Effectiveness of Smart Controllers for Water Conservation in Residential Irrigation

MISSING IMAGE

Material Information

Title:
Effectiveness of Smart Controllers for Water Conservation in Residential Irrigation
Physical Description:
1 online resource (261 p.)
Language:
english
Creator:
Davis, Stacia L
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Agricultural and Biological Engineering
Committee Chair:
DUKES,MICHAEL D
Committee Co-Chair:
MUNOZ-CARPENA,RAFAEL
Committee Members:
HAMAN,DOROTA ZOFIA
KRUSE,JASON KEITH
MIGLIACCIO,KATI WHITE
HEANEY,JAMES

Subjects

Subjects / Keywords:
conservation -- controller -- evapotranspiration -- irrigation -- model -- smart
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 ET controller has the potential to produce water savings without sacrificing landscape quality in Florida when implemented at residences that habitually irrigate excessively and programmed with settings that are appropriate for the landscape. The education provided as part of the education treatments was effective at reducing irrigation application in addition to water savings by the technology without education. Proper controller programming was essential to achieving maximum water savings while maintaining acceptable landscape quality. If the time and effort of proper programming is not achievable, other smart irrigation technologies that do not increase existing irrigation, such as soil moisture sensors, would be more appropriate. The time-based treatments selected to represent actual homeowner practices in Florida resulted in excessive irrigation during the field plot study. Though the ET controllers decreased irrigation compared to 100% replacement of the net irrigation requirement, there were no additional water savings compared to 60% replacement. Additionally, the performance scores for the ET controllers were not different from the 60% replacement with all treatments experiencing over-irrigation. The ET controller should be implemented only in situations of excessive irrigation habits. In efforts to better approximate water conservation potential, an hourly soil water balance (SWB) was developed as an alternative to the industry standard, the daily SWAT test, for evaluating irrigation adequacy, a measure of under-irrigation, and scheduling efficiency, a measure of over-irrigation. The hourly SWB better matched the deterministic soil water model compared to the daily SWAT test thus improving estimates of volumetric water content in a shallow root zone. Additionally, the hourly SWB was preferential to the daily SWAT test based on the combination of more accurate landscape programming required to obtain high performance scores and the realistic representation of soil water movement by hourly SWB resulting in a more accurate translation of SWAT scores to controller performance in the landscape. As a result, there is confidence in implementing the hourly SWB as an alternative method for testing ET controllers that will more accurately reflect true soil-water relationships thus allowing for the estimation of water conservation potential.
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 Stacia L Davis.
Thesis:
Thesis (Ph.D.)--University of Florida, 2014.
Local:
Adviser: DUKES,MICHAEL D.
Local:
Co-adviser: MUNOZ-CARPENA,RAFAEL.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2014
System ID:
UFE0046503:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 EFFECTIVENESS OF SMART CONTROLLERS FOR WATER CONSERVATION IN RESIDENTIAL IRRIGATION By STACIA L. DAVIS 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 2014

PAGE 2

2 2014 Stacia L. Davis

PAGE 3

3 To those who have supported me during this extended adolescence

PAGE 4

4 ACKNOWLEDGMENTS I would not have achieved this accomplishment without the sacrifice of my parents, Charles and Suz Davis, who encouraged me to become a professor when I declared a future profession of teaching at the age of 10. Though they preferred that I abandon that p lan and get a job immediately after obtaining my Bachelor of Science degree, they have truly never held me back from chasing my dream. many top notch professors that have truly encou raged success and my graduate committee was no exception I would like to acknowledge my committee members: Rafael Muoz Carpena, Dorota Haman, Kati Migliaccio, Jason Kruse, and James Heaney. They have provided their time, resources, advice, and opportun anywhere else. I would like to especially acknowledge my graduate advisor and mentor, Michael Dukes, for demanding excellence but ultimately accepting my efforts in substitution. By setting a high bar, I was able to sur pass my personal goals while still having seemingly infinite opportunities for growth and learning. Despite closing this chapter of my life and looking forward to the future, the bounty of experiences received while at UF will not be forgotten. The Irri GATOR research team, past and present members, provided significant support while working toward my Master of Science and Doctor of Philosophy degrees at UF. Though many have come and gone from UF over the last nine years, they are still a part of my life and hold a place in my heart.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 Current Irrig ation Practices ................................ ................................ ..................... 18 Soil Water Modeling ................................ ................................ ................................ 19 Soil Water Storage ................................ ................................ ........................... 22 Soil Parameters ................................ ................................ ................................ 23 Evapotranspiration ................................ ................................ ........................... 26 Precipitation ................................ ................................ ................................ ...... 29 Irrigation ................................ ................................ ................................ ........... 30 Infiltration ................................ ................................ ................................ .......... 32 Richards ................................ ................................ ................................ ..... 33 Green Ampt ................................ ................................ ............................... 35 Surface Runoff ................................ ................................ ................................ 39 Calibration and Validation ................................ ................................ ................. 40 Sensitivity Analysis ................................ ................................ ........................... 42 Summary of Available Models ................................ ................................ .......... 44 CHEMFLO 2000 ................................ ................................ ........................ 44 HYDRUS 1D ................................ ................................ .............................. 45 SWAP ................................ ................................ ................................ ........ 45 WAVE ................................ ................................ ................................ ........ 46 ET Controllers ................................ ................................ ................................ ......... 47 Evaluation Programs ................................ ................................ ........................ 47 Previous Research ................................ ................................ ........................... 50 Bench tests ................................ ................................ ................................ 50 Plot studies ................................ ................................ ................................ 53 Cooperator studies ................................ ................................ ..................... 55 2 PERFORMANCE OF WEATHER BASED IRRIGATION CONTROLLERS IN SOUTHWEST FLORIDA LANDSCAPES ................................ ............................... 68 Introduction ................................ ................................ ................................ ............. 68 Materials and Methods ................................ ................................ ............................ 70 Results and Discussion ................................ ................................ ........................... 77 Conclusions ................................ ................................ ................................ ............ 86

PAGE 6

6 3 IMPLEMENTING SMART CONTROLLERS ON SINGLE FAMILY HOMES WITH EXCESSIVE IRRIGATION ................................ ................................ ........... 90 Introduction ................................ ................................ ................................ ............. 90 Materials and Methods ................................ ................................ ............................ 91 Participant Selection ................................ ................................ ......................... 91 Treatment Description ................................ ................................ ...................... 93 Data Collection and Analysis ................................ ................................ ............ 93 Gross Irrigation Requirement ................................ ................................ ........... 95 Results ................................ ................................ ................................ .................... 97 Weather Patterns ................................ ................................ ............................. 97 Irrigation Application ................................ ................................ ......................... 98 Dry 2012 ................................ ................................ ................................ .... 98 Wet 2012 ................................ ................................ ................................ .... 99 Dry 2013 ................................ ................................ ................................ .. 101 Wet 2013 ................................ ................................ ................................ .. 102 Turfgrass Quality ................................ ................................ ............................ 103 Discussion ................................ ................................ ................................ ............ 104 Conclusions ................................ ................................ ................................ .......... 108 4 COMPARISON OF MODELING METHODS TO ESTIMAT E SOIL WATER CONTENT IN TURFGRASS BASED LANDSCAPES ................................ ........... 118 Introduction ................................ ................................ ................................ ........... 118 Materials and Methods ................................ ................................ .......................... 120 Soil Water Balance ................................ ................................ ......................... 120 Daily SWAT Test ................................ ................................ ............................ 122 Hourly SWB Model ................................ ................................ ......................... 123 HYDRUS 1D ................................ ................................ ................................ .. 125 Field Measured Data ................................ ................................ ...................... 126 PSREU ................................ ................................ ................................ .... 126 GCREC ................................ ................................ ................................ .... 127 Model Parameteriz ation, Calibration, and Validation ................................ ...... 128 Results ................................ ................................ ................................ .................. 130 PSREU ................................ ................................ ................................ ........... 130 Calibration period ................................ ................................ ..................... 131 Validation period ................................ ................................ ...................... 134 GCREC ................................ ................................ ................................ .......... 136 Calibration period ................................ ................................ ..................... 137 Validation period ................................ ................................ ...................... 139 Conclusions ................................ ................................ ................................ .......... 143 5 SIMULATED PERFORMANCE OF EVAPOTRANSPIRATION BASED IRRIGATION CONTROLLERS USING SOIL WATER BALANCE MODELS ........ 164 Introduction ................................ ................................ ................................ ........... 164 Materials and Methods ................................ ................................ .......................... 166

PAGE 7

7 Soil Water Balance ................................ ................................ ......................... 166 Daily SWAT Model ................................ ................................ ......................... 168 Hourly SWB Model ................................ ................................ ......................... 169 Data Description ................................ ................................ ............................. 170 Bench testing study ................................ ................................ .................. 170 Field plot study ................................ ................................ ......................... 172 Statistical analysis ................................ ................................ .................... 173 Results ................................ ................................ ................................ .................. 174 Bench Testing Study ................................ ................................ ...................... 174 Frequent rainfall period ................................ ................................ ............ 174 Infrequent rainfall period ................................ ................................ .......... 175 Low ET O period ................................ ................................ ........................ 177 High ET O period ................................ ................................ ....................... 178 Correlation coefficients ................................ ................................ ............ 179 Field Plot Study ................................ ................................ .............................. 180 Discussion ................................ ................................ ................................ ............ 182 Conclusion ................................ ................................ ................................ ............ 185 6 REGRESSION ANALYSIS FOR SENSITIVITY OF AN H OURLY SOIL WATER BALANCE MODEL ................................ ................................ ............................... 212 Introduction ................................ ................................ ................................ ........... 212 Materials and Methods ................................ ................................ .......................... 214 Hourly SWB ................................ ................................ ................................ .... 214 Case Studies ................................ ................................ ................................ .. 216 Bench test study ................................ ................................ ...................... 216 GCREC field plot study ................................ ................................ ............ 218 Sensitivity Analysis ................................ ................................ ......................... 21 9 Results ................................ ................................ ................................ .................. 220 Irrigation Adequacy ................................ ................................ ........................ 220 B ench test study ................................ ................................ ...................... 220 GCREC field plot study ................................ ................................ ............ 222 Scheduling Efficiency ................................ ................................ ..................... 224 Bench test study ................................ ................................ ...................... 224 GCREC field plot study ................................ ................................ ............ 225 Discussion ................................ ................................ ................................ ............ 227 Conclusion ................................ ................................ ................................ ............ 229 7 CONCLUSIONS ................................ ................................ ................................ ... 245 LIST OF REFERENCES ................................ ................................ ............................. 250 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 261

PAGE 8

8 LIST OF TABLES Table page 1 1 Summary of restrictions on outdoor water use from the posupply for each water management district ................................ ................................ ................. 59 1 2 The parameters of K and M d S av were estimated from So il Conservation Service (1964) cylinder infiltrometer data by Chu (1978) to be used as guidelines for parameter estimates. ................................ ................................ .... 59 1 3 Soil parameters originally put together by Rawls et al. (1982; 1983) and summarized by Ogden and Saghafian (1997). ................................ ................... 60 1 4 Warm season turfgrass crop coefficients calculated from various studies. ......... 61 1 5 Parameters used to calculate the redistribution coeffi cient determined by Gowdish and Muoz Carpena (2009). ................................ ................................ 61 1 6 Ranges of accephydrological parameters for each method for calculati ng soil water retention curves and hydraulic conductivity curves available in the CHEMFLO 2000 model ................................ ................................ ...................... 62 1 7 Description of virtual landscape zones used in the SWAT testing protocol (IA 2008) ................................ ................................ ................................ .................. 63 1 8 Crop coefficients for zones 1, 2, and 6 for the SWAT test (IA 2008). ................. 64 1 9 Summary results from the SWAT test. ................................ ............................... 65 2 1 The ET O and rainfall for the historica l time period (2001 2009) and the study period (2009 2010). ................................ ................................ ............................ 88 2 2 Average monthly irrigation application from February 2009 through January 2011 as compared to average monthly historical irrigation and average gross irri gation determined using Equations 2 1 and 2 3. ................................ ............ 88 2 3 Cumulative irrigation application over the 2009 2010 study period as compare d to cumulative gross irrigation requirement calculated using Equations 2 1 and 2 3. ................................ ................................ ....................... 89 2 4 Cumulative irrigation application as c ompared to cumulative historical irrigation application calculated from 9 years of billing data, when available. ..... 89 2 5 Average monthly irrigation application from February 2009 through January 2011 as compared to average monthly historical irrigation and average gross irrigation where additional cooperators were added to the comparison treatment through billing records. ................................ ................................ ....... 89

PAGE 9

9 3 1 Number of cooperators by location and treatment. ................................ ........... 110 3 2 Description of seasonal periods used to account for various dry/wet periods occurring throughout the study period. ................................ ............................. 110 3 3 Statistics for irrigation application by each treatment over the study period. .... 111 3 4 Turfgrass quality ratings for each treatment a were evaluated seasonally b ........ 112 4 1 Crop coefficients used for each study locati on adapted from Jia et al. (2009). 145 4 2 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 1 (0 22.5 cm) at the PSREU location. ................................ ............. 145 4 3 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 2 (22.5 3 7.5 cm) at the PSREU location. ................................ ........ 146 4 4 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 3 (37.5 52.5 cm) at the PSREU location. ................................ ........ 146 4 5 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 4 (52.5 67.5 cm) at the PSREU location. ................................ ........ 147 4 6 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 5 (67.5 8 2.5 cm) at the PSREU location. ................................ ........ 147 4 7 Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 6 (82.5 9 7.5 cm) at the PSREU location. ................................ ........ 147 4 8 Hydraulic parameters required for each model for the Weathermatic at the GCREC location. ................................ ................................ .............................. 148 4 9 Hydraulic parameters required for each model for the Toro at the GCREC location. ................................ ................................ ................................ ............ 148 4 10 Hydraulic parameters required for each model for TIME at the GCREC location. ................................ ................................ ................................ ............ 149 4 11 Hydraulic parameters required for each model for RTIME at the GCREC location ................................ ................................ ................................ ............. 149 4 12 Evaluation measures of RMSE and NSE used to evaluate the model results during the calibration and validation periods for each layer in the PSREU study. ................................ ................................ ................................ ................ 150 4 13 Evaluation measures o f RMSE and NSE used to evaluate the model results during the calibration and validation periods for each treatment in the GCREC field study. ................................ ................................ .......................... 151

PAGE 10

10 5 1 Crop coefficients for each study. ................................ ................................ ...... 187 5 2 Summary of treatment periods evaluated by the models. ................................ 188 5 3 Cumulative irrigation, water savings compared to GIR, and performance scores during the frequent rainfall period, lasting sixty days, for each controller. ................................ ................................ ................................ .......... 189 5 4 Cumulative irrigation, water savings, and performance scores during the infrequent rainf all period, lasting sixty days, for each controller. ....................... 189 5 5 Cumulative irrigation, water savings, and performance scores during the low ET O period, lasting sixty days, for each controller. ................................ ........... 190 5 6 Cumulative irrigation, water savings, and performance scores during the high ET O period, lasting sixty days, for each controller. ................................ ........... 190 5 7 Correlation coefficients were used to determine dependence between the performance results and irrigation application for each treatment and model . 191 6 1 Summary of data used for the soil water balance for the bench test study during the thirty days ending Jun. 5, 2010, rep resenting the high ET O period. 231 6 2 Statistical results a of irrigation adequacy and scheduling efficiency for each ET controller in the bench test study after performing the global sensitivity analysis using the Monte Carlo simulation. ................................ ...................... 231 6 3 Summary of data used for the soil water balance during each 30 day evaluation period for the GCREC field plot study. ................................ ............ 232 6 4 Statistical results a of irrigation adequacy and scheduling efficiency for the Weathermatic in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation. ................................ ...... 233 6 5 Statistical results a of irrigation adequacy and scheduling efficiency for the Toro in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation ................................ ...................... 234 6 6 Statistical results a of irrigation adequacy and scheduling efficiency for TIME in the GCREC field plot study after perfo rming the global sensitivity analysis using the Monte Carlo simulation. ................................ ................................ .... 235 6 7 Statistical results a of irrigation adequacy and scheduling efficiency for RTIME in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation. ................................ ................................ .... 236

PAGE 11

11 LIST OF FIGURES Figure page 3 1 Monthly ET O and rainfall for all locations during the study period with error bars representing the standard error of measurements across all study locations. ................................ ................................ ................................ .......... 113 3 2 Cumulative irrigation application for all five treatments occurring on the flatwoods locations during the dry 2012 season. ................................ .............. 114 3 3 Cumulative irrigation application for all five treatments occurring on the sand locations during the dry 2012 season ................................ ............................... 114 3 4 Cumulative irrigation application for all five treatments occurring on the flatwoods locations during the wet 2012 season. ................................ ............. 115 3 5 Cumulative irrigation application for all five treatments occurring on the sand locations during the wet 2012 season ................................ .............................. 115 3 6 Cumulative irrigation application for all five treatments occurring on the flatwoods locations during the dry 2013 season. ................................ .............. 116 3 7 Cumulative irrigation application for all five treatments occurring on the sand locations during the dry 2013 season ................................ ............................... 116 3 8 Cumulative irrigation application for all five treatments occurring on the flatwoods locations during the wet 2013 season. ................................ ............. 117 3 9 Cumulative irrigation application for all five treatments occurring on the sand locations during the wet 2013 season ................................ .............................. 117 4 1 Time series comparison of volumetric water content measured at the 15 cm depth to three models in Citra, FL during the calibration period. ...................... 152 4 2 Time series comparison of volumetric water content measured at the 30 cm depth to two models in Citra, FL during the calibration period. ......................... 152 4 3 Time series comparison of volumetric water content measured at the 45 cm depth to two models in Citra, FL during the calibration period. ......................... 153 4 4 Time series comparison of volumetric water content measured at the 60 cm depth to two models in Citra, FL during the calibration period. ......................... 153 4 5 Time series comparison of volumetric water content measured at the 75 cm depth to two models in Citra, FL during the calibration period. ......................... 154

