<%BANNER%>

Assessing Efficiencies in Vegetable Production

Permanent Link: http://ufdc.ufl.edu/UFE0045362/00001

Material Information

Title: Assessing Efficiencies in Vegetable Production Hydrologic Modeling of Soil-Water Dynamics and Estimation of Greenhouse Gas Emissions
Physical Description: 1 online resource (206 p.)
Language: english
Creator: Jones, Curtis D
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2013

Subjects

Subjects / Keywords: agriculture -- dssat -- efficiency -- ghg -- hydrology -- irrigation -- management -- modeling -- production -- soil -- systems -- vegetable
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The vegetable industry is economically important in Florida, with annual production valued at around 1.8 billion dollars spanning 300,000 acres. As such, these crops tend to be intensively managed to ensure optimal yields. However, in an atmosphere of increasing population and resource scarcity, both in Florida and globally, there is increasing pressure for agricultural commodities to be produced efficiently. To this end, cropping systems models (CSMs) are useful tools for identifying efficient production practices as they can be used to help understand complex relationships between management, system drivers, production, and environmental consequences. In this research the commonly used Decision Support System for Agrotechnology Transfer (DSSAT) CSM was amended to allow for water-limited simulation under cultural practices common for vegetable production, including drip irrigation. The modified model was evaluated to assess its utility in simulating soil-water dynamics, and simulation experiments were conducted to identify implications of management choices. Analyses demonstrated the relative importance of different drip irrigation management options as well as the importance of sensor placement for soil moisture sensor controlled irrigation. As an added dimension, estimates of the greenhouse gas (GHG) emissions from typical tomato production systems were calculated. The estimates emphasize the importance of irrigation management, nutrient management, and productivity for minimizing these emissions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Curtis D Jones.
Thesis: Thesis (Ph.D.)--University of Florida, 2013.
Local: Adviser: Fraisse, Clyde William.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2013
System ID: UFE0045362:00001

Permanent Link: http://ufdc.ufl.edu/UFE0045362/00001

Material Information

Title: Assessing Efficiencies in Vegetable Production Hydrologic Modeling of Soil-Water Dynamics and Estimation of Greenhouse Gas Emissions
Physical Description: 1 online resource (206 p.)
Language: english
Creator: Jones, Curtis D
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2013

Subjects

Subjects / Keywords: agriculture -- dssat -- efficiency -- ghg -- hydrology -- irrigation -- management -- modeling -- production -- soil -- systems -- vegetable
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The vegetable industry is economically important in Florida, with annual production valued at around 1.8 billion dollars spanning 300,000 acres. As such, these crops tend to be intensively managed to ensure optimal yields. However, in an atmosphere of increasing population and resource scarcity, both in Florida and globally, there is increasing pressure for agricultural commodities to be produced efficiently. To this end, cropping systems models (CSMs) are useful tools for identifying efficient production practices as they can be used to help understand complex relationships between management, system drivers, production, and environmental consequences. In this research the commonly used Decision Support System for Agrotechnology Transfer (DSSAT) CSM was amended to allow for water-limited simulation under cultural practices common for vegetable production, including drip irrigation. The modified model was evaluated to assess its utility in simulating soil-water dynamics, and simulation experiments were conducted to identify implications of management choices. Analyses demonstrated the relative importance of different drip irrigation management options as well as the importance of sensor placement for soil moisture sensor controlled irrigation. As an added dimension, estimates of the greenhouse gas (GHG) emissions from typical tomato production systems were calculated. The estimates emphasize the importance of irrigation management, nutrient management, and productivity for minimizing these emissions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Curtis D Jones.
Thesis: Thesis (Ph.D.)--University of Florida, 2013.
Local: Adviser: Fraisse, Clyde William.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2013
System ID: UFE0045362:00001


This item has the following downloads:


Full Text

PAGE 1

1 ASSESSING EFFICIENCIES IN VEGETABLE PRODUCTION: HYDROLOGIC MODELING OF SOIL WATER DYNAMICS AND ESTIMATION OF GREENHOUSE GAS EMISSION S By CURTIS DINNEEN JONES 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 2013

PAGE 2

2 2013 Curtis Dinneen Jones

PAGE 3

3 To Callie my family, my friends

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank the members of my supervisory committee : Clyde Fraisse, Lincoln Zotarelli, John Schueller, Chris Martinez, and Kelly Morgan, for their advice and guidance I would also like to thank Cheryl Porter and Jin Wu, for their collaboration an d hel pfulness with the DSSAT project. To Wayne Williams, my hole digging partner, thanks for all the help setting up and maintaining equipment, and for the good talks. I would like to acknowledge Monica Ozores Hampton for being an awesome collaborative par tner Thanks to Buck Nelson and the staff at the PSREU in Citra, for all their help setting up my field experi ment and keeping everything operational Thanks to Robin Robbins and Irrigation Mart for their contribution of the drip tapes used in this study, which was a big help getting this study started. Thanks t o m y colleagues, especially Anna Linhoss, the late Arun Jain, Ahmed Al Jumaili, Mackenzie Boyer, and Stacia Davis, for a ll the support and good company. Lastly, I want to thank all of my family and f riends, for always being there.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIS T OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Back ground ................................ ................................ ................................ ............. 15 Objectives ................................ ................................ ................................ ............... 20 2 ADDITION OF A TWO DIMENSIONAL WATER BALANCE MODEL TO THE DSSAT CSM ................................ ................................ ................................ ........... 22 Background ................................ ................................ ................................ ............. 22 Model Description ................................ ................................ ................................ ... 24 Soil Water Movement Calculations ................................ ................................ .. 24 Parameterization Methodologies ................................ ................................ ...... 27 Implementation within the DSSAT CSM ................................ ........................... 32 Summary ................................ ................................ ................................ ................ 34 3 EVALUATION OF A TWO DIMENSIONAL WATER BALANCE WITHIN THE DSSAT CSM ................................ ................................ ................................ ........... 38 Background ................................ ................................ ................................ ............. 38 Evaluation of the Parameterization Methodology ................................ .................... 40 Benchmark Comparison of the Soil Water Dynamics to HYDRUS 2D ................... 42 Comparison of the Soil Water Dynamics to Field Measurements ........................... 49 Summary ................................ ................................ ................................ ................ 60 4 DRIP IRRIGATION MANGEMENT IMPLICATIONS ................................ ............... 86 Background ................................ ................................ ................................ ............. 86 Irrigation Management Impacts ................................ ................................ .............. 88 Soil Moisture Sensor Con trolled Irrigation ................................ ............................... 93 Summary ................................ ................................ ................................ ................ 99

PAGE 6

6 5 QUANTIFICATION OF GREENHOUSE GAS EMISSIONS FROM OPEN FIELD GROWN FLORIDA TOMATO PRODUCTION ................................ ........... 108 Background ................................ ................................ ................................ ........... 108 Estimation of Greenhouse Gas Emissions ................................ ............................ 110 Desc ription of Typical Tomato Production System in Florida ......................... 110 System Boundaries ................................ ................................ ........................ 111 Input Data ................................ ................................ ................................ ....... 113 Calculations ................................ ................................ ................................ .... 114 Emissions Estimates ................................ ................................ ...................... 117 Implications ................................ ................................ ................................ .... 121 Summary ................................ ................................ ................................ .............. 128 6 CONCLUSIONS ................................ ................................ ................................ ... 135 APPENDIX A DSSAT 2D AND HYDRUS 2D GRID LAYOUTS ................................ .................. 141 B CODE FOR EXTRACTING PROBE VWC FROM GRIDDED DSSAT 2D VWC VALUES ................................ ................................ ................................ ................ 143 C TIME EVOLUTION GRAPHS OF THE MEASURED AND SIMULATED VWC VALUES AT EACH PROBE M EASUREMENT LOCATION ................................ .. 145 LIST OF REFERENCES ................................ ................................ ............................. 191 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 206

PAGE 7

7 LIST OF TABLES Table page 3 1 Error metrics for predicting SMRCs using the three point DSSAT methodology. ................................ ................................ ................................ ...... 63 3 2 Error metrics for predictin g SMRCs using the RETC methodology. ................... 63 3 3 Soil hydraulic parameters for the benchmark comparison of DSSAT 2D with HYDRUS 2D. ................................ ................................ ................................ ...... 63 3 4 Difference metrics and closeness of fit measures between VWC predictions by DSSAT 2D and HYDRUS 2D. ................................ ................................ ....... 64 3 5 Drip tape manufacturer specifications. ................................ ............................... 64 3 6 Field experiment treatment details. ................................ ................................ ..... 65 3 7 Soil particle size, bulk density, and particle density averages by soil depth. ...... 66 3 8 Overall DSSAT 2D error metrics from the average probe measurements. ......... 66 3 9 Overall DSSAT 2D adjusted error metrics from the average probe measurements. ................................ ................................ ................................ ... 66 3 10 DSSAT 2D error metrics from the average probe measurements for each treatment. ................................ ................................ ................................ ........... 67 3 11 DSSAT 2D adjusted error metrics from the a verage probe measurements for each treatment. ................................ ................................ ................................ ... 68 5 1 Diesel fuel use by machine operation for typical Florida tomato production practices. ................................ ................................ ................................ .......... 129 5 2 Agrochemical use for typical Florida tomato production practices. ................... 129 5 3 Plastic use for typical Florida tomato production practices. .............................. 130 5 4 Transportation practices for typical Florida tomato production practices. ......... 130 5 5 Crop production details for typical Florida tomato production. .......................... 130 5 6 Carbon emissions for the manufacturing, transporation, and storage of input materials or agrochemicals. ................................ ................................ .............. 130 5 7 Emission factors used for estimating N 2 O emissions from N fertilizer and lime. ................................ ................................ ................................ .................. 131

PAGE 8

8 LIST OF FIGURES Figure page 2 1 Cross section of half of a raised bed system as modele d by DSSAT 2D. .......... 36 2 2 Definitions of the DSSAT 2D grid cell dimensions and layout. ........................... 37 3 1 Distribution of VWC prediction residua ls using the three point DSSAT methodology. ................................ ................................ ................................ ...... 69 3 2 Distribution of VWC prediction residuals using the RETC methodology. ............ 70 3 3 B oxplot of VWC prediction residuals vs. tension using the three point DSSAT methodology. ................................ ................................ ................................ ...... 71 3 4 Boxplot of VWC prediction residuals vs. tension using the RETC methodology. ................................ ................................ ................................ ...... 72 3 5 Simulated VWCs for the Uniform 1 soil. ................................ ............................. 73 3 6 Simulated VWCs for the Uniform 2 soil. ................................ ............................. 74 3 7 Simulated VWCs for the Layered 1 soil ................................ ............................. 75 3 8 Simulated VWCs for the Layered 2 soil. ................................ ............................. 77 3 9 Plot of residuals betw een DSSAT 2D and HYDRUS 2D VWC predictions. ....... 79 3 10 The RMSD value over time between DSSAT 2D and HYDRUS 2D VWC predictions. ................................ ................................ ................................ ......... 80 3 11 A set of water content reflectometer probes installed with measurement rods inserted parallel to the length of the bed row. ................................ ..................... 82 3 12 Two dimensional representation of the 17 unique probe l ocations and sensing areas. ................................ ................................ ................................ .... 83 3 13 Average bulk density measurements vs. sampling date for each soil depth. ...... 84 3 14 Boxplots of the field measured SMRCs. ................................ ............................. 84 4 1 System response to changes in irrigation amount, rate, and application splitting. ................................ ................................ ................................ ............ 102 4 2 DSSA T 2D predicted water leaching from the root zone at different applications rates and application splitting methods. ................................ ........ 102

PAGE 9

9 4 3 DSSAT 2D predicted water leaching from the root zone under different scenarios. ................................ ................................ ................................ ......... 104 4 4 Impact of probe location on the efficacy of soil moisture sensor controlled irrigation systems. ................................ ................................ ............................. 104 4 5 Root zone VWC over time with different probe locations and thresholds for automated irrigation control. ................................ ................................ ............. 106 4 6 Impact of probe location on the best threshold value for sensor controlled irrigatio n systems. ................................ ................................ ............................. 106 5 1 Greenhouse gas emissions due to production, transportation, and storage of agrochemicals. ................................ ................................ ................................ 131 5 2 Greenhouse gas emissions due to farm machinery operations. ....................... 132 5 3 Greenhouse gas emissions from field losses ................................ .................. 133 5 4 Total greenhouse gas emissions estimates. ................................ ..................... 134 A 1 Grid layout for the HYDRUS 2D simulations that were used for comparison with the DSSAT 2D simulations.. ................................ ................................ ...... 141 A 2 Grid layout for the DSSAT 2D simulations that were used for comparison with the HYDRUS 2D simulations. ................................ ................................ ... 142 C 1 Time evolution at each probe location of the measured and simulated VWCs using DSSAT 2D with the three point DSSAT estimated soil parameters. ....... 145 C 2 Time evolution at each probe location of the measured and simulated VWCs using DSSAT 2D with the RETC estimat ed soil parameters. ........................... 168

PAGE 10

10 LIST OF ABBREVIATIONS T T ime step X G rid cell width Z G rid cell height C C arbon C EFF Nash Sutcliffe c oefficient of efficiency CH 4 M ethane CO 2 C arbon dioxide CO 2 EQ C arbon dioxide equivalent CSM C ropping systems model CVR M SE Coefficient of variation of the root mean square error C VRM S D Coefficient of variation of the root mean square difference D H ydraulic diffusivity DSSAT Decision Support for Agrotechnology Transfer DSSAT 2D The modified two dime nsional DSSAT CSM ET E vapotranspiration GHG Greenhouse gas HDPE Hi gh density polyethylene H IN ( I J ) H orizontal in flow to a grid cell in row i and cell j H OUT ( I J ) H orizontal out flow from a grid cell in row i and cell j K H ydraulic conductivity K A B ulk die lectric permittivity K S S aturated hydraulic conductivity LDPE L ow density polyethylene m Van Genuchten fitting parameter

PAGE 11

11 n Van Genuchten fitting parameter N Nitrogen N 2 O N itrous oxide NH 3 A mmonium NO 3 N itrate P P hosphorus PA P eriod average Q X H orizontal wa ter flux Q Z V ertical water flux R 2 C oefficient of determination RMSD R oot mean squared difference RMSE R oot mean squared error SMRC Soil moisture release curve SSD S um of squares difference V OUT ( I J ) V ertical in flow to a grid cell in row i and cell j V OUT ( I J ) V er t ical out flow from a grid cell in row i and cell j VWC V olumetric water content Van Genuchten fitting parameter N ormalized water content S oil water content 1500 W ater content at 1500 kPa of tension 33 W ater content at 33 kPa of tension D UL D rained upper limit water content LL L ower limit water content R R esidual water content; also Van Genuchten fitting parameter

PAGE 12

12 S S aturation water content; also Van Genuchten fitting parameter S oil tension E A ir entry pressure

PAGE 13

13 Abstract of Dis sertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ASSESSING EFF I CIENCIES IN VEGETABLE PRODUCTION: HYDROLOGIC M ODELING OF SOIL WATER DYN AMICS A ND ESTIMATION OF GREENHOUSE GAS EMISSIONS By Curtis Dinneen Jones May 2013 Chair: Clyde Fraisse Major: Agricultural and Biological Engineering The vegetable industry is economically important in Florida, with annual production valued at around 1. 8 billion dollars spanning 300,000 acres. As such, these crops tend to be intensively managed to ensure optimal yields. However, in an atmosphere of increasing population and resource scarcity, both in Florida and globally, there is increasing pressure for agricultural commodities to be produced efficiently. To this end, cropping systems models (CSMs) are useful tools for identifying efficient production practices as they can be used to help understand complex relationships between management, system driver s, production, and environmental consequences. In this r esearch the commonly used Decision Support System for Agrotechnology Transfer (DSSAT) CSM was amended to allow for water limited simulation under cultural practices common for vegetable production, in cluding drip irrigation. The modified model was evaluated to assess its utility in simulating soil water dynamics, and simulation experiments were conducted to identify implications of management choices. Analyses demonstrated the relative importance of di fferent drip irrigation management options as well as the importance of sensor placement for soil moisture sensor

PAGE 14

14 controlled irrigation. As an ad ded dimension, estimates of the greenhouse gas ( GHG ) emissions from typical tomato production systems were calc ulated. The estimates emphasize the importance of irrigation management, nutrient management, and productivity for minimizing these emissions.

PAGE 15

15 CHAPTER 1 INTRODUCTION Background The vegetable industry is an economically important industry in Florida, with a value of 1.88 billion dollars in 2008 (Olson & Santos, 2011) spanning 300,000 acres (Cantliffe et al., 2009). The management practices of this large industry have many potential environment implications. Vegetables in Florida are typically produced on s andy soils with low water holding capacities (Dukes et al., 2006), commonly with plastic mulched raised beds and either drip or seepage irrigation. The potential to attain high water use efficiencies (Simonne et al., 2004) and fertilizer use efficiencies ( Zotarelli Dukes et al., 2009 ) has been demonstrated within these intensively managed systems given proper management. However, improper management can result in excessive water use and excessive nutrient loss, which are wasteful economically for the produ cer, damaging to the environment, and a drain o n water resources. Water shortages have become an increasing concern globally as naturally occurring rainfall shortages are intensified by population growth, land use change, and et al., 2008). Vorosmarty et al. (2000) believe global water resources are already under significant stress based on the high ratio of water demand to sustainable water supply in many regions. Recent water shortages in the southeastern United States exemp lify this global trend as regional droughts between 2005 and 2007 resulted in water shortages in this fast growing region, which is expected to have the largest net population growth between 2000 and 2030 of any region in the United States (Nagy et al., 20 10). Water scarcity has long been an issue in the arid regions of the southwestern United States, but has become an increasingly

PAGE 16

16 prevalent problem in the more humid southeastern United States, particularly in Florida (Asano et al., 2007) despite average an nual rainfall of 114 to 152 cm (Smajstrla et al., 1999). Increased groundwater withdrawals in Florida have resulted in reduced river, lake, and aquifer levels throughout southwest Florida, saltwater intrusion into the Floridan aquifer in coastal Manatee, S arasota, and Hillsborough counties, and caused the Southwest Florida Water Management District (SWFWMD) to create water use caution areas and tighten water permitting (Romero et al., 2008; SWFWMD, 2006). Lowered groundwater levels have also led to increase d pumping costs, sinkhole formation and land subsidence (Molle & Berkoff, 2009), with pumping used for freeze protection of citrus and strawberries in Hillsborough County identified as causing many sinkholes in the area (Bengston, 1987). Agriculture has a significant role in water resource issue s Agriculture is the largest user of water in the world, with Bruinsma (2003) estimating that 70% of glo bal water use is for irrigation Irrigation is the second largest user of water in the United States behind the rmoelectric power generation accounting for 31% of all water withdrawals and 37% of freshwater withdrawals (Kenny et al., 2009). Irrigation also accounts for the largest withdrawal of groundwater in the United States at around 67% (Kenny et al., 2009). In Florida, irrigation is the largest user of freshwater, accounting for around 45% of all freshwater withdrawals, 62% of all fresh surface water withdrawals, and 35% of all groundwater withdrawals (Kenny et al., 2 009). Rosenzweig et al. (2004) suggest that improvements in irrigation and crop technology will play an integral role in meeting the expected future increases in wat er demand

PAGE 17

17 Another issue important in agriculture that impacts production and the environment is nitrogen loading. Nutrient loss from a griculture has been shown to be significant. A United States Geological S urvey (USGS) study (1998) conducted in Florida and Georgia demonstrated that 20% of water samples taken from surficial aquifers had nitrate nitrogen concentrations in exceedance of th e 10 mg/l United States Environmental Protection Agency (USEPA) drinking water standard, while 33% of samples collected from row cropping areas exceeded this standard. Nitrates are a concern as elevated levels can have adverse health and eco logical effects Ecologically, a causal relationship has been shown between high nitrogen co ncentrations and plant and algal blooms in surface waters (Rosen, 2003) as well as fish poisoning (Di & Cameron, 2002). The problem is widespread, as agricultural nitrate losses h ave been identified as a source related to hypoxic zones in the Gulf of Mexico (Burkart & James, 1999) and elevated nitrate levels in the Baltic Sea (Tiemeyer et al., 2010). From a public health perspective, c ontaminated drinking water can also cause s igni ficant human health impacts especially in infants, by inducing the potentially life threatening condition methemoglobinemia (Mueller & Helsel, 1996). Drip irrigation is a technology which offers the ability to supply water to crops with increased efficien cy, reducing water use as well as nutrient leaching compared to seepage irrigated systems (Dukes et al., 2010; Pitts et al., 1988; Sato et al., 2010) Micro irrigation comprises 45% of the irrigated land in Florida, with most of this area comprised of micr osprinklers used in citrus production while seepage irrigation comprises 44% of irrigated lands (Dukes et al., 2010). However, due to the potential benefits of drip irrigation compared to seepage, there is increasing interest in

PAGE 18

18 conversion of seepage irri gated systems to drip irrigation ( Dukes et al., 2010; Simonne Hutchinson et al., 2010). This could result in an increased prevalence of drip irrigation for vegetable production in Florida, and increases the need to understand the system dynamics under dri p irrigation to ensure the potential benefits of the technology are best realized. In order to reduce nitrogen losses from agricultural production, management practices need to be developed to increase efficien cy. Irrigation management is important for con trolling these losses as irrigation practices have a strong impact on nitrate movement in soils, and thus nitrate loading. Due to the climate in the southeastern United States, most soil nitrogen is in the nitrate form because other forms are rapidly conve rted (Jansson & Persson, 1982). Nitrate moves rapidly with water, and under drip irrigation it has been shown that most of the nitrate collects near the fringe of the wetting front (Li et al., 2003). This behavior is even more significant on sandy soils as the low water holding capacities and high hydraulic conductivities encourage more vertical movement of the wetting front compared to heavier textured soils. Thus, the potential benefit from improved irrigation efficiency is two fold as irrigation manageme nt has a large impact on both absolute water use and nutrient movement in soils. The use of automated soil moisture based irrigation has been shown to allow significant reduction in irrigation without reductions in yield (Smajstrla & Locascio, 1996; Dukes et al., 2003; Zotarelli et al., 2008; Zotarelli Scholberg et al., 2009), with some estimates that irrigation can be cut in half compared to once or twice daily fixed time irrigation approaches (Zotarelli Dukes et al ., 2009 ). However, the performance of t hese systems is dependent on decisions such as the sensor placement (Coelho & Or, 1996),

PAGE 19

19 the soil moisture threshold value (Coelho & Or, 1996; Zotarelli et al., 2010), and the irrigation volume that is applied once the threshold moisture content has been r eached (Zotarelli et al., 2010). Coelho and Or (1996) point out that while much work has been done to identify threshold soil moisture or matric values for optimal crop yields, most recommendations are general and empirical, ignoring site specific soil wat er dynamics. Soil water models can be used to estimate the spatio temporal dynamics of the system and thus make improved and more site specific recommendations. Another area for possible improvement in irrigation and fertilizer efficiency is the crop esta blishment phase of irrigation, during which irrigation is run for long durations after transplanting to encourage root development and prevent transplant shock (Simonne et al., 2004). The establishment of irrigation can account for a large portion of the s easonal nitrate leaching (Vazquez et al., 2005), especially when irrigation water is managed closely during the remainder of the season such that drainage is minimal (Zotarelli et al., 2007). This establishment phase has been identified as an area in which irrigation management efficiency can be improved (Schroder, 2006). Models can be used for estimating the irrigation durations necessary to maintain specific root zone soil moisture levels This would allow water to be supplied in order to sustain young pl ants with poorly distributed and shallow roots while not supplying excess water beyond the shallow rooted depths. The frequency necessary to maintain high moisture levels in this region would also be important, and consideration of the site management capa bilities would be necessary to determine if such irrigation frequencies would be feasible for real production applications

PAGE 20

20 Water use and nutrient losses need to be considered in the production of vegetables due to the effects they have on the environment and water resources locally and regionally. However, their impacts and the impacts of vegetable production as a whole also have global impacts through their influence on global GHG emissions. Atmospheric carbon dioxide levels are increasing, and the agricu ltural sector has been shown to account for 10 12% of an thropogenic GHG emissions (Smith et al., 2007). The Copenhagen Accord set a goal of keeping global temperature increases below 2 C c interference with ntion on Climate Change 2009). However it appears that these incr eases are likely. For example, Rogelj et al. (2010), predict that based on current emissions pledges, there is a greate r than 50% chance t hat an increase of more than 3 C will in fact occur. A reasonable first step for reducing GHG emissions from the agricultural sector is to quantify emissions from specific sources within agricultural production and identify the most eco nomically sensible actions for cutbacks. In similar production systems, irrigation and nitrogen fertilizer have been found to be among the largest sources of GHG emissions (Clyde Fraisse, personal communication, 2011). Thus, improvements in nitrogen and wa ter use efficiencies could have major impacts on GHG emissions. Objectives It has been shown that the management of vegetable production has wide ranging impacts stemming from its impact on water demand nutrient loss, and net GHG emission s Irrigation pr actices have an influence on all of these areas. Frequent, low volume irrigation is considered the best method for providing sufficient water to the crop

PAGE 21

21 while minimizing flow beyond the root zone (Locascio, 2005). However, while general recommendations ar e important, making site specific recommendations that are more considerate of particular site, soil crop, and management parameters can be beneficial Investigations either need to be done site specifically, which can be very costly, or models need to be created that are suitable for enhancing the details of such recommendations at lower cost. Therefore, the overall goal of this research was to create tools to help improve the efficiency of intensively managed vegetable production in order to reduce water use, nutrient leaching, and GHG emissions while maintaining competitive yields as well as to quantify the GHG emissions associated with typical tomato production practices The specific objectives of this research were : To a ssist in the addition of a two dimensional water balance model to the DSSAT CSM which is capable of accounting for plastic mulch, raised beds, and drip irrigation To e valuate the performance of the amended DSSAT CSM through a benchmark comparison with the HYDRUS 2D model To c onduct fiel d experiments to monitor the soil water dynamics under different drip tape specifications and management scenarios To u se the experimental data to evaluate the ability of the amended DSSAT CSM to predict soil water dynamics with different drip tapes and un der different drip irrigation management scenarios To i dentify tangible management implications from simulation experiments using the amended DSSAT CSM To estimate the GHG emissions associated with the production of open field grown tomatoes under typical Florida production practices To i dentify the most promising areas of production for reducing GHG emissions from open field Florida tomato production

