2001. Distributed hydrological modeling o f the Senegal River Basin model construction and validation. Journal of Hydrology 247 200 214. Arnold JG, Srinivasan R, Muttiah RS, Williams JR. 1998. Large area hydrologic modeling and assessment. Part 1: Model development. Journal of the American Water Resources Association 34 : 73 89. Audubon of Florida (AF). 2010. Distributed or dispersed water management and storage in the Lake Okeechobee watershed. http://audubonoffloridanews.org/wp content/uploads/2010/08/Audubon Dispersed Storage Fact Sheet.pdf Audubon of Florida (AF). 2013. Expediting Lake Okeechobee Pollution Control. http://audubonoffloridanews.org/wp content/uploads/2013/02/Audubon_LakeOkeechobee_NutrientStatement_FactSh eet.pdf.

PAGE 210

210 Baldocchi DD, Hicks BB, Meyers TP. 1988. Measuring biosphere atmosphere exchange s of biologically related gases with micrometeorological methods. Ecology 69 : 1331 1340. Bedford BL. 1999. Cumulative effects on wetland landscapes: links to wetland restoration in the United States and southern Canada. Wetlands 19 : 775 788. Bergstrom S, L indstrom G, Pettersson A. 2002. Multi variable parameter estimation to increase conf idence in hydrological modeling. Hydrological Processes 16 : 413 421. Bidlake WR, Woodham WM, Lopez MA. 1996. Evapotranspiration from areas of native vegetation in west central Florida U.S. Geological Survey Water Supply Paper 2430, U.S. Government Printing Office: Washington, DC : 35pp. Bidlake WR. 2000. Evapotranspiration from a bulrush dominated wetland in the Klamath Basin, Oregon. Journal of the American Water R esearch Association 36 : 1309 1320. Blann KL, Anderson JL, Sands GR, Vondracek B. 2009. Effects of agricultural drainage on aquatic ecosystems: a review Critical Reviews in Environmental Science and Technology 39 909 1001. Bohlen PJ, Lynch S, Shabman L, Cl ark M, Shukla S, Swain H. 2009. Paying for environmental services from agricultural lands: An example from the Northern Everglades. Frontiers in Ecology and the Environment 7 : 46 55. Boman B, Tucker D. 2002. Drainage systems for flatwoods citrus in Florida Circular 1412, Agricultural and Biological Engineering Department, Florida Cooperative Extension Service, Institute of Food and Agricultural Sciences, University of Florida, Gainesville. Bottcher AB, Hiscock JG, Jacobson BM. 2002. WAM View, A GIS Approac h to Watershed Assessment Modeling. Proceedings of the Watershed 2002 Pre Conference Modeling Workshop. Ft. Lauderdale, FL. Publisher: Water Environment Federation. Alexandria, VA. Bottcher AB, Jacobson BM, Hiscock JG. 2003. Characterizing Flow and Nutrien t Loads for TMDL Development in Florida Using WAM. Total Maximum Daily Load (TMDL) Environmental Regulations II, Albuquerque, NM, American Society of Agricultural Engineers. Boudreau DL, Rouse WR. 1995. The role of individual terrain units in the water bal ance of wetland tundra. Climate Research 5 : 31 47. Bradley C. 1997. The hydrological basis for conservation of flood plain wetlands: implications of work at Narborough Bog, UK. Aquatic Conservation 7 : 41 62.

PAGE 211

211 Bradley C. 2002. Simulation of the annual water table dynamics of a floodplain wetland, Narborough Bog, UK. Journal of Hydrology 261 : 150 172. Brinson MM. 1993. A hydrogeomorphic classification for wetlands. Vicksburg, Mississippi, U.S. Army Corps of Engi neers, Waterways Experiment Station, Technical Report WRP DE 4, 79pp. Burba GG, Anderson DJ. 2010. A Brief Practical Guide to Eddy Covariance Flux Measurements: Principles and Workflow Examples for Scientific and Industrial Applications LI COR Biosciences, Lincoln, USA. Campbell KL, Capece JC, Tremwel TK. 1995. Surface/subsurface hydrology and phosphorus transport in the Kissimmee River Basin, Florida. Ecological Engineering 5 : 301 330. Carter V. 1986. An Overview of the hydrologic conc erns rel ated to wetlands in the United States. Canadian Journal of Botany 64 : 364 374. Carter V, Bedinger MS, Novitzki RP, Wilen WO. 1979. Water resources and wetlands. In Greeson PE, Clark JR, Clark JE, eds Wetland functions and values The state of our understan ding Minneapolis, Minnesota, Water Resources Association : 344 376. Chambliss C, Miller P, Lord E. 2001. Florida Cow Calf Management, 2nd ed, Forages. UF/IFAS Publication AN118. University of Florida, Gainesville, FL. Chen E, Gerber JF. 1990. Climate. In M yers RL, Ewel JJ, eds. Ecosystem of Florida University Press of Florida, Gainesville, FL. Collins SD. 2005. Utilizing in situ benthic Mesocosms to quantify phosphorus and nitrogen fluxes in South Florida agricultural canals. Master Thesis Gainesville, FL : University of Florida. Department of Agricultural and Biological Engineering. Cowardin LM, Carter V, Golet FC, LaRoe ET. 1979. Classification of wetlands and deepwater habitats of the United States. FWS/OBS 79/31, US Fish and Wildlife Service, Office of Biological Services, Washington, DC. Dai ZH, Trettin CC, Li CS, Amatya DM, Sun G, Li H. 2010. Sensitivity of stream flow and water table depth to potential climatic variability in a coastal forested watershed. Journal of the American Water Resources Associ ation 46 : 1036 1048 Das BM. 2009. Fundamentals of Geotechnical Engineering Third Edition. Chris Carson, United States. Davis FE, Marshall ML. 1975. Chemical and biological investigations of Lake Okeechobee, January 1973 June 1974, Interim Report. Central and Southern Florida Flood Control District Technical Publication 75 1.