PAGE 12

12 4 6 Time series comparison of volumetric water content measured at the 90 cm depth to two models in Citra, FL during the calibration period. ......................... 154 4 7 Time series comparison of volumetric water content measured at the 15 cm depth to the three models in Citra, FL during the validation period. .................. 155 4 8 Time series comparison of volu metric water content measured at the 30 cm depth to the two models in Citra, FL during the validation period. .................... 155 4 9 Time series comparison of volumetric water content measured at the 45 cm depth to the two models in Citra, FL during the validation period. .................... 156 4 10 Time series comparison of volumetric water content measured at the 60 cm depth to the two models in Citra, FL during the validation period. .................... 156 4 11 Time series comparison of volumetric water content measured at the 75 cm depth to the two models in Citra, FL during the validation period. .................... 157 4 12 Time series comparison of volumetric water content measured at the 90 cm depth to the two models in Citra, FL during the validation period. .................... 157 4 13 Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for the Weathermatic during the calibration period. .............................. 158 4 14 Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for the Toro during the calibration period. ................................ ............ 158 4 15 Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for TIME during the calibration period. ................................ ................. 159 4 16 Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for RTIME during the calibration period. ................................ ............... 159 4 17 The volumetric water content was estimated using all three models for the Weathermatic during the validation period at the GCREC. ............................... 160 4 18 The volumetric water content was estimated using all three models for the Toro during the validation period at the GCREC ................................ .............. 161 4 19 The volumetric water content was estimated using all three models for TIME during the validation period at the GCREC ................................ ....................... 162 4 20 The volumetric water content was estimated using all three models for RTIME during the validation period at the GCREC ................................ ........... 163 5 1 Thirty day totals of ET O and rainfall for the frequent rainfall period that must meet or exceed the thresholds for a valid testing period. ................................ 192

PAGE 13

13 5 2 Thirty day performance scores for the frequent rainfall period by the ET Water. ................................ ................................ ................................ ............... 193 5 3 Thirty day performance scores for the frequent rainfall period by the Toro. ..... 194 5 4 Thirty day performance scores for the frequent rainfall period by the Weathermatic. ................................ ................................ ................................ .. 19 5 5 5 Thirty day totals of ET O and rainfall for the infrequent rainfall period that must meet or exceed the thresholds for a valid testing period. ................................ 196 5 6 Thirty day performance scores for the infrequent rainfall period by the ET Water. ................................ ................................ ................................ ............... 197 5 7 Thirty day performance scores for the infrequent rainfall period by the Toro. .. 198 5 8 Thirty day performance scores for the infrequent rainfall period by the Weathermatic. ................................ ................................ ................................ .. 199 5 9 Thirty day totals of ET O and rainfall for the low ET O period that must meet or exceed the thresholds for a valid testing period. ................................ ............... 200 5 10 Thirty da y performance scores for the low ET O period by the ET Water. ......... 201 5 11 Thirty day performance scores for the low ET O period by the Toro. ................. 202 5 12 Thirty day totals of ET O and rainfall for the high ET O period that must meet or exceed the thresholds for a valid testing period. ................................ ............... 203 5 13 Thirty day performance scores for the high ET O period by the ET Water. ........ 204 5 14 Thirty day performance scores for the high ET O period by the Toro. ................ 205 5 15 Thirty day performance scores for the high ET O period by the Weathermatic. 206 5 16 Thirty day totals of ET O and rainfall for the field plot study at the GCREC that must meet or exceed the thresholds for a valid testing period. ......................... 207 5 17 Thirty day performance scores for the Weathermatic evaluated during the field plot study at the GCREC. ................................ ................................ .......... 208 5 18 Thirty day performance scores for the Toro evaluated during the field plot study at the GCREC. ................................ ................................ ........................ 209 5 19 Thirty day performance scores for the TIME treatment evaluated during the field plot study at the GCREC. ................................ ................................ .......... 210 5 20 Thirty day performance scores for the RTIME treatment evaluated during the field plot study at the GCREC. ................................ ................................ .......... 211

PAGE 14

14 6 1 Probability densities used in the bench test study. ................................ ........... 237 6 2 Probability densities used in the GCREC field plot study. ................................ 238 6 3 Response in irrigation adequacy to varying the input factor of maximum allowable depletion during the bench test study ................................ ............... 239 6 4 Response in irrigation adequacy due to varying the input factor of maximum allowable depletion for the Weathermatic for the period ending Nov. 11, 2006 in the GCREC field study. ................................ ................................ ................. 240 6 5 Response in irrigation adequacy due to varying the input factor of maximum allowable depletion for the Toro for the period ending Nov. 11, 2006 in the GCREC field study. ................................ ................................ .......................... 240 6 6 Response in irrigation adequacy due to varying the input factor of maximum allowable depletion for TIME for the pe riod ending May 12, 2007 in the GCREC field study. ................................ ................................ .......................... 241 6 7 Response in irrigation adequacy due to varying the input factor of max imum allowable depletion for RTIME for the period ending May 12, 2007 in the GCREC field study. ................................ ................................ .......................... 241 6 8 Response by the Weathermat ic during the bench test study between s ................................ 242 6 9 Response in scheduling efficiency due to varying the input factor of saturation for the Weathermatic for the period ending Aug. 10, 2007 in the GCREC field study ................................ ................................ ........................... 242 6 10 Response in scheduling efficiency due to varying the input factor of saturation for the Toro for the period ending Apr. 12, 2007 in the GCREC field study ................................ ................................ ................................ ......... 243 6 11 Response in scheduling efficiency due to varying the input factor of saturation for TIME for the period ending Dec. 11, 2006 in the GCREC field study ................................ ................................ ................................ ................. 243 6 12 Response in scheduling efficiency due to varying the input factor of saturation for RTIME for the period ending Dec. 11, 2006 in the GCREC field study. ................................ ................................ ................................ ................ 244

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EFFECTIVENESS OF SMART CONTROLLERS FOR WATER CONSERVATION IN RESIDENTIAL IRRIGATION By Stacia L. Davis May 2014 Chair: Michael D. Dukes Major: Agricultural and Biological Engineering The ET controller has the potential to produce water savings without sacrificing landscape quality in Florida when implemented at residences that habitually irrigate excessively and programmed with settings that are appropriate for the landscape. The education provided as part of the education treatments was effective at r educing irrigation application in addition to water savings by the technology without education. Proper controller programming was essential to achieving maximum water savings while maintaining acceptable landscape quality. If the time and effort of prop er programming is not achievable, other smart irrigation technologies that do not increase existing irrigation, such as soil moisture sensors, would be more appropriate. The time based treatments selected to represent actual homeowner practices in Florida resulted in excessive irrigation during the field plot study. Though the ET controllers decreased irrigation compared to 100% replacement of the net irrigation requirement, there were no additional water savings compared to 60% replacement. Additionally, the performance scores for the ET controllers were not different from the 60% replacement with all treatments experiencing over irrigation. The ET controller should be implemented only in situations of excessive irrigation habits.

PAGE 16

16 In efforts to better ap proximate water conservation potential, an hourly soil water balance (SWB) was developed as an alternative to the industry standard, the daily SWAT test, for evaluating irrigation adequacy, a measure of under irrigation, and scheduling efficiency, a measur e of over irrigation. The hourly SWB better match ed the deterministic soil water model compared to the daily SWAT test thus improv ing estimates of volumetric water content in a shallow root zone Additionally, the hourly SWB was preferential to the daily SWAT test based on the combination of more accurate landscape programming required to obtain high performance scores and the realistic representation of soil water movement by hourly SWB resulting in a more accurate translation of SWAT scores to controlle r performance in the landscape. As a result, the hourly SWB should be used as an alternative method for testing ET controllers that will more accurately reflect true soil water relationships thus allowing for the estimation of water c onservation potential.

PAGE 17

17 CHAPTER 1 INTRODUCTION Historically, t he primary purpose of irrigation was to produce better yield of agricultural crops by supplementing water when rainfall is inadequate to maintain sufficient soil moisture. Over time, irrigation technology has become affordable in the residential sector, thus providing an opportunity for mor e lush and vibrant residential more than 30 million acres of land make up home lawns in the United States (National Turfgrass Federation, Inc. 2003). As of 2005, domest ic water use, which includes both indoor and outdoor water consumption, accounted for 58 % of the public water supply in the United States and 60 % of the public water supply in Florida (Kenny et al. 2009). Florida was ranked fourth in domestic water use; t he only states having higher consumption were California, Texas, and New York It is commonly assumed that access to potable water is a right and should be provided unconditionally with minimal costs. However, using the limited potable water supply to irrigate landscapes, which are typically considered ornamental in nature and not purposeful to sustain life such as in agriculture, may not be socially responsible. Additionally, the tariffs for potable water are generally low with an international averag e of $2.03 per m 3 of combined water and wastewater calculated from a 276 city survey (GWI 2011). This rate does not cover the real costs associated with the water and wastewater treatment processes while maintaining the aging infrastructure used to provid e clean water to most homes in the United States (USEPA 2002). On the other hand, turfgrass can be a functional part of the landscape by reducing soil erosion, preventing particulate inhalation from excess dust, dissipating

PAGE 18

18 heat at the soil surface level and providing a wildlife habitat. There are also recreational and aesthetic reasons for maintaining turfgrass landscapes such as providing cushioned support for sports and playground activities and improving mental health and quality of life (Beard and Green 1994). Finding better ways to preserve turfgrass based landscapes through modernized irrigation techniques may be a good compromise. Current Irrigation Practices The Florida Water Resources Act of 1972 was enacted to address the water quantity and quality problems in the state of Florida (Chapter 373 Florida Statutes). The Water Resources Act specified that water related decisions would be managed in the best interest of the public and established regional jurisdiction to five Water Management Dist ricts supervised by the Florida Department of Environmental Protection (FDEP). In addition to creating the five districts, the Water Resources Act specified the activities of the district s (Carriker 2006). Water related regulations such as consumptive use permitting and surface water management Providing flood control systems Monitoring and researching water related a ctivities paid for by taxes Develop ing regional water supply plans The five Water Management Districts are Northwest Florida (NWFWMD), Suwan nee River (SRWMD), St. Johns River (SJRWMD), Southwest Florida (SWFWMD), and South Florida (SFWMD). The boundaries of the districts were chosen based on hydrological features and did not fall within political boundary lines. A nine to eleven

PAGE 19

19 person board governs each district with members selected by the governor (Carriker 2006). Chapter 373.62 of the Florida Statutes outlines the only irrigation system regulations common to the state with the possibility of additional restrictions at the district, coun ty, municipality, or other local level. Irrigation watering restrictions are one well known regulation mandated by the districts where irrigation is allowed only for certain times of day and days of week (Table 1 1). The enforcement in the form of warnin gs and fines for violating watering restrictions is typically carried out at the county level resulting in varying intensity throughout the state. The American lawn became a fundamental part of the home after the Second World War when suburbanization ex ploded (Robbins and Birkenholtz 2003). In Florida, in ground, automatic irrigation system s ha ve become a standard installation for single family homes with primarily turfgrass based landscapes The most popular type of turfgrass is St. Augustinegrass due to its potential for dark green color and moderate levels of drought tolerance, shade tolerance, and salt tolerance. For homes with automatic irrigation systems, more than half of total household water use can be irrigation In Nevada, Devitt et al. (2 008) found that 66% of the total water used at all sites was outdoors for 27 residential sites. Similar results were also found in central Florida where 64% of the total household water use was determined to be irrigation (Haley et al. 2007). Soil Water M odeling Models are used to solve problems in the real world when actual data cannot be collected. In soil water modeling, mathematical representations are used to model

PAGE 20

20 hydrological processes in the variably saturated shallow soil profile known as the vad ose zone. S oil water models can be used for various purposes. E xploratory tools to determine ce rtain processes or interactions M anagement tools to make real time water management decisions and c ontrol hydrological systems P redictive tools to planning d ecision s and contribute develop policy Models can be described in a number of ways such as physically based vs. empirically based, temporal vs. spatial, or static vs. dynamic. The appropriate model should be created or selected based on the purpose of th e model while factoring the types of available input data and ability to produce meaningful outcomes. Simple models typically use a soil water balance or tipping bucket approach to water movement (evapotranspiration exits, precipitation and irrigation en ter) for a fixed, homogeneous soil profile whereas complex models evaluate continuous water movement through one dimensional or two dimensional flow through a vertically heterogeneous soil profile (Ranatunga et al. 2008). Complex models require intensive input data provided by the user, numerical approximation techniques, and intense computations compared to simple models that are genera lly easier to use (Jeong et al. 2010). However, simple models sometimes lack key biological and hydrological processes m aking the model less realistic (Ranatunga et al. 2008). As watersheds experience urbanization, permeable surfaces become impervious from structures such as buildings, roads, and sidewalks that create small, individualized soil water systems located through o ut the urban area (Jeong et al. 2010). The fragmentation of the watershed to many small, unconnected landscapes renders many large scale models inapplicable. Additionally, building construction changes the natural

PAGE 21

21 soil properties found on soil maps and in other sources to the properties of a generic fill material that may not be represented in the model. For urban residential landscapes, a model that spatially addresses the hydrogeological properties of individual home lots would be most beneficial. S oil water models are calculated on various timesteps depending on the desired outcome. Large watershed models typically run on an annual, monthly, or weekly basis to minimize calculation errors whereas smaller landscape models typically run on daily or su b daily timesteps. For residential turfgrass systems, combining sub daily and daily time intervals can be beneficial to better capture short term hydrological processes such as individual rainfall and irrigation events while still evaluating long te rm wat er movement (Jeong et al. 2010). There are many considerations to modeling the hydrological processes of a turfgrass system such as turfgrass species, plant growth, climatic conditions, plant density, water table depth, and water availability. Turfgrass systems are unique compared to agricultural systems because the goal of turfgrass in the landscape is to be ornamental and not marketable. As a result, it is unfavorable to increase yield of turfgrass due to increased mowing and nutrient requirements with out a return on investment (Rosenberg et al. 2011). Research studies have shown that turfgrass growing in water conservative conditions can provide a minimally accepta ble aesthetic (Bonos and Murphy 1999; Davis et al. 2009; Fu et al. 2004; McCready et al. 2009; Peacock and Dudek 1984) while resulting in les s plant growth (Banuelos et al. 2011; Biran et al. 1981; Carrow 2006; Hsiao and Acevedo 1974). As a result, climatic conditions (evapotranspiration) and water availability (soil water storage, irrigatio n,

PAGE 22

22 rainfall, drainage, and runoff) are the most important parameters in modeling a turfgrass system. Soil Water Storage The amount of water able to be stored in the soil column of interest (root zone) is dependent on the characteristics of the soil. The s oil column consists of soil particles structured with pore spaces sized from soil type and the level of compaction. For example, sandy soils have larger pore spaces compared to clay soils because the soil particles for sand are larger. Air and water exis t in the pore spaces, or storage, of the soil column. Saturation of the storage occurs when the pore spaces consi st of only water and no air (IA 2005). However, the gravitational forces are stronger than the surface tension forces (cohesive and adhesive ) between the water molecules and the soil particles when saturated causing drainage to the next layer of soil assu med to be unsaturated (Richards 1931). At the point when gravitational forces are less than tension forces, drainage due to gravity ceases and the soil column holds the maximum amount of water that can be stored given the soil characteristics. When this phenomenon occurs, the volume o f water in the soil column divided by the total volume of the soil column can be calcul ated as field capacity (FC) (IA 2005). Through evapotranspirative losses, water will continue to leave storage until the work required by the plant material to upta ke w ater is too great (Richards 1931). This occurs at the permanent wilting point (PWP), also calculated as the volume of water in the soil column divided by the tota l volume of the soil column (IA 2005). Typically, plants cannot recover from water stress wh en PWP is reached.

PAGE 23

23 The amount of water that can be stored in a specific soil column is the available water holding capacity (AWHC) calculated as t he difference in FC and PWP (IA 2005). The plant available water (PAW) can be calculated from AWHC by multipl ying by the volume of the soil column to convert to total storage. For plant management purposes, a maximum allowable depletion (MAD) level is selected to ensure that plants do not become water stressed. Readily available water (RAW), calculated by multip lying PAW by MAD, is the amount of water above the MAD level that can be depleted while maintaining well watered conditions (i.e. no water stress). T he amount of water in the soil column c an be calculated using a soil water balance (IA 2005) (1 1) The SWB is the soil water level (mm) for timestep i or i 1, PWR is the plant water requirement (mm) on day i, R e is effective rainfall (mm) on day i, and IWR net is the net irrigation water requirement (mm) on day i. The soil water balance level can fluctuate based on the storage parameters of the soil column that determines the AWHC. Soil Parameters The hydraulic conductivity and soil water retention parameters are important hydraulic properties used to determine water movement through a soil column. Hydraulic conductivity (K) is the rate that water can move through the soil column. Hydraulic conductivity primarily depends on soil moisture and pressure head, but can be affected by temperature and salinity of the soil column (Nielson et al. 1986). Temperature and salinity affect the bonding forces between the soil particles and water molecules, thus affecting capillary f low through the pores (Richards 1931). Soil water retention curves relate soil moisture and m atric potential.