PAGE 22

22 CHAPTER 2 ADDITION OF A TWO DIMENSIONAL WATER BALANCE MODEL TO THE DSSAT CSM Background Agricultural pro duction systems are complex, highly interactive systems. Crop models can be useful tools for understanding the relationships between crop soil, climate, management actions and the environment. An example of such a model is the DSSAT CSM, a widely used mo dular software package that integrates of principles from experts in various fields to provide a unified model of cropping systems. DSSAT has been used in many regions and instances for addressing irrigation and fertilizer management strategies (Jones et a l., 2003), making it a viable tool for improving the efficiencies in crop production. However, while DSSAT has been widely applied to a range of cropping systems, it is in fact poorly suited for representing water or nutrient limited production under the production practices typical for vegetable production in Florida. The typical Florida vegetable production practices which are not captured by DSSAT include the use of plastic mulch, raised beds, and drip irrigation. These production practices result in n on uniformities throughout the field that are not considered by the structure or water balance of DSSAT, and also include management options that are not available within DSSAT. Conversely, HYDRUS 2D is a commonly used software package which can simulate soil water dynamics under a wide range of boundary conditions. It has been shown to accurately predict soil water flow under drip irrigation regimes ( Provenzano, 2007; Roberts et al., 2009; Skaggs et al., 2004). However, unlike DSSAT, HYDRUS has a shortcom ing in that it does not include a crop model, and thus cannot be used to

PAGE 23

23 model interactions between management and crop growth and development. Additionally, HYDRUS is computationally intensive, requiring considerable time to complete simulations. The soi l water balance currently implemented in DSSAT (Ritchie, 1985; Ritchie, 1998) assumes one dimensional, vertical flow. It operates on a daily time step, and considers water applied uniformly as either irrigation or rainfall. Runoff is calculated using the S CS curve number approach ( United States Department of Agriculture, Soil Conservation Service, 1972) with modifications made by Williams et al. (1984) to account for layered soils and their moisture content, which was not considered in the original method. The soil is split into layers, with a maximum allowance of 20 layers, and each layer is chara cterized by its depth, saturation water content, drained upper limit water content, lower limit water content, and saturated hydraulic conductivity. Water which in filtrates the soil is added to the upper most layer, after which it is allowed to approach modified with an empirical hydraulic conductivity by which water drains from upper layers if the water content exceeds the drained uppe r limit. The rate of this drainage is computed based on a general DSSAT drainage parameter which is a constant at all depths ( Hoogenboom et al., 2003 ). This drainage rate is limited by the saturated hydraulic conductivity of that layer. While one dimension al modeling approaches have been shown to adequately represent water infiltration for rainfed, sprinkler, or flood irrigation systems (Brandt et al., 1971; Bresler et al., 1977), such an approach proves insufficient for drip irrigation systems. One reason for this is that under drip irrigation, the irrigation is applied non uniformly. Since a one dimensional reduction must assume uniform moisture

PAGE 24

24 distribution across the field, it fails to correctly represent the system. Also, since the frequency of irrigati on increases and the quantity of each irrigation decreases, the dynamics of the infiltration process become much more important. Thus the daily time step becomes inadequately coarse in most instances. These shortcomings of the one dimensional approach ari se because water movement under drip irrigation systems is truly a transient, three dimensional process (Cote et al., 2003). However, because of difficulties for DSSAT developers to integrat e a three dimensional model (Gowdish & Muoz Carpena, 2009) with in the DSSAT CSM a two dimensional vertical plane flow approximation was made where uniformity was assumed in the direction of the crop row. Such an assumption is reasonable for drip irrigated systems when emitters have a sufficiently small spacing (Bresler 1977; Warrick, 1985), which has been reported to as being between 16 and 50 cm (Skaggs et al., 2004; Wang et al., 2000). This two dimensional approximation has been made in many studies (Elmaloglou & Malamos, 2003; Rubin, 1968), and Schwartzman and Zur ( 1986) note that for row crops, emitters are typically spaced such that a fairly continuous ly wetted soil strip is created along the row, thus reasonably satisfying the assumption of uniformity in the row direction. The following will describe the theory be hind and implementation of a two dimensional water balance which was developed by a group of DSSAT developers, within the DSSAT CSM. Model Description Soil Water Movement Calculations The two dimensional w ater flow theoretically can be computed using Rich ards equation, which is the most widely used model for soil water transport (Pachepsky et al.,

PAGE 25

25 which was derived from saturated flow experiments, could be applied to unsatur ated flows if the saturated hydraulic conductivity parameter was replaced by the unsaturated hydraulic conductivity, which is a reduced value that is a function of the soil tension Richards equation is derived by combining the continuity equation with Dar law. It is the equation which most accurately describes unsaturated water flow (Gowdish & Muoz Carpena, 2009). The original form of Richards equation is referred to as the potential based form and is rep resented by equation (2 1), where is the soil water content (cm 3 /cm 3 ), K is the hydraulic conductivity (cm/h), is the soil suction head (cm), is a vector differential operator representing the gradient in space, t is the time (hr), and z is the vertical position (cm). (2 1) The potential based form is the most commonly used for numerical solutions, and it is applicable for variably saturated and layered soils (Hillel, 1998). However, this form requires short time steps and small spatial interv als (Hillel, 1998). Implementations of this form are computationally intensive, require extensive soil property data, and generally involve fine spatial and temporal discretization. Additionally, Issues of convergence and instability are often found, somet imes impeding a solution to specific sets of initial and boundary conditions. Therefore mathematical manipulations were made to create an alternate water content based form which circumvents some of the issues encountered using the potential based form. Th e form is represented by equations (2 2) and (2 3), where D is the hydraulic diffusivity (cm 2 /hr). (2 2)

PAGE 26

26 (2 3) The water content based form is less sensitive to the non linear relationship of t he soil hydraulic conductivity and pressure, and can tolerate much larger time and space intervals (Hillel, 1998). However, b ecause it is a mathematical manipulation and not a physical representation of the system, it is not applicable for saturated flow, layered soils, or heterogeneous soils (Hillel, 1998). Taking this into consideration, it was decided that in order to create an approximate, practical model, which would have faster computation al times, the water content based form would be implemented wit h some adjustments made to improve the performance under saturated and layered conditions. The shortcoming of this form of Richards equation with saturated conditions is due to the hydraulic diffusivity, which approaches infinity as the water content appro aches saturation. Therefore a hydraulic diffusivity limit of 417 cm 2 /hr was imposed such that the value would not exceed the maximum value found in practice (Hillel, 1998). The shortcoming of this form in dealing with layered soils, as described by Talbot et al. (2004), has to do with discontinuities in the moisture content of different soil layers when in fact the matric potential varies continuous ly This can result in erroneous flow predictions and potentially in predictions of flow in the opposite direc tion of the true flow, an illustration that water flow is dictated by the gradient of the total soil potential rather than by the difference in water contents. While this is a shortcoming of this form of the equation, in order to reduce the error, the norm alized water content was used in place as shown in equation (2 4), where s is the saturated water content ( cm 3 /cm 3 ), r is the residual water content ( cm 3 /cm 3 ), and is the actual water content ( cm 3 /cm 3 ). Thus by using to drive soil water flow, some of the error should be reduced when predicting

PAGE 27

27 flow between differently textured soil layers. The horizontal and vertical flow can then be calculated as described in equations (2 5) and (2 6), where q x is the horizontal flux ( cm/hr ) and q z is the vert ical flux ( cm/hr ). It should be noted that for homogenous soils, modeling soil moisture flow using the normalized water content is equivalent to using the actual water content. As such, horizontal flows are unaffected by the adjustment since soil hydraulic properties are assumed uniform within each layer. (2 4) (2 5) (2 6) Parameterization Methodologies In order to utilize Richards equation, it is necessary to p rovide the soil moisture release curve (SMRC) as well as the hydraulic conductivity and hydraulic diffusivity as functions of T here are many methodologies t hat can be used to relate ( ), K ( ) and D( ) Gardner and Mayhugh (1958) proposed an empirical e quation to estimate the hydraulic diffusivity which is shown in equation (2 7), where a and b are fitting parameters. While this form does generally approximate D( ), its unimodal shape poorly estimates D( ) for low values of (Hillel, 1998) and the model parameters lack a physical basis. (2 7)

PAGE 28

28 Gupta et al. (1974) estimated the hydraulic diffusivity as described in equation (2 8), where D s is the soil diffusivity at saturation and is a parameter with little sensitivity t o the soil pore size distribution. This method was shown to perform well for a range of soil textures (Reichardt et al., 1972). However, the D s and parameters are unknown and must be determined experimentally, which makes it a poor approach for the DSSAT 2D model. (2 8) An additional method used the soil water retention curve predicted using the Saxton and Rawls (2006) methodology to estimate the hydraulic diffusivit y using the defining equation of hydraulic diffusivity p reviously described by equation (2 3). However, the soil water retention was calculated as a piecewise function, and its derivative resulted in a piecewise function which was discontinuous. Ultimately the choice was made to use the Van Genuchten (1980) soi l water retention curve model and the Mualem (1976) pore size distribution model. These models were selected due to their wide use (Schaap & Leij, 2000; Tuller & Or, 2001; Kosugi et al., 2002; Ippisch et al., 2006). They have been found to be applicable to a wider variety of soils than other equations (Van Genuchten & Nielsen, 1985) and have been shown to predict soil water retention and unsaturated hydraulic conductivity better than other models on coarse sand and gravel (Mace et al., 1998). Certain models perform best for certain soil types and applications (Kosugi et al., 2002 ), but there is no model which performs best for all situations. However the Van Genuchten and Mualem models were chosen due to th eir wide use and flexibility for characterizing man y soil types and scenarios. The Van Genuchten parameters can be used to generate K,

PAGE 29

29 and D. The SMRC is described by equation ( 2 9) where s r m and n are the Van Genuchten fitting parameters A common restriction of setting m = 1 1/n was made b ecause it has been shown to perform better for most soils than any of the other common restriction s (Van Genucthen et al., 1991). The restriction was necessary because while the fitting is typically more accurate with no restriction s it can suffer from poo r convergence and correlated parameters with large uncertainty (Van Genucthen et al., 1991). With this restriction, K and D can be described by equations (2 10) and (2 11), where l is the pore connectivity parameter. (2 9) (2 10) (2 11) DSSAT allows a variety of soil properties to be input into the model. In order to operate the two dimensional water balance, users are allowed the option of entering the Van Genuchten pa ramete rs directly for each soil layer If the Van Genuchten parameters can be estimated from expert information or from fitting to SMRC data, this is preferred for improving the accuracy of simulations of the soil water dynamics. However, in order to ensur e that users with knowledge of only the standard soil properties for each soil layer can utilize the model, a methodology was developed to estimate the Van Genuchten parameters from the standard DSSAT soil property inputs. Among the standard soil property inputs to DSSAT are the soil lower limit of plant extractable water ( LL ), drained upper limit ( DUL ), and saturation ( s ) volumetric water

PAGE 30

30 conten ts as well as the soil texture and the sand, silt, clay, and organic matter soil fractions The LL is define d as the soil water content at which a crop can no longer extract water, while the DUL is defined as the soil water content at which a previously saturated soil reaches once it has been allowed to drain until drainage ceases. P ractically speaking, this is considered the point when the soil moisture content changes no more than 0.1% to 0.2% per day (Tsuji et al., 1994). The LL is therefore analogous to the permanent wilting point while the DUL is analogous to the field capacity. Thus in order to estimate using the standard DSSAT model inputs, several steps were necessary, with each of these steps being performed for each soil layer. First, the SMRC was derived using relationships reported by Saxton and Rawls (2006). Many equations have been proposed that relate soil water retention, soil texture, soil hydraulic properties, and other soil properties (Hillel, 1998; Rawls et al., 1992, Schaap & Leij, 2000). The Saxton and Rawls (2006) rela tionships were chosen d ue to their applicability to the DSSAT inputs, previous applications in agricultural hydrology ( Caruso et al., 2013; Looper & Baxter, 2011 ; Shrestha et al., 2010 ), and reports of good performance compared to other methods (Gijsman et al., 2002), altho ugh t he performance of various pedot ransfer f unctions often varies strongly from site to site and thus performance conclusions are difficult outside of the locations at which they have been evaluated (Baroni et al., 2009). Following the Saxton and Rawls (2006) methodology the tension at a particular moisture content ( ) (cm water) was calculated as described by equation (2 12) for tensions between 1500 kPa and 33 kPa and equation (2 13) for tensions between 33 kP a and air entry pressure ( e ) (kPa)

PAGE 31

31 where 33 is the water content at a tension of 33 kPa (cm 3 /cm 3 ), and 1 500 is the water content at a tension of 1500 kPa (cm 3 /cm 3 ) The air entry pressure was calculated using relationships reported by Brooks and Corey (1966) and are described by equation s (2 14) and (2 15) where is a fi tting parameter DUL is the tensio n at DUL DUL is assumed to be10 kPa for coarse ly textured soils and 33 kPa for all other soils based on the recommendations of Cassel and Nielsen (1986). r was estimated using the ratios of r to LL reported by Gowdish (2009) for 11 USDA textural clas sifications and the user inputted value of DUL (2 12) (2 13) (2 14) (2 1 5 ) In order to extract the Van Genuchten parameters from the est imated SMRC, 38 points from the SMRC were computed and used as inputs for fitting the Van Genuchten parameters using the RETC code (Van Genuchten et al., 1991). The code uses linear least squares optimization to fit the retention data to the soil water function described by equation (2 9). Once the fitting process is completed for each soil layer, the estimation of the Van Genuchten parameters is complete

PAGE 32

32 Implementation within the DSSAT CSM T o implement the soil water movement calculations, the two dimensional modeling space had to be discretized. It was decided that the modeling space would be split in to a two dimensional grid cell as depicted in Figure 2 1. The water balance assum es the water content to be uniform along the row length, and thus to only vary with distance perpendicular to the length of the row and depth. Additionally, due to symmetry only half of the bed is simulated because it is assumed that the other half will be a mirror. The model allow s additional management capabilities, including the ability to input the depth and wi dth of a raised bed, the option of applying plastic mulch to the bed and the option of applying irrigation through drip tape with adjustable spe cifications The model delineates the modeling space as either being within the raised bed, under the raised bed, within the furrow, or under the furrow. Cells are then created to fill the modeling space, with smaller cell sizes being used within the raise d bed and at the soil surface. This method was implemented in order to ensure finer discretized cells were placed in locations where greater soil water dynamics is expected, and coarser cells were placed in locations with less soil water dynamics. The mini mum c ell size was 5 cm by 5 cm. In order to operate the water balance, two dimensional flux equations described by equations (2 5) and (2 6) had to be expressed in the context of the two dimensional grid cell. Figure 2 2 depicts the cell layout cell geom etry, and flow s necessary for updating the water balance to a new time step. Within this layout, the horizontal and vertical fluxes can be calculated as described by equations (2 16) and (2 17), where i is the cell row, j is the cell column, z is the cell height (cm), and x is the cell width (cm)

PAGE 33

33 T he horizontal (H out(i,j) ) and vertical (V out(i,j) ) flows can then be cal culated as described by equations (2 18) and (2 19) where t is the time step (hours) The inflows are then related to the outflows as de scribed by equations (2 20) and (2 21). No lateral flow is allowed across the center of the bed, the edge of the bed, or the edge of the furrow. Vertical flow is computed in the deepest soil layer by fixing D( ) (i,j) at zero, limiting deep percolation to g ravitational flow at the limit of the soil extent A special cell is the cell directly beneath the drip emitter. All drip irrigation water is input into this cell, which stores all applied irrigation un til it flows out of the cell. (2 16) (2 17) (2 18) (2 19) (2 20) (2 21) The model oper ates on a variable time step, which is computed based on the dynamics within the system to assure stability and convergence. The minimum required time step is computed for each cell according to equation (2 22), and the overall minimum from all the cells i s accepted as the required time step. Potential evaporation and transpiration demand are computed on a daily basis, using the original DSSAT

PAGE 34

34 methodologies (Hoogenboom et al., 2009) with the daily demand being distributed across the day based on the distri bution of solar radiation throughout a day. Actual evaporation depends on surface cover and water content. Actual root water uptake is limited by water content, root density, root distribution, and soil properties. Root growth is also computed on a daily b asis according to standard DSSAT methodologies, with some adjustments to account for the added dimension. Root growth only occurs in the vertical direction in the first column. Once root growth in a row of the first column is fully penetrated with roots, h orizontal growth begins to occur in that row. Root growth factors within the soil represent soil impacts on root growth, while horizontal and vertical root growth rates provided by the user adjust Daily rainfall infilt ration is computed using the Soil Conservation Service runoff curve number method (United States Department of Agriculture, Soil Conservation Service, 1972). The daily infiltration is distributed across the day and evenly applied to non mulched surface cel ls, with rainfall on mulched areas being routed to the furrowed area. At the end of each time step, the water content in each cell is updated, considering vertical and horizontal inflows and outflows, infiltration, and root water uptake. (2 22) Summary A two dimensional water balance was added to the DSSAT CSM by a group of DSSAT developers using the Richards equation to drive soil water flow and adapting the standard DSSAT methods for estimating evaporation, root water uptake, infiltration, and root growth within a two dimensional framework. Additional management options

PAGE 35

35 were added to DSSAT to allow the model to account for management practices common for vegetable production. An approximate form of the Richards equation was cho sen in order to reduce the computational time required for model simulations. This increases the utility of the model for analyses which require large quantities of model runs, such as parameter estimation, uncertainty analysis, sensitivity analysis, large scale GIS applications and detailed simulation experiments. The model provides a detailed water balance model within the DSS AT modeling framework, allowing investigations of re lationships between crop growth and development soil water dynamics and mana gement actions within these types of production systems.

PAGE 36

36 Figure 2 1. Cross section of half of a raised bed system as modeled by DSSAT 2D Image created by and used with permission of Cheryl Porter.

PAGE 37

37 Figure 2 2 Definitions of the DSSAT 2D grid cell dimensions and layout. Image created by and used with permission of Cheryl Porter.

PAGE 38

38 CHAPTER 3 EVALUATION OF A TWO DIMENSIONAL WATER BALANCE WITHIN THE DSSAT CSM Background As populations grow, wate r resources become increasingly limiting, and atmospheric GHG levels rise, there is increasing pressure for agriculture to produce food more efficiently and with less environmental effects. Thus it is important that tools and methods be developed that help the understanding of relationship s between management actions system drivers, production, and environmental loadings within these production systems Crop models can be useful tools for understanding these relationships and interactions. Vegetable produ ction is an important industry in Florida, both economically and environmentally, in which a lot of efforts have been invested to improve production efficiency. However, there are currently no models available that can sufficiently simulate the soil water dynamics and crop growth and development dynamics under the production practices typical of these production systems. HYDRUS 2D is a commonly used software that can simulate soil water dynamics under a wide range of boundary conditions, including those ty pical of Florida vegetable production systems. It has been shown to accurately predict soil water flow under drip irrigation regimes (Provenzano, 2007; Roberts et al., 2009; Skaggs et al., 2004). However, HYDRUS 2D does not model crop growth or development T hus HYDRUS 2D cannot be used to model interactions between management and crop growth and development. Additionally, HYDRUS 2D is computationally intensive, requiring considerable time to complete simulations which limits its utility for analyses requi ring large numbers of model runs

PAGE 39

39 The DSSAT CSM (Hoogenboom et al., 2009) is a commonly used software that can simulate soil plant atmosphere interactions. Model runs are not computationally intensive, and it has been used in many regions and instances fo r addressing irrigation and fertilizer management strategies (Jones et al., 2003). However, it is limited in its ability to function under the management practices typical of these production systems. Therefore the DSSAT CSM was amended such that it could sufficiently characterize the soil water dynamics occurring under management practices typical for Florida vegetable production. The modeling approach was more approximate compared to HYDRUS 2D in order to allow simple input requirements and shorter compu tational times. To this end, in comparison to HYDRUS 2D, the amended DSSAT CSM operated with a coarser grid size, an approximate form of the Richards equation, and the option of calculating the necessary soil hydraulic parameters from only the standard DSS AT soil inputs. However, this DSSAT 2D model requires evaluation to determine its ability to accurately characterize the soil water dynamics of these systems. Thus, in order to evaluate the DSSAT 2D model, three approaches were used First, the methodolog y for estimating the soil hydraulic parameters from the standard DSSAT inputs was evaluated. Second, the model was assessed through a benchmark comparison against HYDRUS 2D for different theoretical soil profiles Finally, field experiments were conducted in order to evaluate the model s ability to predict the measured soil water dynamics in a field setting

PAGE 40

40 Evaluation of the Parameterization Methodology In order to evaluate the ability of the model to estimate the SMRC using only the standard DSSAT soil inputs, the three point approximation methodology was evaluated against measured SMRC data from the Florida Soil Characterization Database ( University of Florida Department of Soil and Water Science, 2012). Soil profiles were accepted from the database for the evaluation if they were a mineral soil, had SMRC SAT value greater than the DUL value, and DUL value greater than the LL value. The average measured SAT DUL LL values as well as the soil texture were used as inputs for th e approximation procedure. DUL was estimated by the volumetric water content (VWC) at a tension of 10 0 cm water for coarsely textured s oils and the VWC at a tension of 345 cm water SAT was e stimated by the VWC at a tension of 3.5 cm water, LL was estimated by the VWC at a tension of 15295.7 cm water. The SMRC s in the database consisted of soil moisture measurements at tensions of 3.5, 20, 30, 45, 60, 80, 100, 150, 200, 345, and 15295.7 c m water. Thus, once the SMRC was estimated by the DSSAT approximation method for a soil sample, the predicted VWC s at the measured tensions were compared to the measured values to assess the prediction ability of the method. The calculated error metrics ar e listed in Table 3 1. To further evaluate the DSSAT approximation methodology, a benchmark comparison was conducted, comparing the DSSAT approximation methodology to the commonly used hydraulic parameter estimation software RETC (Van Genuchten et al., 19 91). The RETC software can fit a number of different parameter model types, but the Van Genuchten model was chosen with the common restriction of fixing m = 1 1/n because the parameter estimation process with this restriction has been shown to

PAGE 41

41 perform be tter for most soils than with any of the other common restriction s (Van Genucthen et al., 1991). The RETC method method (Marquardt, 1963) to perform a non linear least squares optimization to fit the retention data to the soil water retention function described by equation (2 9). Thus, the RETC method accepts all the SMRC data available, compared to the average values at the three soil tensions used by the DSSAT three point approximation method Once the model is fit, t he predicted VWC s at the measured tensions from the database were compared to the measured values to assess the prediction ability of the method. The calculated error metrics are listed in Table 3 2. Inspection of the error metrics in Tables 3 1 and 3 2 r eveals a clear difference between the three point DSSAT methodology and the RETC methodology with the RETC methodology performing demonstrably better This is to be expected, as the RETC approximation method has the advantage of utilizing a greater amount of input data for fitting the model to that data Both methods have smaller prediction errors and better goodness of fit measures for the coarser soils, although the CVRMSE followed the inverse of this pattern, indicating some of this difference is due to the difference in magnitude of the VWC values among textures, which are higher in heavier textured soils and thus tend to increase error amounts. Figures 3 1 and 3 2 demonstrate that both methods are mostly unbiased, with the VWC prediction residuals appr oximately normally distributed with means very near zero. The figures also further confirm the shortcoming of the three point DSSAT method, with the residuals having greater spread and thus greater magnitudes Figures 3 3 and 3 4 show boxplots of the VWC p rediction residuals at each soil tension for the two parameter estimation methods. These figures

PAGE 42

42 explain the reason for the poorer prediction abilities of the three point DSSAT method. While the RETC method demonstrated fairly consistent boxplots with medi ans near zero at all tensions, the DSSAT method demonstrates good predictions at tensions near the three input tensions of 0, 345, and 15295 cm water, but has VWC prediction residuals straying noticeably from zero at tensions distant from the inputted tens ions. This confirms that the difference in model abilities stems from the inputs not being supplied over a sufficient range of tensions for the three point DSSAT approximation method Benchmark Comparison of the Soil Water Dynamics to HYDRUS 2D In addition to the parameterization methodology evaluation, the two dimensional DSSAT water balance model was further evaluated through a benchmark comparison with HYDRUS 2D ( Simunek et al., 2006) soil water dynamics simulations. HYDRUS 2D computes water flow using t he Galerkin finite element method with linear basis functions to numerically solve the Richards equation for variably saturated flow. The modeling space is discretized into triangular and quadrilateral elements with nodes at each corner. Users can control the size of the elements, and are encouraged to use smaller elements in regions expected to have large hydraulic gradients. The element size is typically orders of magnitude smaller than the elements used in DSSAT 2D HYDRUS 2D allows flexible boundary con dition s and modeling space geometries, thus enabling it to be applied to typical vegetable production systems. For simulating half of a raised bedded system, one notable difference between the HYDRUS 2D and DSSAT 2D approaches is that the shape of the surf ace drip flux boundary can be customized in HYDRUS 2D as a quarter circle of a specified radius, whereas DSSAT 2D applies water from the emitter to the element beneath the emitter mixing the water uniformly to this element Each time step evolves through an iterative process designed to keep the

PAGE 43

43 absolute change in pressure head between iterations less than a small tolerance value. The length of each time step is bounded by minimum and maximum time limits and is a function of the number of iterations requi red to solve the previous time step. Thus comparisons between DSSAT 2D and HYDRUS 2D can be used to understand the combined impact s of the Richards equation form, numerical solving technique, element size, element distribution, and drip emitter boundary s hape on soil water dynamics. In order to compare the models, each model was run using different soil profiles to assess the relative model performance for differently textured uniform and layered soils. Analyses were conducted for four soil profiles consis ting of two uniform soils and two layered soils. The uniform soils were a sand and a sandy clay textured soil chosen from the Florida Soil Characterization Database (University of Florida Department of Soil and Water Science, 2012) to represent a coarse an d fine textured soil. The two layered soils were combinations of the uniform soils, with one soil constituting the upper 20 cm and the other soil constituting the remainder of the soil profile. The hydraulic parameters characterizing these soil profiles ar e listed in Table 3 3. For uniform soils, the two forms of the Richards equation theoretically should essentially function equivalent ly for unsaturated flow, minimizing the importance of the equation form as a source of deviation between the model s The si mulation space and boundary conditions were set up to be equivalent for the DSSAT 2D and HYDRUS 2D model s, with the exception of the emitter boundary condition with an 80 cm wide by 205 cm deep soil space. The 80 cm horizontal surface was set as a zero fl ux boundary to represent a layer of impermeable plastic mulch and to remove the effects of ET estimations from the model comparison The 200 cm vertical edges were set as zero