PAGE 212

212 Demissie M, Akanbi, AA, Khan A. 1997. Hydrologic modeling of landscape. Functions of wetlands. Research Report 125, Illinois State Water Survey Champaign Illinois, 42pp. DHI. 2006. MIKE SHE User Manual DHI Software, Hrsholm, Denmark. Dooge J. 1975. The water balance of bogs and fens. International Symposium on the Hydrology of Marsh Ridden Areas. Proceedings of the Minsk Symposium UNESCO Press, IAHS, Paris: 233 271. Drexler JZ, Be dford BL, DeGaetano AT, Siegel DI. 1999. Quantification of the water budget and nutrient loading in a small peatland. Journal of American Water Resources Association 35 : 753 769. Drexler JZ, Snyder RL, Spano D, Paw U KT. 2004. A review of models and microm eteorological methods used to estimate wetland evapotranspiration. Hydrological Processes 18 : 2071 2101. Drexler JZ, Anderson FE, Snyder RL. 2008. Evapotranspiration rates and crop coefficients for a restored marsh in the Sacramento San Joaquin Delta, Cali fornia, USA. Hydrological Processes 22 : 725 735. Earth Tech, DHI. 2007. Kissimmee Basin Modeling and Operations Study KBMOS, AFET Model Documentation/Calibration Report, West Palm Beach, FL : 318pp. Favero L, Mattiuzzo E, Franco D. 2007. Practical results o f a water budget estimation for a constructed wetland. Wetlands 27 : 230 239. Federico AC, Dickson KG, Kratzer CR, Davis FE. 1981. Lake Okeechobee water quality studies and eutrophication assessment. Technical Publication 81 2. Florida Water Management Dist rict, West Palm Beach. Flaig EG, Havens KE. 1995. Historical trends in the Lake Okeechobee ecosystem. I. Land use and nutrient loading. Archiv fur Hydrobiologie Beihefte 107 1 24. Flaig EG, Reddy KR 1995. Fate of phosphorus in the Lake Okeechobee watershed, Florida, USA: Overview and recommendations. Ecological Engineering 5 127 142. Florida Soil Characterization Data Retrieval System (FSCDRS). 2013. http://flsoils.ifas.ufl.edu [accessed August 2013]. Freeze R A, Harlan RL. 1969. Blueprint of a physically based, digitally simulated hydrologic response model. Journal of Hydrology 9 : 237 258. Fritschen LJ, Simpson JR. 1989. Surface energy balance and ra diation systems: general description and improvements. Journal of Applied Meteorology 28 : 680 689.

PAGE 213

213 Gaff H, DeAngelis DL, Gross LJ, Salinas R, Shorrosh M. 2000. A dynamic landscape model for fish in the Everglades and it application to restoration. Ecologic al Modeling 127 : 33 52. Gary PN, Farrell C, Romine T. 2007. Lake Okeechobee. Audubon of Florida. http://crca.caloosahatchee.org/crca_docs/LakeOReport_1_07.pdf German ER. 2000. Reg ional evaluation of evapotranspiration in the Everglades U.S. Geological Survey Water Resources Investigation Report 00 4217. Gilvear DJ, Andrews R, Tellam JW, Lloyd JW, Lerner DN. 1993. Quantification of the water balance and hydrogeological processes in the vicinity of a small groundwater fed wetland, East Anglia, UK. Journal of Hydrology 144 : 311 334. Gholz H L, Clark KL. 2002. Energy exchange across a chronosequence of slash pine forests in Florida. Agricultural and Forest Meteorology 112 : 87 102. Graf WL. 1999. Dam nation: A geographic census of American dams and their large scale hydrologic impacts. Water Resources Research 35 : 1305 1311. Graham WD. 2009. Evaluating the e ffectiveness of b est m anagement p ractices to r educe nutrient loadings from be ef cattle r anches to Lake Okeechobee Final Report. Florida Department of Environmental Protection, the Florida Department of Agriculture and Consumer Services and the South Florida Water Management District Graham DN, Butts MB. 2005. Chapter 10 Flexible Integrated Watershed Modeling With MIKE SHE. In Singh VP, Frevert DK eds. Watershed Models CRC Press, Boca Raton, Florida: 245 272. Graham WD Donigian AS, Munoz Carpena R, Skaggs W, Shirmohammadi A. 2009. Peer Review of the Watershed Assessment Model (W AM), Final Panel Report. http://waterinstitute.ufl.edu/news/downloads/WAM_final_report.pdf Guardo M. 1999. Hydrologic balance for a subtropical treatment wetland constructed for nutrient removal. Ecological Engineering 12 : 315 337. Gupta HV, Sorooshian S, Yapo PO. 1998. Toward improved calibration of hydrologic models: multiple and noncommensura ble measures of information. Water Resources Research 34 : 751 763. Guzha AC, Shukla S. 2011. Effects of topographic data accuracy on water storage and its Implications for Payment for Environment Services program. ASCE Journal of Irrigation and Drainage En gineering DOI:10.1061/(ASCE)IR.1943 4774.0000430).