PAGE 24

24 Empirical formulas exist to predict the soil water retention curves (Brooks and Corey 1964). (1 2) The is soil moisture content s r is soil moisture content at PWP h is the pressure head (cm) at a certain depth in the soil profile, h b popular formula was de veloped by van Genuchten (1980). (1 3) The m minus n. Hydraulic properties can be measured in the field with equipment such as tensiometers and soil m oisture sensors (Nielson et al. 1986). Though the parameters woul d be most accurate when directly measured, it is typically not feasible due to time and cost. Many have developed theoretical approaches to determining hydraulic conductivity related to pore si ze distribution (Nielson et al. 1986). The equation developed by Childs and Collis George (1950) were modified by Marshall (1958) to b ecome the modified CCG equation. (1 4) The to tal number of divisions of the soil profile. There is an alternate equa tion developed by Mualem (1976) and another equation proposed by Burdine (1953). (1 5)

PAGE 25

25 (1 6) The x is the depth in from 0 at PWP to 1 at saturation. (1 7) In Europe, a widely used equation is the exponential Gardner relationshi p (Gardner 1958). (1 8) Hydraulic conductivity cu rves were further developed by v an Genuchten (1980) using the Mualem relationship it can be expressed a s a function of soil moisture or as a function of pressure head. (1 9) (1 10) The m ranges from 0 to 1 and can b e estimated from n as m=1 1/n. Additionall y, hydraulic conductivity from v an Genuchten using the Burdine relationship can be expressed a s a function of soil moisture or as a function of pressure head. ( 1 11) (1 12) The n is greater than 2 and m ranges from 0 to 1 and can be estimated from n as m=1 2/n. Chu (1978) analyzed cylinder infiltrometer data collected by the Soil Conservation Service from field sites having multiple soil types to determine soil parameters (Table 1 2). Additionally, Rawls et al. (1982 ; 1983) developed various soil parameters for a variety of soil types (Table 1 3).

PAGE 26

26 The soil moisture profile can be affected by the historical wetting patterns due to hysteresis, or the dependence of a system on its current and past environment. As a result, soil hydraulic properties can evolve based on the wetting pattern and effectively change the soil wate r retention curve (Izady et al. 2009). As a result, the wetting process may require a different soil water retention curve than the drying process. It can be attributed to geometric nonuniformit y of pores, differences in spatial connectivity of pores, variation in the liquid solid contact angle, a nd air entrapment (Izady et al. 2009). Though hysteresis can be neglected in soil water modeling, it is an important process for redistribution and lon g term modeling. Preferential flow can occur in soil profiles that have holes, cracks, or other larger pores that result in fast and uneven flow of water through the soil profile. Water movement in soil profiles that experience preferential flow may not be described with the same hydraulic properties as a n isotropic soil profile. Evapotranspiration Evapotranspiration (ET) occurs as part of a balanced energy budget that exchanges energy for outgoing water at the surface of the plant, including bo th the evaporation from the soil surface and the transpiration throu gh plant material (Allen et al. 1998). The components that affect the rate of ET are solar radiation, temperature, relative humi dity, and wind speed (ASCE 2005). Though there are variou s methods of estimating ET including field lysimeter experiments, soil water studies, and a long term inflow outflow water budget for large areas, it is more efficient and economical to predict ET of a reference plant material using climatic based equation s. The reference plant material is located around the

PAGE 27

27 sensors collecting the climatic data used to calculate reference ET (ET o ) and should be an actively growing, well watered, dense green gra ss of uniform height (ASCE 2005). The Penman Monteith equat ion has been endorsed by many organizations including the American Society of Civil Engineers (ASCE), International Commission for Irrigation and Drainage (ICID), Food and Agriculture Organization of the United Nations (FAO), and the Ir rigation Association (ASCE 2005). The most recent form of the Penman Montieth equation is the ASCE standardized refere nce evapotranspiration equation. (1 13) Variables defined. (1 14) (1 15) (1 16) (1 17) (1 18) (1 19) (1 20) (1 21) (1 22) (1 23) (1 24) (1 25) (1 26) (1 27) ET o = reference evapotranspiration, mm/day = slope of the saturation vapor pressure temperature curve, kPa/C T = daily mean air temperature, C

PAGE 28

28 e s = saturation vapor pressure, kPa e o T = saturation vapor pressure function, kPa e a = actual vapor, kPa RH = relative humidity, % R n = net radiation, MJ/m 2 /day R ns = net short wave radiation, MJ/m 2 /day R nl = net outgoing long wave radiation, MJ/m 2 /day R s = incoming solar radiation, MJ/m 2 /day = albedo or canopy reflection coefficient, 0.23 Boltzmann constant, 4.901 x 10 9 MJ/K 4 /m 2 /day f cd cd R so = calculated clear sky radiation, MJ/m 2 /day R a = extraterrestrial radiation, MJ/m 2 /day z = station elevation above sea level, m d r = inverse relative distance factor for the earth sun = solar declination, rad = latitude, rad s = sunset hour angle, rad J = Julian day G sc = solar constant, 4.92 MJ/m 2 /hr G = daily soil heat flux density, 0 MJ/m 2 /day u 2 = wind speed at 2 m height, m/s This equation takes into account net radiation, R n (MJ m 2 d 1 ); heat flux, G ( MJ m 2 d 1 C 1 ), e s (kPa), e a (kPa); temperature, T (C); and wind speed, u 2 (m s 1 ). The measured mean atmospheric pressure (ASCE 2005). A grass reference crop rather than alfalfa is used for r esidential landscapes; for a grass reference, the constants of C n and C d are 900 and 0.34 respectivel y (ASCE 2005). T he ET of the irrigated crop (ET C ) is calculated from the ET of the reference plant material (ET O ) under the same climatic condition s using a ratio called a crop coefficient (K C ) (1 28) The K C accounts for differences in ground cover, canopy characteristics, and aerodynamic resistance from the reference grass (Jenson et al. 1990). Values of K C can vary based on soil water availability, nutrient management, and overall turfgrass health (Gibeault et al. 1989); ideally, K C values should be calculated from well watered,

PAGE 29

29 properly managed turfgrasses. Though annual K C values have been suggested (Gib eault et al. 1989; Allen et al. 1998), monthly K C values should be used when available (Table 1 4). In the soil water balance, the PWR equals the amount of water necessary to main tain healthy plant material (IA 2005). The PWR is calculated as the plant s pecific evapotranspiration (ET C ). In Florida, the maximum ET occurs in May when there is the least cloud cover, maximum vapor pressure deficit conditions, and the highest incom ing solar radiation (Jia et al. 2009). Other summer months are not as high due to the rainy season occurring from Jun through Nov resulting in reduced ET rates due to frequent rainfall events and increased cloud cover. The K C values in South Florida exceeded values for all other locations for all months except in Jun and Jul for Las Vegas, NV. The K C values from California maximized in May, similar to the Florida locations, but at a much lower average. Precipitation Precipitation is considered to be the most variable input to hydrological models. It can come in the form of driz zle, rainfall, freezing rainfall, snow, or hail. In Florida, the primary type of precipitation is rainfall. Rainfall can vary spatially, with localized rainfall events, and temporally, with fluctuating intensities. Effective rainfall is the amount of w ater from a rainfall event that can be stored in the soil profile. In the tipping bucket method of a soil water balance model, effective rainfall was limited to the portion of total daily rainfall that would cause the soil water storage to reach FC after PWR was taken into account (IA 2005). Rainfall that exceeded soil water storage was considered lost due to surface runoff or deep

PAGE 30

30 percolation. In a real soil water storage system, this assumption would be inaccurate because water can still enter the soil profile until reaching saturation, which is above field capacity, and may be able to temporarily store water on the soil surface depending on the landscape features such as depressions and slopes. Urbanization of watersheds affected the way rainfall in teracts with the soil profile. Impermeable surfaces may cause excess water to be directed onto the soil profile causing flooding when the volume exceeds the soil infiltration rate. Chang (2007) found that the impact of urbanization during individual stor m events is more significant than mean annual runoff. This indicates that the short term effects on soil water storage from rainfall may only be apparent at a small timestep. A major effect of whether rainfall enters the soil water profile or becomes ru noff is the intensity of the event. Intensity is determined by measuring the volume of rainfall occurring over a certain time period. Intensity has greater fluctuation and higher peaks when the timestep is small, thus allowing more accurate depiction of the soil water processes. However, when the timesteps are extremely small, such as a minute, it may be difficult to determine separate rainfall events that occur on the same day because intensities are so low. In the Chemicals, Runoff and Erosion from Ag ricultural Management Systems (CREAMS) model, rainfall events were considered independent when more than 180 minutes passed be tween measured rainfall (Knisel 1980). Irrigation Irrigation is the amount of water required to maintain plant health and should b e applied only on days when the water level reaches the MAD level. The depth of irrigation required to increase soil water storage to field capacity, thus satisfying PWR, is the net irrigation (IWR net ) (IA 2005).

PAGE 31

31 Irrigation systems that are perfectly eff icient may apply IWR net directly. However, irrigation systems are not perfectly efficient thus requiring additional irrigation application to ensure IWR net is received in the soil column. Because efficiency is difficult to quantify directly, distribution uniformity (DU) of the irrigation system is typically us ed as a substitute (Burt et al. 1997). The DU is typically measured using catch can testing that involves placing a grid of buckets throughout the irrigated area and measuring the volume of water co llected from an irrigation event. The lower quarter distribution uniformity (DU lq ) can be calculated as the ratio of the average of the lowest 25% of volumes (d lq ) to the average of all volumes measured (d avg ) (Burt et al. 1997) (1 29) The IA (2005) suggested that DU lq overestimates the effect of non uniform landscapes. Therefore, the DU lh should be used for irrigation scheduling purposes and can be estimated from DU lq (1 30) Using the DU lh the amount of additional water required to achieve IWR net can be calculat ed as the RTM (IA 2005). (1 31) Multiplying the RTM by IWR net produces the gross irrigation requirement (GIR), or the amount of water that should be scheduled in order to apply IWR net to the soil column (IA 2005). Efficiency is typically affected by irrigation system design, such as not having head to head coverage, and pressure problems. Additional considerations to take into account when scheduling irrigation include soil type, sprinkler type, and slope. Soil type and sprinkler type are important to

PAGE 32

32 determining the maximum runtime length, especially if the irrigation application rate of the sprinklers exceeds the infiltration rate of the soil type. Also, high slopes can decrease the infiltration rate, which will also affect the maximum runtime length. Irrigation events should be broken into multiple irrigation events where the cycle lengths are less than the maximum runtime, thus minimizing runoff. The length of ti me between irrigation events, or soak times, should also be determined based on these factors. Infiltration Infiltration of water through the soil profile is primarily dependent on soil parameters, pre existing soil moisture conditions, and precipitation intensity. According to Mein and Larson (1973), infiltration is one of the largest influences on runoff, but is one of the most deficient processes in hydrological modeling. In the tipping bucket method of a soil water balance model, infiltration is cons idered instantaneous. Additionally, the total volume of infiltrated water is assumed to be equal to the available water storage of the soil profile and any excess water is lost to surface runoff or deep percolation. Though these assumptions simplify the soil water balance, it fails to follow physical responses to infiltration and can lead to compounding errors if applied over long time periods for small study sites. Infiltration can observe three behaviors based on values of infiltration rate, rainfall intensity, and saturated hydraulic conductivity (Mein and Larson 1973). When rainfall intensity is less than saturated hydraulic conductivity, all water infiltrates into the soil column and there is no runoff. In this case, the soil moisture changes but does not exceed field capacity. When rainfall intensity exceeds saturated hydraulic conductivity but does not exceed the infiltration rate, all of the water infiltrates causing the soil

PAGE 33

33 moisture to exceed field capacity. However, if the rainfall intensity exceeds the infiltration rate, the infiltration rate declines while runoff is generated. Many empirical equations have been suggested to better approximate infiltration processes. An infiltration equation suggested by Kostiakov and Horton was a simplistic representation that had the a bility to fit most data (Childs 1969). The Holtan model (1961) expressed infiltration as a function of pore space and not as a function of time. However, these and other empirically based equations require paramete rs that are difficu lt to estimate (Mein and Larson 1973). The Philip equation (1957) was developed for ponded, homogeneous soils using more predictable parameters. However, this equation can only assume ponding conditions and cannot account for initial c onditions when the infiltration capacity is infinite. An analytic infi ltration model (Parlange et al. 1982) was expanded for redistribution with multipl e rainfall events (Smith et al. 1993). Currently, the two most prominent methods for modeling infiltrati on are Richards equation (1931) and Green Ampt equation (1911). Richards Richards equation (1931) was developed for the flow of water through porous media under satura of liquids in capillary tubes. (1 32) The q is flow rate over the surface area of the soil column, K is the hydraulic conductivity, and is the total water moving field that is based on the gravita tional and pressure forces in the three dimensional plane. The primary difference between

PAGE 34

34 saturated flow and unsaturated flow is the dependency of water movement on capillary forces and hydraulic conductivity of the soil column. Richards equation can be expressed based on pressure head (h), soil moisture on of both (Celia et al. 1990). (1 33) (1 34) (1 35) The and z denotes the depth of the soil column. The structure of these equations assumes positive vertical dimension is upward and the soil column is isotropic. Richards eq uation is a nonlinear partial differential equation that cannot be solved analytically. Instead, numerical approximations such as the finite difference or finite element methods combined with iteration techniques such as Picard or Newton methods are requi red for d iscrete solutions (Celia et al. 1990). Celia et al. (1990) found that despite widespread use, the h based equation combined with a backward Euler time discretization, using any iteration method, produced large mass balance errors. Using the h bas ed equation, smaller timesteps are necessary to obtain convergence and reduce mass balance errors. A timestep of based equation, though perfectly mass conservative, produces discontinuous soil moi sture content profiles that are not easily used. It was also found that the mixed form of Richards equation combined with a backward Euler time approximation and Picard iteration would alleviate these problems. Additionally, finite difference approximati ons were suggested as preferable to finite element approximations due to lack of mass lumping. Finite element

PAGE 35

35 approximations produced oscillatory solutions, larger mass balance errors, underpredicting infiltration depth, and exhibiting undershoot errors ah ead of the infiltration front. Green A mpt The Green Ampt (GA) equation (1911) was developed to model infiltration in a homogeneous and isotropic soil profile during ponded conditions. It was a physically based empirical model that was derived from Darcy the soil surface. Infiltration occurs as a sharp wetting front, sometimes referred to as a rectangular piston front, with the wetting front advancing through the lower portion of the soil profile. Water infiltrates com pletely until the soil surface reaches saturation, also known as infiltration capacity (Mein and Larson 1973). At infiltration capacity, conditions change from unponded to ponded. In 1973, Mein and Larson contributed to the GA model by accommodating tim e periods where ponding was not instantaneous. The new model, referred to as GAML, had the ability to change from unponded to ponded conditions during steady rainfall events. The GAML was evaluated at several rainfall intensities and soil moisture conten ts on five types of soils (Mein and Larson 1973). Cumulative infiltration predicted by the Green Ampt equation did not deviate significantly from cumulative infiltration equation for 47 rainfall events in seven locations, the most accurate results were obtained when analyzed at a ten minute time interval (King 2000). Continued development to GAML occurred when Chu (1978) further enhanced GAML to account for unsteady rainfall eve nts that occur from multiple periods of rainfall intensity exceeding the infiltration rate. Chu suggested that the improved equation

PAGE 36

36 adequately modeled infiltration for a watershed due to total estimated runoff approximated measured runoff when replicated for three separate storm events. In 1997, Ogden and Saghafian included the option of having two wetting fronts to account for hiatuses between pulses of a single rainfall event. They suggest that infiltration is significantly underestimated during the second pulse by not redistributing Ampt with redistribution (GAR) performed well with only 2% error in cumulative infiltration for the first wetting front and 4.5% error for the second wetting front when evaluated for 6 h. However, coarser soils that have saturated hydraulic conductivity values greater than 1 cm/h had errors that increased logarithmically. Gowdish and Muoz Carpena (2009) expanded on GAR proposed by Ogden and Saghafian (1997) with modifications to reduce the error seen when adding multiple rainfall events. The new method, titled MGAR, increased the total number of rainfall events possible from two to infinity thus creating the ability for long term analyse s of infiltration. Convergence of the model was obtained for all soil types when using the van Genuchten (1980) retention curves and Brooks and Corey (1964) unsaturated hydraulic conductivity curve. The relationship between error and soil texture was det ermined using a method for non linear least squares curve fitting resulting in the development of a correction factor termed the redistribution coefficient. The MGAR textu red soils, over the 365 h evaluation. The Green Ampt equation (1911) requires an iterative solution to calculate the infiltration capacity.

PAGE 37

37 (1 36) f p = infiltration capacity K s = saturated hydraulic conductivity M d = initial soil moisture deficit S av = capillary suction at the wetting front F = cumulative infiltration Integration of the above equation results in the expression for cumulative infiltration (Mein and Larson 1973). The capillary suction at the wetting front varies as the wetting front moves through the soil during a rain event. Though S av can be determined experimentally, it can also be estimated from inte grating Brooks and Corey (1964). (1 37) The b parameters of K s M d and S av are considered to be constant during a rainfall event because diffusion is assumed to be negligible (Chu 1978). For calculation purposes, a tempo rary shift in the time scale so that ponding begins when time equals zero is necessary to account for cumulative infiltration and is expressed as a pseudotime (t s ) (1 38) The F tp is the cumulative infi ltration at the time ponding occurs and t p is the time at the start of ponding. (1 39) The I (mm) is the rainfall intensity and t n 1 is the time at the beginning of t he current rainfall intensity. T he time to ponding calculation was substituted into an integrated form of Equation 1 36 to get the equation for cumulative infiltration.

PAGE 38

38 (1 40) After ponding, this equation requires an iterative procedure to obtain a solution because the cumulative infiltration term cannot be isolated on one side of the equation. However, t s and t p can be calculated implicitly. Ponding will continue until a specified time when the rainfall intensity falls below the inf iltration capacity and all of the ponded water has infiltrated or become runoff. The time this occurs is called time to end ponding (t j ). (1 41) F t 1 = last calculated cumulative infiltration t t 1 = the time when cumulative infiltration was last calculated These equations are only valid when there are ponding conditions. If there are no ponding conditions because the rainfall intensity is less than the infiltration rate, then the infiltration rate for unponded conditions (f u ) equals the rainfall intensity. Additionally, cumulative infiltration (F u ) would be equal to the amount of rainfall occurring over time. During periods of no rainfall, a diff erential equation for r edistribution of water in the soil column was pro posed by Ogden and Saghafian (1997) (1 42) The Z is the depth to the wetting front, r h is the rainfall flux during hiatus, K i is the o ) is the unsaturated i o ) is the integral of the capillary i o ) and continued deri vations from assumptions from van Genuchten (1980) and Brooks and Corey (1964) relationships, wh

PAGE 39

39 (1 43) was used to correct for errors occurring o ver time for a wide range of soils. (1 44) The T R is the redistribution time and N R is the redistribution number determined from the number of wetting fronts (Gowdish and Muoz Carpena 2009). The values of a 1 a 2 and a 3 were calculated using three intervals ( Table 1 5 ). (1 45) Surface Runoff In many hydrological models, the curve number method developed by the Natural Resources Conservation Service has been implemented to estimate s urface runoff (Garen and Moore 2005). The curve number method is an empirical method designed for flood models, but has expanded for use in continuous models and varies based on soil moisture. It was developed to predict the amount of streamflow in addit ion to baseflow of a water body and does not correspond to overland flow. Because the soil water balance does not have a spatial component, estimating surface runoff based on the curve number would be inappropriate (Jeong et al. 2010). Surface runoff, i n terms of overland flow, can occur when the precipitation rate exceeds the infiltration rate of the soil column or the soil moisture is already at saturation (Garen and Moore 2005). As a result, surface runoff can be quantified based on the infiltration calculations in the model.