PAGE 44

44 flux boundaries due to symmetry The 80 cm horizontal bottom edge was set as fre e drainage. For HYDRUS 2D, the drip emitter flux boundary was represented by a quarter circle with a 1 cm radius. Since DSSAT 2D and HYDRUS 2D operate with different element size s, shape s an d layout s it is not possible to compare the model results direc tly. The grid layout for both models can be seen in Appendix A. Thus for both models the simulated VWC values from all the elements were interpolated onto a uniform grid via linear point kriging using Surfer ( Golden Software Inc., 2011 ) software with a s lope of one, anisotropy ratio of one, and anisotropy angle of zero Contoured values were allowed to vary between 0.05 and 0.45 cm 3 /cm 3 with an interval of 0.001 cm 3 /cm 3 The grid was split such that there was a row and a column every two cm. Thus the gri d split the simulation spa ce into 100 rows and 40 columns for the creation of a total of 4 000 points that were evenly spaced throughout the grid. In this manner the points created from the DSSAT 2D and HYDRUS 2D simulations could b e directly compared at each time step However, since much of the soil will receive no changes in water content, the similarity between the models could be overestimated due to comparison of points in the simulation space that remain static during the entirety of the simulation period. To make a more informative comparison, the extent of the wetting front was calculated from each model. This was done by calculating the deepest grid node at each node distance from the row center that increased from its initial value by at least 0. 001 cm 3 /cm 3 The greate st depth at each distance calculated from the two models was used to determine the comparison points, such that if one model had a very different shape from the other model, the

PAGE 45

45 union of the space of the two wetting fronts would be c onsidered in the comparison calculations. For each analysis, the soil was set to an initial soil tension of 3 36.5 cm water. Simulations were then conducted for a five day period. Water was applied daily for four days beginning at 1200 and being applied f or 30 minutes at a rate of 1.617 lh 1 m 1 For DSSAT 2D, this was expressed as a drip tape with an emitter spacing of 30.5 cm and an emitter rate of 0.1367 ml/s. For HYDRUS 2D, this was expressed as a line source represented by a quarter circle with a one c m radius and a flux rate of 123.34 cm/day. After the fourth i rrigation event, the irrigation was cutoff for the remainder of the simulation period to allow soil water redistribution To analyze the difference s in soil water dynamics between the two models the simulated VWCs were compared at the nodes in the dynamic zone of the two models. The initial 12 hours were excluded from the comparison calculations because this was before the start of the first irrigation event and thus the soils were assumed to be at steady state. Comparisons were made at a 15 minute temporal resolution from 0.5 days to 5 days. This temporal resolution and time period were chosen in order to capture the dynamics of the system, including the rapid changes near the drip tape during i rrigation events and the slow changes over time during soil water redistribution The overall results of the analyses for the four soils are listed in Table 3 4. For the Uniform 1 soil, the DSSAT 2D simulations were very similar to the HYDRUS 2D simulation s. The RMSD between the two models was very small at 0.0010 cm 3 /cm 3 and the correlation coef ficient was very large at 0.985 The contour map in Figure 3 5 illustrates the two dimensional model predictions at four different times, providing a

PAGE 46

46 visual demons tration of the model similarity at each time. Analysis of the Uniform 2 soil, which is a heavier textured soil, also indicates strong similarity between the two models with a very small RMSD of 0.0013 cm 3 /cm 3 a very large c orrelation coefficient of 0.993 and visual similarity at four time steps as seen in Figure 3 6. However, while the two models created very similar simulations for the uniform soils, a greater deviation was observed for the layered soils. For the Layered 1 soil with the DSSAT 2D model b eing driven by the normalized water content, the RMSD was markedly higher at 0.0228 cm 3 /cm 3 and the correlation coeff icient was notably lower at 0.958. V isual observation of the two dimensional VWC snapshots in Figure 3 7 revealed strong differences betwee n the DSSAT 2D and HYDRUS 2D simulations. Primarily, the DSSAT 2D model failed to maintain the strong difference in VWC between the two layers that the HYDRUS 2D model predicted due to the sharp difference in soil properties. As such, the DSSAT 2D model pr edicted excessively high moisture in the sandy soil at the top of the profile and excessively low moist ure in the clay soil in the remainder of the profile, especially near the interface of the two soils. This vertical movement of water was not associated with the drip irrigation events, but rather driven by the water content differential between the differently textured soils, which is why water movement was seen all the way out to 80 cm from the row center in the layered soils while neither of the uniform soils had movement beyond 59.5 cm from the row center For the Layered 1 soil with the DSSAT 2D model driven by the actual water content, the deviations were greater than with the normalized water content drive n model. The RMSD rose to 0.0253 cm 3 /cm 3 th e correlati on coefficient dropped to 0.947

PAGE 47

47 and the visual differences in Figure 3 7 were greater. Figure 3 7 again revealed a difficulty for the DSSAT 2D model to maintain the difference in VWC between the two layers, with an increased amount of vertical water flow that was not simulated by the HYDRUS 2D model. This beha vior was even more pronounced for the actual water content driven DSSAT 2D model than the normalized water content driven DSSAT 2D model. The analysis of the Layered 2 soil was similar to t he Layered 1 soil analysis, but with a greater agreeance with HYDRUS 2D. The normalized water content driven DSSAT 2D model had a smaller RMSD of 0.0157 cm 3 /cm 3 a greater coef ficient of correlation of 0.975 and greater visual similarity in Figure 3 8. Th e actual water content driven DSSAT 2D mo del had a greater RMSD of 0.0176 cm 3 /cm 3 a smaller coef ficient of correlation of 0.968 and less ability to maintain a VWC differential between the soil layers compared to the normalized water content driven DSSAT 2D simulations in Figure 3 8. In order to better understand the nature of the model differences, residuals were computed and plotted (Figure 3 9). The residual plots first demonstrate a clear deterioration in agreement between the HYDRUS 2D and DSSAT 2D V WC predictions for the layered soils. The prediction residuals are distributed much more closely to zero for the uniform soils than the layered soils Further, this fit is much more consistent across VWC values. While the Uniform 1 soil does show some unde r predictions at some of the higher VWC values, which is likely a function of the differences in grid sizes between the two models directly under the emitter during irrigation events both uniform soils have similar distribution s of residuals at all VWC va lues. Conversely, in all analyses the layered soils have much larger prediction differences, with consistent over

PAGE 48

48 predictions at lower water contents and consistent under predictions at higher water contents. This trend is indicative of the DSSAT 2D model predicting flow of water from soil of high VWC to soil of low VWC at the interface of the differently textured soils, regardless of the orientation of the two layers, and is further indication of the inability of the water content based form of the Richard s equation to accurately predict soil water dynamics for layered soils. The use of the normalized water content as the system driver compare d to the actual water content resulted in some reduction in prediction residuals. However, the general pattern of er roneous flow predictions for sharply layered soils persisted despite this amendment. F urther to analyze the differences between the models over time, the RMSD was calculated at each time step ( Figure 3 10 ) It is first apparent that the RMSD values for t he uniform soils are drastically smaller than for the layered soils. Further, for the uniform soils the model differences peak during irrigation application, likely due to the different emitter boundary conditions for the DSSAT 2D and the HYDRUS 2D models and the greater difference in grid size between the two models near the emitter These differences then lessen during water redistribution as the soil water distributions of the two models become increasingly similar. This indicates that the models are dri ving the soil water dynamics to a similar steady state if no further perturbations occur within the system. For the layered soils, however, the errors are the least at the beginning of the simulation, increasing during the simulation until the lack of dyna mics in the system allows the error to remain at or near the maximum difference. For these layered soils, the system is largely insensitive to the irrigation events, as the error is being caused by the predicted water flow between the two layers. This is t rue for both the actual and

PAGE 49

49 normalized water content driven models, although the errors are greater for the actual water content driven model. This is an indication that the DSSAT 2D model is consistently forcing the system to a different steady state than the HYDRUS 2D model. It is also notable that while the DSSAT 2D model driven by the normalized water content has less error compared to the actual water content driven model, this difference is much more pronounced early in the simulation period. Later in the simulation period, the errors become increasingly similar Thus it appears that while driving the DSSAT 2D model with the normalized water content is an improvement compared to using the actual water content as the system driver, the improvement is t he result of slowing the speed with which the erroneous water flow between the layers is predicted. Both methods, however, eventually result in a similar steady state water distribution which fails to maintain the desired distinct difference in water conte nt between layers. It is thus apparent that for sharply layered soils, the form of Richards equation used in the DSSAT 2D model forces considerably different soil water flow predictions than the HYDRUS 2D model. While t he use of the normalized water conten t as the driver of flow in the DSSAT 2D model reduces these differences noticeably the differences are still considerable. Comparison of the Soil Water Dynamics to Field Measurements To further evaluate the model performance against field measured data, a field experiment was conducted at the Plant Science Research and Education unit in Citra, FL. A plastic mulched raised bed was constructed with an 80 cm width and 20 cm height Additional p lastic was laid to extend 1.2 m from the edge of each side of the bed in order to reduce the impact of rainfall on the soil moisture pattern and thus better isolate the soil moisture dynamics to being controlled by the drip tape specifications and

PAGE 50

50 irrigation management rather than ET and rainfall The main drip tape spe cifications and irrigation management variables in the study were flow rate, emitter spacing, number of daily irrigation applications, and daily irrigation amount. Different drip tapes were installed to obtain a variety of flow rates and emitter spacings but within the restriction of drip tape availability The six tapes selected for the experiment ( Table 3 5 ) consist of three emitter spacings with two emitter rates at each spacing There were two treatments for e ach drip tape with one treatment applying the entire daily irrigation amount in one application starting at 1200 and the other applying the same daily irrigation amount in two split applications starting at 1000 and 1400 Thus, the whole experiment consisted of 12 unique treatments. Following comp letion of the 12 treatments, half of the treatments were repeated. Each of the first 12 treatment s was conducted with four days of operation, followed by four days without irrigation This was done to allow enough days of irrigation for the soil moisture to adjust to a treatment, and enough non irrigated days for the following treatment to begin at a similar soil moisture state as the preceding treatment. Several of the repeated treatments were applied for additional days to allow for a longer model evalua tion period. Daily irrigation amounts were set at 5 mm, or 3 .8 l/m per day Thus, the duration of irrigation events were set such that each drip tape would apply this amount daily assuming application at the manufacturer specified flow rate. However, flow rates were measured in drip pans for each drip tape to more accurately estimate their flow rates under the experimental conditions As such, the daily irrigation amounts varied between drip tapes. The experimental details for each treatment can be seen in Table 3 6. Water

PAGE 51

51 use was measured using a DLJSJ50C water meter (Daniel L. Jerman Co., Hackensack, NJ) with a 1 gallon pulse output that was recorded using an H07 002 04 HOBO event logger (Onset Computer Corporation, Bourn, MA) Manual water meter readings were recorded with at least a weekly frequency, and manual readings were always taken at the start and end of a treatment Soil moisture was monitored by three sets of soil moisture probes installed at three distances along the row. The first set of probes were installed at the beginning of the experiment, with 12 CS616 (Campbell Scientific, Logan, UT) water content reflectometers installed at depths of 7.5, 22.5, 37.5, and 70 cm and at lateral distances from the row center of 0, 15, and 30 cm. The probe me asurement rods were installed parallel to the row length, with an installed se t of probes shown in Figure 3 11 The probe measurements were recorded every 15 minutes using a CR10X (Campbell Scientific, Logan, UT) datalogger with the measured probe period average (PA; s) converted to VWC by equation (3 1) (Campbell Scientific, 2011). This method is reported to provide accurate VWC conversions for mineral soils with bulk densities less than 1.55 g/cm 3 clay contents less than 30%, and bulk electrical conduc tiv it ies less than 0.5 dS/m (Campbell Scientific, 2011). After each experimental treatment had been conducted once, two sets of 16 CS650 (Campbell Scientific, Logan, UT) water content reflectometer probes were installed with each set being installed at di fferent distances along the bed row. From each set of probes, 12 were installed at depths of 7.5, 22.5, 37.5, and 52.5 cm and lateral distances from the row center of 0, 15, and 30 cm. The remaining four probes from each set were installed at depths of 70 and 100 cm and at lateral distances from the row center of 0 and 15 cm. The probe measurements were

PAGE 52

52 recorded every 15 mi nutes using a CR1000 (Campbell Scientific, Logan, UT) datalogger with the measured bulk dielectric permittivity (K a ) converted to VWC v ia Topps equation (Topp, 1980), as shown in equation (3 2) which has been shown to work well in most mineral soils. Both the CS616 and CS650 probes have 30 cm measurement rods and a 7.5 cm sensing radius. A two dimensional view of the 17 unique probe loca tions and approximate sensing volumes is shown in Figure 3 12 (3 1) (3 2) In order to characterize the soil, several soil properties were measured. The soil was characterized as a Gainesville loamy sand ( Soil Survey Staff 2011) Bulk density measurements were taken at three different dates during the experiment at depths of 7.5, 22.5, 37.5, and 70 cm and lateral distances from the row center of 0, 15, and 30 cm. A total of 12 samples were collected for each unique sampling location with the exception of the 70 cm depth, for which only six samples were collected during the first sampling date since this depth was undisturbed by the bed formation process and thus not expected to change over time. Sampl es consisted of undisturbed soil cores of 5.4 cm diameter and 3 cm depth with the bulk de nsity determined according the f ield core method (Klute, 1986) The average bulk density from each measurement date and soil depth is shown in Figure 3 13 The averag e bulk density from all the measurements was 1.55 g/cm 3 Bulk density values remained fairly consistent over time, with an increase from the May 22 sampling at the 22.5 cm depth as the only marked temporal change. The 7.5 cm and 70 cm depth had nearly iden tical bulk densities that were

PAGE 53

53 considerably less than the middle sampled depths. This is indicative of a plow pan layer somewhere above the 70 cm depth. The particle density was determined by the Pycnometer method (Klute, 1986) using the May 22 samples, w ith four samples being used for each sampling depth. Results are shown in Table 3 7. The average particle density for all the samples was 2.63 g/cm 3 It was apparent that the particle density was very consistent at all the sampling depths. The particle si ze distribution was determined according to the USDA Soil Survey Lab Method (Soil Survey Staff, 2004) using four samples for each sampling depth from the May 22 sampling. Results are shown in Table 3 7. The overall average particle size distribution consis ted of 91.3 percent sand, 5.5 percent silt, and 3.2 percent clay. The particle size distribution was also quite uniform with depth. Finally, SMRCs and saturated hydraulic conductivities were determined u sing six samples from a 10 cm depth and five samples from a 45 cm depth taken on June 30 Soil samples were saturated with de ionized water for 72 hours, then placed in Tempe cells (Soil Measurement Systems, Tucson, Arizona) that were connected to adjustable water columns for soil moisture determination at a range of pressures per Klute (1986). Moisture determina tions were made at tensions of 10 102, 183, 326, 571, 1019, 3220, 10190, and 15285 cm water. Samples were then re saturated and the saturated hydraulic conductivities were determined according to Kl ute (1986). The geometric average of the saturated hydraulic conductivities was 25.4, 19.3, and 22.4 cm/hr for the 10 cm samples, 45 cm samples, and all samples, respectively, with average values of 26.1, 19.6, and 23.1 cm/hr. The r esults of the SMRC measu rements are shown in

PAGE 54

54 Figure 3 14 The average saturation water content, drained upper limit water content, and soil lower limit water content were 0.489, 0.172, and 0.058 cm 3 /cm 3 for the 10 cm soil samples, 0.469, 0.170, and 0.060 cm 3 /cm 3 for the 45 cm, an d 0.480, 0.171, and 0.059 cm 3 /cm 3 for all the samples. Both the RETC and the three point DSSAT parameter estimations methods were used to estimate the Van Genuchten parameters from the SMRCs. For the 10 cm samples, the RETC method 1 r of 0.062 cm 3 /cm 3 while the three point DSSA T cm 1 an n of 1.58 r of 0.0 35 cm 3 /cm 3 For the 45 cm samples, the RETC code estimated 1 an n of 2 r of 0.061 cm 3 /cm 3 while the three point DSSAT cm 1 r of 0.036 cm 3 /cm 3 For all the samples t he R cm 1 an n of 2. 2 5 and a r of 0.062 cm 3 /cm 3 while t he th ree 0432 cm 1 an n of 1.57 r of 0.036 cm 3 /cm 3 The DSSAT 2D model was run for each treatment using parameter estimates from each of the parameter estimation methodologies. The Van Genuchten parameters and the saturated hydraulic conductivity in the first 20 cm of the soil profile were set in the DSSAT soil file to be equal to the values estimated from the 10 cm samples The parameters in the DSSAT soil profile from 20 to 35 cm depths were set based on the valu es estimated from all the samples, as this was in between the 10 and 45 cm sampling depths. Finally, the parameters in the DSSAT soil profile from 35 cm to the bottom of the profile were set based on the values estimated from the 45 cm samples.

PAGE 55

55 To initial ize the model for each treatment, the soil water content was initialized based on the measured probe values at the beginning of the day before the first irrigation event This allowed the model to be given an accurate representation of the actual initial c onditions in the soil as well as time for the model to settle before beginning model evaluation. Since the initial water content was s et based on the probe measurements, it had to be set considering the 12 unique probe locations for the first 12 treatments and based on the 17 unique pro be locations for the repeated treatments. In both cases, a grid within the soil with horizontal bounds that fall entirely within 7.5 cm horizontally from a probe column will have their initial VWC determined by a probe in tha t column only. Alternatively, if the horizontal bounds of a soil grid fall within 7.5 cm horizontally from two probe columns, its initial VWC will be determined by an average of probes from the two probe columns. Vertically, if a soil grid has its upper ve rtical extent above the line 7.5 cm vertically below a row of probes, it will have its initial VWC determined by the highest row of probes for which that is true. If, however, the highest vertical extent of a soil grid falls more than 7.5 cm below the deep est row of probes, the initial VWC will be determined by the deepest set of probes available. Weather data was obtained from the Florida Automated Weather Network for the Citra location. The data was formatted and units were converted for use with in the D SSAT CSM. In order to account for the effect of the rain shelter, the recorded rainfall was changed to zero. The weather, however, was quite un impactful on the probe measurements, as the plastic minimized the effects that rainfall or potential ET would ha ve on the soil moisture regime.

PAGE 56

56 For each treatment, the model was evaluated against the 15 minute probe measurements from the beginning of the day of the first irrigation until the end of the day after the last irrigation. This was done in order to evalua te the model while the system was dynamic, including the days with irrigation as well as a day of soil water redistribution with no irrigation events During the experiment, there were several occasions where probe errors occurred or conditions existed whi ch made the probe data erroneous. These issues were caused due to probe exposure, elevated salinity, and rainfall infiltration through the rain shelter. The probe exposure always occurred with the probe directly beneath the drip emitter, as the high flows and shallow probe placement resulted in inadequate soil cover age once enough soil was washed away. As soon as a probe was thought to be exposed, the plastic was removed to inspect, and additional soil was added and attempted to be packed to a representativ e density. Elevated salinity also only occurred in the probe directly beneath the emitter, because for some of the treatments solute was injected through the drip tape, and it had the highest concentration right after injection directly beneath the tape. A t elevated electrical conductivity, the CS650 probes will not return VWC values. The CS616 probes, however, cannot measure the electrical conductivity and thus return erroneously elevated VWC values at high salinity levels. Measurements were no longer cons idered erroneous once irrigation commenced the following day, as the salinity was flushed from the system and readings returned to the expected levels. When rainfall infiltrated the rain shelter, the matrices that were affected were removed from the evalua tion for the remainder of that treatment to allow time for the levels to return to normal. Probes which were not included in the evaluation for each treatment are shown in Table 3 6.

PAGE 57

57 The evaluation was conducted by comparing the average probe measurements to the predicted probe measurements. However, since DSSAT 2D breaks the soil into rectangular grids of uniform VWC, the gridded DSSAT 2D VWC distribution had to be converted into a VWC value representative of the probe sensing area. Thus, a program was wr itten to compute the percentage of the probe sensing area that intersected with each grid rectangle. This code can be seen in Appendix B. These percentages were then used to weigh the intersected grid VWCs in order to obtain an estimated probe VWC. These c alculations were computed in R (R Development Core Team, 2011) using the gpclib package to compute the intersection areas and were used as weights for estimating a probe VWC These weights and grid cell locations were read into DSSAT 2D at the beginning of a simulation and used to calculate the probe VWC at each time step. This was conducted for each of the unique probe locations, which are shown in Figure 3 12. The overall results of the model evaluation using the average probe measurements at each pro be location for comparison, are shown in Table 3 8. Error metrics are different for the first repetition of treatments compared to the additional repetitions in part due to the installation of additional probes and at additional measurement locations. Cons idering all the treatments, the model performs fairly well, with RMSE values of 0.0290 and 0.0248 cm 3 /cm 3 and correlation coefficients of 0.590 and 0.567 for the simulations using the three point DSSAT estimation method and the RETC estimation method, resp ectively. For both parameter estimation techniques, the model made more accurate simulations for the repeat treatments. Using the three point DSSAT parameter estimation method, the model had RMSE values of 0.0305 and

PAGE 58

58 0.0280 cm 3 /cm 3 for the first repetition treatments and repeated treatments, respectively. Similarly, with the RETC estimated parameters, the model had RMSE values of 0. 0267 and 0.0235 cm 3 /cm 3 for the first repetition treatments and repeated treatments, respectively. The repeated treatments may be more representative due to the increased number of measurement probes allowing the characteristic response to be better approximated Further, the additional probes allowed for estimation of the uncertainty in the VWC measurements at each probe locatio n. Using this information, it is possible to calculate uncertainty adjusted metrics. Thus, 95 percent confidence intervals were computed for the VWC at each time step and for each probe location. Then error was assumed to be zero if the predicted value fel l within this uncertainty range, and as the difference from the nearest outer bound of the uncertainty range if the predicted value falls outside. The results of this analysis (Table 3 9) yielded reduced error evaluations. Using the three point DSSAT param eter estimation method, the model predicted VWCs with an RMSE of 0.0212 cm 3 /cm 3 and a correlation coefficient of 0.787. With the RETC estimated parameters, the model predicted VWCs with an RMSE of 0.0150 cm 3 /cm 3 and a correlation coefficient of 0.775. The evaluation metrics for individual treatments using the average probe measurements (Table 3 10) reveal the range of prediction error for the different treatments. For the three point DSSAT estimated parameters, the RMSE ranged from 0.0220 to 0.0378 cm 3 /cm 3 the correlation coefficient ranged from 0.286 to 0.661, and the Nash Sutcliffe model efficiency coefficient ranged from 0.287 to 0.587. For the RETC estimated parameters, the RMSE ranged from 0.0209 to 0.0302 cm 3 /cm 3 the

PAGE 59

59 correlation coefficient ranged from 0.213 to 0.631, and the Nash Sutcliffe model efficiency coefficient ranged from 0.371 to 0.623. While the model appeared to perform reasonably well and robustly across all the treatments, there was a noticeable range in model performance. Two of the t reatments that were simulated using the three point DSSAT parameter estimation method made predictions with a negative valued Nash Sutcliffe coefficient, which indicates average predictions would have better predicted the soil water dynamics than the model predictions. Uncertainty adjusted evaluation metrics were also computed for the individual treatments (Table 3 11). These are again based on the 95 percent confidence interval VWC at each time step for each probe location, and thus are only computed for t he repeated treatments which had the additional probe installations. For the three point DSSAT estimated parameters, the RMSE ranged from 0.0145 to 0.0271 cm 3 /cm 3 the correlation coefficient ranged from 0.712 to 0.855, and the Nash Sutcliffe model efficie ncy coefficient ranged from 0.504 to 0.748. For the RETC estimated parameters, the RMSE ranged from 0.0099 to 0.0142 cm 3 /cm 3 the correlation coefficient ranged from 0.749 to 0.850, and the Nash Sutcliffe model efficiency coefficient ranged from 0.683 to 0 .829. The comparison of the model predictions to the probe measurements for each of the treatments can be visualized in Figure C 1 and Figure C 2 for the three point DSSAT estimated parameters and the RETC estimated parameters, respectively. In general, th e model predictions appear to respond similarly to the probe measurements, having the appropriate response to changes in flow rate, irrigation durations, and application splitting. It is notable, however, that the model does consistently over predict the V WC values at the three measurement locations at the 7.5 cm depth. There were no noticeable patterns of

PAGE 60

60 model prediction error relating to the emitter spacing, flow rate, application splitting, or daily irrigation amount which varied between the treatments Thus, field evaluations of the DSSAT 2D model indicate the model as robust in its ability to adequately simulate the soil water dynamics under various drip irrigation regimes. Summary The amended DSSAT model was evaluated to provide some assessment of i ts ability to accurately characterize the soil water dynamics of these systems. First, the methodology for estimating the soil hydraulic parameters from the standard DSSAT inputs was evaluated. The model was then evaluated against the HYDRUS 2D model for f our hypothetical soil profiles. Finally, a field experiment was conducted in order to The DSSAT three point soil parameter estimation method was evaluated against 4447 SMRCs from soil profiles in a Florida soil database. Analyses revealed that the SMRCs were estimated fairly well by this method, although with notable error. Further, the method seemed to perform better for estimating coarser soils than for heavier textured soils. Overall, however, VWC prediction residuals were normally distributed with a mean near zero and the model bias was very near zero, indicating the methodology was functioning properly The estimation methodology was further evaluated against the RETC estimation methodology, w hich is a commonly used method for estimating soil hydraulic parameters and can thus be used as a benchmark for model comparison. The RETC method showed a simila r trend to the DSSAT approximation method in predicting the coarser soils more accurately than the more finely textured soils, indicating that some of this error is due to inherent sampling uncertainty. However, the RETC method predicted the SMRCs with notably improved

PAGE 61

61 accuracy, which is to be expected since the methodology used all the available da ta rather than the limited data used for the approximate DSSAT method. It is thus concluded that the approximate DSSAT method is a functional solution for widening the model applicability for scenarios where the soil hydraulic parameters have not been well characterized. However, this comes at the cost of reduced model representation of the soil properties. Thus, it is recommended to use more detailed soil hydraulic measurements when possible. Next, t he DSSAT 2D soil water flow calculations were evaluated against the HYDRUS 2D model to allow evaluation of the model performance under specific theoretical soil profiles The two models produced very similar simulations for uniformly coarsely textured and uniformly fine ly textured soils under drip irrigation. T his indicates that the grid design and numerical solving technique used in DSSAT 2D perform comparably to those of HYDRUS 2D. However, the model simulations diverged notably for layered soils, with significantly higher error metrics and divergence between models that grew largely monotonically with simulation time. This indicates that the approximate form of the Richards equation used in the DSSAT 2D model results in questionable soil flow predictions for sharply layered soils. The analysis also demonstrate d that driving the DSSAT 2D model by the normalized water content rather than the actual water content reduced the deviations from the HYDRUS 2D model, although the deviations were still considerable Further, the improvements appeared to be the result of the adjustment slowing the development of the deviation s, not eliminating them The DSSAT 2D soil water flow calculations were further evaluated through comparison with field measured soil moisture data. The soil was fairly uniform, with

PAGE 62

62 largely uniform t exture and some deviation with depth in hydraulic conductivity and soil water retention likely due to structural compaction. The model demonstrated the ability to simulate the soil water flow patterns fairly adeptly under different drip irrigation manageme nt regimes and drip tape specifications. Additionally, simulations using the three point DSSAT soil hydraulic parameter estimation method were able to provide representative simulations, but with error notably greater than simulations with the soil hydraul ic parameters estimated by the RETC method.