PAGE 214

214 Harmel RD, Cooper RJ, Slade RM, Haney RL, Arnold JG. 2006. Cumulative uncertainty in measured streamflow and water quality data for small watershed. Transactions of the American Society of Agricultural Eng ineers 4 9 : 689 701. Harvey R, Havens KE. 1999. Lake Okeechobee Action Plan Hatala JA, Detto M, Sonnentag O, Deverel SJ, Verfaillie J, Baldocchi DD. 2012. Greenhouse gas (CO2, CH4, H2O) fluxes from drained and flooded agricultural peatlands in the Sacramen to San Joaquin Delta. Agriculture, Ecosystems & Environment 150 : 1 18. Havens KE, Aumen NG, James RT, Smith VH 1996. Rapid ecological changes in a large subtropical lake undergoing cultural eutrophication. Ambio 25 : 150 155. Havens KE, James RT, 2005. The phosphorus mass balance of Lake Okeechobee, Florida: implications for eutrophication management. Lake and Reservoir Management 21 : 139 148. Hayashi M Rosenberry DO. 2001. Effects of groundwater exchange on the hydrology and ecology of surface waters. Journal of Groundwater Hydrology 43 : 327 341. Hayashi M, van der Kamp G, Rudolph DL. 1998. Water and solute transfer between a prairie and adjacent uplands, 1. Water balance. Journal of Hydrology 207 : 42 55. Hernandez T, Nachabe M, Ross M, Obeysekera J. 2003. Modeling runoff from variable source areas in humid, shallow water table environments. Journal of the American Water Resources Association 39 : 75 85. Hill AR. 2000. Stream chemistry and riparian zones. In Streams and ground waters, Jones JB and Mullholland PJ (eds.) Academic Press: San Diego 83 110. Horst TW, Weil JC. 1992. Footprint estimation for scalar flux measurements in the atmospheric surface layer. Boundary Layer Meteorology 69 : 417 448. Hsieh CI, Katul G, Chi T. 2000. An approximately analytical model for footprint estimation of scalar fluxes in thermally stratified atmospheric flows. Advances in Water Resources 23 : 765 772. Hunt RJ, Krabbenhoft DP, Anderson MP. 1996. Groundwater inflow m easurements in wetland systems. Water Resources Research 32 : 495 507. Hyatt RA, Brook GA. 1984. Ground water flow in the Okefenokee Swamp and the hydrologic and nutrient budgets for the period August, 1981 through July, 1982. In Cohen AD, Casagrande DJ, Andrejko MJ, Best GR, eds. The Okefenokee Swamp Wetland Surveys, Los Alamos, New Mexico: 229 245. Ingram HAP. 1983. Hydrology. Ecosystems of the world Vol. 4, Mires: Swamp, Bog, Fen and Moor, Gore AJP eds Elsevier: Amsterdam, 67 158.