PAGE 40

40 Calibration and Validation The accuracy of a model is evaluated through calibration and validation procedures. Calibration is performed to minimize the differences between observed and predicted results and optimize models. A dditionally, validation of the model can be completed to quantify its performance compared to the physical system being modeled. Data used to calibrate a model should not be used to validate it, thus requiring multiple datasets. The combination of goodnes s of fit measures, absolute error measures, and graphical assessments can provide an adequate evaluation of model performance for calibration and validation p urposes (Legates and McCabe Jr. 19 99; Willmott et al. 1985; McCuen et al. 2006; Krause et al. 2005 ). Combining these three methods will result in overcoming the limitations inherent in each individual measure. Goodness of fit measures summarized here include the coefficient of determination (R 2 ) and Nash Sutcliffe coefficient of efficiency ( NSE ). Abs olute error measures include mean absolute error (MAE) and root mean square error (RMSE). Graphical assessments, though not detailed here, include 1:1 graphs and time series graphs. The coefficient of determination, calculated as the square of the Pearson product moment correlation coefficient, is the proportion of the variability that can be explained in the model. It is calculated as a unitless measure of the colinearity of the residuals (Legates and McCabe, Jr. 1999; Jeong et al. 2010). (1 46) O = observed value P = predicted value = average of the observed values = average of the predicted values

PAGE 41

41 N = number of observed or predicted values Values of the coefficient of determination r ange from 0, indicating no agreement, to 1, indicating perfect linear agreement. Limitations to this model include inability to describe nonlinear systems, sensitivity to outliers, and biased results when the model accurately predicts extremes but not valu es near th e mean (Legates and McCabe, Jr. 1999; McCuen et al. 2006). Models that systematically over or under predict measured data can result in high coefficients of determination (Kr ause et al. 2005). The NSE is an alternate goodness of fit parameter t o the coeff icient of determination. (1 47) The numerator of the equation measures the variation in the data that was not explained by the model and the denominator measures the total variation in the observed data (McCuen et al. 2006). Values of the coefficient can range from where values of 0 to 1 indicate that the model, if linear, is unbiased. According to Jeong et al. (2010), NS E larger than 0 indicate minimally acceptable performance and greater than 0.65 on a daily basis is preferred. However, the NS E is subjective and may be lower for subdaily results and higher for monthly or annual outputs. Similar to the coefficient of dete rmination, limitations to the NS E include sensitivity to sample size, outliers, magnitude bias, and time offset bias. The mean absolute error is a model prediction error that is expressed in the units of the variable of interest. It can be calculated to d etermine the magnitude of the differences between th e predicted and observed values. (1 48)

PAGE 42

42 T he optimal value of MAE is 0 indicating that there is no error between the predicted and observed values Willmott and Matsuura (2005) suggested that the MAE is the most natural and unambiguous measure of the magnitude of average error. The root mean square error, also a dimensioned model prediction error, is a function of average error, squared errors, and n 1/2 (1 49) Though frequently used as a measure for environmental models, its meaning is often misinterpreted by res earchers (Willmott and Matsuura 2005). Squaring the residual errors allows the magnitude to affect the metric and not the positive or n egative difference. However, this exacerbates the influence of large errors allowing a few large errors to increase the RMSE value when there may be few small errors. As a result, RMSE will always be greater than MAE unless all residual errors are equal, in which case RMSE equals MAE. Due to MAE being only a function of average error, it is the preferred metric over RMSE for evaluating model performance (Legates and McCabe, Jr. 1999; Willmott and Matsuura 2005). Sensitivity Analysis A sensitivity an alysis is performed to determine the influence of various parameters on the outcome of the model. It can also be used to test the logic and stability of the model by testing for extreme values, thus determini ng its limits (Jones and Luyten 1998; Saltelli et al. 2000). Sensitivity analysis can be local by evaluating an outcome based on varying input parameters individually. It can also be global by evaluating many parameters simultaneously.

PAGE 43

43 Local sensitivity analysis, also called one at a time (OAT) analy sis, can be performed by varying the value of a single parameter while maintaining the original values for all other parameters and quantifying the difference in the outcome. Absolute sensitivity is a measure of the variation of the outcome of interest re lated to the c hange in the selected parameter. (1 50) The y is the outcome variable and k is the parameter selec ted to change (Jones and Luyten 1998). Normalizing the measure to other variables in the model can be done by calculating the relative sensitivity (1 51) Local sensitivity measures are appropriate for problems where parameters cannot be obs erved directly (Saltelli et al. 2000) or if the parameters produce linear output responses (Muoz C arpena et al. 2007). Global sensitivity analysis involves varying many parameters and quantifying variation through probability distribution functions. Analytic approach es to calculating the variation include, but are not limited to, Taylor series approximations, First Order Approximations ([FOA], Morgan and Henrion 1992), and Monte Carlo si mulations (Muoz Carpena et al. 2007). Global sensitivity methods can be used for determining the portions of the model that do not provide an adequate contribution to the results (Saltelli et al. 2000). This aids in simplifying the model to parameters that affect the outcome of interest and verifying the representation of the physica l system. Additional benefits to global analysis include testing many parameters simultaneously, the ability to

PAGE 44

44 rank parameter importance, and determining interactio n effects (Muoz Carpena et al. 2007). Summary of Available Models There are thousands o f hydrological models in existence with a wide range of purposes such as vadose zone water movement, groundwater movement, and overland flow. Typically, hydrological models are used to predict agricultural water use, water availability, and water managemen t issues like flooding and stream flow levels. Due to these large scale issues, many models evaluate large scale areas such as catchments, basins, and watersheds with varying timesteps such as daily, weekly, monthly, or even annually. Also, many models h ave multiple objectives in addition to the results of the hydrological model such as sediment transport, nutrient transport, and crop growth. CHEMFLO 2000 This program, available through the EPA, uses Richards equation to model one dimensional water move ment in soil columns of finite length or semi infinite length with uniform initial conditions (Nofziger and Wu 2003). Partial differential equations are solved using finite differences with backward Euler approximation and a Picard iteration scheme. Soil water retention curves can be selected from Brooks and Corey (1964), Simmons et al. (1979), and van Genuchten (1980) using the Mualem (1974) relationship. Hydraulic conductivity curves can be calculated using Brooks and Corey (1964), Gardner (1958), and van Genuchten (1980), also using the Mualem (1974) relationship. Acceptable ranges for parameters for soil water retention curve s and hydraulic conductivity curves can be determined through extensive soil testing ( Table 1 6 ) The user must specify the i nitial matric potential and boundary conditions at the soil surface expressed as a constant matric potential, constant flux density, or falling head.

PAGE 45

45 Additional boundary conditions at some depth of the soil column for a finite soil system are expressed as constant matric potential, constant flux density, or free drainage. There are various l imitations specific to the model. No source or sink terms for root water uptake No provisions for shrink/swell, preferential flow, and hysteresis Sensitivity to initia l and boundary conditions Discretization errors increase as the mesh size increases HYDRUS 1 D This model can be used to analyze one dimensional water and solute movement in variably saturated porous media by numerically solving Richards equation for wate r flow and the convection dispersion equation for solute transport (USDA 2008). Richards equation is solved by finite element schemes using either Gaussian elimination for banded matrices, conjugate gradient method for symmetric matrices, or the ORTHOMIN method (Mendoza et al. 1991) for asymmetric matrices. Implementing automatic time step adjustments, error checking, and minimizing numerical oscillations improve solution efficiency of the finite element method. Flow regions can be delineated by irregula r boundaries and composed of non uniform soils that have a variety of boundary conditions. Hydraulic properties were described by van Genuchten (1980), Brooks and Corey (1964), and modified van Genuchten functions where modifications were made to improve hydraulic properties near saturation. SWAP The Soil Atmosphere Plant (SWAP) model simulates the transport of water, solutes, and heat in variably saturated zone over growing seasons. SWAP consists of clearly defined modules for soil water flow, soil heat flow, solute transport, crop growth, macropore flow and interaction with groundwater and surf ace water systems (Kroes et

PAGE 46

46 al. 2008). It has been derived from previous versions, such as SWATR(E), SWACROP, and SWAP93, and integrated into other models for nut rient management, such as PEARL, ANIMO, and STONE. The upper system boundary is defined by the soil surface with or without a crop and the atmospheric conditions. The bottom is bounded by the unsaturated zone and describes the interaction with regional gr oundwater. Richards equation is applied for water movement throughout the variably saturated zone, allowing the use of soil hydraulic functions from databases and simulation of all kinds of management options. Root water extraction at various depths is c alculated from potential transpiration, root length density, and possible reductions due to wet, dry, or saline conditions. Though not standard, SWAP can be implemented on a regional scale within GIS. WAVE This FORTRAN based model can be used to evaluate water movement using the SWATRER code, nitrogen movement using SOILN code, heat and solute modeling using LEACHN code, and crop growth using SUCROS. Soil water movement is determined using a soil water balance where one dimensional flow is determined from Richards equation solved using the finite difference method. The upper boundary condition is specified by flux density or pressure head. The lower boundary can be specified based on the presence of a water table, pressure head, flux density, or free dra inage. Hydraulic conductivity values are fitted using Gardner (1958) when measured values are not available. Additionally, if a soil water retention curve is not specified, van Genuchten (1980) with the Mualem (1974) relationship can be used. Hysteresis can be considered using Mualem second parametric model (Mualem 1974). Preferential flow can be considered by summing multiple soil water retention curves.

PAGE 47

47 ET Controllers ET controllers use evapotranspiration (ET) to estimate plant water needs. Though they vary in functionality based on model, most ET controllers use a soil water balance to determine how much irrigation is necessary and when to apply it. The controllers are designed to either replace the typical timer or act as an amendment to the time r. The primary differences in types of controllers are due to the methods for obtaining ET information that include: 1) historical pre programming, 2) on site calculation, and 3) provided by a signal service. More information about the types of ET contro llers can be found in Davis (2008). Evaluation Programs The Irrigation Association developed the Smart Water Application Technologies (SWAT) Testing Protocol for Weather Based Irrigation Controllers as a method to evaluate the ability of ET controllers to use ET for irr igation scheduling purposes (IA 2008). This testing protocol was established in 2002 and is now in its 8 th draft. The protocol specifies a 30 day test to determine if ET controllers can maintain theoretical soil moisture for six diverse virtual landscapes (Table 1 7 ; Table 1 8 ). The controllers were evaluated against the soil water characteristics of the virtual landscapes to determine performance measured through scheduling efficiency and irrigation adequacy. Scheduling efficiency is a measure of over irrigation and is quantified as scheduling losses combined from three sources: soil moisture rising abov e field capacity, too long cycle times, and too short soak times. It i s calculated from 30 day totals. (1 52)

PAGE 48

48 The E is scheduling efficiency (%), I n is net irrigation (mm), and SL is scheduling losses (mm). Similarly, irrigation ade quacy is a measure of under irrigation and can be quantified from the periods when the virtual landscape would not be considered as well watered conditions. It i s calculated from 30 day totals. (1 53) The A is irrigation adequacy (%), ET C is crop specific evapotranspiration (mm), and D is the amount of deficit (mm). Deficit is calculated as the cumulative depth below well watered conditions, or depth below the maximum allowable depletion level. Results of 100% for either score wo uld indicate that the irrigation scheduling performance by the ET controller was effectively perfect over the 30 day test. Manufacturers program their own ET controllers and choose when to publish the results, sometimes resulting in repeated testing until nearly perfect scores are produced (Table 1 9 ). In June 2006, the United States Environmental Protection Agency (USEPA) started the WaterSense program to add labels on product packaging of certified water saving devices to aid consumers in choosing better products. By Jan 2007, the first devices were labeled as a part of this program included high efficiency toilets that use less than 4.9 L per flush. Other products with WaterSense specifications include faucets, showerheads, and urinals. The first out door products eligible for the WaterSense label were ET controllers with the final specification released on Nov. 3 2011 and the first ET controllers were officially labeled as water saving devices in May 2012 (USEPA 2012).

PAGE 49

49 The WaterSense testing specific ation was developed by expanding and adapting testing protocol include: Minimum of four rainfall events that exceed 2.54 mm Irrigation occurs before rainfall in the soil water balance Runtimes less than or equal to three minutes are not considered No more than 2 consecutive days and 3 total days of missing ET O data in any 30 day testing period Backup rain gauge is only acceptable substitute for missing rainfall data Manufa cturers must become WaterSense partners, indicating a commitment to a high standard in water efficiency and performance, and work directly with an accredited certified licensing body (LCB) to have their ET controllers tested. The LCB inspects the overall production capabilities of each manufacturer to ensure products can be mass produced and available nationally. The LCB selects a single random sample from the production line after final packaging and cannot receive any additional materials to aid in inte racting with the ET controller since those materials will not be available to the end user. Manufacturers must have no interaction with the product during testing. Unlike the SWAT test, scores are counted from the first 30 day period that meets the weathe r requirements. The criteria for achieving the WaterSense label include: Irrigation adequacy score greater than or equal to 80% for every zone Scheduling efficiency score is greater than or equal to 90% for every zone Scheduling efficiency score is great than or equal to 95% averaged across zones If these score thresholds are achieved, the LCB notifies the USEPA of the brand and model to add to the registry of WaterSense labeled devices (Table 1 10 ); however, quantitative score results are not made available to the USEPA or the public.

PAGE 50

50 Previous Research Multiple research studies in the forms of bench tests, plot studies, and cooperator studies have been performed in the U.S. The bench tests were conducted to determine the functionality of the ET controllers such as how the program settings determine the irrigation schedule, requirements for installation, and accuracy compared to theoretical water needs. The evaluation programs described above were performed as bench tests. Plot studies were conducted to assess the functionality as well, but to also evaluate the effect of the irrigation schedule calculated by the ET controller on actual landscape s Cooperator studies were more complex and variable due to differences in landscapes and homeowner interactions, but were the most applicable for determining the performance of an ET controller in the real world. Bench t ests In addition to the evaluation programs already discussed, there have been multiple bench tests performed to determine the effectiveness of ET controllers. Bench tests are good for determining the nuances of the controller model such as responses to different types of programming and irrigation scheduling techniques. However, i t is impossible to determine the effects of the nuances because the test is conducted virtually where only the controller valve output signals are recorded which would result in irrigation if connected to an irrigation system. In 2002, the Metropolitan Wat er District of Southern California (MWD 2004) conducted a bench test to compare irrigation scheduling of ET controllers to theoretical water needs for three types of landscapes (Table 1 11 ). The ET controllers selected for soil moisture depletion analyses were AquaConserve, WeatherTRAK, and Weatherset. All three controllers exhibited good performance, normally falling in the acceptable

PAGE 51

51 range of 30% to 70% depletion of the available water holding capacity (AWHC). However, the Weatherset controller irrigate d for longer runtimes with fewer cycles per month that may have resulted in exceeding the available water holding capacity on irrigation days, causing runoff or deep percolation, and caused deficit conditions between irrigation days. A bench test was con ducted in 2003 using Aqua Conserve, WeatherSet, WeatherTRAK, and Calsens e controllers (Pittenger et al. 2004) to determine the programming requirements, ease of setup and operation, and accuracy of matching irrigation needs to five types of landscapes: tur fgrass, high water use mixed plantings, low water use mixed plantings, trees, and annuals. A real time reference landscape was used to calculate ET C for the turfgrass directly and was adapted using plant factors for the other virtual landscapes. Though a ll ET controllers adjusted irrigation schedules based on the ET C trends, the magnitudes of the adjustments were not consistent. Consequently, none of the ET controllers achieved high accuracy in predicting irrigation needs for all types of virtual landsca pes with the error usually resulting in too much irrigation. Programming required professional setup and monitoring with significant manual adjustments, thus making it difficult to properly program all of the controllers despite the level of technological complexity. In 2008, Texas A&M University bench tested six ET controllers over a four week 1 7) (Swanson and Fipps 2009). Comparisons were made to ET O directly and irrigation rec ommendations provided by TexasET Network Landscape Plant Water Requirem ent Calculator (TexasET Network 2012). Irrigation application by the ET controllers

PAGE 52

52 exceeded ET O at a frequency of 58%. Additionally, irrigation application exceeded the irrigation re commendations with 100% frequency for signal based controllers and 75% frequency for standalone controllers with on site sensors. All controllers failed to adequately adjust for rainfall, incorporating only 48.53% of the time for signal based controllers and 56.25% of the time when sensed by onsite sensors. Texas A&M University repeated testing for Year 2 on ten ET controller models tested for 13 weeks (Swanson and Fipps 2010). Over the season, only two ET controller models irrigated within the Texas r ecommendations with the remaining ET controllers applying 3 5 times the irrigation volumes. Over Year 2, the ET controllers exceeded ET O 20% of the time, which is significantly less than Year 1 results, but the frequency of exceeding recommended irrigatio n amounts was 97% for signal based controllers and 76% for standalone controllers which matched Year 1 results. The third year of bench testing by Texas A&M University consisted of eight controllers tested for 34 weeks that were programmed with landscape characteristics found in Texas and not with the SWAT virtual landscapes (Swanson and Fipps 2011). Additional changes included testing two signal based controllers with rain sensors and were no longer tested indoors. Though none of the controllers adequate ly met plant water requirements for the entire study period, all controllers still applied excessive amounts of irrigation during sporadic portions of the study period. However, ET controllers with onsite sensors typically performed better than signal bas ed controllers by applying irrigation similar to the recommendation more frequently. In 2009 and 2010, UF IFAS implemented a bench test to determine the reproducibility and transferability of the SWAT test if adopted for the EPA Wate rSense