PAGE 63

63 Table 3 1 Error metrics for predicting SMRCs using the three point DSSAT methodology SMRCs are from the Florida Soil Characterization Database. Texture Samples RMSE (cm 3 /cm 3 ) CVRMS E (cm 3 /cm 3 ) Ceff R 2 Bias (cm 3 /cm 3 ) Fine sand 2319 0.0393 0.1959 0.923 0.889 0.00355 Sand 1072 0.0528 0.3274 0.820 0.857 0.02218 Sandy loam 419 0.0576 0.2044 0.705 0.842 0.03487 Sandy clay 532 0.0463 0.1375 0.673 0.73 1 0.00233 Clay 105 0.0619 0.133 9 0.501 0.630 0.00900 All 4447 0.0458 0.2090 0.899 0.900 0.00517 Table 3 2. Error metrics for predicting SMRCs using the RETC methodology SMRCs are from the Florida Soil Characterization Database. Texture Samples RMSE (cm 3 /cm 3 ) CVRMSE (cm 3 /cm 3 ) Ceff R 2 Bias (cm 3 /cm 3 ) Fine sand 2319 0.0275 0.1369 0.962 0.967 0.00952 Sand 1072 0.0244 0.1515 0.962 0.964 0.00674 Sandy loam 419 0.0330 0.1171 0.903 0.90 4 0.00109 Sandy clay 532 0.0298 0.0885 0.864 0.866 0.00044 Clay 105 0.0354 0.0767 0.836 0.838 0 .00232 All 4447 0.0277 0.1262 0.963 0.966 0.00686 Table 3 3. Soil hydraulic parameters for the benchmark comparison of DSSAT 2D with HYDRUS 2D. Analysis ID Layer Depth (cm) Texture K s (cm/hr) r (cm 3 /cm 3 ) s (cm 3 /cm 3 ) (cm 1 ) n Uniform 1 1 0 200 Sand 7. 7 0. 050 0.360 0.024 2.123 Uniform 2 1 0 200 Sandy clay 1.7 0.163 0.426 0.052 1.270 Layered 1 1 0 20 Sand 7. 7 0.050 0.360 0.024 2.123 Layered 1 2 20 200 Sandy clay 1.7 0 .163 0.426 0.052 1.27 0 Layered 2 1 0 20 Sandy clay 1.7 0.163 0.426 0.052 1.270 Layered 2 2 20 200 Sand 7. 7 0.050 0.360 0.024 2.123

PAGE 64

64 Table 3 4. Difference metrics and closeness of fit measures between VWC p redictions by DSSAT 2D and HYDRUS 2D. Analysis ID Water Content For m RMSE (cm 3 /cm 3 ) CVRMSE (cm 3 /cm 3 ) Ceff R 2 Bias (cm 3 /cm 3 ) X max (cm) Z max (cm) Uniform 1 Normalized 0.0010 0.0124 0.982 0.985 0.00007 59.5 72.7 Uniform 2 Normalized 0.0013 0.0044 0.993 0.993 0.00012 43.1 48.5 Layered 1 Normalized 0.0228 0.1413 0.944 0.958 0.00150 80.0 46.5 Layered 1 Actual 0.0253 0.1507 0.932 0.947 0.00039 80.0 46.5 Layered 2 Normalized 0.0157 0.1123 0.972 0.975 0.00059 80.0 78.8 Layered 2 Actual 0.0176 0.1262 0.965 0.968 0.00058 80.0 78.8 Table 3 5. Drip tape manufacturer specific ations. Emitter Spacing (cm) Flow Rate Level (High (H) or Low (L)) Emitter Rate (ml/s) 30.5 H 0.3155 30.5 L 0.1367 20.3 H 0.2804 20.3 L 0.1402 10.2 H 0.2804 10.2 L 0.1402

PAGE 65

65 Table 3 6. Field experiment treatment details Treatment Spacing (cm) Rate (H/L) Splits Emitter Spacing (cm) Pan Measured Emitter Rate (ml/s) Pan Measured Irrigation Amount (l/m/day) Missing Matrix 1 Probes Missing Matrix 2 Probes Missing Matrix 3 Probes 30 H 1 30.5 0.3354 4.03 None All All 30 H 2 30.5 0.3354 4.09 None All A ll 30 L 1 30.5 0.1819 5.01 None All All 30 L 2 30.5 0.1819 5.01 None All All 20 H 1 20.3 0.3537 4.80 None All All 20 H 2 20.3 0.3537 4.80 None All All 10 H 2 10.2 0.3070 3.99 Probe 1 (all days) All All 10 H 1 10.2 0.3070 3.99 Probe 1 (all days) All All 20 L 1 20.3 0.1828 4.86 None All All 20 L 2 20.3 0.1828 4.86 None All All 10 L 2 10.2 0.1368 3.72 None All All 10 L 1 10.2 0.1368 3.72 None All All 10 H 1 Rep 2 10.2 0.2944 3.82 None None None 10 H 2 Rep 2 10.2 0.2944 3.82 None None None 10 H 2 Rep 3 10.2 0.2944 3.82 None None None 10 H 1 Rep 3 10.2 0.2944 3.82 Probe 1 (all days) None None 30 L 1 Rep 2 30.5 0.1917 5.28 Probe 1 (all days) None None 30 L 2 Rep 2 30.5 0.1917 5.28 Probe 1 (1 day) All ( 2 days) None 30 L 1 Rep 3 30.5 0.1917 5.2 8 Probe 1 (1 day) All None 30 L 2 Rep 3 30.5 0.1917 5.28 Probe 1 (1 day) All None 20 H 2 Rep 2 20.3 0.3278 4.45 Probe 1 (1 day) None None 20 H 1 Rep 2 20.3 0.3278 4.45 Probe 1 (1 day) None None Probe 1 is the probe inserted at a distance of 0 cm f rom the row center and a depth of 7.5 cm

PAGE 66

66 Table 3 7. Soil particle size, bulk density, and particle density averages by soil depth Samples taken for the field experiment. Depth (cm) Sand (%) Silt (%) Clay (%) Bulk density (g/cm 3 ) Particle density (g/ cm 3 ) 7.5 91.9 5.3 2.8 1.47 2.62 22.5 92.0 5.0 3.0 1.60 2.62 37.5 91.1 6.3 2.6 1.58 2.63 70.0 90.4 5.4 4.2 1.48 2.64 Table 3 8. Overall DSSAT 2D error metrics from the average probe measurements Treatments Parameter Estimation Method RMSE (cm 3 /cm 3 ) CVRMSE (cm 3 /cm 3 ) Ceff R 2 Bias (cm 3 /cm 3 ) All Three point DSSAT 0.0290 0.2209 0.350 0.590 0.01240 Repetition 1 Three point DSSAT 0.0305 0.2235 0.269 0.518 0.01314 Repeat Three point DSSAT 0.0280 0.2189 0.386 0.626 0.01193 All RETC 0.0248 0.1888 0.525 0 .567 0.00729 Repetition 1 RETC 0.0267 0.1954 0.442 0.504 0.00855 Repeat RETC 0.0235 0.1838 0.567 0.600 0.00650 Table 3 9. Overall DSSAT 2D adjusted error metrics from the average probe measurements Treatments Parameter Estimation Method RMSE adj (cm 3 / cm 3 ) CVRMSE adj (cm 3 /cm 3 ) Ceff adj R 2 adj Bias adj (cm 3 /cm 3 ) Repeat Three point DSSAT 0.0212 0.1622 0.627 0.787 0.00927 Repeat RETC 0.0150 0.1162 0.730 0.775 0.00580

PAGE 67

67 Table 3 10. DSSAT 2D error metrics from the average probe measurements for each treatmen t Treatment Spacing (cm) Rate (H/L) Splits Parameter Estimation Method RMSE (cm 3 /cm 3 ) CVRMSE (cm 3 /cm 3 ) R 2 Ceff Bias (cm 3 /cm 3 ) 30 H 1 Three point DSSAT 0.0267 0.2018 0.435 0.331 0.01060 30 H 2 Three point DSSAT 0.0262 0.1933 0.416 0.402 0.01180 30 L 1 Three point DSSAT 0.0335 0.2490 0.404 0.019 0.01880 30 L 2 Three point DSSAT 0.0319 0.2430 0.445 0.012 0.01960 20 H 1 Three point DSSAT 0.0324 0.2438 0.395 0.056 0.01780 20 H 2 Three point DSSAT 0.0306 0.2266 0.418 0.221 0.01530 10 H 2 Three point DS SAT 0.0254 0.1911 0.341 0.495 0.00820 10 H 1 Three point DSSAT 0.0271 0.2045 0.326 0.436 0.01100 20 L 1 Three point DSSAT 0.0337 0.2342 0.286 0.114 0.00790 20 L 2 Three point DSSAT 0.0337 0.2312 0.310 0.236 0.01150 10 L 2 Three point DSSAT 0.0309 0.215 1 0.341 0.389 0.01020 10 L 1 Three point DSSAT 0.0317 0.2308 0.346 0.307 0.01440 10 H 2 Rep 2 Three point DSSAT 0.0232 0.1926 0.629 0.544 0.00590 10 H 2 Rep 3 Three point DSSAT 0.0247 0.2028 0.596 0.459 0.00790 10 H 1 Rep 2 Three point DSSAT 0.0220 0.1 924 0.653 0.580 0.00510 10 H 1 Rep 3 Three point DSSAT 0.0221 0.1790 0.652 0.587 0.00570 30 L 1 Rep 2 Three point DSSAT 0.0260 0.2040 0.638 0.461 0.00890 30 L 1 Rep 3 Three point DSSAT 0.0319 0.2459 0.614 0.025 0.01850 30 L 2 Rep 2 Three point DSSAT 0. 0265 0.2046 0.629 0.434 0.00990 30 L 2 Rep 3 Three point DSSAT 0.0378 0.2786 0.585 0.287 0.02380 20 H 1 Rep 2 Three point DSSAT 0.0257 0.1966 0.661 0.545 0.01080 20 H 2 Rep 2 Three point DSSAT 0.0263 0.1983 0.648 0.546 0.01050 30 H 1 RETC 0.0224 0.169 6 0.330 0.527 0.00690 30 H 2 RETC 0.0243 0.1793 0.305 0.485 0.00770 30 L 1 RETC 0.0262 0.1947 0.306 0.377 0.01320 30 L 2 RETC 0.0250 0.1901 0.334 0.395 0.01400 20 H 1 RETC 0.0260 0.1961 0.299 0.389 0.01220 20 H 2 RETC 0.0250 0.1849 0.319 0.481 0.00980 10 H 2 RETC 0.0250 0.1884 0.249 0.509 0.00590 10 H 1 RETC 0.0270 0.2037 0.225 0.440 0.00850 20 L 1 RETC 0.0281 0.1956 0.213 0.382 0.00280 20 L 2 RETC 0.0294 0.2016 0.229 0.419 0.00550 10 L 2 RETC 0.0300 0.2091 0.228 0.422 0.00560 10 L 1 RETC 0.0302 0.2197 0.227 0.371 0.01010 10 H 2 Rep 2 RETC 0.0220 0.1825 0.600 0.591 0.00320 10 H 2 Rep 3 RETC 0.0226 0.1851 0.570 0.550 0.00490 10 H 1 Rep 2 RETC 0.0209 0.1822 0.631 0.623 0.00290 10 H 1 Rep 3 RETC 0.0216 0.1746 0.619 0.607 0.00260 30 L 1 Rep 2 RET C 0.0225 0.1765 0.615 0.596 0.00480

PAGE 68

68 Table 3 10. Continued Treatment Spacing (cm) Rate (H/L) Splits Parameter Estimation Method RMSE (cm 3 /cm 3 ) CVRMSE (cm 3 /cm 3 ) R 2 Ceff Bias (cm 3 /cm 3 ) 30 L 1 Rep 3 RETC 0.0245 0.1886 0.570 0.427 0.01210 30 L 2 Rep 2 RE TC 0.0224 0.1730 0.620 0.595 0.00550 30 L 2 Rep 3 RETC 0.0254 0.1869 0.562 0.421 0.01250 20 H 1 Rep 2 RETC 0.0245 0.1876 0.617 0.585 0.00580 20 H 2 Rep 2 RETC 0.0254 0.1915 0.612 0.577 0.00530 Table 3 11. DSSAT 2D adjusted error metrics from the aver age probe measurements for each treatment Treatment Parameter Estimation Method RMSE adj (cm 3 /cm 3 ) CVRMSE adj (cm 3 /cm 3 ) R 2 adj Ceff adj Bias adj (cm 3 /cm 3 ) 10 H 2 Rep 2 Three point DSSAT 0.0162 0.1340 0.818 0.695 0.00550 10 H 2 Rep 3 Three point DSSAT 0.0168 0.1354 0.816 0.674 0.00620 10 H 1 Rep 2 Three point DSSAT 0.0157 0.1379 0.840 0.684 0.00560 10 H 1 Rep 3 Three point DSSAT 0.0145 0.1172 0.855 0.748 0.00550 30 L 1 Rep 2 Three point DSSAT 0.0190 0.1472 0.820 0.664 0.00740 30 L 1 Rep 3 Three point DSSAT 0.0236 0.1700 0.733 0.531 0.00940 30 L 2 Rep 2 Three point DSSAT 0.0190 0.1438 0.814 0.668 0.00740 30 L 2 Rep 3 Three point DSSAT 0.0271 0.1839 0.712 0.504 0.01190 20 H 1 Rep 2 Three point DSSAT 0.0176 0.1305 0.829 0.735 0.00680 20 H 2 Rep 2 Three poi nt DSSAT 0.0175 0.1280 0.827 0.739 0.00660 10 H 2 Rep 2 RETC 0.0114 0.0945 0.819 0.791 0.00260 10 H 2 Rep 3 RETC 0.0117 0.0940 0.805 0.774 0.00290 10 H 1 Rep 2 RETC 0.0107 0.0936 0.840 0.808 0.00250 10 H 1 Rep 3 RETC 0.0099 0.0798 0.850 0.829 0.00240 30 L 1 Rep 2 RETC 0.0118 0.0916 0.832 0.802 0.00300 30 L 1 Rep 3 RETC 0.0137 0.1001 0.759 0.683 0.00470 30 L 2 Rep 2 RETC 0.0118 0.0897 0.823 0.798 0.00290 30 L 2 Rep 3 RETC 0.0142 0.0989 0.749 0.688 0.00480 20 H 1 Rep 2 RETC 0.0124 0.0932 0.800 0.786 0.00280 20 H 2 Rep 2 RETC 0.0130 0.0960 0.785 0.777 0.00240

PAGE 69

69 Figure 3 1 Distribution of VWC prediction residuals using the three point DSSAT methodology. Residuals are derived from predictions of the SMRCs from the Florida Soil Characterization Data base

PAGE 70

70 Figure 3 2 Distribution of VWC prediction residuals using the RETC methodology. Residuals are derived from predictions of the SMRCs from the Florida Soil Characterization Database

PAGE 71

71 Figure 3 3 Boxplot of VWC prediction residuals vs. tensio n using the three point DSSAT methodology. Residuals are derived from predicting SMRCs from the Florida Soil Characterization Database T he horizontal axis is not to scale.

PAGE 72

72 Figure 3 4 Boxplot of VWC prediction residuals vs. tension using the RETC met hodology. Residuals are derived from predicting SMRCs from the Florida Soil Characterization Database The horizontal axis is not to scale.

PAGE 73

73 Figure 3 5 Simulated VWCs for the Uniform 1 soil VWC values are plotted versus depth and distance for A) HYDRU S 2D at the end of the first irrigation event B) HYDRUS 2D at the end of the fourth irrigation event C) HYDRUS 2D four hours after the day four irrigation event D) HYDRUS 2D one day after the day four irrigation event E) DSSAT 2D with normalized water cont ent at the end of the first irrigation event F) DSSAT 2D with normalized water content at the end of the fourth irrigation event G) DSSAT 2D with normalized water content four hours after the fourth irrigation event H) DSSAT 2D with normalized water conten t one day after the fourth irrigation event.

PAGE 74

74 Figure 3 6 Simulated VWCs for the Uniform 2 soil VWC values are plotted versus depth and distance for A) HYDRUS 2D at the end of the first irrigation event B) HYDRUS 2D at the end of the fourth irrigation event C) HYDRUS 2D four hours after the day four irrigation event D) HYDRUS 2D one day after the day four irrigation event E) DSSAT 2D with normalized water content at the end of the first irrigation event F) DSSAT 2D with normalized water content at the end of the fourth irrigation event G) DSSAT 2D with normalized water content four hours after the fourth irrigation event H) DSSAT 2D with normalized water content one day after the fourth irrigation event.

PAGE 75

75 Figure 3 7 Simulated VWCs for the Layered 1 so il VWC values are plotted versus depth and distance for A) HYDRUS 2D at the end of the first irrigation event B) HYDRUS 2D at the end of the fourth irrigation event C) HYDRUS 2D four hours after the day four irrigation event D) HYDRUS 2D one day after the day four irrigation event E) DSSAT 2D with normalized water content at the end of the first irrigation event F) DSSAT 2D with normalized water content at the end of the fourth irrigation event G) DSSAT 2D with normalized water content four hours after the fourth irrigation event H) DSSAT 2D with normalized water content one day after the fourth irrigation event I) DSSAT 2D with actual water content at the end of the first irrigation event J) DSSAT 2D with actual water content at the end of the fourth irrig ation event K) DSSAT 2D with actual water content four hours after the fourth irrigation event L) DSSAT 2D with actual water content one day after the fourth irrigation event.

PAGE 76

76

PAGE 77

77 Figure 3 8 Simulated VWCs for the Layered 2 soil VWC values are plotted ver sus depth and distance for A) HYDRUS 2D at the end of the first irrigation event B) HYDRUS 2D at the end of the fourth irrigation event C) HYDRUS 2D four hours after the day four irrigation event D) HYDRUS 2D one day after the day four irrigation event E) DSSAT 2D with normalized water content at the end of the first irrigation event F) DSSAT 2D with normalized water content at the end of the fourth irrigation event G) DSSAT 2D with normalized water content four hours after the fourth irrigation event H) DS SAT 2D with normalized water content one day after the fourth irrigation event I) DSSAT 2D with actual water content at the end of the first irrigation event J) DSSAT 2D with actual water content at the end of the fourth irrigation event K) DSSAT 2D with a ctual water content four hours after the fourth irrigation event L) DSSAT 2D with actual water content one day after the fourth irrigation event.

PAGE 78

78

PAGE 79

79 Figure 3 9 Plot of residuals between DSSAT 2D and HYDRUS 2D VWC predictions. Plots are for A) Uniform 1 soil, B) Uniform 2 soil, C) Layered 1 soil with DSSAT 2D driven by normalized water content, D) Layered 1 soil with DSSAT 2D driven by actual water content, E) Layered 2 soil with DSSAT 2D driven by normalized water content, F) Layered 2 soil with DSSAT 2 D driven by actual water content.

PAGE 80

80 Figure 3 10 The RMSD value over time between DSSAT 2D and HYDRUS 2D VWC predictions. This includes the RMSD value at each simulation time step between the VWC predictions of HYRDUS 2D and A) DSSAT 2D with normalized wat er content for the Uniform 1 soil, B) DSSAT 2D with normalized water content for the Uniform 2 soil, C) DSSAT 2D with normalized water content for the Layered 1 soil D) DSSAT 2D with actual water content for the Layered 1 soil, E) DSSAT 2D with normalize d water content for the Layered 2 soil, F) DSSAT 2D with actual water content for the Layered 2 soil.

PAGE 81

81

PAGE 82

82 Figure 3 11 A set of water content reflectometer probes installed with measurement rods inserted parallel to the length of the bed row.

PAGE 83

83 Figure 3 1 2 Two dimensional representation of the 17 unique probe locations and sensing areas The + at the center of a circle represents the probe installation location, and the 7.5 cm radius circle represents its approximate sensing area.

PAGE 84

84 Figure 3 13 A verage bulk density measurements vs. sampling date for each soil depth. Samples taken for the field experiment. Figure 3 14 Boxplot s of the field measured SMRCs This includes boxplots for A) the samples at a 10 cm depth B) the sample s at a 45 cm depth, C) all the samples. Note that the horizontal axis is not to scale.

PAGE 85

85

PAGE 86

86 CHAPTER 4 DRIP IRRIGATION MANGEMENT IMPLICATIONS Background Vegetable production in Florida is typically done on sandy soils with low water holding capacities (Dukes et al., 2006). Production on these soils necessitates close management to maintain soil moisture and nutrient levels conducive for optimal crop production The combination of coarse soils requiring intensive irrigation and nutrient applications presents th e potential for high water use and nutrient leaching, for which there is increasing pressure to minimize It thus becomes increasingly complicated to manage these production systems to meet both economic and environmental goals. Vegetable production syste ms are typically irrigated using either seepage or drip irrigation. Drip irrigation has been shown to have the potential to increase water and nutrient use efficiencies and reduce nutrient leaching compared to seepage irrigation (Dukes et al., 2010; Pitts et al., 1988; Sato et al., 2010). Around 44 percent of irrigated lands in Florida are irrigated via seepage (Dukes et al., 2010). Thus, conversion of irrigation systems from seepage to drip has been suggested as a possible measure for reducing water use an d nutrient leaching (Dukes et al., 2010; Sato et al., 2010). However, while drip systems offer the potential for improved production efficiencies, system management is the key for realizing these potential reductions. Surveys have shown that it is common industry practice to apply fertilizer at rates in excess of the recommended amounts (Cantliffe et al., 2009). One facet of drip applications throughout the growi ng season through fertigation. Thus, fertilizer can be applied more synchronously with the crop demand, resulting in less nutrients in the soil

PAGE 87

87 at any one time and a system that is less susceptible to leaching during heavy rainfall events. Ironically, howe ver, improper management can result in excessive nutrients to be applied as fertigation. This has been observed in surveys that have shown instances of producers using drip systems using more fertilizer than those using seepage systems (Jones et al., 2012) With the potentially increasing prevalence of drip systems in the future, it is essential that these systems be managed properly. Research has shown that, given proper management, high water use efficiencies (Simonne et al., 2004) and fertilizer use effi ciencies (Zotarelli Dukes et al., 2009 ) can be attained within these intensively managed systems. Nutrient and irrigation management are intrinsically linked, as nutrient movement is driven by soil water dynamics. Nitrates especially move with soil water flow, tending to collect near the fringe of wetting fronts under drip irrigation regimes (Li et al., 2003). Simonne and Ozores Hampton (2006) argue that nutrient management should be approached by first attempting to improve the irrigation management, sinc e nutrient requirements cannot be properly established until proper irrigation management is in place. It is apparent that there is a need for proper irrigation management of drip irrigated systems. Models can be useful tools for identifying the impacts o f management decisions. Models can relate the complex system interactions between management, soil, climate, crop, and environment. DSSAT 2D is a modified version of the DSSAT CSM that was designed and evaluated for simulating the soil water dynamics under management practices and conditions typical of vegetable production systems. It was therefore the goal of this research to apply the DSSAT 2D model to analyze the impact

PAGE 88

88 of different drip tape designs and irrigation management options on the soil water re gime and the implications on root zone moisture and water loss from the root zone. Irrigation Management Impacts Assessing the impacts of different irrigation management options requires the establishment of quantitative indicators of eff ectiveness With a reliable crop model, the natural indicators would be simulated crop water stress, crop yield, and water use. In the absence of a validated crop component, however, it is necessary to consider the soil moisture conditions that are conducive to crop growth This scenario is analogous to irrigation during crop establishment, when root water uptake is essentially zero. The indicators selected were the average root zone moisture content and the water flowing out of the root zone, with higher root zone moisture content considered better for crop growth and less water flow out of the root zone considered better for minimizing water use and nutrient loss. Analyses were conducted for a specific field layout. A plastic mulched raised bedded system of 20 cm height a nd 80 cm width with a row spacing of 180 cm was simulated. The soil was uniform, with a lower limit water content of 0.047 cm 3 /cm 3 drained upper limit water content of 0.170 cm 3 /cm 3 saturation water content of 0.480 cm 3 /cm 3 saturated hydraulic conductiv ity of 23.1 cm/hr, 3.2 percent clay, and 5.2 percent sil 1 r of 0.036 cm 3 /cm 3 The initial soil water content was set at 0.10 cm 3 /cm 3 Model implications from these analyses can be generalized, but site specific applic a bility is a strength of physically based models The main drip irrigation management options are flow rate, daily irrigation amount, and method of splitting irrigation applications. Application splitting is the