PAGE 215

215 Irmak S, Kabenge I, Rudnick D, Knezevic S, Woodward D, Moravek M. 2013. Evapotranspiration crop coefficients for mixed riparian plant community and transpiration crop coefficients for Common reed, Cottonwood and Peach leaf willow in the Platte River Basin, Nebraska, USA. Jou rnal of Hydrology 481 : 177 190. Jaber FH, Shukla S. 2004. Simulating water dynamics in agricultural stormwater impoundments for irrigation water supply. Transactions of the American Society of Agricultural Engineers 47 : 1465 1476. Jaber FH, Shukla S. 2012. MIKE SHE: model use, calibration, and validation. Transactions of the American Society of Agricultural and Biological Engineering 55 : 1479 1489. Jacobs JM, Mergelsberg SL, Lopera AF, Myers DA. 2002. Evapotranspiration from a wet prairie wetland under drough t conditions: Paynes Prairie Preserve, Florida, USA. Wetlands 22 : 374 385. Jacobs J, Jia X, Choi M, Sumner D, Douglas E. 2007. Net radiation estimations in evapotranspiration research over different canopy surfaces in a humid climate. World Environmental and Water Resources Congress 2007 : 1 9. DOI: 10.1061/40927(243)228 James RT, Jones BL, Smith VH. 1995. Historical trends in the Lake Okeechobee ecosystem. II. Nutrient budgets. Archiv fur Hydrobiologie Beihefte 107 25 47. Jensen ME. 1973. Consumptive use of water and irrigation water requirements Irrigation and Drainage Division of the American Society of Civil Engineers New York: 215pp. Jensen ME, Burman RD, Allen RG. 1990. Evapotranspiration and irrigation water requirements. ASCE Manuals and Reports on Engineering Practice No. 70 New York : 332pp. Jia X, Swancar A, Jacob JM, Dukes MD, Morgan K. 2007. Comparison of evapotranspiration rates for flatwoods and ridge citrus. Transactions of the American Society of Agricultural and Biological Engineering 50 : 83 94. Jia X, Dukes MD, Jacobs JM. 2009. Bahiagrass crop coefficients from eddy correlation measurements in Florida. Irrigation Science 28 : 5 15. Kadlec RH, Knight RL. 1996. Treatment Wetlands CRC Lewis Publishers, Boca Raton, FL. Kang SZ, Gu BJ, Du TS, Zhang JH. 2003. Crop coefficient and ratio of transpiration to evapotranspiration of winter wheat and maize in a semi humid region. Agricultural Water Management 59 : 239 254.

PAGE 216

216 Kenward T, Lettenmaier DP, Wood EF, Fielding E 2000 Effects of Digital Elevati on Model accuracy on hydrologic predictions. Remote Sensing of Environment 74 : 432 444 Khu ST, Madsen H, di Pierro F. 2008. Incorporating multiple observations for distributed hydrologic model calibration: An approach using a multi objective evolutionary algorithm and clustering. Advances in Water Resources 31 : 1387 1398. Kienzle SW, Schmidt J. 2008. Hydrological impacts of irrigated agriculture in the Manuherikia Catchment, Otago, New Zealand. Journal of Hydrology (NZ) 47 :67 83. Kiker GA, Clark DJ, Martinez CJ, Schulze RE. 2006. A Java based, object oriented modeling system for southern Afri can hydrology. Transactions of the ASABE 49 : 1419 1433. Kljun N, Calanca P, Rotach MW, Schmid HP. 2004. A simple parameterization for flux footprint predictions. Boundary Layer Meteorology 112 : 503 523. Kristensen KJ, Jensen SE. 1975. A Model for Estimatin g Actual Evapotranspiration from Potential Evapotranspiration. Nordic Hydrology 6, Royal Veterinary and Agricultural University : 170 188. Kustas WP, Stannard DI, Allwine KJ 1996. Variability in surface energy flux partitioning Monteith and Priestley Taylor parameters. Agricultural and Forest Meteorology 82 : 171 193. LaBaugh JW. 1986. Limnological characteristics of selected lake s in the Nebraska sandhills, U.S.A., and their relation to chemical characteristics of adjacent ground water. Journal of Hydrology 86 : 279 298. Leclerc MY, Thurtell GW. 1990. Footprint prediction of scalar fluxes using a Markovian analysis. Boundary Layer Meteorology 52 : 247 258. Lee CH, Chen WP, Lee RH. 2006. Estimation of groundwater recharge using water balance coupled with base flow record estimation and stable base flow analysis. Environmental Geology 51 : 73 82. Linacre, E.T., 1976. Swamps. Vegetation and the atmosphere Vol. 2, Case studies, Monteith JL ed s Academic Press: London, 329 348. Linacre ET, 1968. Estimating the net radiation flux. Agricultural Meteorology 5 : 49 63. Li Y, Zhou L, Xu Z, Zhou G. 2009. Comparison of water vapor, heat and energ y exchanges over agricultural and wetland ecosystems. Hydrological Processes 23 : 2069 2080.