PAGE 53

53 program (Davis and Dukes 2012). Though not reproducible in the current form, UF IFAS was able to recreate the testing software into a more understandable spreadsheet and provide training materials for future testing labs. The SWAT test was not considered transferrable due to the decreases in scheduling efficiency scores during periods of increased rainfall frequency. Some recommended changes to the SWAT test include longer testing periods, disclosing program settings with test results, and increasing the rainfall thre shold. Plot s tudies Three ET controller models (Weathermatic Smartline 1600, Toro Intelli Sense, and ETWater Smart Controller 100) were implemented on 60 m 2 St. Augustinegrass plots in southwest Florida (Davis et al. 2 009; Davis and Dukes 2012). Two tim e based treatments with rain sensors were also implemented to simulate UF IFAS recommendations for Florida. The TIME treatment replaced 100% of the net irrigation requirement and the RTIME applied the same schedule with 60% replacement, both treatments r estricted to two days per week. All treatments were compared to a simulated timed based treatment without a sensor ( TIME WOS ) that included irrigation events bypassed by the rain sensor for the TIME treatment, which is closer to the irrigation habits of a Florida residential irrigator. Overall, the ET controllers produced 43% annual water savings with savings fluctuating throughout the year based on overall irrigation demand. The ET controllers had the highest water savings compared to TIME WOS in winter, 50% 60%, during the period of lowest irrigation demand. The least savings occurred in the spring, 9% 15%, during the highest irrigation demand. Rainfall features for these three ET controllers were not sophisticated enough to account for localized rain fall typi cal to Florida (Davis and Dukes 2010). Similar to the

PAGE 54

54 results of the Flo rida SWAT test (Davis and Dukes 2012), periods of more frequent rainfall resulted in lower scheduling efficiency scores. Additionally, the values for various program setting s were not available from the controller manufacturers. Therefore, it was difficult for the controllers to perform well against a soil water balance because the settings could not be matched. A similar plot study was conducted in north Florida where th e Toro Intelli Sense and Rain Bird ET Manager ET controllers were used on 18.2 m 2 St. Augustinegrass plots using a two day per week irrig ation schedule (McCready et al. 2009; Davis and Dukes 2012). Water savings were also calculated compared to a TIME WOS treatment with savings ranging from 25% during the spring to 63% during the fall. The ET Manager had an incorrect rainfall setting that caused the controller to under irrigate in the spring, resulting in higher water savings, but at the expense of landsc ape quality. The researchers found that the default settings were not appropriate for all landscapes and were not representative of the study site. Programming, especially application rates, should be done by a qualified professional to maximize the bene fits of using an ET controller. Rutland and Dukes (2012) conducted a study using the Toro Intelli Sense on 60 m 2 St. Augustinegrass plots in southwest Florida to test the effectiveness of the rainfall features used by th e controller This controller acce pts a standard rain sensor as well as offers a rain pause feature that pauses irrigation for a number of days based on the depth of rainfall received. The algorithm used to calculate number of days for rain pause is proprietary and not clearly described i n manufacturer documentation. Maximum water savings was achieved by combining a rain sensor and the rain pause

PAGE 55

55 feature, totaling 41%. Each rain feature that was independently tested achieved only 25% water savings. The rain sensor typically bypassed for localized rainfall events not sensed at the weather station(s) used by the ET controller. Conversely, the rain pause feature bypassed for longer time periods based on the amount of rainfall received whereas a rain sensor dries out within 24 48 hour s (Car denas Lailhacar and Dukes 2008). Based on the current technology level, including rain sensors on ET controllers is preferable in Florida. Grabow et al. (2012) reported on an evaluation of the Toro Intelli Sense ET controller on 16 m 2 St. Augustinegrass plots planted on clay soils in North Carolina The ET controller over irrigated in 2007 1 1% compared to a time based treatment, causing the soil water content to reach levels frequently above field capacity. They estimated that over irrigation occurred due to high ET O values broadcasted to the controller that averaged 30% higher than on site calculations. Though not as drastic as the 2007 results, the ET controller over irrigated compared to the timer treatment in 2008 as well. Cooperator s tudies A thre e year study in three Colorado cities from 2000 to 2002 was conducted to determine the reliability and effectiveness of a WeatherTRAK enabled ET controller (Aquacraft, Inc. 2002; Aquacraft, Inc. 2003). Participation included nine residences and one commer cial property in 2001 with seven of the ten sites remaining in 2002. Due to voluntary participation in the program, five of the seven sites were historical under irrigators that maintained their historical average during the study, thus resulting in no ac tual water savings. However, the ET controllers captured 88% (2001) and 92%

PAGE 56

56 (2002) of potential water savings compared to evapotranspirative demand and was reliable in obtaining ET O responding to drought restrictions, and equipment durability. In 2002, the Los Angeles Department of Water and Power assessed the performance of ET controllers by professionally installing 18 WeatherTRAK and 7 Water2Save controllers, totaling 25 sites (Bamezai 2004). Historical irrigation was determined from two years of wa ter use prior to installation and was normalized for differences in weather during the study period. Water savings were calculated from data collected up to one year after installation. Combined, the ET controllers had 78% potential water savings compare d to the theoretical irrigation demand and the water savings compared to historical irrigation application averaged 17.4% for the WeatherTRAK controllers and 28.3% for the Water2Save controllers. Although most cooperating homeowners reacted positively to the technology, education and outreach programs for landscapers and homeowners were strongly suggested. A residential runoff reduction (R3) study was conducted in Irvine, CA to evaluate changes in water usage, dry weather runoff, water quality, and custo mer attitudes and awareness from implementing WeatherTRAK enabled ET controllers or an education only program (MWDOC and IRWD 2004). Treatments were separated into neighborhoods resulting in 112 residential and 15 commercial ET controller installations (o ne neighborhood), 225 homes and one school participating in the education group (one neighborhood), and the remaining three neighborhoods were selected as blind comparisons with no known changes to their irrigation practices. Each neighborhood had a singl e point storm drain separate from other communities that was monitored for 12 to 18 months beginning in 2001. The educational materials consisted of postcards

PAGE 57

57 sent with change in weather conditions suggesting days of week and minutes per day of irrigation. Results showed that the ET controller neighborhood decreased water consumption by 156 L d 1 and dry weather runoff by 50% with significantly lower total nitrogen levels in runoff compared to all other treatments. The education households decreased water consumption by only 97 L d 1 and increased dry weather runoff during the study period. A Las Vegas Valley study occurring over a 12 to 18 month period, ending in 2005, consisted of installing WeatherTRAK enabled ET controllers at 17 residential sites and Rain Bird ESP timers at 10 control sites where 5 control sites received seasonal irrigation schedu ling information (Devitt et al. 2008). On average, there was a 20% reduction from historical irrigation application by implementing an ET controller, which w as significantly more water savings than both control groups that had no water savings, 9% increase by the control with educational programming and 2% increase by the control without educational programming. Mayer et al. (2009) implemented a study in Cal ifornia that involved installing 3,112 ET controllers of varying models at residential and commercial properties through rebate and voucher programs, exchange programs, and direct installation programs. Overall, ET controllers reduced outdoor water use af ter one year by 6.1% compared to one year of pre smart controller outdoor water use that was normalized for differences in weather. Though water savings were statistically significant, the resulting savings was lower than expected due to historical under irrigators experiencing increases in irrigation application by implementing a smart controller. This indicates that smart controllers may not be applicable for broad implementation to the entire population.

PAGE 58

58 Using a subset of 384 sites that had three year s of data, a multi year analysis showed that water savings increased over time from 6% after the first year to + 16.4% after the third year. Davis and Dukes (2012) performed an ET controller cooperator study in three cities located in Hills borough County The Toro Intelli Sense was selected as the ET controller based on the positive results of the SWF plot study. There was a total of 21 ET controllers professionally installed and programmed by research personnel and 15 comparison homes that were represe ntative of the area. Using only one year of data collected in 2009 2010, ET controller performance was compared to irrigation application from 15 representative comparison homes, historical irrigation application calculated from five years of billing reco rds, and theoretical irrigation needs. The ET controllers applied significantly more monthly irrigation than the comparison groups in two locations, averaging 32 mm and 42 mm, while the comparison groups significantly decreased their irrigation applicatio n compared to their historical practices. At the third location, there were not significant differences from the comparison group, averaging 38 mm, or from the historical average. Cumulatively, the ET controllers ranged from 25.8% water savings to 1.1% i ncrease from the historical totals over the same months. Also, the ET controllers reduced irrigation application in all locations compared to the theoretical r equirement by 23.4% to 49.9%.

PAGE 59

59 Table 1 1. Summary of restrictions on outdoor water use from the potable supply for each water management district Water Management District Restriction Days during Daylight Savings (d wk 1 ) Restriction Days during Eastern Standard Time (d wk 1 ) Restriction Times Source Northwest Florida NA NA NA NA Suwannee River 2 1 10 am to 4 pm SRWMD 2012 St. Johns River 2 1 10 am to 4 pm SJRWMD 2012 Southwest Florida 2 1 8 am to 6 pm SWFWMD 2013 South Florida 2 2 10 am to 4 pm SFWMD 2012 Note: NA indicates that restrictions were not applicable to that district. Table 1 2. The parameters of K and M d S av were estimated from Soil Conservation Service (1964) cylinder infiltrometer data by Chu (1978) to be used as guidelines for parameter estimates. Soil Type K a (m/h) M d S av b (m) Clay <0.0021 >0.061 Clay loam 0.0021 0.0067 0.043 0.0061 Silt loam 0.0067 0.0142 0.036 0.043 Sandy loam 0.0142 0.0247 0.027 0.036 Sand >0.0247 <0.027 a Estimation of the hydraulic conductivity for specific soil types b Initial soil moisture deficit and capillary suction at the wetting front, commonly expressed as one term

PAGE 60

60 Table 1 3. Soil parameters originally put together by Rawls et al. (1982 ; 1983) and summarized by Ogden and Saghafian (1997). Soil Type Total Porosity Effective Porosity Residual water content Wiltin g point water content b (cm) K s (cm/h) H c (cm) Sand 0.437 0.417 0.020 0.033 7.26 0.694 23.56 9.62 Loamy Sand 0.437 0.401 0.035 0.055 8.69 0.553 5.98 11.96 Sandy Loam 0.453 0.412 0.041 0.095 14.66 0.378 2.18 21.53 Loam 0.463 0.434 0.027 0.117 11.15 0.252 1.32 17.50 Silt Loam 0.501 0.486 0.015 0.133 20.79 0.234 0.68 32.96 Sandy clay loam 0.398 0.330 0.068 0.148 28.08 0.319 0.30 42.43 Clay Loam 0.464 0.390 0.075 0.197 25.89 0.242 0.20 40.89 Silty clay loam 0.471 0.432 0.040 0.208 32.56 0.177 0.20 53.83 Sandy clay 0.430 0.321 0.109 0.239 29.17 0.223 0.12 46.65 Silty clay 0.479 0.423 0.056 0.250 34.19 0.150 0.10 57.77 Clay 0.475 0.385 0.090 0.272 37.30 0.165 0.06 62.25

PAGE 61

61 Table 1 4. Warm season turfgrass crop coefficients calculated from various studies. Month North Florida a South Florida a California b Las Vegas, NV c Beijing, China d January 0.35 0.71 0.55 0.44 NA February 0.35 0.79 0.54 0.43 NA March 0.55 0.78 0.76 0.67 NA April 0.80 0.86 0.72 0.76 0.66 May 0.90 0.99 0.79 0.75 0.79 June 0.75 0.86 0.68 0.89 1.06 July 0.70 0.86 0.71 0.89 1.15 August 0.70 0.90 0.71 0.82 1.09 September 0.75 0.87 0.62 0.82 0.97 October 0.65 0.86 0.54 0.77 0.78 November 0.60 0.84 0.58 0.81 0.84 December 0.45 0.71 0.55 0.51 NA a Values were found in Jia et al. (2009) b Values were found in Gibeault et al. (1989) c Values were found in Devitt et al. (1992) d Values were found in Zhang et al. (2007) Table 1 5. Parameters used to calculate the redistribution coefficient determined by Gowdish and Muoz Carpena (2009). Iteration b i c i d i 1 4.2952 154.6101 1.0 2 0.0020 0.0010 0.5 3 14.0032 61.5429 1.0

PAGE 62

62 Table 1 6. Ranges of acceptable hydrological parameters for each metho d for calculating soil water re tention curves and hydraulic conductivity curves avai lable in the CHEMFLO 2000 model Parameter Symbol Brooks and Corey (1964) van Genuchten (1980) Simmons et al. (1979) Gardner (1958) Min Max Min Max Min Max Min Max Residual water content r 0 0.15 0 0.15 0 0.15 NA NA Saturated water content s 0.25 0.60 0.25 0.60 0.25 0.60 NA NA Air entry value h b 60.0 1 NA NA NA NA NA NA Constant 0.05 1.7 NA NA NA NA NA NA Constant NA NA 0.002 0.15 1.15 1.25 NA NA Constant n 2 6 1.25 3 0.025 0.1 1.8 3.5 Saturated hydraulic conductivity K s 0 NA 0 NA NA NA 0 NA Matric potential constant h c NA NA NA NA NA NA 50 5

PAGE 63

63 Table 1 7. Description of virtual landscape zones used i n the SWAT testing protocol (IA 2008) Description Units Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Soil Texture Loam Silty Clay Loamy Sand Sandy Loam Clay Loam Clay Slope % 6 10 8 12 2 20 Exposure 75% Shade Full Sun Full Sun 50% Shade Full Sun Full Sun Vegetation Tall Fescue Bermuda Ground Cover Woody Shrubs Trees & Ground Cover Bermuda Landscape Coefficient a,b Varies Monthly (see Table 1 8) Varies Monthly (see Table 1 8) 0.55 0.40 0.61 Varies Monthly (see Table 1 8) Irrigation System Pop Up Spray Heads Pop Up Spray Heads Pop Up Spray Heads Pop Up Spray Heads Surface Drip Rotors Precipitation Rate c in/h 1.60 1.60 1.40 1.40 0.20 0.35 Application Efficiency % 55 60 70 75 80 65 Gross Area ft 2 1,000 1,200 800 500 650 1,600 Allowable Depletion % 50 40 50 55 50 35 Available Water a,d in./in. 0.17 0.17 0.09 0.13 0.18 0.17 Root Zone Depth a,d in. 10.0 8.1 20.0 28.0 25.0 9.2 RZWWS e in. 0.85 0.55 0.90 2.00 2.25 0.55 Soil Intake Rate d in./h 0.35 0.15 0.50 0.40 0.20 0.10 Allowable Surface Accumulation d,f in. 0.25 0.16 0.26 0.24 0.26 0.10 a Values selected based on vegetation description. b Values selected based on exposure description. c Values selected based on irrigation system description. d Values selected based on soil texture description. e Root zone working water storage (RZWWS) is calculated using the available water, root zone depth, and allowable depletion val ues. f Values selected based on slope description.