PAGE 89

89 practice of applying a desired irrigation am ount in multiple applications instead of in a single application. On coarse soils with low water holding capacities and at periods of high demand, this is considered a good practice for keeping irrigation water in the root zone. This is because at times of high ET demand, daily irrigation requirements can exceed the water holding capacity of the root zone. Thus, the necessary irrigation amount must be applied in two or more split applications, with each application targeted to not exceed the ro ot zone water holding capacity. The daily irrigation amount was varied between 9 and 189 cm 2 in increments of 9 cm 2 The irrigation rate w as varied between 0.1 and 0.8 lpm/m in increments of 0.1 lpm /m. Applications were split as either one application at 1200, two appl ications at 1000 and 1400, or four applications at 0800, 1000, 1200, and 1400. Results are shown in Figure 4 1, with only the extreme application splitting and irrigation rates being shown for simplicity of visual interpretation. It is apparent that the ir rigation amount is the dominant management factor for influencing the soil water regime, with increased irrigation amounts resulting in greater water loss from the root zone and water storage in the root zone. The response of wate r leaching to irrigation a mount is such that with 47 cm 2 of daily irrigation or less, nearly no water leaches from the root zone. Beyond this amount however, leaching responds more rapidly to increases in irrigation amount until the response becomes approximately linear. Increases in irrigation amount result in increases in root zone storage as well. However, the response is much more rapid at lower application amounts, with increased irrigation amounts resulting in only slight root zone storage increases in regions of high irrigat ion amounts. At equivalent irrigation amounts the system showed increased root zone storage and reduced leaching for irrigation applied

PAGE 90

90 at lower rates and increased application splitting. These effects were minor, however, relative to the impacts of the ir rigation amount. To assess the response of the system to application rate and application splitting in greater detail, the model was run for a range of irrigation rates and application splitting methods while maintaining fixed daily irrigation amounts. Si mulations were run with daily irrigation amounts of 20 cm 2 Flow rates were set to 0.100, 0.150, 0.200, 0.250, 0.300, and 0.375 cm 2 /min. Application splitting methods consisted of, as before, one application at 1200 or two applications at 1000 and 1400. Ho wever, in addition to the number of splits, the distribution of the splits throughout the day is another factor that determines the effectiveness of splitting applications. Thus, applications were also split into three applications at 1000, 1200, and 1400, which has the same first and last irrigation start times as the two split application, as well as a more spread out splitting of applications into three applications at 0900, 1300, and 1700. The three split application s are thus referred to as the 3 narro w and 3 wide split applications. In order to assess any interactive response with application amount, the same analysis was also conducted with daily irrigation amounts of 40 and 60 cm 2 The results of this analysis, which are shown in Figure 4 2, demonst rate consistent reduction of water leaching from the root zone for lower application rates A general pattern of reduced water leaching was also observed for greater split applications. However, while the 3 wide split applications was the best application splitting method in almost all the considered instances, the 3 narrow split method in fact consistently resulted in greater water leaching from the root z one than the two split method. It should be noted that while the 3 narrow split method utilizes an add itional

PAGE 91

91 application event, if one considers the total daily irrigation application time as the difference between the start of the first irrigation event and the end of the last irrigation event, the two split method actually applies the irrigation water o ver a larger application time compared to the 3 narrow split method. This is because the first and last daily irrigation events start at the same time for the two splitting methods, but each event is longer for the two split method. This trend of irrigatio n time truly influencing soil water flow rather than phenomena associated with irrigation rates or application splitting is similar to observations by Skaggs et al. (2010). By this way of thinking, the root zone leaching reductions associated with lower ap plication rates are in fact due to the longer application times required to apply the fixed amounts at slower rates. This concept is further sup ported by the variable effect that changing application rate s had under different split methods. The application rate had notably greater impact on the root zone leaching for the splitting methods that resulted in shorter irrigation durations. Conversely, changes in application rate had the least impact on root zone leaching for the 3 wide splitting method. This sup ports the application time concept because for single daily applications, the application rate has a large influence over the total irrigation time, thus making changes in application rate very influential. However, if the splitting method is already servi ng to separate the first and last irrigation events by a large amount of time, changes in application rate will have a relatively small influence over the application time. It should also be noted that the application rate and the application splitting me thod had different influences at the different daily application amounts. At low daily application amounts, there are clear distinctions in root zone leaching between the

PAGE 92

92 application splitting methods, and smaller differences between the application rates. At higher daily application amounts, however, the daily application ra tes become much more impactful. This in turn results in variable effectiveness of different splitting methods depending on application rate, as a splitting method may outperform another at a certain application rate, but perform worse at a different application rate. The increased impact of application rates and reduced impact of application splitting method on root zone leaching at higher daily application amounts is to be expected. Wit h greater daily flow, irrigation durations must increase to reach the flow target. As such, changes in flow rate result in increasing changes in application time. The reduced impact of the splitting method also stands to reason because as irrigation times increase, the periods of rest between irrigations imposed by the splitting method become increasingly small. For example, if the splitting method dictates a two hour difference between the start of two irrigation events, if the irrigation amount dictates t hat water be applied for two hours in each split, the split method becomes rendered meaningless as the two applications have merged into a single continuous application. However, the logical extension would indicate that by a certain irrigation amount and at a fixed application rate all the split methods would observe equivalent root zone water leaching. However, it is observed that in several instances, and at equivalent application rates, the single daily application method observed reduced leaching comp ared to other split methods. An additional analysis was conducted in which, in addition to varying the application rate and application splitting method, different combinations of daily ET rate and daily irrigation amount were imposed. Daily ET rates were set by simulating tomato

PAGE 93

93 crop growth and varying daily solar radiation until the 10 day average of solar radiation between 45 and 55 days after transplanting reached the desired ET rate. It should be noted that the DSSAT 2D model has not been properly cal ibrated or evaluated for simulating crop growth and development. As such, this analysis should be considered as a proof of concept analysis. Results of the analysis, which can be seen in Figure 4 3, indicate that the ET rate has a similar impact on root zo ne water leaching as the irrigation amount. The relative insignificance of the application rate and application splitting method are also very noticeable. The analysis most notably underlines the importance of matching irrigation amount with ET demand in o rder to meet the crop water requirements while minimizing root zone water leaching. Soil Moisture Sensor Controlled Irrigation S oil moisture sensor based irrigation has been shown as a technology capable of allow ing significant reduction in irrigation with out reductions in yield compared to traditional irrigation management practices (Smajstrla & Locascio, 1996; Dukes et al., 2003; Zotarelli et al., 2008; Zotarelli et al., 2009b) However, th e performance of these systems depends on decisions such as the se nsor placement (Coelho & Or, 1996) and soil moisture threshold value for triggering irrigation (Coelho & Or, 1996; Zotarelli et al., 2010) Coelho and Or (1996) point out that while much work has been done to identify threshold soil moisture or matric valu es for optimal crop yields, most recommendations are general and empirical, ignoring site specific soil water dynamics. Thus, it is attempted to use the DSSAT 2D model to evaluate the impact of sensor location and threshold value on the soil water dynamics T he analysis is conducted for the same field layout described previously, but can be applied to any field setup to provide site specific recommendations.

PAGE 94

94 To evaluate the impact of different soil moisture sensor locations and threshold values, the DSSAT 2D code first had to be modified to simulate soil moisture sensor based irrigation. This was done by allowing the depth and distance of the probe to be inputted into the model as well as the threshold VWC and the probe sensing radius. Different soil moistu re probes have varying soil moisture sensing radiuses, but for this study the radius was set to 7.5 cm, as this is a common probe specification Since DSSAT 2D breaks the soil into rectangular grids of uniform VWC, this grid information had to be translate d into a representative probe VWC. Thus a program was written to compute the percentage of the probe sensing area that intersected with each grid area This code can be seen in Appendix B. These percentages were then used as weight s for the intersected gr id VWCs to obtain an estimated probe VWC. These calculations were computed in R, using the gpclib package to compute the intersection areas. These weights and grid cell locations were read into DSSAT 2D at the beginning of a simulation and used to calculat e the probe VWC at each time step. Irrigation events were thus triggered to occur at any time the probe VWC dropped below the threshold value. Additionally, in order to conduct this analysis measurements had to be established on which to evaluate the pro be performance. T he main goals of an irrigation system are to maintain a certain level of root zone soil moisture with a minimum amount of water use. Thus, the closeness of fit of the root zone soil moisture to a desired moisture level was considered to ev aluate the adequacy of an irrigation setup for meeting the crop water requirement. As such, the desired root zone soil moisture was set as 0.15 cm 3 /cm 3 and the closeness of fit was measured as the SSD between the

PAGE 95

95 simulated root zone soil moisture and 0.15 The root zone was considered the soil within 30 cm from the row center and at a depth of no more than 45 cm. The irrigation amount for a particular irrigation setup was then considere d for assessing the water use with that setup. To conduct the analysi s probe locations were considered within the entire space of the root zone. A grid of locations was created to cover this space, starting at 7.5 cm depth because of the sensing radius and going down to 45 cm deep, and ranging from 0 cm from the row center to 30 cm from the row center. In each direction, the probe location was moved in increments of 2.5 cm for a total of 208 probe locations. The soil moisture thresholds were varied from 0.10 to 0.18 cm 3 /cm 3 Thus, in order to evaluate all combinations of lo cation and threshold 1 664 unique irrigation management scenarios were created For all analyses, irrigation events were fixed to apply 7.5 cm 2 of water, which represents a small application amount such that irrigation can be more finely controlled by th e sensors. Additionally, irrigation was applied with a 30 cm emitter spacing and at an emitter rate of 0.2 ml/s. Irrigation was started on day 100 and terminated on day 220 to represent a standard planting date and growing season for tomatoes grown in Nort h Central Florida. Weather data was used from the Florida Automated Weather Network d atabase for the Citra weather station. The first anal ysis was conducted with no crop and with rainfall set to zero. This isolated the effect of probe location from rainfa ll and ET effects, resulting in a highly controlled system with minimal noise and strong repeatability throughout the season Preliminary analysis demonstrated precise system repeatability between years. Since

PAGE 96

96 this system was so unaffected by weather, simu lations were only run for a single year of weather, for a total of 1,664 simulations. The second analysis also had no crop, but considered the impact of rainfall on the system. Thus, simulations were run for 12 years of weather data, from 2001 to 2012. Thi s necessitated 19,968 simulations. For each of the 1,664 unique automated irrigation setups, total SSD and irrigation water use were computed from the 12 years of simulations to evaluate the cumulative performance of each system setup The final analysis c onsidered rainfall as well as a planted tomato crop, which was trans planted on day 100 with a density of 1.20 plants/m 2 initial planting depth of 1 cm, transplant dry weight of 3 kg/ha, transplant age of 28 days, and transplant greenhouse temperature of 2 5 C. This analysis also considered simulations from years 2001 to 2012. It should again be noted that since the DSSAT 2D model has not been properly calibrated or evaluated for simulat ing crop growth and development, this particular analysis should be con sidered as a proof of concept analysis. In order to visualize the implications of probe location s on the ability to meet the target root zone soil moisture, at each probe location the simulation with the VWC threshold value that resulted in the best SSD wa s accepted as the best possible SSD for a system with a probe at that location. T he se SSD values were then interpolated onto a uniform grid via linear point kriging using Surfer (Golden Software Inc., 2011) software with a slope of 1, an anisotropy ratio o f 1, and an anisotropy angle of 0. Contour graphs from these analyses (Figure 4 4) reveal different implications of probe location choices, and how these impacts depend on the production system conditions. With no crop and no rain, the system can be quite precisely controlled, with RMSD values ranging between 0.0031 and 0.0173 cm 3 /cm 3 The optimal probe placement area

PAGE 97

97 is also fairly well defined, with an arced shape reminiscent of the characteristic flow pattern of the wetting front of drip regimes ranging from around 17 to 26 cm depths directly below the emitter to between 16 and 26 cm from the row center at 7.5 cm below the surface. There is a clear pattern that locations very close to the emitter are sub optimal, and locations beyond the optimal range inc reasingly deteriorate in quality as well. With the addition of rainfall to the un cropped system, the ability of the system to control the root zone moisture regime is hampered considerably, with RMSD values ranging between 0.0293 and 0.0335 cm 3 /cm 3 This jump in deviation from the desired moisture levels is due to large rainfall events, which increase the moisture in the root zone far beyond the desired levels This effect is exacerbated by the plastic mulch and lack of crop, which essentially eliminate E T within the root zone and result in the system remaining at these elevated moisture levels for long periods of time. With the spring planting, these rainfall events are much more prevalent during the later months. The optimal probe location area is more p oorly defined, but remains in a similar location as the non rainfall scenario, though favoring locations slightly more distant from the emitter. Locations very close to the emitter and very distant from the emitter again perform the worst. However, with th e addition of rainfall to the un cropped system, the difference between probe locations becomes less impactful. When considering a cropped system, however, the system regains some ability to control the root zone moisture near the desired level, with RMSD values ranging between 0.0128 and 0.031 cm 3 /cm 3 This improved ability is due to the effect of root water uptake, which allows the moisture levels in the root zone to drop to threshold

PAGE 98

98 values more quickly following large rainfall events. This increased de pendence on the irrigation regime allows the system to exact more control over the soil moisture levels. This results in a more clearly defined optimal probe location area, which extends a bit further from the emitter, ranging from beneath the emitter at d epths of around 22 to 30 cm and 18 to 26 cm from the row center at 7.5 cm below the soil surface. The cropped system results in the strongest negative effect of probe locations that are very close to the emitter. This is illustrated in Figure 4 5, which co mpares the root zone VWC over time with a probe placed in a well suited location and a probe placed directly below the emitter. The well placed probe controls the irrigation fairly similarly throughout the season, consistently remaining near the 0.15 cm 3 /c m 3 target with the exception of dur ing a few large rainfall events where the root zone VWC peaks. Conversely, control with the probe under the emitter results in over irrigation in the beginning of the season, when root water uptake is concentrated near th e emitter and the probe VWC is not representative of the root zone VWC. Later in the season, the system is strongly under irrigated, as the roots extract water from a greater portion of the root zone and the area near the emitter now under estimates the ne eds for moisture in the root zone. In order to evaluate the impact of the probe location on the proper probe threshold value, the optimal threshold value at each probe location was also in terpolated via kriging These graphs, which can be seen in Figure 4 6, demonstrate the general need to set thresholds to higher values nearer the emitter. This higher threshold is necessitated due to the quicker feedback with probes near the emitter. Irrigation events reach the probes near the emitter more quickly, result ing in frequent, short irrigations. Thus to supply water to an adequate level in the root zone, higher

PAGE 99

99 thresholds are necessitated. It should also be noted that the system with rainfall and no crop required much lower thresholds than the system with no rai nfall and no crop, as the rainfall events supplied moisture to the system and reduced thresholds allowed the system to accept more rainfall with smaller increases beyond the desired levels. Similarly, adding the crop to the system requ ired increased thresh old levels to meet the increased demand for water. Importantly, however, the greatest changes in threshold values between these scenarios occurred at locations near the emitter, while locations in the optimal range showed the least changes in threshold val ues between scenarios. This further supports these locations as optimal, as their functionality is not dependent on altering the threshold value to fit varying system demand scenarios, but instead are more flexibly applicable. Summary The DSSAT 2D model was applied to evaluate the impacts of different drip irrigation management options on irrigation efficiency. Model experiments implied that irrigation amount, especially as it relates to ET rates, is very impactful on root zone storage and water leaching from the root zone. Increasing application amounts increased root zone storage and root zone leaching. However, root zone storage and root zone leaching responded inversely to increases in daily application amounts, with root zone storage responding with d iminishing increases to increases at high daily application amounts, while root zone leaching responded strongly at high daily application amounts. Root zone leaching and storage are inherently linked, as minimizing leaching will maximize storage increases However, at lower daily application amounts, increases in application amount are able to result in increased

PAGE 100

100 root zone storage with minimal increases in root zone leaching. This is an optimal region to be in for efficient irrigation. Although considerab ly less impactful, lower application rates and increased spreading of applications through application splitting were shown to result in increased root zone storage and reduced root zone leaching. Model responses generally coincided with the belief that th e effect of application splitting and application rate was not due to flow phenomena associated with the se management practices, but instead due to the impact each practice has on spreading the total irrigation application duration. Model experiments were also conducted with DSSAT 2D to consider the impact of different management options for soil moisture sensor automated irrigation systems. A methodology was demonstrated for a site specific evaluation that can be applied for other specific production site s and setups using automated drip systems. A clear area of optimal probe locations was identified for the site specific analysis. This area was additionally shown to have the most consistent optimal soil moisture threshold values under differing system sce narios, similar to its ability to represent the whole of the root zone during both early season low ET periods and late season high ET periods. Conversely, probe locations near the emitter resulted in early season over irrigation a nd late season under irri gation. P robe locations far from the emitter were insensitive to the root zone ET resulting in few and large irrigation events While rainfall events introduce some randomness to the system and thus reduce the ability of the irrigation system to control t he moisture levels precisely crop growth restored some control to the system as root water extraction allowed the root zone to more rapidly fall to desirable moisture

PAGE 101

101 levels following these events This is additionally beneficial for spring plantings, as the Florida climate is conducive to a greater prevalence of large summer rainfall events, which coincides with late season high ET demand.

PAGE 102

102 Figure 4 1 System response to changes in irrigation amount, rate, and application splitting. This includes A) DS SAT 2D predicted leaching from root zone for various irrigation amounts, rates, and application splitting B) DSSAT 2D root zone storage for various irrigation amounts, rates, and application splitting Figure 4 2 DSSAT 2D predicted water leaching fro m the root zone at different applications rates and application splitting methods Analyses are shown with daily irrigation amounts fixed at A) 20 cm 2 B) 40 cm 2 C) 60 cm 2

PAGE 103

103

PAGE 104

104 Figure 4 3 DSSAT 2D predicted water leaching from the root zone under differ ent scenarios. Scenarios include various application rates, split methods, and combinations of daily ET rate and irrigation amount. Figure 4 4 Impact of probe location on the efficacy of soil moisture sensor controlled irrigation systems Efficacy here is considered the ability to control the root zone soil moisture content with A) no rain and no crop, B) rain and no crop, C) rain and crop.

PAGE 105

105

PAGE 106

106 Figure 4 5 Root zone VWC over time with different probe locations and thresholds for automated irrig ation control Figure 4 6 Impact of probe location on the best threshold value for sensor controlled irrigation systems Threshold values are assessed based on the system ability to control the root zone soil moisture content with A) no r ain and no crop, B) rain and no crop, C) rain and crop.

PAGE 107

107

PAGE 108

108 CHAPTER 5 QUANTIFICATION OF GREENHOUSE GAS EMISSIONS FROM OPEN FIELD GROWN FLORIDA TOMATO PRODUCTION Background Atmospheric carbon dioxide (CO 2 ) concentrations have been rising over the past seve ral centuries from pre industrial levels (World Meteorological Organization, 2006). The atmospheric concentration of CO 2 had risen to 380 ppmv in 2006 compared to 280 ppmv in the 1700s (World Meteorological Organization, 2006). Existing projections consist ently estimate 2030 levels 25 90% higher than in the year 2000 (Rogner et al., 21st century projections by Bernstein et al. (2007) show a likelihood for increased f requency of drought, heavy rainfall events, heat waves, and cyclones due to elevated atmospheric levels of GHGs. Hanse n et al. (2006) considered a 1 C temperature increase above 2000 levels to be dangerous based on predictions that this would cause sea le vel rise that could damage human populations and cause species extinction. The Copenhagen Accord set a goal of keeping global temperature increases within 2 C of climate based on current emissions pledges, there is a greater than 50% probability that an increase o f greater than 3 C will occur. While the details of predictions are a subject for d ebate, due to uncertainty of and differences among climate projections, most in the scientific community agree that temperature increases of this magnitude would result in significant negative impacts and stresses on human developments and ecosystems (Fisc hlin et al., 2007; Meehl et al., 2007). Nevertheless, it is believed that these occurrences can be avoided with significant reductions in GHG emissions

PAGE 109

109 (Meinshausen et al., 2009). It is therefore important to understand emissions from various actions and r esources and to identify potential areas for emissions reductions. The primary source of GHG emissions is fossil fuel used for transportation, construction, and energy (Del Grosso et al., 2008). Fossil fuel burning accounts for more than 75% of global CO 2 emissions (Snyder et al., 2009). However, the agricultural sector is considered a significant contributor (Smith et al., 2007). In 2005, agriculture was estimated to produce 10 12% of the global anthropogenic GHG emissions (Smith et al., 2007). However, a relatively low proportion of agricultural emissions (13%) are as CO 2 (Del Grosso et al., 2008). Instead, agriculture contributed to approximately 60% of global anthropogenic nitrous oxide (N 2 O) emissions and 50% of global anthropogenic methane (CH 4 ) emissi ons (Smith et al., 2007). While other sectors contribute substantially greater total GHG emissions, mitigation in agriculture offers more cost effective options (Smith et al., 2007). Therefore, a reasonable first step in GHG reductions in agriculture is to quantify emissions from specific sources in production and identify the most economically sensible options for reduction. Studies have been conducted on the GHG emissions from many different agricultural products including wheat ( Triticum aestivum L.; Bre ntrup et al., 2004), sugar cane ( Saccharum officinarum L.; Barretto de Figueiredo et al., 2010), cotton ( Gossypium hirsutum L.; Weinheimer et al., 2010), milk (Casey and Holden, 2003), beef (Beauchemin et al., 2010), and corn ( Zea mays L.; Kendall and Chan ge, 2009). Most of these studies involved agronomic crops or livestock, while a limited number of them have focused on high value specialty vegetable crops, especially in the US.

PAGE 110

1 10 Fresh field grown tomato ( Solanum lycopersicum ) production has historically b agricultural receipts value (Florida Department of Agriculture and Consumer Servic es, 2010). Florida ranks second nationally in fresh market vegetable production with 92,000 ha planted, depending on the season, with a farm value of US $1.9 billion in 2008 2009 (Florida Department of Agriculture and Consumer Services, 2010). Reports of G HG emissions from tomato production typically describe indoor greenhouse production (Muoz et al., 2004; Antn et al., 2005), emissions from processing tomatoes (Andersson et al., 1998), or European production systems. Greenhouse tomato production is drast ically different from field grown tomato production. Processing tomato production differs from fresh market production by its use of a single mechanized harvesting event, lack of plastic mulched beds, and absence of plant staking or tying. However, reports of emissions for US field grown tomato production are currently unavailable in the literature. Greenhouse gas emissions have also been shown to vary largely depending on location and implementation of varying management practices. Thus, the objective of t his study was to quantify the GHG emissions of field grown Florida tomatoes for the fresh market under typical production practices. Additionally, potential areas for emissions reductions were identified. Estimation of Greenhouse Gas Emissions Description of Typical Tomato Production System in Florida Florida is an important production area for winter fresh market tomat oes in the US with more than 13 000 ha planted annually (Florida Department of Agriculture and Consumer Services, 2010). Depending on market conditions, statewide production

PAGE 111

111 value ranges from US $400 to $600 million annually. The tomato production system in Florida, which generally includes raised beds, polyethylene mulch, and seepage or drip irrigation, has been very effective in producing hi gh tomato yields with production costs of US $37 000 per hectare (Olson et al., 2010). In seepage production, fertilizer is applied pre plant during bed formation, while with drip it is applied both pre plant and during the season through the drip tubing a added to seepage irrigated production systems but can be labor intensive and costly. The crop is planted from transplants produced within the State near tomato producing areas. Transplants are four to five week s old depending on rate of development. Tomato beds are typically formed with 1.8 m spacing, with a plant spacing ranging from climate results in high insect, weed, and disease pressure that necessitate significant pesticide applications for protection of the high value crop. Transplants are in the ground for 16 19 weeks, although extended seasons occur under certain circumstances. Fresh tomato harvesting is done manually, with workers filling buckets and carrying them to harvest trucks. At the end of the season the polyethylene mulch is removed, herbicide is applied to the crop, and the desiccated crop is disked into the soil to ensure decomposition. System Boundaries In order t o quantify the GHG emissions from a season of typical Florida tomato production, the system boundaries needed to be determined. The emissions considered included those associated with the use of material inputs and farm operations, and include both direct emissions, which are emissions occurring on the farm due to production activities, and indirect emissions, which are released off site but occur due to

PAGE 112

112 the crop production. Farm operations account for both mobile and stationary tasks, while material emissi ons account for the GHGs released during the manufacturing, storage, packaging, and transportation of the material inputs. Thus, this analysis included primary and secondary emissions. For practicality purposes, tertiary emissions, which would include emis sions involved in the manufacturing of equipment and structures, were not considered. This was for several reasons. For one, it could be difficult to characterize the typical amounts of machinery, building area, and irrigation infrastructure typically used Second, appropriate emission factors that are accurate and updated are difficult to find in the literature (Roos et al., 2010). Lastly, many studies do not consider these tertiary emissions (Hillier et al., 2009; Kim et al., 2009; Robertson et al., 2000; Spreen et al., 2010), and in many cases, the tertiary emissions do not contribute significantly to overall emissions. Graboski (2002) estimated that less than 1% of the energy used in the production of corn ethanol is from the manufacturing of equipment a nd structures, such as irrigation systems and tractors. Ceschia et al. (2010) conducted a study of 11 crops in nine European countries and estimated the primary and tertiary emissions of various field operations. The tertiary emissions accounted for 3.3% o f the total emissions from the field operations, which accounts for an even smaller percentage of the total emissions. Considering the input intensiveness of Florida tomato production, it is expected that the tertiary emissions in these systems would be mi nor relative to the other emissions and high crop yields. The analysis took into account the delivery of transplants to the farm, the manufacturing, transportation, and storage of material inputs, farm operations including field preparation, planting, spra ying, irrigation, harvesting, terminating the crop, and transportation of seedlings to

PAGE 113

113 the farm and of fruit to the packing house. The analysis did not consider the production of the transplants, the packaging of fruits, or the delivery of fruits to market Input Data A survey of the Florida tomato industry was conducted from January to July of 2011 in order to characterize the typical production practices in the region and the fuel usage associated with these practices. The producers surveyed constituted 6 5% of the average planted area for tomato production in Florida, with 65% of the area under seepage irrigation and 35% of the area under drip irrigation. The survey was conducted through personal interviews, phone interviews, and mailed questionnaires. Gro wer information was grouped by irrigation method, with seepage and drip systems being the predominant methods, since the irrigation method has an impact on management factors and pumping characteristics. Pumping fuel requirements were also affected by the water source, as surface water required less energy compared to ground water per volume of water pumped. The information collected consisted of the diesel fuel used in mobile management activities involving farm machinery, fuel usage for the transportation of tomato seedlings and fruit, irrigation pumping fuel requirements, fertilizer and lime application rates, standard cultural practices, and crop yields This information is listed in Tables 5 1, 5 2, 5 3, 5 4, and 5 5. Fuel used for transportation was ba sed on vehicle fuel efficiency, transport distance, carrying capacity of the vehicle, and amount of material to be transported. The length of polyethylene mulch and drip tubing used was computed based on a 1.8 m bed spacing, which equates to 5,468 meters o f linear bed (lbm) per ha. The typical polyethylene mulch width used is 51 to 61 cm greater than the bed width to ensure adequate covering of the side of the beds and sufficient material for anchoring the mulch (Olson, 2011). The estimates used in these