PAGE 217

217 Lott RB, Hunt RJ. 2001. Estimating evapotranspiration in natural and constructed wetlands. Wetlands 21 : 614 628. Lu J, Sun G, McNulty SG, Comerford NB. 2009. Sensitivity of pine flatwoods hydrology to climate change and forest management in Florida, USA. Wetlands 29 : 826 836. Mallows CL. 1973. Some comments on Cp. Technometrics 15 : 661 675. Mansell R S, Bloom SA, Sun G. 2000. A model for wetland hydrology: description and validation. Soil Science 1656 : 384 397. Mamillapalli S, Srinivasan R, Arnold JG, Engel BA. 1996. Effect of Spatial Variability on Basin Scale Modeling. In: Proceedings of the Third International Conf erence/Workshop on Integrating GIS and Environmental Modeling National Center for Geographic Information and Analysis, Santa Barbara, California. Mao LM, Bergman MG, Tai CC. 2002. Evapotranspiration measurement and estimation of three wetland environments in the upper St. Johns River basin, Florida. Journal of the American Water Resources Association 38 : 1271 1285. Martinez CJ, Campbell KL, Annable MD, Kiker GA. 2008. An object oriented hydrologic model for humid, shallow water table environments. Journal of Hydrology 351 : 368 381. McCormick P, James RT, Zhang J, 2010. Chapter 10: Lake Okeechobee Protection Program State of the lake and watershed. In: 2009 South Florida Environmental Report Volume I. South Florida Water Management District, West Palm Be ach, FL. McMichael CE, Hope AS, Loaiciga HA. 2006. Distributed hydrological modeling in California semi arid shrublands: MIKE SHE model calibration and uncertainty estimation. Journal of Hydrology 317 : 307 324. Meile T, De Cesare G, Blanckaert K, Schleiss AJ. 2008. Improvement of acoustic Doppler velovimetry in steady and unsteady turbulent open channel flows by means of seeding with hydrogen bubbles. Flow Measurement and Instrumentation 19 215 221. Migliaccio KW, Shoemaker WB. 2013. Estimation of urban subtropical bahiagrass (Paspalum notatum) evapotranspiration using crop coefficients and the eddy covariance method. Journal of Hydrological Processes doi: 10.1002/hyp.9958. Mitsch WJ, Gosselink JG 2000. Wetlands John Wiley & Sons: New York. Monteith JL. 1965. Evaporation and Environment. In Proceedings of the 19th Symposium of the Society for Experimental Biology Cambridge University Press: New York: 205 233.

PAGE 218

218 Moore CJ. 1976. Eddy flux measuremen ts above a pine forest. Quarterly Journal of the Royal Meteorological Society 102 : 913 918. Moriasi DN, Arnold JG, Van Liew MW, Bingner RL, Harmel RD, Veith TL. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simula tions. Transactions of the ASABE 50 : 885 900. Mutiibwa D, Irmak S. 2013. AVHRR NDVI based crop coefficients for analyzing long term trends in evapotranspiration in relation to changing climate in the U.S. High Plains. Water Resources Research 49 : 231 244. Nachabe M, Shah N, Ross M, Vomacka J. 2005. Evapotranspiration of two vegetation covers in a shallow water table environment. Soil Science Society of American Journal 69 : 492 499. Nash JE, Sutcliffe JV. 1970. River forecasting through conceptual models: P art I. A discussion of principles. Journal of Hydrology 10 : 282 290. National Climatic Data Center (NCDC). 2012 South Central Florida regional weather data http://www.ncdc.noaa.gov/oa/ncdc.html [accessed March 2012]. National Oceanic and Atmospheric Administration (NOAA). 2010. Climatological data,Florida National Oceanic and Atmospheric Administration, National Climatic Center Environmental Data Service. Asheville, NC. National Research Counci l (NRC). 2008. Progress toward Restoring the Everglades: The Second Biennial Review. National Academies Press, Washington, DC. Natural Resources Conservation Service (NRCS). 2003. Soil survey of Okeechobee County, Florida. Natural Resources Conservation Se rvice, United States Department of Agriculture. National Research Council (NRC), 2001. Compensating for Wetland Losses under the Clean Water Act Committee on Mitigating Wetland Losses, Board on Environmental Studies and Toxicology, Water Science and Techn ology Board, Division on Earth and Life Studies, National Research Council. National Academy Press, Washington, DC ( http://books.nap.edu/books/0309074320/html/R1.html#pagetop ). Novick KA, Oren R, Stoy PC, Siqueira MBS, Katul GG, 2009. Nocturnal evapotranspiration in eddy covariance records from three co located ecosystems in the Southeastern U.S.: Implications for annual fluxes. Agricultural and Forest Meteorology 149 : 1491 1504. NRCS, 2010. Soil survey of Okeechobee County, Florida Natural Resources Conservation Service, United States Department of Agriculture.