PAGE 64

64 Table 1 8. Crop coefficients for zones 1, 2, and 6 for the SWAT test (IA 2008). Month Zone 1 Zone 2 Zone 6 Jan. 0.41 0.52 0.52 Feb. 0.46 0.64 0.64 Mar. 0.52 0.7 0 0.7 0 Apr. 0.56 0.73 0.73 May 0.6 0 0.73 0.73 Jun. 0.62 0.71 0.71 Jul. 0.62 0.69 0.69 Aug. 0.6 0 0.67 0.67 Sep. 0.56 0.64 0.64 Oct. 0.5 0 0.6 0 0. 0 6 Nov. 0.45 0.57 0.57 Dec. 0.4 0 0.53 0.53

PAGE 65

65 Table 1 9 Summary results from the SWAT test. Brand Name Model Name Published Date Scheduling Efficiency (%) Irrigation Adequacy (%) Min Max Mean Min Max Mean ETwater Smart Controller 9/9/05 100 100 100 93.7 100 98.5 Irritrol Smart Dial 1/9/06 100 100 100 100 100 100 Toro Intelli Sense 1/9/06 100 100 100 100 100 100 Hydropoint WeatherTRAK 1/9/06 100 100 100 100 100 100 Rain Bird ET Manager 6/27/06 100 100 100 100 100 100 Calsense ET2000e 4/11/07 100 100 100 100 100 100 Hunter ET System 4/11/07 100 100 100 97.7 100 99.5 Weathermatic SL1600 4/11/07 100 100 100 97.7 100 99.6 Alex Tronix Enercon Plus 5/17/07 100 100 100 96.4 100 99.0 Alex Tronix Smart Clock 5/17/07 100 100 100 98.9 100 99.8 Aqua Conserve Aqua ET 9 9/10/07 100 100 100 98.7 100 99.8 Hydrosaver ETIC 10/8/07 94.8 100 98.9 93.1 100 98.5 Cyber Rain XCI 4/11/08 100 100 100 79.5 100 88.4 Rain Master RME Eagle 4/11/08 100 100 100 100 100 100 SMG Superior Controls Sterling 8 7/15/08 79.5 100 92.9 100 100 100 Rain Bird ESP LX with ET Manager 5/11/09 100 100 100 95.5 100 98.5 WaterOptimizer WaterOptimizer 5/11/09 100 100 100 100 100 100 Tucor RKS with Tipping Rain Bucket 5/13/09 68.0 100 92.0 100 100 100 Rain Bird ESP SMT 6/1/09 100 100 100 93.2 100 98.5 NDS Raindrip WeatherSmartPro 1/5/10 96.2 100 98.4 100 100 100 Alex Tronix Universal Smart Module 4/15/10 99.1 100 99.9 94.4 100 97.8 ETwater Hermit Crab 3/22/11 100 100 100 100 100 100 Hunter Solar Sync Module 3/22/11 100 100 100 87.8 100 95.7 Hunter Solar Sync Sensor 3/22/11 100 100 100 95.4 100 98.8 Rain Bird SST Smart Controller 4/5/11 99.0 100 99.8 97.8 100 99.3 Toro EC XTRA w Smart Pod 4/5/11 100 100 100 87.0 100 94.7

PAGE 66

66 Table 1 9 Continued Brand Name Model Name Published Date Scheduling Efficiency (%) Irrigation Adequacy (%) Min Max Mean Min Max Mean Irritrol KwikDial Climate Logic 4/20/11 100 100 100 94.4 100 97.8 Irritrol MC48E Climate Logic 4/20/11 100 100 100 95.8 100 98.6 Irritrol RainDial Climate Logic 4/20/11 100 100 100 98.2 100 99.4 Irritrol Total Control Climate Logic 4/20/11 100 100 100 92.1 100 95.5 Toro TMC 212 Climate Logic 4/20/11 100 100 100 87.7 100 94.2 Toro TMC 424E ID Climate Logic Kit 4/20/11 100 100 100 94.7 100 98.4 DIG Leit 2 ET 6/21/11 99.0 100 99.7 100 100 100 Rain Master RME Eagle Plus 6/21/11 89.5 100 95.4 96.7 100 98.9 Tucor RKD WS 5/9/12 96.5 100 98.9 100 100 100

PAGE 67

67 Table 1 10 List of EPA WaterSense labeled ET controllers. Brand Name Model Name Maximum Zone Capacity Weather Data Source Cyber Rain Pro System 16 Signal Cyber Rain Standard System 16 Signal Cyber Rain Long Range 24 Signal Irritrol Rainmaster RME Eagle 36 Signal Rain Bird ESP SMT 13 Sensor Rain Bird Simple To Set 12 Sensor Raindrip Weather Smart Pro 6 Sensor Toro XTRA Smart EC XTRA 10 Sensor Table 1 1 1 Landscape descriptions adapted from MWD (2004). Characteristic Scenario A Scenario B Scenario C Plant Type Cool Season Turfgrass Annuals Acacia and Myoporum Crop Coefficient 0.8 1.0 0.3 Root Depth (in) 4 4 12 Soil Type Loam Sand Clay Slope Flat Flat >20 Exposure Full Sun Shady Most of Day Full Sun Sprinkler Type Pop up Sprays Pop up Sprays Rotors Application Rate (in/hr) 1.58 1.94 0.56 Irrigation Efficiency 0.625 0.55 0.70

PAGE 68

68 CHAPTER 2 PERFORMANCE OF WEATHER BASED IRRIGATION CONTROLLERS IN SOUTHWEST FLORIDA LANDSCAPES Introduction In 2005, Florida ranked fourth in domestic water use trailing only California, Texas, and Michigan (Kenny et al. 2009). Daily per capita indoor water use has been found to be constant across the United States with a mean of 0.262 m 3 d 1 (Mayer et al. 1999). Conservation programs targeted to outdoor water use should be able to achieve reductions in do mestic water use in Florida, where annual rainfall averages 1,340 mm (Romero and Dukes 2013). A study in central Florida found that 64% of domestic water use was for irrigation (Haley et al. 2007). The potential for reduction in the public water supply d emand by better irrigation scheduling techniques is substantial. The Irrigation Association (IA 2008) defined smart controllers as technologies an irrigation system One type of smart controller is the weather based or evapotranspiration (ET) controller that determines irrigation schedules based on estimates of plant water needs determined from weather measurements. This information can be obtained in one of three ways: through satellite signal, measured on site, or programmed with historical data (Dukes 2012). The Municipal Water District of Orange County and Irvine Ranch Water District (IRWD) con ducted a study in Irvine, CA to evaluate WeatherTRAK enabled ET controllers compared to an education only program (MWDOC and IRWD 2004). With permission from ASCE

PAGE 69

69 rain pauses to the ET controller on a daily basis. Treatments were separated into neighborhoods resulting in 112 residential and 15 commercial ET controller installations (one neighborhood), 225 homes and one school participating in the education group (one neighborhood), and the remaining th ree neighborhoods were selected as blind comparisons with no known changes to their irrigation practices. The educational materials consisted of postcards sent that suggested days of week and minutes per day of irrigation. Results showed that the ET cont roller decreased neighborhood wide water consumption by 0.156 m 3 d 1 whereas water consumption in the education households decreased by only 0.097 m 3 d 1 during the twelve to eighteen month study period. A Las Vegas Valley study occurring over a 12 to 18 month period, ending in 2005, consisted of installing WeatherTRAK enabled ET controllers at 17 residential sites and Rain Bird ESP timers at 10 control sites where half of the controls received seasonal ir rigation scheduling information (Devitt et al. 2008). On average, the ET controller treatment reduced irrigation by 20% compared to historical irrigation application. This was the only treatment to have water savings; there was a 9% increase by the contr ol with educational programming and 2% increase by the control without educational programming. Mayer et al. (2009) analyzed data from 3,112 ET controllers of varying models at residential and commercial properties throughout California. Overall, ET con trollers reduced irrigation use after one year by 6.1% compared to one year of pre smart controller outdoor water use that was normalized for differences in weather. Though water savings were statistically significant, the resulting savings was lower than

PAGE 70

70 expected due to historical under irrigators that had increased irrigation with the ET controllers. This result indicates that smart controllers may not be applicable for broad implementation to the entire population with the expectation of irrigation red uction. Using a subset of 384 sites that had three years of data, a multi year analysis showed that irrigation use decreased over time from +6% after the first year to 16.4% after the third year. In 2006 2007, a field plot study of ET controllers in southwest Florida resulted in irrigation savings of 43% compared to a time based schedule of 100% replacement of the net irrigation requirement without a rain sensor (Davis et al. 2009; Davis and Dukes 2012). Higher water savings were seen during the fal l and winter seasons with a maximum of 60% savings. However, this plot study was conducted in a controlled setting and may not translate to actual savings for homeowners. The objective of this study was to determine the effectiveness of an ET controller in real world conditions, based on adequate performance during the field plot study. Materials and Methods Potential participants located in the southwest Florida communities of Apollo Beach, Riverview, and Valrico and in the top 50th percentile of water users based on a county wide analysis completed by Romero and Dukes (2010) were solicited by a letter from Hillsborough County Water Resource Services. Potential participants were directed from the mailed letter to an online questionnaire as the primary method for volunteering for the study. Additionally, potential participants could respond to the questionnaire by phone if averse to providing personal information on the internet. Participation in the study was completely voluntary.

PAGE 71

71 Thirty six cooperato rs were selected across the three communities after undergoing irrigation evaluations conducted by University of Florida, Institute of Food and Agricultural Sciences (UF IFAS). Evaluations were conducted to assess study eligibility based on the following: presence of a functional automatic irrigation system, good landscape quality, and commitment to long term participation to the study. As part of the on site evaluation, irrigation distribution uniformity was measured in one representative zone of the la ndscape by the catch can test (IA 2013) using 16 24 collectors when wind conditions were low and soil moisture measurements from a FieldScout TDR 300 (Spectrum Technologies, Inc.) when soil moisture was below field capacity. Cooperators were separated into two treatment groups: 21 cooperators were outfitted with Toro Intelli 612 (Riverside, CA) ET controllers (ET treatment) Inc., Petaluma, CA), and the remaining 15 formed a comparison group with no added technology. In addition to ET controllers, rain sensors were installed on the ET treatment homes. The Intelli the ET controller based on r esults from a field plot study also conducted in southwest Florida (Davis et al. 2009; Davis and Dukes 2012). (ET O ) and rainfall information to the Intelli Sense that required an ann ual subscription. According to the manufacturer, daily ET O and rainfall information were interpolated by modeling software using data collected at multiple off site weather stations triangulated in proximity to the controller location (The Toro Company 201

PAGE 72

72 signal service used the American Society of Civil Engineers ET O equation (ASCE 2005) for calculating ET O and required multiple weather parameters including temperature, relative humidity, solar radiation, and wind speed. We program med the ET controllers using the landscape and irrigation system information obtained during the irrigation evaluation. Most settings were chosen based on descriptions, thus using default values, except for application rates that were customized by zone a nd irrigation system efficiency that were based on measured lower half distribution uniformity (DU lh ) values (IA 2005) based on the catch can test results. Cooperators that received the ET controller also received exemptions from day of the week watering restrictions. The ET controllers were programmed to allow irrigation with a 12 am to 6 am water window unless otherwise specified by the cooperator. Automatic meter reading (AMRs) devices (FIREFLY, Datamatic, LTD., Plano, TX) were installed on household water meters in January 2009 and collected sub hourly water meter data through January 2011. The AMRs were equipped with an optical sensor that attached to the face of the water meter to record every sweep of the dial, signifying 0.038 m 3 of water consum ption. Indoor water use was separated from irrigation by removing any water accumulated for a single sub hourly time period that was less than the minimum volume capable by the irrigation system which was determined during the initial irrigation system eva luation. The average per capita indoor water use for the cooperators ranged from 0.054 m 3 d 1 to 0.530 m 3 d 1 averaging 0.260 m 3 d 1 which matches the average that was found in a U.S. study covering twelve cities (Mayer et al. 1999).

PAGE 73

73 Irrigation was conv erted from volume to depth using an estimated irrigated area. In the initial analysis presented in Davis and Dukes (2012), property appraiser landscape square footage totals were used as estimated irrigated areas for all cooperating homes. However, estab lished ornamental plants do not require irrigation in southwest Florida to maintain acceptable quality (Schieber et al. 2008; Gilman et al. 2009) allowing for decreases from total landscape area to turfgrass area only. The percentage of turfgrass was esti mated visually during the irrigation evaluation. Additionally, the cooperators in Riverview were not traditionally landscaped with parts of the properties left as natural, non irrigated areas. As a result, all property borders were redefined using Geograp hical Information Systems (GIS) to determine more accurate irrigated areas. The adjusted turfgrass area estimates in Riverview resulted in higher irrigation depths. New area estimates using GIS in the other two locations resulted in relatively no change i n estimations. The presented results were calculated from areas found using GIS for the Riverview location and property appraiser areas for the other two locations. Historical irrigation use (2001 to 2009) was determined for each cooperator from total me tered data provided by the regional water wholesaler, Tampa Bay Water. Irrigation water use was estimated for this period by subtracting the estimated average monthly indoor estimate from the sub hourly AMR data during the study period from the total meter ed values. This method assumes that indoor water use generally remains constant throughout the year (Dziegielewski and Keifer 2010). Additional comparison homes were identified and billing records from 2001 to 2011 were obtained for 46 neighbors with co mparable landscapes to the ET controller

PAGE 74

74 treatment. There were 3 5 neighbors identified for every cooperator that received an ET controller. Neighbors were chosen if residing on the same or immediately adjacent street, similar size of irrigable area, and had an acceptable landscape quality. These billing records were included as additional comparisons in the C+E treatment. Sub daily water meter data was not available for these homes; instead, indoor water use was assumed as the calculated average of 0.2 60 m 3 d 1 and 2.59 persons per household obtained from the 2007 2011 American Community Survey for Hillsborough County, FL (United States Census Bureau 2013). Using these assumptions, indoor water use totaled 20.2 m 3 mo 1 which was similar to the average monthly volume of 20.4 m 3 mo 1 calculated from the AMR data. The method of using billing records, assumed indoor water use, and property appraiser data for estimating irrigation was used by Friedman et al. (2013). The GIR was used as a comparison for irr igation application by the treatments over the study period. The GIR is an estimate of theoretical irrigation needs calculated by multiplying the net irrigation water requirement (IWR net ) by a scheduling multiplier (SM). The IWR net is defined as the amou nt of irrigation required to increase soil water storage to the maximum water level that can be stored before gravitational drainage (IA 2005). The IWR net was determined from mass conservation of soil water content (IA 2005). (2 1) The PWR is the plant water requirement (mm) and R e is effective rainfall (mm). The IWR net was accumulated daily, but was applied only on days when the soil water level fell below management allowable depletion (MAD), taken as 50% of the available water

PAGE 75

75 holding capacity determined for each soil type from the soil survey map for Hillsborough County (National Resource Conservation Service 1989). Values for the available water holding capacity ranged from 23 mm to 28 mm per 305 mm of soil, which was t he assumed root zone for turfgrass. Once IWR net was applied, the soil water level increased to the maximum storage capacity and IWR net was reset to zero. Deep percolation and surface runoff, also typically part of the soil water balance, were considered negligible since deep percolation and runoff can be avoided with proper design and management of the irrigation system. The PWR is the amount of water necessary to maintain healthy plant material (IA 2005). The PWR is calculated as the plant specific evap otranspiration (ET C ) for any given plant material by applying a crop coefficient (K C ) using Equation 1 28 (Allen et al. 1998). T he K C values for turfgrass ranged monthly from 0.45 (December through February) to 0.90 (May) based on the location of the fiel d study (Jia et al. 2009). It was assumed that established ornamental plants did not require irrigation in southwest Florida (Schieber et al. 2008; Gilman et al. 2009) thus using turfgrass K C values was appropriate for this analysis. The ET O is reference ET calculated using Equation 1 13 (ASCE 2005). Effective rainfall was limited to the portion of total daily rainfall that caused the soil water level to reach the maximum soil storage capacity after PWR was taken into account. Rainfall that exceeded the s oil storage capacity was considered lost due to surface runoff or deep percolation. Weather data were collected from three weather stations installed in the communities associated with the study. The weather stations were equipped to provide temperature relative humidity, solar radiation, and wind speed at 15 minute intervals to calculate ET O Weather station distances from the cooperating homes varied but were

PAGE 76

76 not greater than 4 km. Weather stations were also equipped with rain gauges to determine rai nfall totals. Two additional rain gauges were added to ensure rainfall measurements occurred within 500 m of all cooperating homes. Historical weather data was collected from two Florida Automated Weather Network (FAWN) stations located in the communitie s of Balm and Dover within Hillsborough County, FL. When data was available for both stations, ET O and rainfall values were averaged. The SM was determined using an approximation of the lower half distribution uniformity (DU lh ) from the lower quarter distr ibution uniformity (DU lq ) determined from catch can tests using Equation 1 30 (IA 2013) The measured DU lq for all 36 cooperating homes ranged from 0.45 to 0.79. There was no discernible pattern to distribution uniformity values of homes assigned to spec ific treatments indicating that treatment results were not biased based on this factor. The average DU lh of 0.80 obtained from on site irrigation evaluations was assumed as the average efficiency for all 46 neighbors that did not receive evaluations. Turf grass quality ratings were taken a minimum of every three months using National Turfgrass Evaluation Program (NTEP) procedures (Shearman and Morris 1998). The rating scale ranged from 1 to 9, where 1 represents dead turfgrass or bare ground, and 9 represen ts the highest possible quality and 5 was the minimal level for acceptable landscape quality. A baseline turfgrass quality rating was taken prior to the implementation of treatments to determine the change in quality during the study period. Turfgrass qu ality was judged primarily based on color and density. Though turfgrass quality can be affected by factors other than irrigation, the practice of assessing turfgrass quality as a way to validate that the decline in quality was not due to irrigation

PAGE 77

77 has be en used in multiple peer reviewed studies (Cardenas Lailhacar and Dukes 2012; Davis et al. 2009; Davis and Dukes 2012; Haley et al. 2007; Haley and Dukes 2012). Of the 21 cooperators that received ET controllers as a part of the study, there were 15 respon dents to a post study survey concerning their opinion of the technology and how it affected their irrigation system and landscape. The survey was administered online using the same methods employed for the sign up questionnaire. There were approximately twenty questions that addressed various topics such as their knowledge of the functionality of the controller, overall satisfaction with performance, and likelihood of continuing to use the controller. Statistical analyses were performed using Statistical Analytical Systems (SAS) software (Cary, NC). Irrigation applied and turfgrass quality were analyzed using the glimmix procedure in which irrigation application required a log transformation, and comparisons were made using the least mean square differen ces by treatment and location. Significance was determined at a 95% confidence level. Results and Discussion The annual ET O averaging 1,483 mm for the two year study period, was not significantly different from the previous nine years of annual ET O avera ging 1,423 mm (Table 2 1). Rainfall was significantly lower during the study period with an annual average of 1,210 mm, compared to the previous nine years average of 1,306 mm. Rainfall during the study period was the second lowest observed year from the years of data analyzed. A severe drought was declared by the U.S. Geological Survey from 1998 to 2002 (Verdi et al. 2006), which includes the period of lowest annual rainfall observed. Given that annual rainfall was significantly less than the average a nd the only other period with less rainfall was under drought conditions, the study period was

PAGE 78

78 considered a dry period. However, the frequency of rainfall events, averaging 112 events per year, was not different from the historical average of 117 events pe r year. Total irrigation applied varied across the treatment areas with significance in the statistical model based on location (P value < 0.0001) and location by treatment (P value = 0.0001) effects. Irrigation application by the ET treatment, averagi ng 38 mm mo 1 was significantly less than the comparison treatment that averaged 51 mm mo 1 in Apollo Beach (Table 2 2). In contrast, the ET treatment in Riverview applied significantly more than the comparison treatment, 36 mm mo 1 compared to 24 mm mo 1 respectively. Results were not significantly different in Valrico where the ET treatment applied 41 mm mo 1 and the comparison treatment applied 36 mm mo 1 during the study period. The average monthly gross irrigation requirement (GIR) varied across t he study locations from 51 mm to 65 mm depending on soil type and irrigation DU lh measured at each cooperating home (Table 2 2). Only the Apollo Beach comparison homes had no significant difference between irrigation application (51 mm mo 1 ) and the GIR ( 53 mm mo 1 ). As a result, comparison homes in this location applied 3% more than the GIR over the study period (Table 2 3), which was a negligible increase considering the monthly averages. All other treatments had significantly less irrigation applicati on than their respective GIR, with differences of 23% to 57%, indicating deficit irrigation conditions during the study period. The historical average irrigation calculated from 2001 to 2008 ranged from 46 mm mo 1 in Riverview to 70 mm mo 1 in Apollo Beach (Table 2 2). All treatments applied significantly less irrigation than their historical average with cumulative