PAGE 114

114 ca lculations assumed the use of polyethylene mulch with a width of 147 cm, 1 mm thickness, and weight of 0.023 kg/m 2 The drip tape was assumed to weigh 0.010 kg/m. The surveyed fresh market yields were in the range of attainable yields reported in the liter ature (Scholberg, 1996). Average amounts of insecticide, herbicide, fungicide, and fumigant were estimated based on interviews conducted prior to this study by Glades Crop Care, Inc. (1999) with growers in tomato producing areas of Florida. Bacteriophage w as not included in these calculations, as it is a biological control not a manufactured chemical, which would require specialized calculation of GHG emissions associated with its production. Calculations Due to the complexity and heterogeneity of vegetable production systems, detailed modeling requires detailed site specific information. The goal of this study was to provide a useful approximation of GHG emissions for a typical field tomato production operation in Florida while requiring only commonly avail able inputs. Since tomato production in Florida occurs with a range of growing conditions and production practices, general methods were used to quantify GHG emissions from the various sources. Emissions were expressed as CO 2 equivalents (CO 2 eq), which co nsiders the global warming potential of various emission forms using a common unit. Conversion between emission gases was made according to the standard suggested by the Intergovernmental Panel on Climate Change (IPCC; Eggleston et al., 2006). One unit of CO 2 emission was defined as one unit of CO 2 eq, while one unit of CH 4 emitted was equivalent to 25 units of CO 2 eq, and one unit of N 2 O was equivalent to 298 units of CO 2 eq. Absolute emissions represent the emissions associated with one tomato

PAGE 115

115 growing sea fruit yield. All resulting emissions estimates are calculated using average input, operation, and emission factor values. Total emissions, however, were computed for low, average, and high values, thus giving a range of estimated emissions. The emissions from crop agrochemical inputs were estimated using documented emissions values and input quantities. The agrochemicals, operations, and materials considered in the calculations are incl uded in Table s 5 1 5 2, 5 3, 5 4, and 5 5 The emissions occurring due to the manufacturing, storage, and transportation of these materials were estimated using emission factors presented by Lal (2004) and can be seen in Table 5 6 Uncertainties in these values are represented with the provision of low, average, and high values. Diesel fuel usage for farm management and transportation was converted into carbon (C) content and then CO 2 eq using conversion factors defined by the United States Environmental P rotection Agency (USEPA; 2005). Therefore, 1 L of diesel was assumed to have 0.73 kg of C, and one unit of C was assumed to have 3.67 units of CO 2 Carbon dioxide emissions from the burning of polyethylene mulch and drip tape, which is a common practice in the industry, were calculated using the USEPA (2010) methodology for estimating CO 2 emissions from plastic combustion. By this method, high density polyethylene (HDPE) or low density polyethylene (LDPE) were both assumed to be constituted of 86% C and to have an oxidation fraction of 98%. The CO 2 emitted was then calculated as the product of the amount of plastic, the C content, the fraction oxidized, and a C to CO 2 conversion factor. Additionally the manufacturing of

PAGE 116

116 plastic contributes GHG emissions in t he forms of CO 2 and CH 4 Emissions were computed according to IPCC guidelines (Eggleston et al., 2006), with emissions factors of 3 kg CO 2 /kg plastic and 1 kg CH 4 /kg plastic. In addition to the GHG emissions associated with the manufacturing and acquiremen t of N fertilizer, the application of N to the soil also results in the emission of GHG in the form of N 2 O. These emissions occur both directly (on site) through nitrification and denitrification, and indirectly (off site) following leaching, runoff, and a mmonia (NH 3 ) volatilization. While the emission of N 2 O due to applied N has been shown to vary site specifically (Snyder et al., 2009) as well as non linearly with application rate (McSwiney and Robertson, 2005), a linear response of emissions to N applica tion was assumed following the tier 1 IPCC guidelines (Eggleston et al., 2006). The emission factors used by these guidelines are shown in Table 5 7 Direct emissions were computed as the product of the direct N 2 O emissions factor and the amount of N appli ed. Indirect N 2 O emissions were broken down into those due to volatilization and those due to leaching or runoff. Each indirect emission path was calculated as the product of the amount of N applied, the fraction of N lost through that emission path, and t he emission factor for that path. The only applied N considered was as applied fertilizer. While N in irrigation water can be a significant N source, its contribution is considered minor when irrigation water has a nitrate (NO 3 ) concentration of less than 45 ppm (Tak et al., 2012). A study of shallow groundwater samples, which generally have significantly higher NO 3 concentrations than surface water samples (Mueller and Helsel, 1996), from agricultural lands in Florida and Georgia found NO 3 concentrations generally well below 45 ppm, with a median concentration of 4.2 ppm (Berndt et al.,

PAGE 117

117 1998). Thus, irrigation water was ignored as a source of N in the GHG emissions calculations. Due to typically intense tillage practices and low initial soil organic matter that is typical of tomato production in this area, it was assumed that the soil neither sequestered nor released C. Thus, all tomato stem and root biomass was not considered as a C sink. It was assumed that tomato fruit would be respired following consump tion. Therefore, C captured as fruit was also not considered a C sink in the calculations. Emissions Estimates The total average estimated GHG emissions due to the manufacture, storage, and transport of agrochemicals for seepage and drip irrigation systems were 9,728 kg CO 2 eq ha 1 and 9,889 kg CO 2 eq ha 1 respectively as shown in Figure 5 1. The largest contributors were N fertilizer, fungicide, and soil fumigant, accounting for 16.7%, 16.9%, and 33.8% of the average agrochemical emissions in drip system s, and 15.4%, 17.2%, and 34.4% of average emissions in seepage systems. Phosphorus (P) and potassium fertilizer resulted in low GHG emissions, accounting for 0.7% and 3.1% of the average agrochemical related emissions in drip systems, and 0.8% and 3.2% in seepage systems. All of the agrochemical GHG emissions were relatively equivalent between the two irrigation systems except the N fertilizer, which was 1,656 kg CO 2 eq ha 1 in the drip system compared to 1,496 kg CO 2 eq ha 1 in the seepage system. This is due to the typically higher application rates of N observed in drip compared to seepage systems, despite the potential for greater N use efficiency in drip systems compared to seepage systems.

PAGE 118

118 The total GHG emissions from the operation of farm machinery wa s 1,276 kg CO 2 eq ha 1 for seepage irrigation systems and 1,180 kg CO 2 eq ha 1 for drip sys tems, as can be seen in Figure 5 2 The difference in emissions between the two irrigation systems is due to ditch maintenance being required in seepage systems whil e it is unnecessary in drip systems, which do not use ditches for supplying irrigation water. The field leveling, fungicide and insecticide spraying, and bed lying are the largest GHG emission categories, accounting for 19.7%, 16.8%, and 11.8% of the total machine operation emissions in seepage systems, and 21.3%, 18.1%, and 12.8% of the emissions in drip systems. Additional GHGs were emitted because of chemical conversions in the field. The total GHGs released as field emissions were 3,089 kg CO 2 eq ha 1 f or drip systems and 2,865 kg CO 2 eq ha 1 for seepage systems. The distribution of these emissions through different pathways is shown in Figure 5 3. The largest amount of GHGs was emitted through nitrification and denitrification, which accounted for 51.3% of all field emissions in average seepage systems and 52.7% in drip systems. Emissions that resulted from the field were considered direct emissions, while losses of substances from the field that are assumed to result in emissions away from the field are considered indirect emissions. Since lime dissolution results in the emission of CO 2 from the field, GHGs emitted through this pathway were considered direct emissions. Nitrification and denitrification occur within the soil and result in losses of N 2 O di rectly from the field; therefore, they were also considered a direct emission. Volatilization results in the loss of NH 3 from the field. Ammonia is not a GHG, but some of this N in the atmosphere can return to the soil through atmospheric deposition, of wh ich a certain amount will be

PAGE 119

119 nitrified, denitrified, or lost as N 2 O. Since this release is occurring off the farm, this is considered an indirect emission. Similarly, N lost through leaching, erosion, or runoff is not directly lost from the field as GHGs. However, some of this N will eventually be returned to a field, where some of it will be nitrified or denitrified, resulting in the release of N 2 O. Thus, this pathway also results in indirect emissions. Indirect emissions accounted for 21.8% of the field e missions in seepage systems and 22.4% of these emissions in drip systems. The main source of field emissions was N fertilizer, as 73.1% and 75.1% of the field emissions were due to N fertilizer in the seepage and drip irrigation systems, respectively. The difference in emissions between the two irrigation systems was due to the typically higher use of N fertilizer in drip irrigation systems, as mentioned previously. The use of pumps for supplying irrigation water resulted in GHG emissions dependent on both the irrigation type and water source used. The overall average emissions results are shown in Figure 5 4. Systems with drip irrigation and surface water had emissions of 996 kg CO 2 eq ha 1 Systems with drip irrigation and well water had emissions of 5,976 kg CO 2 eq ha 1 Systems with seepage irrigation and surface water had emissions of 449 kg CO 2 eq ha 1 while systems with seepage irrigation and well water had emissions of 2,696 kg CO 2 eq ha 1 The use of plastic also results in the emission of GHGs. The se emissions are due to both the manufacturing of the plastic and the disposal of the plastic (i.e., burning). This resulted in the emission of 735 kg CO 2 eq ha 1 in the manufacture of plastic for drip irrigation systems compared to 571 kg CO 2 eq ha 1 for seepage irrigation systems as well as the emission of 751 kg CO 2 eq ha 1 from burning of the plastic for

PAGE 120

120 drip irrigation systems compared to 583 kg CO 2 eq ha 1 in seepage irrigation systems. Overall, plastic manufacturing and burning results in the emissio n of 1,486 kg CO 2 eq ha 1 in drip irrigation systems compared to 1,154 kg CO 2 eq ha 1 in seepage irrigation systems. The difference in emissions between the two systems is due to the absence of drip tape in seepage systems, which results in less plastic ma nufacturing and burning, while both systems utilize polyethylene mulch. The transport of transplants from the nursery to the field accounted for a negligible amount (2 kg CO 2 eq ha 1 ) of GHG emissions due to the large number of seedlings that could be carr ied on each truck as well as the proximity of transplant houses to tomato production areas. The transport of the fruit from the field to the packinghouse, however, resulted in the emission of 708 kg CO 2 eq ha 1 Overall, the least GHGs were emitted from th e seepage irrigation system with a surface water supply, which resulted in an average estimated emission of 16,183 kg CO 2 eq ha 1 or 0.19 kg CO 2 eq kg fruit 1 Seepage irrigation systems with well water supplies resulted in greater emissions of 18,429 kg C O 2 eq ha 1 or 0.22 kg CO 2 eq kg fruit 1. Drip irrigation systems with surface water supplies resulted in comparable emissions of 17,446 kg CO 2 eq ha 1 or 0.21 kg CO 2 eq kg fruit 1 while drip irrigation systems with well water supplies resulted in the larg est emissions of 22,426 kg CO 2 eq ha 1 or 0.27 kg CO 2 eq kg fruit 1 Thus the average total GHG emissions from the different irrigation type water source combinations ranged from 16,183 kg CO 2 eq ha 1 (0.19 kg CO 2 eq kg fruit 1 ) to 22,426 kg CO 2 eq ha 1 (0 .27 kg CO 2 eq kg fruit 1 ). Considering the uncertainty range of the emission factors, the total GHG emissions ranged from 8,267 kg CO 2 eq ha 1 (0.10 kg CO 2 eq kg fruit 1 ) to 40,307 kg CO 2 eq ha 1

PAGE 121

121 (0.48 kg CO 2 eq kg fruit 1 ). Finally, when accounting for th e uncertainty range of the emission factors and of the production practices, the total GHG emissions ranged from 6,318 kg CO 2 eq ha 1 (0.06 kg CO 2 eq kg fruit 1 ) to 52,813 kg CO 2 eq ha 1 (0.75 kg CO 2 eq kg fruit 1 ). Implications The GHG emissions estimated from this study for fresh field grown Florida tomato production are consistent with emissions from production of other fruit and vegetable crops. The emissions estimates of this study are similar to estimates of 0.24 0.48 kg CO 2 eq kg pineapple 1 for pine apple ( Ananas comosus ) produced in Costa Rica (Ingwersen, 2012), 0.27 1.36 kg CO 2 eq kg strawberry 1 for strawberries ( Fragaria ananassa ) produced in Spain and the United Kingdom (Mordini et al., 2009), and 0.08 0.33 kg CO 2 eq kg orange 1 for oranges ( Cit rus sinensi ) produced in Spain, Italy, and of 0.10 0.16 kg CO 2 eq kg potato 1 for potato ( Solanum tuberosum ) produced in Sweden (Roos et al., 2010) and 0.10 0.14 kg CO 2 eq kg carrot 1 for carrot ( Daucus carota ) produced in Denmark, the Netherlands, Germany, Great Britain, Italy, and grain and oilseed crops, with estimates of 0.45 0.52 k g CO 2 eq kg seed 1 for oilseeds ( Brassica napus Brassica rapa Brassica juncea Brassica juncea Sinapis alba ) grown in Canada (Gan et al., 2012), 0.80 kg CO 2 eq kg wheat 1 for wheat produced in the United Kingdom (Williams et al., 2006), 1.7 kg CO 2 eq kg seed 1 for rape seed ( Brassica napus ) produced in the United Kingdom (Williams et al., 2006), and 0.25 0.82 kg CO 2 eq kg corn 1 for corn produced in the United States (Kim et al., 2009). Studies of tomatoes grown in many European countries reported emissi ons of 0.43 9.4 kg CO 2 eq

PAGE 122

122 kg tomato 1 (Carlsson, 1997; Williams et al., 2006). The upper end of this emissions having occurred in greenhouses in cold climates where the domi nant emissions source was heating. The lower end of this emissions range was estimated from tomato production in Spain, where open field production is used. These estimates were similar missions range. Results indicated that irrigation and agrochemicals were the leading categories for GHG emissions, with irrigation accounting for 2.8% to 26.6% of average emissions estimates and agrochemicals accounting for 44.1% to 60.1%, depending on irr igation method and water source. For systems with lower pumping demand, agrochemicals become an increasingly dominant factor for GHG emissions. For systems with high pumping demand, irrigation becomes a greater emissions source. Field GHG emissions were th e third largest emissions category, ranging from 13.8% to 17.7% of the average estimated total GHG emissions depending on irrigation method and water source. Considering that 73.1% to 75.1% of these GHG emissions were due to N fertilizer, and that 15.4% to 16.7% of the average agrochemical related emissions are due to N fertilizer, it should be noted that the impact of N fertilizer could be significant. This underlines the importance of efficient N and irrigation management. Recent surveys show that common industry practice is to apply N at significantly higher rates than recommended, especially in southern areas of Florida, on seepage irrigated fields, and during the fall season, which is characterized by high temperatures and rainfall (Cantliffe et al., 20 09 ). The current N rate recommendation by the University of Florida/Institute of Food and Agricultural Sciences (UF/IFAS) is 224 kg N/ha plus a

PAGE 123

123 13.7 kg N/ha supplemental N application for tomato grown at any location, using any irrigation method and any pl anting season (Olson et al., 2010). N application rates observed in this study exceeded this recommendation. Considering that between 17.7% and 22.8% of the average total GHG emissions in this study were due to N fertilizer, it is evident that major emissi on reductions could be obtained by more efficient N fertilizer management. The potential for savings can be significant as the recommended rates are intended to reflect the crop nutrient requirement. This means that any residual plant available N in the so il or supplied from irrigation water or other non fertilizer sources will contribute an amount that does not need to be supplied as fertilizer and thus can be subtracted from the recommended N rate application. Thus, the excess fertilizer applied is likely underestimated by the simple difference between the actual application rate and the maximum recommended rate. Additionally, there is evidence that N 2 O emissions increase rapidly when fertilizer is applied beyond yield response (McSwiney and Robertson, 200 5), thus the potential for savings may be underestimated by this study. Realistically, the N application rates cannot be expected to match the recommended rates as it is uncertain whether recommendations accurately reflect the dynamics occurring under part icular growing conditions. Numerous studies have reported that temporary water table rises during heavy rainfall periods and/or for frost protection events, different lengths of growing season, the use of vigorous hybrid cultivars, and/or denitrification j ustify the use of fertilizer beyond the recommended rates (Simonne and Ozores Hampton, 2006; Fraisse et al., 2010). Additionally, extra fertilizer offers an inexpensive insurance at a cost that is offset by very small yield increases (Cantliffe et al., 200 9 ). It has been suggested that the situation can be improved,

PAGE 124

124 gradually and iteratively, through the creation of flexible recommendations reflective of farm operational realities and continual communication with producers (Simonne and Ozores Hampton, 2006) Therefore, improved management could result not only in the typically motivating reductions in N runoff and leaching, but also in reduced GHG emissions to mitigate the severity of adverse impacts of global climate change. While N management is impactful on GHG emissions from tomato production, irrigation management is even more so, as its influence is two fold. First, irrigation accounted for significant GHG emissions, ranging from 2.8% to 26.6% of average estimated total emissions. Second, irrigation has a fundamental effect on nutrient management, as an ample amount of nutrients will be rendered insufficient if the water is improperly managed. Simonne and Ozores Hampton (2006) suggest that pollution reductions should be approached with the goal of improv ing water management rather than reducing N fertilizer application. Nitrate moves rapidly with water, and under drip irrigation it has been shown that most of the nitrate collects near the fringe of the wetting front (Li et al., 2003). In Florida, where sa ndy soils are common, the low water holding capacities and high hydraulic conductivities encourage more vertical movement of the wetting front. Therefore, drip systems must be managed to supply adequate amounts of water while containing NO 3 within the root zone and minimizing leaching. Under seepage systems, the water table should be maintained at appropriate depths throughout the season based on the crop development stage. Shukla and Jaber (2005) demonstrated that under seepage irrigation, the water table in soils typical to South Florida raised an average of 16 times the rainfall amount. Pollutant outflows from such production systems have become an increasing public concern as such production

PAGE 125

125 areas have been considered a major contributor to surface water quality degradation (Breve et al., 1997). Appropriate irrigation management can help reduce some of these undesirable contributions, as well as reducing GHG emissions from pumping and N fertilizer application. In such a highly interactive system, it is in teresting to examine the consequences of particular management decisions. The use of polyethylene mulch, for example, accounts for between 5.1% and 7.1% of the average estimated total GHG emissions. However, in many other ways, polyethylene mulch applicati ons help reduce GHG emissions. Polyethylene mulched beds have been shown to reduce N leaching due to heavy rain events by promoting water flow around, rather than through, the fertilized bed (Hochmuth et al., 2008). This can reduce the fertilizer rate need ed and reduce the indirect GHG emissions from leached N. Polyethylene mulch has also been shown to reduce evapotranspiration (ET) 10% to 30% (Simonne Dukes et al., 2010), which allows for reduced irrigation requirements and thus reduced GHG emissions due to pumping. Additionally, polyethylene mulch reduces weed growth and increases crop yields, allows for reduced herbicide application and greater yield per unit of emission. Thus, overall, the use of polyethylene mulch represents a net GHG emission savings despite the emissions associated with its manufacturing and disposal. Nevertheless, approximately half of the GHG emissions associated with polyethylene mulch were released from the burning of the polyethylene mulch for disposal. Alternative methods of dis posal, especially ones that could recycle or re purpose it, could further reduce production related GHG emissions.

PAGE 126

126 Intensive agricultural production results in large emissions per unit area of production. However, when intensive production results in eleva ted yields, it can result in more efficient crop production from an emissions standpoint. The impact of high yields is two fold, as higher yields also led to a reduction of the emissions per unit weight of fruit produced. It has been shown that intensive h igh yielding production systems result in reductions in net reduction in GHG emission per unit of production compared to less intensive production (Burney et al., 2010). Maintaining highly productive land or increasing the productivity of land can also red uce the need to develop new lands. This can prevent significant amounts of GHG emissions, as typically between 30 and 50% of the carbon present in native ecosystems is lost when converted to agriculture (Guo and Gifford, 2002). Thus, management that mainta ins optimal yields should be essential, from both an economic and a GHG emission perspective. These calculations were intended to provide a simplified, practicable estimate of the GHG emissions from these production systems. Yet, there is certainly a great degree of uncertainty in many of the estimates. For example, the GHG emissions estimates for the manufacturing, transportation, and storage of N used estimates from Lal (2004), which were based on a range of values reported in the literature. However, the se estimates may not accurately reflect other systems due to the particular N fertilizer type, the factory efficiency, and locations of the factory, distributor, and farming operation specific to each system. For particular systems, these details could be determined more accurately, and could be entered into the framework of this methodology on a site specific basis.

PAGE 127

127 However, the GHG emissions due to the application of N were calculated based on a simplified representation of the system dynamics, and a more complex methodology would be needed to capture the dynamics that vary site specifically. Denitrification typically converts NO 3 to N 2 but an amount of N is released as N 2 O gas at a rate depending on dynamics involving NO 3 concentration, water content, temperature, C and other factors (Bedard Haughn, 2006). Nitrification typically converts NH 4 + into NO 3 but when nitrifying bacteria is oxygen limited, nitrite is used as a terminal electron acceptor instead, resulting in the release of nitrous oxide an d nitric oxide as well ( International Fertilizer Industry Association and the Food and Agriculture Organization of the United Nations 2001). Further, indirect GHG emissions can be created from N volatilization that will eventually be returned to soils thr ough deposition and a portion eventually emitted as N 2 O, and through NO 3 runoff and leaching, which will result in N 2 O emissions offsite through denitrification (Del Grosso et al., 2006). The overall rate of N 2 O production can vary largely and depends on site specific conditions including climate, N fertilizer management, soil drainage, soil texture, soil pH, soil organic matter, crop type, tillage, and irrigation ( International Fertilizer Industry Association and the Food and Agriculture Organization of t he United Nations 2001). As such, site specific estimates would need to consider more detailed information than was intended for the purpose of this study. Thus a simplified approach was adopted, as has been done in similar studies (e.g., Hillier et al., 2011), in order to provide a reasonable yet feasible estimate. However, it could be argued that higher GHG emission factors should be chosen due to the warm, humid climate, intensive fertilizer rates, elevated water tables, and sandy soils prevalent in typ ical Florida growing conditions. It could

PAGE 128

128 also be argued that denitrification rates should be higher under seepage systems compared to drip irrigation due to the greater volume of soil being held at high water contents. Nevertheless, detailed measurements would be necessary to dictate such adjustments. Summary The average estimated total GHG emissions associated with typical production of fresh open field grown Florida tomatoes was estimated to range from 16,183 kg CO 2 eq ha 1 (0.19 kg CO 2 eq kg fruit 1 ) to 22,426 kg CO 2 eq ha 1 (0.27 kg CO 2 eq kg fruit 1 ), depending on irrigation type and water source. Considering uncertainties in the emission factors, the average estimated emissions ranged from 8,267 kg CO 2 eq ha 1 (0.10 kg CO 2 eq kg fruit 1 ) to 40,307 kg CO 2 eq ha 1 (0.48 kg CO 2 eq kg fruit 1 ). While these represent typical GHG emissions, site specific information could be used within this methodology to estimate more accurately the GHG emissions of a particular tomato producer. Emissions estimates consi dering the range of production practices and the emission factor uncertainty ranged from 6,318 kg CO 2 eq ha 1 (0.06 kg CO 2 eq kg fruit 1) to 52,813 kg CO 2 eq ha 1 (0.75 kg CO 2 eq kg fruit 1 ). Similar methodologies could also be applied to other crops and production areas. Improvements in irrigation management, N management, and polyethylene mulch disposal were identified as areas with the greatest and most feasible potential for GHG emissions reductions.