PAGE 219

219 Nungesser MK, Chimney MJ. 2006. A hydrological assessment of the Everglades Nutrient Removal Project, a subtropical con structed wetland in South Florida, USA. Ecological Engineering 4 : 331 344. Obeysekera J, Browder JA, Hornung L, Harwell MA. 1999. The natural South Florida system I: climate, geology, and hydrology. Urban Ecosystems 3 : 223 244 OECD/IUCN (Organization for Economic Co operation and Development/ International Union for Conservation of Nature ). 1996. Guidelines for aid agencies for improved conservation and sustainable use of tropical and sub tropical wetlands. Paris. Owen CR. 1995. Water budget and flow patter ns in an urban wetland. Journal of Hydrology 169 : 171 187. Pauwels VRN, Samson R. 2006. Comparison of different methods to measure and model actual evapotranspiration rates for a wet sloping grassland. Agricultural Water Management 82 : 1 24. Peacock CE, H ess TM. 2004. Estimating evapotranspiration from a reed bed using the Bowen ratio energy balance method. Hydrological Processes 18 : 247 260. Penman H L. 1948. Natural evaporation from open water, bare soil, and grass. Proceeding Royal Society London A 193: 120 146. Priestley CHM, Taylor RJ. 1972. On the assessment of surface heat flux and evaporation using large scale parameters. Monthly Weather Review 100 : 81 92. Raisin G, Bartley J, Croome R. 1999. Groundwater influence on the water balance and nutrient bu dget of a small natural wetland in Northeastern Victoria, Australia Ecological Engineering 12 : 133 147. Read KE, Hedges PD, Fermor PM. 2008. Monthly evapotranspiration coefficients of large reed bed habitats in the United Kingdom. In Vymazal J eds Waste water Treatment, Plant Dynamics And Management in Constructed And Natural Wetlands : 99 109. Reese RS. 2004. Hydrogeology, Water Quality, and Distribution of Salinity in the Floridan Aquifer System, Martin and St. Lucie Counties, Florida. USGS Water Resourc es Investigations Report 03 4242:96. U.S. Geological Survey, Tallahassee, FL. Refsgaard JC. 1997. Parameterisation, calibration and validation of distributed hydrological models. Journal of Hydrology 198 : 69 97. Refsgaard JC, Storm B. 1995. MIKE SHE. Computer models of watershed hydrology In: Singh VP eds. Water Resources Publications, Colorado: 809 846.

PAGE 220

220 Richter BD, Baumgartner JV, Powell J, Braun DP. 1996. A method for assessing hydrological alteration within ecosystems. Conservation Biology 10 : 1163 1174. Rijal I, Jia X, Zhang X, Steele D, Scherer T, Akyuz A. 2012. Effects of subsurface drainage on evapotranspiration for corn and soybean in Southeastern North Dakota. Journal of Irrigation and Drainage Engineering 138 :1060 1067. doi:10.1061/(ASCE)IR.1 943 4774.0000508. Sahoo, GB, Ray C, De Carlo EH. 2006. Calibration and validation of a physically distributed hydrological model, MIKE SHE, to predict streamflow at high frequency in a flashy mountainous Hawaii stream. Journal of Hydrology 327 : 94 109 Sa mani Z., 2000. Estimating solar radiation and evapotranspiration using minimum climatological data. ASCE Journal of Irrigation and Drainage Engineering 126 : 265 267. Sanchez Carrillo S, Angeler DG, Sanchez Andres R, Alvarez Cobelas M, Garatuza Payan J. 200 4. Evapotranspiration in semi arid wetlands: relationships between inundation and the macrophyte cover: open water ratio. Advances in Water Resources 27 : 643 655. Schmid HP. 1994. Source areas of scalar and scalar fluxes. Boundary Layer Meteorology 67 : 293 318. Schmid HP. 1997. Experimental design for flux measurements: matching scales of observations and fluxes. Agricultural and Forest Meteorology 87 : 179 200. Schotanus P, Nieuwstadt FTM, de Bruin HAR. 1983. Temperature measurement with a sonic anemometer and its application to heat and moisture fluxes. Boundary Layer Meteorology 50 : 81 93. Schuepp PH, Leclerc MY, MacPherson JI, Desjardins RL. 1990. Footprint pr ediction of scalar fluxes from analytical solutions of the diffusion equation. Boundary Layer Meteorology 50 : 355 373. Sellers PJ, Randall DA, Collatz GJ, Berry JA, Field CB, Dazlich DA, Zhang C, Collelo GD, Bounoua L. 1996. A revised land surface parameterization (SiB2) for atmospheric GCMs part 1: model formulation. Journal of Climate 9 : 676 705. SFWMD (South Florida Wat er Management District) 1989. Interim Lake Okeechobee Surface Water Improvement and Management Plan. South Florida Water Management District, West Palm Beach, FL, USA. SFWMD, 1997. Surface water improvement and management (SWIM) plan update for Lake Okeec hobee. Report, South Florida Water Management District, West Palm Beach, FL, USA.