PAGE 79

79 reductions ranging from 23% to 48% (Table 2 4). Reductions in average irrigation application by comparison treatments in all three communities indicated that there was a fundamental change in irrigation habits, but the reason is unknown. Many possibilities exist for this fundamental change that may include any combination of: the decline in economic conditions causing more eff icient water use practices, well publicized drought conditions in the area, or initial homeowner interaction with researchers that included suggested maintenance fixes that turned cooperator attention to the irrigation system. Irrigation was influenced by drought conditions due to temporary extreme irrigation res trictions (4 hours only, 1 d w k 1 ) that were in place from Apr 3, 2009 through Jun 30, 2009. It is possible that the perception of the drought and increased watering restrictions during spring 20 09 could have promoted deficit irrigation practices at the beginning of the study and a shift in priorities from high landscape quality to reducing water use. The ET controller homes were exempt from day of the week restrictions but were asked to change th eir irrigation start times to correspond with the water window of the comparison homes. Despite the possibility for effects due to researcher interaction, typical interaction with the cooperators in the comparison treatment occurred only once during the i nitial evaluation in 2008 and this extreme of a response (significant decline from historical average) was not observed in other homeowner cooperative studies (Haley et al. 2007; Haley and Dukes 2012). The comparison treatments historically over irrigate d in Valrico and Apollo Beach when compared to the GIR. The historical irrigation of the ET treatment homes in these two locations were not different from the GIR indicating that they were already irrigating similarly to plant water needs. Despite accept able irrigation practices, the addition of

PAGE 80

80 the ET controller further decreased their monthly average irrigation application. Additionally, water savings of 24% occurred in Apollo Beach (Table 2 3). Irrigation on ET homes in Valrico was not significantly different than their respective comparison treatments. In Riverview, the cooperators in both treatments were already historical under irrigators compared to the GIR. The most significant reduction in historical irrigation from GIR occurred by the compari son treatment in Riverview with an additional significant decrease in irrigation over the study period from the historical average. Due to this significant reduction, ET controllers resulted in a 54% increase in irrigation relative to the comparison treat ment. Despite this increase, significant reductions in average irrigation from historical averages occurred by using an ET controller (Table 2 2). It was determined that the cooperators in Riverview and the ET controller treatments in Apollo Beach and Val rico historically maintained or applied less than the GIR, thus were not considered over irrigators prior to the start of the study. It was not the intention to have under irrigating cooperators in the study due to the limitations of potential water savin gs as a result of imposed treatments. These three locations were initially selected due to identifying a large number of potential participants in the top 50th percentile of water users compared to other communities in the county (Romero and Dukes 2010). H owever, Romero and Dukes (2010) showed that utility customers who meet these criteria may still under irrigate compared to GIR determined for well watered conditions. In Valrico, the turfgrass quality of the comparison homes averaged 5.1 whereas the ET con troller group was able to maintain a significantly better turfgrass quality

PAGE 81

81 rating, averaging 7.4 (Table 2 2). Despite significant differences in turfgrass quality, average monthly irrigation application by each treatment was not significantly different. This may indicate that the ET controllers were more effective at applying irrigation when necessary based on plant water needs rather than using a rigid day of the week restricted schedule. However, irrigation application in Apollo Beach by the compariso n treatment was higher than its corresponding ET treatment with no difference in turfgrass quality, averaging 5.7. Thus, this observation was not conclusive and could be due to other uncontrolled factors affecting a landscape such as fertilizer application and pest Turfgrass quality ratings in Riverview were not different and were well above the acceptability threshold of 5, averaging 6.0 and 6.7 for the comparison and ET treatments, respectively (Ta ble 2 2). The decline in irrigation application in Riverview failed to negatively impact turfgrass quality despite a noticeable decrease in irrigation application by both treatment groups. The comparison group consisted of only three cooperators who did n ot have the full sun landscapes common to many planned development communities in Florida. These homes had established, tree covered landscapes that would have contributed to lower irrigation and acceptable quality ratings. High water tables could have bee n present at two of the homes; one home continually experienced periods of poor drainage, with standing water in the vegetated culvert, and another home was located along a creek. As a result, turfgrass quality was unaffected by the decline in irrigation a pplication during the study period. Changes in irrigation behavior due to implementing an ET controller caused a reduction of 23% to 34% compared to the historical average (Table 2 4). These

PAGE 82

82 reductions were generally less than the water savings observed i n field plot studies conducted in central and southwest Florida that showed 25% 63% seasonal savings (McCready et al. 2009) and 43% annual savings (Davis et al. 2009). Although the cooperators ultimately had complete control over the ET controllers and wer e not in a managed research environment, these cooperators received complimentary site specific adjustments of their ET controllers from UF IFAS when requested without relying on irrigation contractors or other technical support options. If this service ha d not been provided, the reductions might have been even less substantial. These results suggest that implementing ET controllers in the real world without considering previous irrigation habits may result in lower water savings than expected. Given the d eficit irrigation of the comparison homes, we were concerned that the study areas. Thus, additional homes in close proximity to cooperators in the ET treatment that e xhibited comparable landscapes were analyzed from billing records as of the small num ber of comparison homes that were recruited for this study. Using these new observations, average irrigation for the comparison treatment in Apollo Beach declined from 51 mm mo 1 to 39 mm mo 1 whereas the comparison treatment in Riverview increased from 24 mm mo 1 to 33 mm mo 1 (Table 2 5). Average historical irrigation application declined in all three locations with cooperators in Apollo Beach, averaging 47 mm mo 1 and Riverview, averaging 39 mm mo 1 applying significantly less than the GIR in those lo cations (59 mm mo 1 for Apollo Beach and 60 mm mo 1 for

PAGE 83

83 Riverview). Despite the reduced historical averages, irrigation application during the study period was still significantly less than the GIR and historical averages, further reinforcing the fundamen tal change in irrigation habits during the study period that was originally observed with the initial comparison homes. In Valrico, the comparison plus estimations treatment group irrigated more than the comparison treatment group, averaging 43 mm mo 1 which was significantly different from both updated averages of the GIR and the historical average. There were no differences in irrigation application between the comparison plus estimations group and ET controller treatments in any location during the study period despite significant decline in irrigation application compared to the historical average for the ET treatment. Fifteen of the 21 cooperators with ET controllers participated in a post test survey concerning their experiences. There were two respondents (13%) that were unable to identify the name of the smart technology installed at their home based on multiple choice answers. Of these two respondents, one indicated that they were unsure of the name of the technology, but knew they were usin g a smart controller and one indicated that they were unaware of a smart technology at their home at all. The respondents were asked questions to determine if they understand the basis of the ET controller technology. When asked to select the major inf luence to determining irrigation runtimes by an ET controller, 80% correctly selected weather conditions whereas three respondents (20%) incorrectly thought that ET controllers were based on soil moisture. The two respondents who were unable to identify a n ET controller installed at their home were also a part of the 20% that made this incorrect choice. The respondents were also asked about their trust in the ET controller and

PAGE 84

84 whether they felt comfortable interacting with the technology. Twelve responde nts (80%) trusted that the ET controller irrigated properly and 11 respondents (73%) felt that they could make changes to the controller settings. Two of the three respondents that did not trust the ET controller also responded with lack of trust of rain sensors indicating that the respondents were possibly leery of overall irrigation automation. To determine overall satisfaction from using an ET controller, respondents were asked about their satisfaction with their irrigation system, landscape, and spec ifically about their controller. All respondents were indifferent, somewhat satisfied, or extremely satisfied with their irrigation practices, which included the use of an ET controller. More specifically, all respondents rated the performance of the ET c ontroller on a 1 to 5 scale as a 3 or higher except for one respondent who also indicated extreme dissatisfaction with their landscape appearance. This one respondent was also the only cooperator that provided a negative answer when asked if the irrigatio n system was adequately irrigating the landscape. Communication with the cooperator that responded as extremely dissatisfied occurred throughout the study period where the cooperator was convinced that the controller did not apply irrigation at all. Thei r irrigation system and ET controller were checked multiple times for functionality and their sub daily water data was verified through manual meter readings to ensure irrigation was occurring. The most prevalent issue with turfgrass quality was weed mana gement, which cannot be fixed by using a smart controller. There were 12 respondents (80%) that indicated they were at least somewhat satisfied with their landscape appearance with the remaining respondents indicating some dissatisfaction and indifference

PAGE 85

85 Overall, a majority of the cooperators were satisfied with the ET controller and felt it maintained their irrigation scheduling practices or improved them. Eleven cooperators (73%) responded that they plan to continue using the ET controller and eigh t of those respondents planned to pay the annual $48 signal fee for the ET controller to a basic time clock with user selected runtimes instead of paying the fee. Despite the rela tively dry conditions and the design of an ET controller to maintain well watered conditions, the ET controllers decreased outdoor water use compared to the historical trends. The comparison treatment also reduced water use from historical trends, but the comparison homeowners accomplished reductions manually whereas the ET controller performed automatically and without water restrictions. An ET controller should produce significant savings if used by a homeowner who historically irrigated more than 576 mm yr 1 in this region. This value was estimated from annual historical irrigation of the ET controller treatment that had the lowest average irrigation application and was significantly less than the historical average (Table 2 2). Based on the results of t his study, cooperators historically irrigating similarly to or more than GIR in southwest Florida with a minimum annual estimate of 696 mm (based on the minimum estimate of 58 mm mo 1 ) would produce substantial irrigation savings using the Intelli Other brands of ET controllers use their own proprietary algorithms for scheduling irrigation that may have different performance. However, most of the cooperators in this study only irrigated a fraction of the gross requirement on average.

PAGE 86

86 ET controllers are designed to maximize efficient irrigation for landscapes without sacrificing quality in the name of water conservation. Mayer et al. (2009) found that homes already irrigating less than GIR saw an increase in irrigation application when implementing a smart controller, whereas those irrigating more than GIR saw a significant decrease in irrigation application. In this study, irrigation savings were not as significant due to cooperators who did not necessarily irrigate more than GIR prior to the study. Conclusions When irrigation above the estimated GIR is typical this study showed that ET controllers have the potential for water conservation in Florida. I rrigation applied was 23% 41% less than GIR and 23% 34% less than the historical average irrigatio n Irrigation amounts applied by the ET controllers were adequate to maintain good landscape quality. Homeowners who already irrigate less than GIR (i.e. well watered conditions ) or accept declines in landscape quality on a regular basis will not benefi t from using an ET controller in terms of water savings. In fact, their irrigation use may increase with an ET controller. However, homeowners who set their time clocks to a peak irrigation schedule without updating the controller settings throughout the y ear may see significant reductions in irrigation by implementing an ET controller, specifically during the fall and winter seasons. In southwest Florida, an Intelli residential properties that irrigate at least the GIR and w ould produce significant savings if irrigating more than 696 mm annually. Adequate performance by other ET controllers may occur at different thresholds due to the variability in irrigation scheduling techniques across brands.

PAGE 87

87 Proper controller programmin g was essential to achieving maximum water savings while maintaining acceptable landscape quality. Although the cooperators ultimately had complete control over the ET controllers, their primary interaction with UF IFAS through the study period was to rec eive complimentary programming adjustments. If this service had not been provided, any lasting effect of the ET controllers may have been less substantial. If irrigation was conservative prior to ET controller implementation, the ET controller increased irrigation. In addition to seeing this trend in the cooperator study, this conclusion was also found by Mayer et al. (2009), in which the sites already irrigating less than the GIR did not experience water savings with an ET controller. Other smart irrigat ion technologies that do not increase existing irrigation, such as soil moisture sensors, would be more appropriate in that situation.

PAGE 88

88 Table 2 1. The ET O and rainfall for the historical time period (2001 2009) and the study period (2009 2010). Period ET O (mm) Rainfall (mm) Days with Rainfall 2001 1,436 764 82 2002 1,388 1,425 154 2003 1,318 1,334 132 2004 1,438 1,900 114 2005 1,401 1,764 141 2006 1,481 1,499 112 2007 1,519 1,393 116 2008 1,453 1,261 117 Historical Average 1,423 a 1,306 a 118 a 2009 1,471 1,280 117 2010 1,495 1,141 107 Study Period 1,483 a 1,210 b 112 a Note: Values followed by different letters within the same columns are different at the 95% confidence level. Table 2 2. Average monthly irrigation application from February 2009 through January 2011 as compared to average monthly historical irrigation and average gross irrigation determined using Equations 2 1 and 2 3 Location Treatment a Number of Homes Irrigation Applicat ion b ( mm mo 1 ) Gross Irrigation Requirement ( mm mo 1 ) Historical Irrigation Application (mm mo 1 ) Turfgrass Quality c Apollo Beach C 6 51 c 53 c 70 a 5.7 BC ET 7 38 d 58 bc 60 b 5.6 BC Riverview C 3 24 e 57 b 46 c 6.0 ABC ET 5 36 d 65 a 48 c 6.7 AB Valrico C 6 36 c 51 b 62 a 5.1 C ET 9 41 c 58 ab 60 a 7.4 A a Labels of C and ET represent the comparison treatment and ET controller treatment, respectively. b Values within each location followed by lowercase letters are different at the 95% confidence level; c Values followed by uppercase letters are different at the 95% confidence level.

PAGE 89

89 Table 2 3. Cumulative irrigation application over the 2009 2010 study p eriod as compared to cumulative gross irrigation requirement calculated using Equations 2 1 and 2 3 Location Treatment Cumulative Irrigation (mm) Difference from C (%) Cumulative Gross Irrigation Requirement (mm) Difference from Gross Irrigation Requirement (%) Apollo Beach C 1,166 0 1,132 3 ET 891 24 1,240 28 Riverview C 531 0 1,225 57 ET 820 54 1,395 41 Valrico C 830 0 1,086 24 ET 950 14 1,238 23 Note: Labels of C and ET represent the comparison treatment and ET controller treatment, respectively. Table 2 4. Cumulative irrigation application as compared to cumulative historical irrigation application calculated from 9 years of billing data, when available. Location Treatment Cumulative Irrigation (mm) Cumulative Historical Irrigation (mm) Difference from Historical (%) Apollo Beach C 1,166 1,580 26 ET 891 1,350 34 Riverview C 531 1,013 48 ET 820 1,068 23 Valrico C 830 1,397 41 ET 950 1,338 29 Note: Labels of C and ET represent the comparison treatment and ET controller treatment, respectively. Table 2 5. A verage monthly irrigation application from February 2009 through January 2011 as compared to average monthly historical irrigation and average gross irrigation where additional cooperators were added to the comparison treatment through billing records. Location Treatment a Number of Homes Irrigation Application b (mm/month) Gross Irrigation Requirement (mm/month) Historical Irrigation Application (mm/month) Apollo Beach C+E 18 39 c 59 a 47 b ET 7 38 c 58 a 60 a Riverview C+E 13 33 d 60 a 39 c ET 5 36 cd 65 a 48 b Valrico C+E 30 43 d 50 c 52 bc ET 9 41 d 58 ab 60 a a Labels of C+E and ET represent the comparison plus estimations treatment and ET controller treatment, respectively. b Values within each location followed by different letters are different at the 95% confidence level.

PAGE 90

90 CHAPTER 3 IMPLEMENTING SMART CONTROLLERS ON SINGLE FAMILY HOMES WITH EXCESSIVE IRRIGATION I ntroduction Previous research has shown that at least half of total household water use goes toward irrigation when homes have automatic irrigation systems. In Nevada, Devitt et al. (2008) found that outdoor water use was 66% of total water used at 27 residential sites. Similar results were also found in central Florida where 64% of the total household water use was irrigation (Haley et al. 2007). When faced with potential population growth, limited water resources, and installation of automatic irrigation systems becoming the standard practice, efficient automatic irri gation has become increasingly important. Smart controllers are technologies designed to adjust or override irrigation based on weather or soil conditions thus limiting over irrigation (Dukes 2012). Based on the current products available on the market, s mart controllers include weather based irrigation controllers, or evapotranspiration (ET) controllers, and soil moisture sensor (SMS) controllers. Though there are many variations, ET controllers typically use weather information, user selected program se ttings, and proprietary algorithms to determine the irrigation schedule instead of relying on manually selected runtimes. Soil moisture sensor controllers bypass irrigation events when the measured soil moisture is greater than a threshold, generally sele cted based on available water holding capacity. A cooperator study conducted in Pinellas County, FL using Acclima TDT RS500 (SMS treatment) showed reductions in irrigation application of 65% after 26 months compared to homes with typical automatic irriga tion systems. Additional treatments of

PAGE 91

91 automatic irrigation system with rain sensor and automatic system with rain sensor and educational materials did not have statistically significant reductions. Recent research has shown that smart technologies are most effective for homeowners that have excessive irrigation. A two year study in Colorado showed no water savings due to five of the seven cooperators being historical under i rrigators (Aquacraft, Inc. 2002; Aquacraft, Inc. 2003). Mayer et al. (2009) fo und only 6% water savings from 3,112 ET controllers in California using one year of pre installation to one year of post installation. When further breaking down the analysis, 42% of the participants were historical under irrigators that experienced incre ased irrigation application as a result of the technology thus decreasing overall savings. Davis and Dukes (in press) recommended that the Toro Intelli Sense ET controller would be appropriate for homeowners with average irrigation totals greater than 696 mm annually in southwest Florida. As evidenced by previous research studies, smart controllers are most effective at water conservation when implemented by historical over irrigators. However, relying on volunteers that may have conservative irrigation tendencies prior to study participation can dampen the impact of smart controllers when targeted to the correct audience. The objective of this study was to determine the water conservation potential of implementing smart controllers on single family ho mes with excessive irrigation in Orange County, FL. M aterials and Methods Participant Selection Historical monthly water billing records were provided by Orange County Utilities for all single family residential accounts in their service area over a seven year period.