PAGE 129

129 Table 5 1 Diesel fuel use by machine operation f or t ypical Florida tomato production practices Data d etermined from Florida tomato grower survey Machine O perations Diesel ( l/ha ) Low Average High Disking 18.7 35.5 35.5 False bedding 18.7 37.4 37.4 Bottom fertilizer mix application 16.8 22.4 22.4 Making beds 35.5 56.1 56.1 Plastic application 9.4 18.7 18.7 Cutting ditches 9.4 11.2 11.2 Punching holes 7.0 9.4 9.4 Planting 7.0 9.4 9.4 Driving stakes 15.0 15.0 28.1 Ditch Maintenance 14.0 35.5 35.5 Fungicide & insecticide 53.0 79.5 79.5 Herbic ide 9.2 13.1 13.1 Drip Surface Water 251.7 370.1 575.8 Drip Well Water 1,510.1 2,220.7 3,455.1 Seepage Surface Water 133.6 167.0 225.9 Seepage Well Water 801.4 1,001.8 1,355.4 Leveling 74.8 93.5 93.5 Mowing 7.0 9.4 9.4 Lifting plastic 25.7 2 8.1 28.1 Table 5 2. Agrochemical use for t ypical Florida tomato production practices Data determined from Florida tomato grower survey Agrochemical Amount Applied (kg/ha) Low Average High Nitrogen (Drip) 291 347 404 Nitrogen (Seepage) 269 314 359 Phosphorus 67 101 135 Potassium 448 560 673 Fungicide 45 a 117 a 204 a Insecticide 41 a 47 a 55 a Herbicide 39 a 42 a 45 a Fumigant 117 a 234 a 336 a Lime 1121 1681 2242 a Denotes active ingredient

PAGE 130

130 Table 5 3. Plastic use for t ypical Florida tomat o production practices Data determined from Florida tomato grower survey Plastic material Amount Meters of linear bed per hectare 5,468 Polyethylene mulch width (cm) 147 Polyethylene mulch weight (kg/m 2 ) 0.023 Drip tube weight (kg/m) 0.010 Plastic disposal method Burning Table 5 4. Transportation practices for t ypical Florida tomato production practices Data determined from Florida tomato grower survey Transportation detail Amount Number of seedlings per truck 169,400 Distance from nursery t o field (km) 64 Truck fuel efficiency (km/l) 5.1 Bins of fruit shipped per truck (1 bin = 227 kg) 48 Distance from field to packing house (km) 72 Truck fuel efficiency (km/l) 2.1 Pallets Shipped per truck 24 Boxes of fruit per pallet (1 box = 11.3 kg ) 70 Truck fuel efficiency (km/l) 2.1 Table 5 5. Crop production details for t ypical Florida tomato production Data determined from Florida tomato grower survey Production Low Average High Plant density (plants/ha) 8,270 9,630 11,860 Crop yield (k g/ha) 70,000 84,000 98,000 Growing period (weeks) 16 17 20 Table 5 6 C arbon emissions for the manufacturing, transporation, and storage of input materials or agrochemicals. Includes low, average, and high value estimates according to Lal (2004). Mate rial Carbon emission (kg C/kg substance) Low Average High Nitrogen fertilizer 0.9 0 1.3 0 1.8 0 Phosphorus fertilizer 0.1 0 0.2 0 0.3 0 Potassium fertilizer 0.1 0 0.15 0.2 0 Lime 0.03 0.16 0.23 Herbicide 1.7 0 6.3 0 12.6 Insecticide 1.2 0 5.1 0 8.1 0 Fungicide /fumigant 1.2 0 3.9 0 8.0 0

PAGE 131

131 Table 5 7 E mission factors used for estimating N 2 O emissions from N fertilizer and lime Includes low, average, and high estimates derived from the IPCC (Eggleston et al., 2006) guidelines. Emission Factor Low Average High Em ission factor for direct N 2 O emissions 0.003 0 0.010 0 0.030 0 Fraction of N fertilizer that volatilizes 0.05 00 0.20 00 0.50 00 Emission factor for volatilized N 0.002 0 0.010 0 0.050 0 Fraction of N fertilizer that leaches 0.1 000 0.3 000 0.8 000 Emission factor for leached N 0.0005 0.0075 0.0250 Emission factor for applied lime 0.120 0 0.125 0 0.130 0 Figure 5 1 Greenhouse gas emissions due to production, transportation, and storage of agrochemicals Emissions are estimated for typical drip and seepage irri gated tomato production systems in South Florida.

PAGE 132

132 Figure 5 2 Greenhouse gas emissions due to farm machinery operation s. Emissions are estimated for typical tomato production systems in South Florida.

PAGE 133

133 Figure 5 3 Greenhouse gas emissions from field losses Emissions are for typical drip and seepage irrigated tomato production systems in South Florida.

PAGE 134

134 Figure 5 4 Total greenhouse gas emissions estimates. Emissions are reported for drip and seepage irrigated tomato productions systems with surfac e and well water sources.

PAGE 135

135 CHAPTER 6 CONCLUSIONS This research consisted of the development and utilization of methods for quantitatively estimating the impacts of agricultural production and management on production efficiency and environmental impacts A n emphasis was placed on drip irrigation management, which is an already common technology in Florida vegetable production that is likely to become more widespread as pressure grows to improve production efficiency. While much of this research focused on i rrigation management, emphasis was also placed on the complex interactions within these systems, including the intrinsic link between soil water dynamics and soil nitrogen dynamics, both of which have major impacts on the GHG emissions associated with prod uction. Specifically, a two dimensional water balance model was implemented within the DSSAT CSM by a group of DSSAT developers in order to allow this widely used software to simulate wat er limited production under drip irrigation and production practices typical for vegetable production in Florida The water balance was designed to provide a practical, simple, and computationally efficient solution. As such, flexibility was allowed for the input requirements to operate the model, with an approximate parame terization method being created to allow the model to be run using only the standard DSSAT soil hydraulic property inputs. The DSSAT 2D model was evaluated against lab measurements, benchmark models, and field measured data to assess its ability to simula te soil water dynamics under drip irrigation regimes The evaluation of the three point DSSAT parameter estimation methodology indicated that overall the methodology provided unbiased characterization of SMRCs. However, the limited range of measurements ut ilized for

PAGE 136

136 these estimates reduced the predictive ability As such, predictions were shown to perform more accurately at tensions near the three measurements points and to perform much less accurately at tensions distant from these measurement points. Char acterization of SMRCs was improved when estimated using the RETC software, which utilized all available SMRC measurements. Thus, detailed measurements of soil hydraulic properties would be expected to improve the accuracy of the model parameter estimations and, consequentially, soil moisture simulations. Benchmark comparison of DSSAT 2D against HYDRUS 2D revealed great similarity between the two models for uniformly textured soils These similarities indicated that the different numerical solving technique s, boundary conditions for representing drip tape, and discretization of the soil simulation space of the two models resulted in comparable predicted outcomes However, the large and persistent divergence between the model simulations for layered soils ind icated that for sharply layered soils, DSSAT 2D simulations should be considered with caution. Steps were taken to improve simulation performance for layered soils. However, while these modifications were shown to improve the model performance, the adjuste d model still produced simulations that were consistently divergent from the expected system behavior in layered soils. Field trials were conducted in order to evaluate the combined parameterization and soil water flow calculation methodologies of DSSAT 2 D in a n applied setting. The e valuation indicated that overall, the model sufficiently simulates the soil water dynamics of an unplanted, drip irrigated and fairly uniform soil. Model runs using the parameters estimated from more detailed soil hydraulic property measurements were

PAGE 137

137 shown to improve the accuracy of the soil moisture simulations compared to simulations using the parameters estimated using the three point DSSAT methodology. However, simulations using both parameter sets were representative of the measured soil water dynamics. When the uncertainty in the soil water measurements is considered, the model simulations appear increasingly apt. Model simulation experiments yielded many implications for drip irrigation mana gement. M odel simulations ind icated that the soil moisture dynamics of a particular soil are dominated by the irrigation amount and ET demand Thus managers will increase their water use efficiency the greatest if they can apply irrigation in amounts synchronous with crop demand and in consideration of the soil conditions and hydraulic properties Simulation experiments did indicate that, while of much less import ance water application splitting and irrigation application rate do have an effect on soil moisture dynamics L eaching of water from the root zone was consistently reduced at lower emitter rates. Splitting applications generally reduced water loss from the root zone, but the effect depended on the manner in which the splitting method distributed the irrigation events rather t han the actual number of irrigation events over which the method split the total irrigation Simulation experiments were also conducted to consider the impact of different implementations of automated soil moisture sensor based drip irrigation systems. A methodology was demonstrated for the site specific identification of optimal probe locations and soil moisture threshold values. This methodology can be applied for other specific production sites and setups using automated drip systems. The analysis also identified the impacts that rainfall, ET rate and distribution of root water uptake have on

PAGE 138

138 the ability of automated irrigation to control the soil moisture within desirable ranges. The impact of the implementation of automated soil moisture sensors on wa ter use and ability to maintain desirable root water content is further indication that these systems must be managed properly in order to maximize the potential benefits. Whi le the focus of this research was on irrigation efficiency, agricultural producti on systems are very interactive. Soil water flow is known to have a strong effect on soil nutrient movement. This is especially true for nitrate N, which moves rapidly with water and has been shown to tend to collect around the extent of the wetting front (Li et al., 2003). Irrigation has a strong impact on soil water dynamics. This is especially true under plastic mulched raised beds, which reduce the influence of rainfall on the soil water regime. Thus, irrigation management practices which more efficient ly provide the necessary root available water for optimal crop growth while reducing the flow of water beyond the root zone will meet producer needs while reducing N leaching. This will result in reduced water use and reduced N loss, which in turn will all ow crop nutrient requirements to be met with lower nutrient application rates. While soil nutrient movement is strongly related to soil water movement, GHG emissions from agricultural production are a function of all aspect s of production. M ethodologies w ere aggregated for estimating GHG emissions from typical open field tomato production. This provided a framework for estimating the emissions impact of different production management options Based on the typical Florida producer, irrigation and nitrogen fertilization were identified as the most impactful management areas that could be improved in order to reduce emissions while maintaining optimal yields. Reductions in water application will result in direct emissions reductions due to

PAGE 139

139 reduced pumping req uirements Similarly, reductions in nitrogen fertilizer application will result in reduced emissions due to the production, transport, and storage of the fertilizer. However, improved irrigation management can additionally result in reduced N leaching. The se reduced N losses allow N applications to be reduced while maintaining similar soil N levels, which reduces the direct emissions from the production, transport, and storage of the fertilizer. Additionally, reduced N losses will result in reduced indirect GHG emissions associated with the eventual conversion of some of the leached N to N 2 O. This underscores the importance of efficient irrigation for improving agricultural efficiency by many measures Overall, the DSSAT 2D CSM provides a tool that can be u sed to improve irrigation management as well as address various production issues that require a more detailed water balance model for adequate system characterization It is hoped that the model can aid in the process of improving the efficiency of vegeta ble production. Applications of the model in this research focused mainly on investigating the impact of irrigation management practices on soil water dynamics, identifying practices which are believed to allow water to be applied in a manner that more eff iciently maintain desirable root zone soil moisture levels with minimal flow of water beyond the root zone Further, improvements in irrigation efficiency that could be realized based on these identified practices should result in benefits beyond reduced w ater use. Improvements in irrigation efficiency should also result in reduced nutrient losses, reduced nitrogen application rates, and reduced GHG emissions. Future development of the DSSAT 2D model should include refinement of the crop growth and developm ent routines within the 2D model as well as developing and evaluating a model for soil

PAGE 140

140 nitrogen dynamics This would allow the model to be applied for further investigation of irrigation and fertilizer management as they relate to crop prod uction and envir onmental impact

PAGE 141

141 APPENDIX A DSSAT 2D AND HYDRUS 2D GRID LAYOUTS Figure A 1 Grid layout for the HYDRUS 2D simulations that were used for comparison with the DSSAT 2D simulations. Note the grid goes from 0 to 80 cm horizontally and from 0 to 205 cm ver tically.

PAGE 142

142 Figure A 2 Grid layout for the DSSAT 2D simulations that were used for comparison with the HYDRUS 2D simulations. Note the grid goes from 0 to 80 cm horizontally and from 0 to 205 cm vertically.

PAGE 143

143 APPENDIX B CODE FOR EXTRACTING PROBE VWC FROM GRIDDED DSSAT 2D VWC VALUES The following is the R code used for computing the inte rsection weights between the sensing area of a probe and the grids in the DSSAT 2D grid space. This example is for a probe placed 0 cm from the row center and 7.5 cm from th e soil surface with a sensing volume of 7.5 cm. library(gpclib) xGrids < c( 0, 5, 10, 15, 20, 25, 30, 35, 40 ) yGrids < c( 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 75, 85, 95, 105, 115, 125, 135, 145, 155, 165, 175, 185, 195, 205 ) probeX < 0.0 probeY < 7.5 radius = 7.5 getCircle < function(myX, myY, myR) { x < array(0,360) y < array(0,360) for(i in 0:360) { th=i*2*pi/360 x[i] < myX + myR*cos(th) y[i] < myY + myR*sin(th) } return(cbind(x,y)) } fillPercentage < array( 0, c( length(yGrids) 1, length(xGrids) 1 ) ) intersectedArea < array( 0, c( length(yGrids) 1, length(xGrids) 1 ) ) probeAreaPercentage < array( 0, c( length(yGrids) 1, length(xGrids) 1 ) ) myCircle < getCircle(probeX, probeY, radius) myCircle < as(myCircle, "gpc.poly") myPlot < myCircle for(j in 1:(length(yGrids) 1)) { for(k in 1:(length(xGrids) 1)) { myRectangle < cbind( c(xGrids[k],xGrids[k],xGrids[k+1],xGrids[k+1]) c(yGrids[j],yGrids[j+1],yGrids[j+1],yGrids[j]) ) myRectangle < as(myRe ctangle, "gpc.poly") intersectionArea < area.poly(intersect(myCircle,myRectangle)) intersectedArea[j,k] < intersectionArea rectangleArea < area.poly(myRectangle) fillPercentage[j,k] < intersectionArea/rectangleArea

PAGE 144

144 myPlot < append.poly(myPlot,myRectan gle) } } for(j in 1:(length(yGrids) 1)) { for(k in 1:(length(xGrids) 1)) { probeAreaPercentage[j,k] < ( intersectedArea[j,k] / sum(intersectedArea[,]) ) } } sink("AutoProbeWeights.txt", append=FALSE, split=FALSE) cat("Row,Column,Weight") for(j in 1:(lengt h(yGrids) 1)) { for(k in 1:(length(xGrids) 1)) { if(probeAreaPercentage[j,k]>0){ cat(" \ n") cat(j,k,probeAreaPercentage[j,k],sep=",") } } } sink()

PAGE 145

145 APPENDIX C TIME EVOLUTION GRAPHS OF THE MEASURED AND SIMULATED VWC VALUES AT EACH PROBE MEASUREMENT LOCATION Figure C 1 Time evolution at each probe location of the measured and simulated VWCs using DSSAT 2D with the three point DSSAT estimated soil parameters. Plots for each probe location include the predicted VWCs, average probe measured VWCs, and 95% prob e measured VWC confidence intervals vs. time for treatments A) 30 H 1 rep 1, B) 30 H 2 rep 1, C) 30 L 1 rep 1, D) 30 L 2 rep 1, E) 20 H 1 rep 1, F) 20 H 2 rep 1, G) 10 H 2 rep 1, H)10 H 1 rep 1, I) 20 L 1 rep 1, J) 20 L 2 rep 1, K) 10 L 2 rep 1, L) 10 L 1 rep 1, M) 10 H 2 rep 2, N) 10 H 2 rep 3, O) 10 H 1 rep 2, P) 10 H 2 rep 3, Q) 30 L 1 rep 2, R) 30 L 1 rep 3, S) 30 L 2 rep 2, T) 30 L 2 rep 3, U) 20 H 1 rep 2, V) 20 H 2 rep 2.

PAGE 146

146

PAGE 147

147

PAGE 148

148

PAGE 149

149

PAGE 150

150

PAGE 151

151

PAGE 152

152

PAGE 153

153

PAGE 154

154

PAGE 155

155

PAGE 156

156

PAGE 157

157

PAGE 158

158

PAGE 159

159

PAGE 160

160

PAGE 161

161

PAGE 162

162

PAGE 163

163

PAGE 164

164

PAGE 165

165

PAGE 166

166

PAGE 167

167

PAGE 168

168 Figure C 2 Time evolution at ea ch probe location of the measured and simulated VWCs using DSSAT 2D with the RETC estimated soil parameters. Plots for each probe location include the predicted VWCs, average probe measured VWCs, and 95% probe measured VWC confidence intervals vs. time for treatments A) 30 H 1 rep 1, B) 30 H 2 rep 1, C) 30 L 1 rep 1, D) 30 L 2 rep 1, E) 20 H 1 rep 1, F) 20 H 2 rep 1, G) 10 H 2 rep 1, H)10 H 1 rep 1, I) 20 L 1 rep 1, J) 20 L 2 rep 1, K) 10 L 2 rep 1, L) 10 L 1 rep 1, M) 10 H 2 rep 2, N) 10 H 2 rep 3, O) 10 H 1 rep 2, P) 10 H 2 rep 3, Q) 30 L 1 rep 2, R) 30 L 1 rep 3, S) 30 L 2 rep 2, T) 30 L 2 rep 3, U) 20 H 1 rep 2, V) 20 H 2 rep 2.

PAGE 169

169

PAGE 170

170

PAGE 171

171

PAGE 172

172

PAGE 173

173

PAGE 174

174

PAGE 175

175

PAGE 176

176

PAGE 177

177

PAGE 178

178

PAGE 179

179

PAGE 180

180

PAGE 181

181

PAGE 182

182

PAGE 183

183

PAGE 184

184

PAGE 185

185

PAGE 186

186

PAGE 187

187

PAGE 188

188

PAGE 189

189

PAGE 190

190

PAGE 191

191 LIST OF REFERENCES Andersson, K., Ohlsson, T., Olsson, P., 1998. Screening life cy cle assessment (LCA) of tomato ketchup: a case study. Journal of Cleaner Production 6, 277 288. Antn, A., Montero, J.I., Muoz, P., Castells, F., 2005. LCA and tomato production in Mediterranean greenhouses. International Journal of Agricultural Resources Governance and Ecology 4, 102 112. Asano, T., Burton, F.L., Leverenz, H.L., Tsuchihashi, R., Tchobanoglous, G., 2007. Water Reuse: issues, Technologies and Applications. Metcalf and Eddy, Inc. Baroni, G., Facchi, A., Gandolfi, C., Ortuani, B., Horeschi, D ., Van Dam, J.C., 2009. Uncertainty in the determination of soil hydraulic parameters and its influence on the performance of two hydrological models of different complexity. Hydrology and Earth System Sciences Discussions 6, 4065 4105. Barretto de Figueir edo, E., Panosso, A.R., Romao, R., La Scala Jr, N., 2010. Greenhouse gas emission associated with sugar production in southern Brazil. Carbon Balance Management 5, 1 7. Beauchemin, K.A., Janzen, H.H., Little, S.M., McAllister, T.A., McGinn, S.M., 2010. Lif e cycle assessment of greenhouse emissions from beef production in western Canada: a case study. Agricultural Systems 103, 371 379. Bedard Haughn, A., Matson, A.L., Pennock, D.J., 2006. Land use effects on gross nitrogen mineralization, nitrification, and N 2 O emissions in ephemeral wetlands. Soil Biology and Biochemistry 38, 3398 3406. Bengston, T. O., 1987. The hydrologic effects from intense ground water pumpage in east central Hillsborough County, Florida. In: Karst hydrogeology: engineering and environm ental applications. Beck, B.F., Wilson, W.L. (editors), Proceedings of the Second Multidisciplinary Conference on Sinkholes and the Environmental Impacts of Karst. A. A. Balkema Publishers, Rotterdam, Netherlands. Berndt, M.P., Hatzell, H.H., Crandall, C.A ., Turtora, M., Pittman, J.R., Oaksford, E.T., 1998. Water quality in the Georgia Florida coastal plain, Georgia and Florida, 1992 96. United States Geological Survey Circular 1151. United States Geological Survey, Reston, VA, USA. Bernstein, L., Bosch, P. Canziani, O., Chen, Z., Christ, R., Davidson, O., Hare, W., Huq, S., Karoly, D., Kundzewicz, Z., Liu, J., Lohmann, U., Manning, M., Matsuno, T., Menne, B., Metz, B., Mirza, M., Nicholls, N., Nurse, L., Pachauri, R., Palutikof, J., Parry, M., Qin, D., Rav indranath, N., Reisinger, A., Ren, J., Riahi, K., Rosenzweig, C., Rusticucci, M., Schneider, S., Sokona, Y., Solomon, S., Stott, P., Stouffer, R., Sugiyama, T., Swart, R., Tirpak, D., Vogel, C., Yohe, G., 2007. Climate change 2007 synthesis report. An as sessment of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK.

PAGE 192

192 Brandt, A., Bresler, E., Diner, N., Ben Asher, I., Heller, J., Goldberg, D., 1971. Infiltration from a trickle source: I. Mathematical models. Soil Scienc e Society of America Journal 35, 675 682. Brentrup, F., Ksters, J., Kuhlmann, H., Lammel, J., 2004. Environmental impact assessment of agricultural production systems using the life cycle assessment methodology: I. Theoretical concept of a LCA method tail ored to crop production. European Journal of Agronomy 20, 247 264. Bresler, E., 1977. Trickle drip irrigation: principles and application to management. Advances in Agronomy 29, 343 393. Bresler, E., 1977. Trickle drip irrigation: Principles and applicatio n to management. Advances in Agronomy 29, 343 393. Bresler, E., Heller, J., Diner, N., Ben Asher, I., Brandt, A., Goldberg, D., 1971. Infiltration from a trickle source: II. Experimental data and theoretical predictions. Soil Science Society of America Jou rnal 35, 683 689. Breve, M.A., Skaggs, R.W., Parsons, J.E., Gilliam, J.W., 1997. DRAINMOD N: A Nitrogen Model for Artificially Drained Lands. Transactions of the American Society of Agricultural Engineers 40, 1067 1075. Brooks, R.H., Corey, A.T., 1966. Pro perties of porous media affecting fluid flow. Proceedings of the American Society of Civil Engineering, Journal of Irrigation Drainage Division IR2, 61 68. Bruinsma, J. (editor), 2003. World agriculture: towards 2015/2030: an FAO perspective. Earthscan, Lo ndon, UK. Burkhart, M. R., James, D. E., 1999. Agricultural nitrogen contributions to hypoxia in the Gulf of Mexico. Journal of Environmental Quality 28, 850 859. Burney, J.A., Davis, S.J., Lobell, D.B., 2010. Greenhouse gas mitigation by agricultural inte nsification. Proceedings of the National Academy of Sciences of the United States of America 107, 12052 12057. Campbell Scientific, 2011. CS616 and CS625 water content reflectometers instruction manual. Campbell Scientific, Logan, UT, USA. Cantliffe, D., G ilreath, P., Haman, D., Hutchinson, C., Li, Y., McAvoy, G., Migliaccio, K., Olczyk, T., Olson, S., Parmenter, D., Santos, B., Shukla, S., Simonne, E., Stanley, C., Whidden, A., 2009. Review of nutrient management systems for Florida vegetable producers: a white paper from the UF/IFAS vegetable fertilizer task force. Florida Cooperative Extension Service Document HS1156, University of Florida, Gainesville, FL, USA.

PAGE 193

193 Carlsson, A., 1997. Greenhouse gas emissions in the life cycle of carrots and tomatoes. IMES/E ESS Report No. 24, Lund University, Lund, Sweden. Caruso, G., Rapoport, H.F., Gucci, R., 2013. Long term evaluation of yield components of young olive trees during the onset of fruit production under different irrigation regimes. Irrigation Science 31, 37 47. Casey, J.W., Holden, N.M., 2003. A systematic description and analysis of GHG Proceedings of the Fourth International Conference on Life Cycle Assessment in the Agri Food Sec tor, Bygholm, Denmark. Cassel, D.K., Nielsen, D.R., 1986. Field capacity and available water capacity. In: Agronomy 9 1. Methods of Soil Analysis, 2nd ed. Klute, A., Dinauer, R.C. (editors), American Society of Agronomy, Madison, WI, USA, pp. 493 544. Cesc hia, E., Beziat, P., Dejoux, J.F., Aubinet, M., Bernhofer, Ch., Bodson, B., Buchmann, N., Carrar, A., Cellier, P., Di Tommasi, P., Elbers, J.A., Eugster, W., Grunwald, T., Jacobs, C.M.J., Jans, W.W.O., Jones, M., Kutsch, W., Lanigan, G., Magliulo, E., Marl oie, O., Moors, E.J., Moureaux, C., Olioso, A., Osborne, B., Sanz, M.J., Saunders, M., Smith, P., Soegaard, H., Wattenbach, M., 2010. Management effec ts on net ecosystem carbon and G HG budgets at European crop sites. Agriculture, Ecosystems, and Environmen t 139, 363 383. Coelho, E.F., Or, D. ,1996. Flow and uptake patterns affecting soil water sensor placement for drip irrigation management. Transactions of the American Society of Agricultural Engineers 39, 2007 2016. Copenhagen Accord, 2009. Decision 2/CP. 15. United Nations Climate Change Conference 2009, Copenhagen, Denmark. Cote, C.M., Bristow, K.L., Philip, C.B., Freeman, J.C., Thorburn, J., 2003. Analysis of soil wetting and solute transport in subsurface trickle irrigation. Irrigation Science 22, 143 1 56. Davies, J.N., Hobson, G.E., 1981. The constituents of tomato fruits the influence of the environment, nutrition and genotype. CRC Critical Reviews in Food Science and Nutrition 15, 205 280. Del Grosso, S.J., Ogle, S., Wirth, J., Skiles, S., 2008. U.S. Agriculture and Forestry Greenhouse Gas Inventory: 1990 2005. United States Department of Agriculture Technical Bulletin 1921, United States Department of Agriculture, Washington, DC, USA. Del Grosso, S.J., Parton, W.J., Mosier, A.R., Walsh, M.K., Ojima, D .S., Thornton, P.E., 2006. DAYCENT National scale simulations of nitrous oxide emissions from cropped soils in the United States. Journal of Environmental Quality 35, 1451 1460.

PAGE 194

194 Di, H. J., Cameron, K.C., 2002. Nitrate leaching in temperate agroecosystems: sources, factors and mitigating strategies. Nutrient Cycling in Agroecosystems 46, 237 256. Dukes, M.D., Simonne, E.H., Davis, W.E., Studstil, D.W., Hochmuth, R., 2003. Effect of sensor based high frequency irrigation on bell pepper yield and water use. In : Proceeding of 2nd International Conference of Irrigation and Drainage, Phoenix, AZ, USA, pp. 665 674. Dukes, M.D., Zotarelli, L., Morgan, K.T., 2010. Use of irrigation technologies for vegetable crops in Florida. HortTechnology 20, 133 142. Dukes, M.D., Zotarelli, L., Scholberg, J.M., Munoz Carpena, R., 2006. Irrigation and nitrogen best management practices under drip irrigated vegetable production. In: Proceedings of the World Environmental & Water Resources Congress. Omaha, NE, USA, May 21 25, 2006. D utton, R., Jiao, J., Tsujita, M.J., Grodzinsk, B., 1988. Whole plant CO 2 exchange measurements for nondestructive estimation of growth. Plant Physiology 86, 355 358. Eggleston, H.S., Buendia L., Miwa K., Ngara T., Tanabe K., 2006. 2006 IPCC guidelines for national greenhouse gas inventories. Institute for Global Environmental Strategies, Hayama, Japan. Elmaloglou, S.T., Malamos, N., 2003. A method to estimate soil water movement under a trickle surface line source, with water extraction by roots. Irrigation and Drainage 52, 273 284. Fischlin, A., Midgeley, G., 2007. Ecosystems, their properties, goods and services. In: Climate change 2007: impacts, adaptation and vulnerability. Contribution of working group II to the fourth assessment report of the intergove rnmental panel on climate change. Parry, M.L., Canziani, O.F., Palutikof, J.P., Hanson, C.E., van der Linden, P.J. (editors), Cambridge University Press, Cambridge, U.K. Fisher, P.D., 1995. An alternative plastic mulching system for improved water manageme nt in dryland maize production. Agricultural Water Management 27, 155 166. Florida Department of Agriculture and Consumer Services, 2010. Florida agriculture statistical directory 2010. Florida Department of Agricultural and Consumer Services, Division of Marketing and Development, Tallahassee, FL, USA. Fraisse, C.W., Hu, Z., Simonne, E.H., 2010. Effect of El Nio Southern Oscillation on the number of leaching rain events in Florida and implications on nutrient management for tomato. HortTechnology 20, 120 132.