PAGE 221

221 SFWMD, 2000. A study to manage phosphorus in Lake Okeechobee. http://mytest.sfwmd.gov/portal/page/portal/pg_grp_sfwmd_watershed/sediment_ management_194/tab744041/factsheet.pdf SFWMD, 2008. Lake Okeechobee Watershed Construction Project Phase II Technical Plan South Florida Water Management Distri ct, Florida Department of Environmental Protection and Florida Department of Agriculture and Consumer Services, West Palm Beach, FL. SFWMD, 2013. South Florida environmental report. South Florida Water Management District, West Palm Beach, FL, USA. Shukla S, Graham WD, Dunne EJ, Reddy KR, Clark MW 2007. Ranch scale evaluation of water retention BMPs in the Lake Okeechobee basin using wetlands and ditches. Research project proposal of DACS Office of Agricultural Water Policy. Shukla S, Graham WD, Goswami D, Knowles JM, Hodges AW, Nair VD, Christman MC, Wu C. 2011. Evaluation of cow calf best management practices with regards to nutrient discharges in the Lake Okeechobee Basin Technical Report: Florida Department of Agricultural and Consumer Services: Tallah assee, FL : 136pp. Shukla S, Jaber FH, Goswami D, Srivastava S. 2012. Evapotranspiration losses for pepper under plastic mulch and shallow water table condition. Irrigation Science DOI 10.1007/s00271 012 0327 3. Shuttleworth WJ. 1989. Insight from large sc ale observation studies of land/atmosphere interactions. Surveys in Geophysics 12 : 3 30. Shuttleworth WJ. 1992. Evaporation. In Maidment DR. e d s Handbook of Hydrology McGraw Hill, New York, NY : 4.1 4.53. Skaggs RW. 1978. A Water Management Model for Shallow Water Table Soils. Technical Report #134. Water Resources Research Institute of the University of North Carolina, NCSU, Raleigh, N.C. Skaggs RW. 1999. Drainage simulation models. In Skaggs RW, van Schilfgaarde J eds. Agricultural Drainage. Monogra ph No. 38 ASA, CSSA, and SSSA, Madison, WI: 469 500. Skaggs RW, Gilliam JW, Evans RO. 1991. A computer simulation study of pocosin hydrology. Wetlands 11 : 399 416. Skaggs RW. 1978. A water management model for shallow water table soils. Technical report N o. 134. Water Resources Research Institute of the University of North Carolina State University, Raleigh, N.C.

PAGE 222

222 Smith M, Allen RG, Monteith J, Perrier A, Pereira L, Segeren A. 1991. Report of the expert consultation on FAO methodologies for crop water requi rements Food and Agricultural Organization Land and Water Development Division, Rome. Souch C, Wolfe CP, Grimmond CSB. 1996. Wetland evaporation and energy partitioning: Indiana Dunes National Lakeshore. Journal of Hydrology 184 : 189 208. Souch C, Grimmon d CSB, Wolfe CP. 1998. Evapotranspiration rates from wetlands with different disturbance histories: Indiana Dunes National Lakeshore. Wetlands 18 : 216 229. Srinivasan MS, Gerard Marchant P, Veith TL, Gburek WJ, Steenhuis TS. 2005. Watershed scale modeling of critical source areas of runoff generation and phosphorus transport. Journal of the American Water Resources Association 41 : 361 375. Steinman AD, Conklin J, Bohlen PJ, Uzarski DG. 2003. Influence of cattle grazing and pasture land use on macroinvertebr ate communities in freshwater wetlands. Wetlands 23 : 877 889. Stannard DI. 1993. Comparison of Penman Monteith, Shuttleworth Wallace, and modified Priestley Taylor evapotranspiration models for wildland vegetation in semiarid rangeland. Water Resources Res earch 29 : 1379 1392. Sumner DM. 1996. Evapotranspiration from successional vegetation in a deforested area of the Lake Wales Ridge, Florida U.S. Geological Survey Water Resources Investigations Report 96 4244. Sumner DM. 2001. Evapotranspiration from a cy press and pine forest subjected to natural fires in Volusia County, Florida, 1998 1999 U.S. Geological Survey Water Resource s Investigations Report 01 4245: 56pp. Sumner DM, Jacobs JM. 2005. Utility of Penman Monteith, Priestley Taylor, reference evapotra nspiration, and pan evaporation methods to estimate pasture evapotranspiration. Journal of Hydrology 308 : 81 104. Sun G, Riekerk H, Comerford NB. 1998. Modeling the forest hydrology of wetland upland ecosystems in Florida. Journal of the American Water Resources Association 34 : 827 841. Sutula M, Day JW, Cable J, Rudnick D. 2001. Hydrologocal and nutrient budgets for freshwater and estuarine wetlands of Taylor Slough in Southern Everglades, Florida (U.S.A). Biogeochemistry 56 : 287 310. Tanner BD, Green J P. 1989. Measurement of sensible heat flux and water vapor fluxes using eddy correlation methods Final report to U.S. Army Dugway Proving Grounds. DAAD 09 87 D 0088.