PAGE 195

19 5 Gan, Y., Liang, C., Huang, G., Malhi, S.S., Brandt, S.A., Katepa Mupondwa, F., 2012. Carbon footprint of canola and mustard is a function of the rate of N fertilizer. International Journal of Life Cycle Assessment 17, 58 68. Gardner, W.R., Mayhugh, M. S., 1958. Solutions and tests of the diffusion equation for the movement of water in soil. Soil Science Society of America Proceedings 22, 197 201. Gijsman, A.J., Jagtap, S.S., Jones, J.W., 2002. Wading through a swamp of complete confusion: how to choose a method for estimating soil water retention parameters for crop models. European Journal of Agronomy 18, 75 105. Glades Crop Care, Inc., 1999. Crop profile for south Florida tomatoes. Glades Crop Care, Inc., Jupiter, Florida, USA. Golden Software, Inc., 2 011. Surfer Version 10.1. Golden Software, Inc., Golden, CO, USA. Gowdish, L., Muoz Carpena, R., 2009. An improved Green Ampt infiltration and redistribution method for uneven multistorm series. Vadose Zone Journal 8, 470 479. Gowdish, L.C., 2007. An impr oved Green Ampt soil infiltration and redistribution method and its application to 1 dimensional and quasi 3 dimensional (point source) flow domains. Doctoral dissertation, University of Florida, Gainesville, FL, USA. Graboski, M., 2002. Fossil energy use in the manufacture of corn ethanol. National Corn Growers Association, Chesterfield, MO, USA. Guo, L.B., Gifford, R.M., 2002. Soil carbon stocks and land use change: a meta analysis. Global Change Biology 4, 345 360. Gupta, S.C., Farrell, D.A., Larson, W., E., 1974. Determining effective soil water diffusivities from one step outflow experiments. Soil Science Society of America Journal 38, 710 716. Hansen, J., Sato, M., Ruedy, R., Lo, K., Lea, D.W., Medina Elizade, M., 2006. Global temperature change. Procee dings of the National Academy of Sciences of the United States of America 103, 14288 14293. Hillel D., 1998. Environmental Soil Physics. San Diego, CA: Academic. Hillier, J., Hawes, C., Squire, G., Hilton, A., Wale, S., Smith, P., 2009. The carbon footprin ts of food crop production. International Journal of Agricultural Sustainability 7, 107 118.

PAGE 196

196 Hillier, J., Walter, C., Malin, D., Garcia Suarez, T, Mila i Canals, L., Smith, P., 2011. A farm focused calculator for emissions from crop and livestock productio n. Environmental Modeling and Software 26, 1070 1078. Hochmuth, G. C., Hochmuth, R. C., Olson, S. M., 2008. Polyethylene Mulching for Early Vegetable Production in North Florida.Florida Cooperative Extension Service Circular 805, University of Florida, Gai nesville, FL, USA. Hoogenboom, G., Jones, J.W., Wilkens, P.W., Porter, C.H., Hunt, L.A., Boote, K.J., Singh, U., Uryasev, O., Lizaso, J.I., Gijsman, A.J., White, J.W., Batchelor, W.D., Tsuji, G.Y., 2009. Decision Support System for Agrotechnology Transfer Version 4.5 [CD ROM]. University of Hawaii, Honolulu, HI, USA. Hoogenboom, G., Jones,J. W., Porter, C. H., Wilkens, P. W., Boote, K. J., Batchelor, W. D., Hunt, L. A., Tsuji, G. Y. (editors), 2003. Decision Support System for Agrotechnology Transfer Versio n 4.0. University of Hawaii, Honolulu, HI. Ibarra, L., Flores, J., Diaz Perez, J.C., 2001. Growth and yield of muskmelon in response to plastic mulch and row covers. Scientia Horticulturae 87, 139 145. Ingwersen, W.W., 2012. Life cycle assessment of fresh pineapple from Costa Rica. Journal of Cleaner Production 35, 1 12. Intergovernmental Panel on Climate Change, 2006. 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme, Eggleston, H.S. Buendia, L., Miwa, K., Ngara, T., and Tanabe, K. (editors),IGES, Japan. International Fertilizer Industry Association and the Food and Agriculture Organization of the United Nations, 2001. Global estimates of gaseous emissions of NH 3 NO, and N 2 O from ag ricultural land. International Fertilizer Industry Association and the Food and Agriculture Organization of the United Nations, Rome, Italy. Ippisch, O., Vogel, H.J., Bastian, P., 2006. Validity limits for the Van Genuchten Mualem model and implications fo r parameter estimation and numerical simulation. Advances in Water Resources 29, 1780 1789. Jansson, S. L., Persson, J., 1982. Mineralization and immobilization of soil nitrogen. In: Nitrogen in Agricultural Soils. Stevenson, F. J. (editor), Agronomy Monog raph 22. American Society of Agronomy, Madison, WI.pp. 229 248. Jones, C.D., Fraisse, C.W., Ozores Hampton, M., 2012. Quantification of greenhouse gas emissions from open field grown Florida tomato production. Agricultural Systems 113, 64 72. Jones, J. W., Hoogenboom, G., Porter, C. H., Boote, K. J., Batchelor, W. D., Hunt, L. A., Wilkens, P. W., Singh, U., Gijsman, A. J., Ritchie, J. T., 2003. The DSSAT cropping system model. European Journal of Agronomy 18, 235 265.

PAGE 197

197 Jones, J.W., Hoogenboom, G., Porter, C. H., Boote, K.J., Batchelor, W.D., Hunt, L.A., Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT cropping system model. European Journal of Agronomy 18, 235 265. Kasperbauer, M.J., Hunt, P.G., 1989. Mulch surface color affects yield of fresh market tomatoes. Jounral of the American Society of Horitcultural Science 114, 216 219. Kendall, A., Chang, B., 2009. Estimating life cycle greenhouse gas emissions from corn ethanol: a critical review of current U.S. practices. Journal of Cleaner P roduction 17, 1175 1182. Kenny, J. F., Barber, N. L., Hutson, S. S., Linsey, K. S., Lovelace, J. K., Maupin, M. A., 2009. Estimated use of water in the United States in 2005.US Geological Survey Circular 1344. Reston, VA: US Geological Survey. Kim, S., Dal e, B.E., Jenkins, R., 2009. Life cycle assessment of corn grain and corn stover in the United States. The International Journal of Life Cycle Assessment 14, 160 174. Klute, A., 1986. Methods of soil analysis, part 1. Physical and Mineralogical Methods, 2nd ed. American Society of Agronomy, Soil Science Society of America, Inc., Madison, WI, USA. Kosugi, K., Hopmans, J.W., Dane, J.H., 2002. Parametric models. In: Methods of Soil Analysis, part 4, Physical Methods. Dane, J.H., Topp, G.C. (editors). Soil Scie nce Society of America, Madison, WI, USA, pp. 739 755. Lal, R., 2004.Carbon emission from farm operations. Environment International 30, 981 990. Li, F., Guo, A., Wei, H., 1999. Effects of clear plastic film mulch on yield of spring wheat. Field Crops Rese arch 63, 79 86. Li, J. S., Zhang, J. J., Ren, L., 2003. Water and nitrogen distribution as affected by fertigation of ammonium nitrate from a point source. Irrigation Science 22, 19 30. Locascio, S. J., 2005. Management of irrigation for vegetables: past, present, and future. HortTechnology 15, 482 485. Looper, J.P., Baxter, E.V., 2011. An assessment of distributed flash flood forecasting accuracy using radar and rain gauge input for a physics based distributed hydrologic model. Journal of Hydrology 412 413 114 132. Mace, A., Rudolph, D.L., Kachanoski, R.G., 1998. Suitability of parametric models to describe the hydraulic properties of an unsaturated coarse sand and gravel. Ground Water 36, 465 475.

PAGE 198

198 Machado, R.M.A., do Rosario, M., Oliveira, G., Portas, C.A .M., 2003. Tomato root distribution, yield and fruit quality under subsurface drip irrigation. Plant and Soil 255, 333 341. Marquardt, D.W., 1963. An algorithm for least squares estimation of nonlinear parameters. Journal of the Society for Industrial and Applied Mathematics 11, 431 441. McSwiney, C.P., Robertson, G.P., 2005. Nonlinear response of N 2 O flux to incremental fertilizer addition in a continuous maize ( Zea mays L.) cropping system. Global Change Biology 11, 1712 1719. Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T., Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper, S.C.B., Watterson, I.G., Weaver, A.J., Zhao, Z.C., 2007. Global climate projections. In: Climate change 2007: the physical science basis. Co ntribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (editors), Cambridge University Press, Cambridg e, UK. Meinshausen, M., Meinshausen, N., Hare, W., Raper, S. C. B., Frieler, K., Knutti, R., Frame, D. J., Allen, M. R., 2009. Greenhouse gas emission targets for limiting global warming to 2 C. Nature 458, 1158 1162. Molle, F., Berkoff, J., 2009. Cities vs. agriculture: a review of intersectoral water re allocation. Natural Resources Forum 33, 6 18. Mordini, M., Nemecek, T., Gaillard, G., 2009. Carbon & water footprint of orange and strawberries. Federal Department of Economic Affairs, Zurich, Switzerland Mualem, Y., 1976. A new model predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12, 513 522. thing? United States Geological Survey Circular 1136. United States Geological Survey, Reston, VA, USA. Muoz, P., Antn, A., Montero, J.I., Castells, F., 2004. Using LCA for the improvement of waste management in greenhouse tomato production. In: Proceedings of the Fourth International Conference on Life Cycle Assessment in the Agri Food Sector, Bygholm, Denmark. Nagy, R.C., Lockaby, B.G., Helms, B., Kalin, L., Stoeckel, D., 2010. Water resources and land use and cover in a humid region: the Southeastern United States. Journal of Enviro nmental Quality 40, 867 878. Journal of Environmental Quality 37, S 157 S 168.

PAGE 199

199 Olson, S.M., 2011. Mulching. In: Vegetable production handbook for Florida 2010 2011. Olso n, S.M., Santos, B. (editors). University of Florida, Gainesville, FL, USA. Olson, S.M., Santos, B., 2011. Vegetable production handbook for Florida, 2010 2011. University of Florida Institute of Food and Agricultural Sciences, Gainesville, FL, USA. Olson, S.M., Stall, W.M., Vallad, G.E., Webb, S.E., Smith, S.A., Simonne, E.H., McAvoy, E.J., Santos, B.M., 2010. Tomato production in Florida. In: Vegetable production handbook for Florida 2010 2011. Olson, S.M., Santos, B. (editors), University of Florida, Gai nesville, FL, USA. water transport in unsaturated soils. Journal of Hydrology 272, 3 13. Pitts, D.J., Clark, G.A., Alvarez, J., Everett, P.H., Grimm, J.M., 1988. A compa rison of micro to subsurface irrigation of tomatoes. Proceedings of the Florida State Horticultural Society 101, 393 397. Provenzano, G., 2007. Using HYDRUS 2D simulation model to evaluate wetted soil volume in subsurface drip irrigation systems. Journal o f Irrigation and Drainage Engineering 133, 342 349. R Development Core Team, 2011. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rawls, W.J., Ahuja, L.R., Brakensiek, D.L., 1992. Estimatin g soil hydraulic properties from soils data. In: Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils. Van Genuchten, M.Th., Leij, F.J., Lund, L.J. (editors). University of California, Riverside, CA, USA, pp. 329 340. Reichardt, K. Nielsen, D.R., Biggar, J.W., 1972. Scaling of horizontal infiltration into homogenous soils. Soil Science Society of America Proceedings 36, 241 245. Richards, L.A. 1931. Capillary conduction of liquids in porous mediums. Physics 1, 318 333. Ritchie, J.T ., 1985. An user orientated model of the soil water balance in wheat. In: Wheat Growth and Modeling. Day, W., Atkin, R.K., (editors). Plenum Press, New York, NY, USA, pp. 293 305. Ritchie, J.T., 1985. An user orientated model of the soil water balance in w heat. In: Wheat Growth and Modeling. Day, W., Atkin, R.K. (editors). Plenum Press, New York, NY, USA, pp. 293 305. Ritchie, J.T., 1998. Soil water balance and water stress. In: Understanding Options for Agricultural Production. Tsuji, G.Y., Hoogenboom, G., Thornton, P. K. (editors). Kluwer Academic Publishers, Dordrecht, Netherlands, pp. 99 128.

PAGE 200

200 Ritchie, J.T., 1998. Soil water balance and water stress. In: Understanding Options for Agricultural Production. Tsuji, G.Y., Hoogenboom, G., Thornton, P.K. (editor s). Kluwer Academic Publishers, Dordrecht, Netherlands, pp. 99 128. Roberts, R., Lazarovitch, N., Warrick, A.W., Thompson, T.L., 2009. Modeling salt accumulation with subsurface drip irrigation using HYDRUS 2D. Soil Science Society of America Journal 73, 2 33 240. Robertson, G.P., Paul, E.A., Harwood, R.R., 2000. Greenhouse gases in intensive agriculture: contributions of individual gases to the radiative forcing of the atmosphere. Science 289, 1922 1925. Rogelj, J., Nabel, J., Chen, C., Hare, W., Markmann, K., Meinshausen, M., Schaeffer, M., Macey, K., Hhne, N., 2010. Copenhagen Accord pledges are paltry. Nature 464, 1126 1128. Rogner, H. H., Zhou, D., Bradley, R., Crabb, P., Edenhofer, O., Hare, B., Kuijpers, L., Yamaguchi, M., 2007. Introduction. In: Cli mate change 2007: mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Metz, B., Davidson, O.R., Bosch, P.R., Dave, R., Meyer, L.A. (editors), Cambridge University Press, Cambridge, UK. Romero, C.C., Dukes, M.D., Baigorria, G.A., Cohen, R.C., 2008. Comparing theoretical irrigation requirement and actual irrigation for citrus in Florida. Agricultural Water Management 96, 473 483. Roos, E., Sundberg, C., Hansson, P., 2010. Uncertaintie s in the carbon footprint of food products: a case study on table potatoes. The International Journal of Life Cycle Assessment 15, 478 488. Rosenzweig, C., Strzepek, K.M., Major, D.C., Iglesias, A., Yates, D.N., McCluskey, A., Hillel, D., 2004. Water resou rces for agriculture in a changing climate: international case studies. Global Environment Change Part A 14, 345 360. Rubin, J., 1968. Theoretical analysis of two dimensional, transient flow of water in unsaturated and partly unsaturated soils. Soil Scienc e Society of America Proceedings 32, 607 615. Sato, S., Morgan, K.T., Ozores Hampton, M., Mahmoud, K., Simonne, E.H., 2010. Nutrient balance and use efficiency in sandy soils cropped with tomato under seepage irrigation. Soil Science Society of America Jou rnal 76, 1867 1876. Saxton, K.E., Rawls, W.J., 2006. Soil water characteristics estimates by texture and organic matter for hydrologic solutions. Soil Science Society of America Journal 70, 1569 1578.

PAGE 201

201 Schaap, M.G., Leij, F.J., 2000. Improved prediction of unsaturated hydraulic conductivity with the Mualem Van Genuchten Model. Soil Science Society of America Journal 64, 843 851. Scholberg, J.M.S., 1996. Adaptive use of crop growth models to simulate the growth of field grown tomato. Doctoral dissertation, Un iversity of Florida, Gainesville, FL, USA. Schroder, J. H., 2006. Soil moisture based drip irrigation for efficient use of water and nutrients and sustainability of vegetables cropped on coarse soils. Masters thesis, University of Florida, Gainesville, FL USA. Schwartzman, M., Zur, B., 1986. Emitter spacing and geometry of wetted soil volume. Journal of Irrigation and Drainage Engineering 112, 242 253. Shrestha, N., Geerts, S., Raes, D., Horemans, S., Soentjens, S., Maupas, F., Clouet, P., 2010. Yield res ponse of sugar beets to water stress under Western European conditions. Agricultural Water Management 97, 346 350. Shukla, S., Jaber, F.H., 2005. Groundwater recharge in shallow water table conditions. Paper No. 05 2088, presented at the 2005 ASAE Annual I nternational Meeting, Tampa, FL, USA, July 17 20, 2005. The American Society of Agricultural Engineers, St. Joseph, MI, USA. Simonne, E., Studstill, D., Hochmuth, R., Olczyk, T., Dukes, M., Munoz Carpena, R., Li, Y., 2003. Drip Irrigation: The BMP Era An Integrated Approach to Water and Fertilizer Management for Vegetables Grown with Plasticulture. Florida Cooperative Extension Service, IFAS, University of Florida HS917, Gainesville, FL, USA. Simonne, E.H., Dukes, M.D., Haman, D.Z., 2004. Principles and p ractices of irrigation management for vegetables. In: Vegetable production guide for Florida. Olsen, S. M., Simonne, E. H. (editors). University of Florida, Gainesville, FL, USA, p. 33 39. Simonne, E.H., Dukes, M.D., Zotarelli, L., 2010. Principles and pra ctices of irrigation management for vegetables. In: Vegetable production handbook for Florida 2010 2011. Olson, S.M., Santos, B. (editors), University of Florida, Gainesville, FL, USA. Simonne, E.H., Ozores Hampton, M., 2006. Challenges and opportunities f or extension research educators involved in best management practices. HortTechnology 16, 403 407.

PAGE 202

202 Simmone, E., Hutchinson, C., DeValerio, J., Hochmuth, R., Treadwell, D., Wright, A., Sants, B., Whidden, A., McAvoy, G., Zhao, X., Olczyk, T., Gazula, A., Ozores Hampton, M., 2010. Current knowledge, gaps, and future needs for keeping water and nutrients in the root zone of vegetables grown in Florida. HortTechnology 20, 143 152. Simunek, J., Van Genuchten, M.Th., Sejna, M., 2006. The HYDRUS software package for simulating the two and three dimensional movement of water, heat, and multiple solutes in variably saturated media, Version 1.0. PC Progress, Prague, Czech Republic. Skaggs, T.H., Trout, T.J., Rothfuss, Y., 2010. Drip irrigation water distribution p atterns: Effects of emitter rate, pulsing, and antecedent water. Soil Science Society of America Journal 74, 1886 1896. Skaggs, T.H., Trout, T.J., Simunek, J., Shouse, P.J., 2004. Comparison of HYDRUS 2D simulations of drip irrigation with experimental obs ervations. Journal of Irrigation and Drainage Engineering 130, 304 310. Smajstrla, A.G., Locascio, S.J., 1996. Tensiometer controlled, drip irrigation scheduling of tomato. Applied Engineering in Agriculture 12, 315 319. Smith, P., Martino, D., Cai, Z., Gw ary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S., Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Metz, B., Davidson, O.R., Bosch, P.R., Dave, R., Meyer, L.A. (editors). Cambridge University Press, New York, NY, USA. Snyder, C.S., Bruulsema, T.W., Jensen, T.L., Fixen, P.E., 2009. Review of greenhouse gas emissions from crop producti on systems and fertilizer management effects. Agriculture, Ecosystems and Environment 133, 247 266. Soil Survey Staff, 2004. Particle size analysis method 3A1a1. In: Soil survey laboratory methods manual. Burt, R. (editor). Soil survey investigations repor t no. 42, version 4. Natural Resources Conservation Service, United States Department of Agriculture, National Soil Survey Center, Lincoln, NE, USA. Soil Survey Staff, 2011. Web soil survey. Natural Resources Conservation Service, United States Department of Agriculture. Retrieved 06.01.2011 from http://websoilsurvey.nrcs.usda.gov/. Southwest Florida Water Management District, 2006. Southern water use caution area recovery strategy. Retrieved 05.01.2011 from http://www.swfwmd.state.fl.us/documents/plans/swu ca_recovery_strategy.pdf Spreen, T., Dwivedi, P., Goodrich Schneider, R., 2010. Estimating the carbon footprint of Florida orange juice. Proceedings in Food System Dynamics, 95 101.

PAGE 203

203 Tak, H.I., Bakhtiyar, Y., Ahmad, F., Inam, A., 2012. Effluent quality para meters for safe use in agriculture. In: Water quality, soil and managing irrigation of crops. Lee, T.S. (editor), InTech, Rijeka, Croatia. Advances in Water Resources 27, 1041 1042. Tiemeyer, B., Kahle, P., Lennartz, B., 2010. Designing monitoring programs for artificially drained catchments. Vadose Zone Journal 9, 14 24. Topp, G.C., 1980. Electromagnetic determination of soil water content measurements in coaxial transmission lines. Water Resources Research 16, 574 582. Tsuji, G.Y., Uehara, G., Balas, S., (editors), 1994. DSSAT Version 3. University of Hawaii, Honolulu, HI, USA. Tuller, M., Or, D., 2001. Hydraulic conductivity of variably saturated porous media: film and corner flow in angular pore space. Water Resources Research 37, 1257 1276. United Nations Framework Convention on Climate Change, 2009. Decision 2/CP.15 Copenhagen Accord, pp. 4 10. United States Department of Agriculture, Soil Conservation Service, 1972. National Engineering Handbook, Hydrology, Section 4. Washington, DC, USA. United States Environmental Protection Agency, 2005. Emission facts: average carbon dioxide emissions resulting from gasoline and diesel fuel. U.S. E PA # 420 F 05 001, Environmental Protection Agency, Washington, DC, USA. United States Environmental Protection Agency, 2010. 2010 US Greenhouse Gas Inventory Report: Inventory of U.S. Greenhouse Gas Emissions and Sinks, 1990 2008. U.S. EPA # 430 R 10 006, Environmental Protection Agency, Washington, DC, USA. University of Florida Department of Soil and Water Science, 2012. Florida Soil Characterization Database. Retrieved 09.01.2012 from The Florida Soil Characterization Retrieval System http://soils.ifas. ufl.edu/flsoils/. Van Genuchten, M.Th., 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44, 892 898. Van Genuchten, M.Th., Leij, F., Yates, S., 1991. The RETC code for qua ntifying the hydraulic functions of unsaturated soils.Technical Report EPA/600/2 91/065, US Environmental Protection Agency.

PAGE 204

204 Van Genuchten, M.Th., Nielsen, D.R., 1985. On describing and predicting the hydraulic properties of unsaturated soils. Annales Geop hysicae 3, 615 628. Vazquez, N., Pardo, A., Suso, M.L., Quemada, M., 2005. A methodology for measuring drainage and nitrate leaching in unevenly irrigated vegetable crops. Plant and Soil 269, 297 308. Vorosmarty, C. J., Green P., Salisbury J., Lammers, R.B., 2000. Global water resources: vulnerability from climate change and population growth. Science 289, 284 288. Wan, Y., El Swaify, S.A., 1999. Runoff and soil erosion as affected by plastic mulch in a Hawaiian pineapple field. Soil and Tillage Research 52, 29 35. Wang, D., Shannon, M.C., Grieve, C.M., Yates, S.R., 2000. Soil water and temperature regimes in drip and sprinkler irrigation, and implications to soybean emergence. Agricultural Water Managem ent 43, 15 28. Warrick, A.W., 1985. Point and line infiltration calculation of the wetted soil surface. Soil Science Society of America Journal 49, 1581 1583. Weinheimer, J., Rajan, N., Johnson, P., Maas, S., 2010. Carbon footprint: a new farm management consideration in the Southern High Plains. Selected paper prepared for presentation at the Agricultural and Applied Economics 27, 2010, Denver, Colorado, USA. Williams, A.G., Audsley, E. Sandars, D.L. 2006. Energy and environmental burdens of organic and non organic agriculture and horticulture. Aspects of Applied Biology 79, 19 23. Williams, J.R., Jones, C.A., Dyke, P.T., 1984. A modeling approach to determining the relationships betwee n erosion and soil productivity. Transactions of American Society of Agricultural Engineers 27, 129 144. World Meteorological Organization, 2006. Greenhouse gas bulletin: the state of greenhouse gases in the atmosphere using global observations up to Decem ber 2004. World Meteorological Organization, Geneva, Switzerland. Zotarelli, L., Dukes, M.D., Scholberg, J.M., Hanselman, T., Le Femminella, K., Munoz Carpena, R., 2008. Nitrogen and water use efficiency of zucchini squash for a plastic mulch bed system on a sandy soil. Scientia Horticulurae 116, 8 16. Zotarelli, L., Dukes, M.D., Scholberg, J.M.S., Munoz Carpena, R., Icerman, J., 2009. Tomato nitrogen accumulation and fertilizer use efficiency on a sandy soil, as affected by nitrogen rate and irrigation sc heduling. Agricultural Water Management 96, 1247 1258.

PAGE 205

205 Zotarelli, L., M.D. Dukes, Morgan, K.T., 2010.Interpretation of Soil Moisture Content to Determine Soil Field Capacity and Avoid Over Irrigating Sandy Soils Using Soil Moisture Sensors. Florida Coopera tive Extension Service AE 460. Zotarelli, L., Scholberg, J.M., Dukes, M.D., Munoz Carpena, R., 2007. Monitoring of nitrate leaching in sandy soils: comparison of three methods. Journal of Environmental Quality 36, 953 962. Zotarelli, L., Scholberg, J.M., D ukes, M.D., Munoz Carpena, R., Icerman, J., 2009. Tomato yield, biomass accumulation, root distribution and irrigation water use efficiency on a sandy soil, as affected by nitrogen rate and irrigation scheduling. Agricultural Water Management 96, 23 34.

PAGE 206

206 BIOGRAPHICAL SKETCH Curtis Jones graduated magna cum laude with a Bachelor of Science degree in agricultural and biological engineering from the University of Florida in 2006. During this time he was introduced to crop modeling and worked on projects relat ed to irrigation management and climate impacts on agricultural risk. After graduating, he worked for Water & Air Research Inc., an environmental consulting firm, focusing on water quality monitoring projects. In 2008 he began graduate school at the Univer sity of Florida, where he was awarded an Alumni Fellowship by the Department of Agricultural and Biological Engineering to pursue a Doctoral degree.