PAGE 223

223 Tanner CB, Thurtell GW. 1969. Anemoclinometer measurements of Reynolds stress and heat tr ansport in the atmospheric boundary layer United States Army Electronics Command, Atmospheric Sciences Laboratory, Fort Huachuca, Arizona, TR ECOM 66 G22 F, Reports Control Symbol OSD 1366 : 10pp. Tasumi M, Allen RG, Trezza R, Wright JL. 2005. Satellite ba sed energy balance to assess within population variance of crop coefficient curves. Journal of Irrigation and Drainage Engineering 131 : 94 109. Thomann RV. 1982. Verification of water quality models. Journal of Environmental Engineering Division 108 : 923 9 40. Tibbals CH. 1990. Hydrology of Floridan Aquifer System in East Central Florida. U.S. Geological Survey Water Resources Investigations: 76 97. Towler B, Cahoon J, Stein O. 2004. Evapotranspiration crop coefficients for cattail and bulrush. Journal of Hy drologic Engineering 9 : 235 239. Tweel AW, Bohlen PJ. 2008. Influence of soft rush (Juncus effuses) on phosphorus flux in grazed seasonal wetlands. Ecological Engineering 33 : 242 251. Twine TE, Kustas WP, Norman JM, Cook DR, Houser PR, Meyers TP, Prueger J H, Starks PJ, Wesely ML. 2000. Correcting eddy covariance flux underestimates over a grassland. Agricultural and Forest Meteorology 103 : 279 300. Vazquez RF, Feyen J. 2003. Effect of potential evapotranspiration estimates on effective parameters and perfor mance of the MIKE SHE code applied to a medium size catchment. Journal of Hydrology 270 : 309 327. Vazquez RF, Feyen L, Feyen J, Refsgaard JC. 2002. Effect of grid size on effective parameters and model performance of the MIKE SHE code applied to a medium sized catchment. Hydrological Processes 16 : 355 372. Verry ES, Boelter D. 1979. Wetland functions a nd values: the state of our understanding. In Greenson PE, Clark JR, Clark JE eds. Peatland hydrology. American Water Resources Association, Minneapolis, Minnesota : 389 402. Vesala T, Kljun N, Rannik U, Rinne J, Sogachev A, Markkanen T, Sabelfeld K, Foken TH, Leclerc MY. 2008. Flux and concentration footprint modeling: state of the art. Environmental Pollution 152 : 653 666. Wang MN, Qin DY, Lu CY, Li YP. 2010. Modeling anthropogenic impacts and hydrological processes on a wetland in China. Water Resource Ma nagement 24 : 2743 2757. Wang MN, Wang H, Qin DY, Lu CY, Li YP. 201 1 Modeling the artificial recharge of a wetland and its influence on regional hydrological process in China: a case study. Ecohydrology 4 589 596

PAGE 224

224 Wang S, Zhang Z, Sun G, Strauss P, Guo J, Tang Y, Yao A. 2012. Multi site calibration, validation, and sensitivity analysis of the MIKE SHE Model for a large watershed in northern China. Hydrology and Earth System Sciences 16 : 4621 4632. Walter IA, Allen RG, Elliott R, Jensen MR, Itenfisu D, Mech am B, Howell TA, Snyder R, Brown P, Eching S, Spofford T, Hattendorf M, Cuenca RH, Wright JL, Martin D. th National Irrigation Symposium, ASAE, Phoenix, AZ. Webb EK, Pearman GI, Leun ing R. 1980. Correction of flux measurements for density effects due to heat and water vapor transfer. Quarterly Journal of the Royal Meteorological Society 106 : 85 100. Wilcox DA, Sweat MJ, Carlson ML, Kowalski KP. 2006. A water budget approach to restoring a sedge fen affected by diking and ditching. Journal of Hydrology 320 : 501 517. Willmott, C.J. 1981. On the validation of models. Physical Geography 2 :184 194. Winter TC. 1981. Uncertainties in estimating the water balance of lakes. Water Resources Bulletin 17 : 82 115. Winter TC. 1988. A Conceptual Framework for Assessing Cumulative Impacts on the Hydrology of Nontidal Wetlands. Environmental Management 12 : 605 620. Win ter TC, LaBaugh JW. 2003. Hydrologic considerations in defining isolated wetlands. Wetlands 23 : 532 540. Wu C, Shukla S. 2013. Eddy covariance based evapotranspiration for a subtropical wetland. Hydrological Processes DOI: 10.1002/hyp.10075 Yan J, Smith KR. 1994. Simulation of integrated surface water and ground water systems model formulation. Water Resource Bulletin 30 : 879 890. Zhao H, Zhang J, James RT, Laing J. 2012. Application of MIKE SHE/MIKE 11 model to structural BMPs in S191 basin, Florida. Journal of Environmental Informatics 19 : 10 19. Zhou L, Zhou G. 2009. Measurement and modeling of evapotranspiration over a reed ( Phragmites australis ) marsh in Northeast China. Journal of Hydrology 372 : 41 47. Zoltai SC. 1998. Wetland environments and classification. In Rubec CDA. (eds). Wetland of Canada Polyscience, Montreal, QC, Canada: 3 26.

PAGE 225

225 BIOGRAPHICAL SKETCH Chin Lung Wu was born in 1978, the first of the three chi ldren of Ching Chih Wu and Li Sin Lin Chin Lung was born and raised in Taiwan. At a young age, he was always interested in mathematics and sciences. During his undergraduate years, he found passion in studying the intricate water system of the earth He completed his b achelor of e ngineering in h ydraulics and o cean e ngineering at National Cheng Kung University Tainan, Taiwan. He decided to continue his area of study by completing a m aster of e ngineering in b ioenv ironmental s ystems e ngineering at National Taiwan University, Taipei, Taiwan. In 2006, he began his work toward a doctorate degree in the a gricultural and b iological e ngineering at the University of Florida under the direction of Dr. Sanjay Shukla and Dr. Wendy D. Graham