PAGE 92

92 Since actual indoor water use was unknown prior to the study period, it was estimated using the per capita method (Mayer et al. 1999) with the assumptions of 0.25 m 3 d 1 per person and 2.25 persons per account (U.S. Census Bureau 2009), totaling 16.9 m 3 mo 1 This method is frequently used to separate indoor and outdoor water uses when only total account data area available (Friedman et al. 2013 ; Ma yer et al. 1999; Romer o and Dukes 2013; Davis and Dukes, in review). Estimated irrigation was evaluated against monthly gross irrigation requirement (GIR) totals determined from a daily soil water balance, thus creating monthly landscape irrigation ratios (LIR ) ( Davis and Duke s, in review). Frequent LIR values greater than 1 indicate over irrigation practices whereas frequent LIR values less than 1 indicate under irrigation practices. Accounts with LIR values that were greater than 1.5 for at least 3 months per year for three consecutive years were considered over irrigators. The threshold of 1.5 was chosen to account for the variability inherent in the assumptions used for the calculations and to increase the probability to identifying customers with excessive irrigation. T he 7,408 accounts that met the criteria as an excessive irrigator were solicited by Orange County Utilities through a mailed letter to respond via an on line questionnaire provided by University of Florida Institute of Food and Agricultural Sciences (UF IF AS). There were 843 unique responses to the solicitation letter. To aid in participant selection, irrigation system evaluations were performed in geographic areas with many respondents to increase the likelihood of meeting the statistical requirements for treatment replications while minimizing spatial variability. Irrigation systems were checked for functionality and problems were reported to the

PAGE 93

93 homeowner. Due to the large number of evaluations required for this study, visual inspections of irrigation system efficiency were substituted for quantitative distribution uniformity measurements. Treatment Description A total of 167 participants were selected in nine locations across Orange County, FL (Table 3 1). Five treatments were replicated four times in each location except for three locations where securing 20 participants was not a viable option Pointe location had only three replications of one treatment (19 participant s). Only three of the five treatments were implemented in Apopka (13 participants) and North Tanner Road (15 participants) locations to maintain sufficient replications. Of the five treatments, four treatments included the addition of an irrigation technolog y to the already existing irrigation system. Two treatment gr oups received ESP SMT ET controllers (Rain Bird Corporation, Azusa, CA) and two treatment groups received WaterTec S100 SMSs (Baseline Inc., Boise, ID). Two treatments, one for each technology, were installed using methods determined solely by the install ing contractor without UF IFAS intervention. The remaining two technology treatments included UF IFAS training for the contractor prior to installations, site specific programming of the smart technology, and cooperator education of the technology install ed at their home. All participants that received a technology also received a variance for day of the week water restrictions. The final treatment was a comparison treatment that did not receive interventions to their normal irrigation practices. Data Co llection and Analysis Weather data were collected from three weather stations installed by UF IFAS in

PAGE 94

94 data for the cooperators in Apopka were collected from the Florida Automat ed Weather Network (FAWN) weather station located at the Mid Florida Research and Education Center in Apopka, FL. The weather stations provided temperature, relative humidity, solar radiation, and wind speed at fifteen minute intervals. Weather stations w ere also equipped with rain gauges to measure rainfall totals. Maximum distances between the representative station and cooperating homes was 10.8 km. Two rain gauges monitored by UF IFAS and one rain gauge monitored by Orange County Utilities were used t o better approximate rainfall in Turtle Creek, Waterford Lakes, and North Tanner Road locations. Historical average rainfall and ET O were determined using thirty years of data (1980 2010) from a weather station located at the Orlando International Airport ( NCDC 2009). All 167 participants received E Coder R900i (Neptune Technology Group, Inc., Tallassee, AL) flow meters installed on dedicated irrigation lines. These flow meters had automatic meter recording (AMR) capability built into the meter that record ed irrigation volumes per hour throughout the study. The minimum volume recognized by the meter was 379 cm 3 Volumes were converted to depth of irrigation applied using irrigated areas measured during the initial on site evaluation. Turfgrass quality ratings were taken seasonally using National Turfgrass Evaluation Program (NTEP) procedures (Shearman and Morris 1998). The rating scale ranged from 1 to 9, where 1 represents dead turfgrass or bare ground, 9 represents the highest possible quality, and 5 was the minimal level for an acceptable single family landscape. Turfgrass quality was judged primarily based on color and density.

PAGE 95

95 Statistical analyses were performed using Statistical Analytical Systems (SAS) software (Cary, NC). Irrigation applicatio n and turfgrass quality were analyzed using the glimmix procedure and comparisons were made using the least mean square differences by treatment, soil type, and season. Significance was determined at a 95% confidence level. The study was conducted from No v. 2011 through Aug. 2013. Gross Irrigation Requirement The GIR, an estimate of theoretical irrigation needs calculated using a daily soil water balance, was used as a comparison to irrigation by the treatments in this study. The GIR was calculated by multiplying the net irrigation water requirement (IWR net ) by a scheduling multiplier (SM). The IWR net is defined as the amount of irrigation required to increase soil water storage to field capacity (FC), or the maximum water level that can be stored befo re gravitational drainage (IA 2005). The IWR net was determined from mass conservation of soil water content using equation 2 1 (IA 2005). The PWR is the plant water requirement (mm) and R e is effective rainfall (mm). The IWR net was accumulated daily, bu t was applied only on days when the soil water level fell below management allowable depletion (MAD), calculated as 50% of the difference between FC and permanent wilting point (PWP) where PWP is the water level where plants can no longer extract water fro m the root zone (IA 2005). The PWP and FC were selected as 9 mm and 5 2 mm (43 mm available water) for the locations classified as flatwoods soils and 9 mm and 33 mm (24 mm available water) for the locations classified as sandy soils (USDA 1989a, USDA 1989 b). Both PWP and FC were calculated based on a root zone depth of 305 mm for turfgrass. Once IWR net was applied, the soil water level increased to field capacity and IWR net was reset to zero. Deep percolation and surface runoff, also typically part of th e soil water balance, were considered negligible since

PAGE 96

96 deep percolation and runoff can be avoided with proper design and management of the irrigation system. The PWR equals the amount of water necessary to maintain healthy plant material (IA 2005) and wa s calculated as the plant specific evapotranspiration (ET C ) using Equation 1 28 (Allen et al. 1998). Reference evapotranspiration (ET O ) is the estimated evapotranspiration of a short reference crop assumed to be a dense, well watered, cool season turfgrass maintained at a 0.12 m height. The ET O was calculated using Equation 1 13 using the collected weather station data. The crop coefficient (K C ) values are ratios of average crop specific evapotranspiration to average reference evapotranspiration. These values incorporate distinguishing characteristics of the specific crop to the reference crop such as crop height, crop soil surface resistance, and albedo of the crop soil surface (Allen 2000). The K C values selected for these studies were updated monthly for turfgrass with values of 0.45 (December February), 0.65 (March), 0.80 (April), 0.90 (May), 0.75 (June), 0.70 (July August), 0.75 (September), 0.70 (October), and 0.60 (November) (Jia et al. 2009). Effective rainfall was limited to the portion of total daily rainfall that caused the soil water level to reach the maximum soil storage capacity after PWR was taken into account. Rainfall that exceeded the soil storage capacity was considered lost due to surface runoff or deep percolation. A scheduling multiplier (SM) based on the average distribution uniformity of the irrigation system was used to convert IWR net to GIR. The SM was estimated as 1.00, 1.25, and 1.67 to represent irrigation system efficiencies of 100%, 80%, and 60%, respectivel y. The GIR was calculated as a range of acceptable irrigation application by

PAGE 97

97 multiplying IWR net (Equation 2 1) by each SM (IA 2013) resulting in two irrigation target ranges of achievable efficiency as 60% to 80% and high efficiency as 80% to 100%. Result s Weather Patterns The monthly ET O and rainfall totals fluctuated throughout the study period with minimums of 72 mm (Jan. 2013) and 1.4 mm (Nov. 2012) and maximums of 177 mm (May 2012) and 236 mm (Aug. 2012), respectively (Figure 3 1). There was little v ariation in ET O calculated across the location clusters as evidenced by relatively narrow error bars throughout the study period. Rainfall was much more variable across locations, especially during periods of frequent rainfall, indicating the occurrence o f some localized rainfall events. Using Figure 3 1, periods of similar weather patterns were identified to account for differences in treatments based on weather. Though ET O fluctuated throughout the study period, differences in ET O between winter and summer months were not as extreme as differences in rainfall. As a result, four periods of wet (summer) and dry (winter) seasons were classified based on rainfall patterns (Table 3 2). The dry seasons were selected as months where rainfall did not meet E T O indicating very little rainfall. The wet seasons were selected as months where rainfall met or exceeded ET O considering the high variability in the rainfall measurements. Further analysis of historical rainfall confirmed that the seasons defined as dry received less rainfall than periods defined as wet (Table 3 2). Due to average rainfall having such high variability, there was no significant difference at the 95% confidence level between rainfall during the study period and historical average for each season. This was also true for the number of rainfall events.

PAGE 98

98 Irrigation Application In all weekly statistics evaluated, location cluster was not significant to the statistical model during the study period. However, the soil type and seasonal period we re significant resulting in the need to separate results by these factors. The treatment effect was also significant (P value < 0.0001) indicating differences in irrigation applied by the various technologies and implementation approaches. Dry 2012 Weekl y irrigation application ranged from 1 2 mm (SMS+Edu) to 2 5 mm (MO) in the flatwoods locations (Table 3 3). There was no significant difference in weekly irrigation application between SMS and ET treatments, applying 1 9 mm and 20 mm, respectively The MO applied the most irrigation per event, averaging 1 2 mm. The ET applied the least per event, averaging 5 mm, but applied irrigation the most often with 3.6 events per week. Only 1. 2 events occurred per week for SMS+Edu combined with only 6 mm per event re sulting in the low weekly application. In these same locations, all treatments applied more irrigation than GIR, representing the irrigation required, regardless of efficiency (Figure 3 2). Cumulative irrigation application ranged from 291 mm (SMS+Edu) to 460 mm (SMS). Both technology treatments that received education applied less irrigation compared to the other three treatments, totaling 1% (SMS+Edu) and 11% (ET+Edu) more than the achievable GIR range of 60% to 80% efficiency. However, there was still uncaptured potential water savings by all treatments during this season with over irrigation by as much as 52% (SMS). Irrigation was the highest in the sand locations with weekly irrigation application ranging from 2 0 mm ( SMS, SMS+Edu ) to 3 2 mm (MO). The MO treatment also applied

PAGE 99

99 the most irrigation per event, averaging 1 6 mm, which was applied 1.9 times per week. There was no significant difference in weekly irrigation application between any of the technology treatments during this period. However, ir rigation per event was 9 mm for ET+Edu SMS, and SMS+Edu which was significantly different from 7 mm for ET. The ET treatment had more events per week, averaging 3.1 which was more than the other treatments resulting in similar weekly totals to the other technology treatments. Cumulative irrigation in the sand locations ranged from 383 mm (SMS+Edu) to 592 mm (MO) (Figure 3 3). Both the SMS and ET+Edu treatments irrigated similarly to SMS+Edu with totals of 416 mm and 385 mm, respectively, whereas ET irrigated similarly to MO, totaling 553 mm. The MO and ET treatments over applied by 34% and 25%, respectively, considering GIR with 60% efficiency. The three treatments with the lowe st cumulative irrigation application still had over irrigation during the months of February and March, but fell within the achievable GIR range assuming 60% to 80% efficiency. Wet 2012 The weekly irrigation average for the flatwood locations was signifi cantly different for all treatments, rang ing from 11 mm (SMS+Edu) to 2 7 mm (MO) (Table 3 3). The technology with education treatments tended to irrigate less than the technology only treatments on a weekly basis. The MO treatment applied 1 3 mm per event, similar to the previous season, which was significantly more than all other treatments. The ET treatment applied the least irrigation per event at 5 mm, but applied the most times per week with 3.8 events. Cumulative irrigation application in the flatwoo ds locations ranged from 304 mm (SMS+Edu) to 629 mm (MO) (Figure 3 4). Compared to the previous season, the

PAGE 100

100 technology treatments began to separate from MO with decreases of 14% by ET and 26% 30% by SMS and ET+Edu. Though SMS+Edu had the best results, fa lling within the preferred GIR range assuming 80% to 100% efficiency at the end of the season, some under irrigation occurred by this treatment in July. The ET+Edu began the season in the acceptable GIR range, but steadily increased during the last three months of the season resulting in over irrigation. The MO treatment for the sand locations continued to irrigate similarly to the dry 2012 period, applying 36 mm per week at 1 6 mm per event and 1.8 events per week, despite frequent rainfall. Weekly irrig ation statistics for the technology treatments also had similar trends to dry 2012 in irrigation per event, ranging from 6 mm to 9 mm. The educati on treatments and SMS applied 17 mm ( SMS ) to 21 mm (SMS +Edu ) across 1 .9 ( both SMS treatments ) to 2 2 ( ET +Edu) events per week. The biggest difference from dry 2012 was that the technology treatments began to have significant differences with the ET treatment applying more per week than the other three technology treatments. Irrigation per event remained relat ively the same for all treatments between these two seasons, but events per week increased to 4. 1 events for ET resulting in higher overall irrigation In the sand locations, cumulative irrigation ranged from 457 mm (SMS+Edu) to 858 mm (ET) (Figure 3 5). This is the first season and soil type where a technology treatment applied more irrigation than MO, resulting in an 8% increase. The remaining three technology treatments irrigated similarly throughout the entire season resulting in a maximum differenc e of 2% and a minimum reduction of 42% from MO. Assuming no residential irrigation system is 100% efficient, slight under irrigation occurred by these

PAGE 101

101 three treatments during the beginning of the season, but increased in July to just outside the high effi ciency GIR range thus following the 80% efficiency trend. The ET treatment also increased steeply in July resulting in 43% more irrigation than the irrigation required using an irrigation system with 60% efficiency. Dry 2013 In the flatwoods locations, weekly irrigati on application by MO averaged 26 mm, which was applied in 1 5 mm events on an average of 1.7 events per week (Table 3 3). Education was important to the SMS technology for this season, with significantly less irrigation by SMS+Edu (1 1 mm) compared to SMS (1 8 mm). Both ET treatments applied 15 mm per week which were significantly different from both SMS treatments After MO, SMS applied the most irrigation per event, averaging 10 mm, with the remaining treatments averaging 5 6 mm Once again, t he ET treatment applied the least per event, 5 mm, but irrigated the most times per week at 3. 8 events. All treatments irrigated more than what was required throughout these winter months in the flatwoods locations, ranging from 264 mm to 55 1 mm by SMS+Edu and MO, respectively (Figure 3 6). There was no difference in cumulative irrigation between the ET treatments, applying 328 mm (ET+Edu) to 335 mm (ET) with reductions of 39% 40% compared to MO. The SMS applied the most out of the technolo gy treatments, resulting in 29% reduction from MO compared to a 52% reduction by SMS+Edu. Both ET treatment s applied the least on the sand soil, averaging 17 mm for ET and 18 mm for ET+Edu. The remaining three treatments were not significantly different from each other with averages ranging from 21 mm (SMS, SMS+Edu) to 23 mm (MO) Irrigation application by MO was significantly less than previous periods, indicating that there was a change in behavior by this treatment group due to an unknown outside

PAGE 102

102 fact or. This treatment still applied the most irrigation per event, 1 5 mm, but was only applied 1.4 times per week. The ET treatment continued to apply the least per event, 5 mm, but irrigated most often at 3.1 events per week. Similar to the flatwoods locat ions, cumulative irrigation in the sand locations was much higher than required during these winter months, applying 370 mm (ET+Edu) to 515 mm (MO) (Figure 3 7). Though ET+Edu had the largest reduction in irrigation application from MO (28%), this treatme nt still over irrigated by 37% assuming 60% irrigation system efficiency. All other technology treatments reduced irrigation application from MO by 4% to 17%, indicating that there were uncaptured potential savings during the period. Wet 2013 In the flatw oods locations, weekly irrigation application for MO, SMS, and both ET treatments was not different with averages ranging from 21 mm to 22 mm (Table 3 3). Similar to other seasons, the application schedule within the week varied across all treatments. Ir rigation per event ranged from 5 mm (ET ) to 9 mm (SMS) and events per week ranged from 1.2 events (SMS+Edu) to 3.6 events (ET). The SMS+Edu treatment applied the least amount of irrigation per week (13 mm). All treatments fell within or just outside of th e achievable and high efficiency GIR ranges during this season for the flatwoods soils, ranging from 264 mm to 472 mm by SMS+Edu and ET, respectively (Figure 3 8). All treatments except SMS+Edu applied irrigation within the achievable efficiency GIR range with only the ET treatment applying 2% more than the range maximum. The SMS+Edu constantly fell just below the high efficiency range indicating that there were deficit conditions by this treatment throughout this season.

PAGE 103

103 Weekly irrigation application in the sand locations was highest by MO and ET, applying 28 29 mm, but had very different scheduling techniques as was seen in other seasons (Table 3 3). The MO treatment averaged 1 4 mm per event with 1.6 events per week comp ared to 6 mm per event by ET with 4. 7 events per week. Similar to most other seasons, there were no differences in irrigation application between both SMS treatments, applying 1 6 1 7 mm per week The SMS treatments applied similar irrigation per event at 8 9 mm compared to ET+Edu but weekly irrigation was different for ET+Edu due to averaging 2.6 events per week compared to 1. 3 events (SMS+Edu) and 1.6 events (SMS) Cumulative irrigation ranged from 328 mm (SMS+Edu) to 570 mm (ET) in the sand locations (F igure 3 9). The two SMS treatments applied just less than the achievable GIR range for the first half of the season, but declined in irrigation application over the second half of the season resulting in final irrigation totals falling in the high efficie ncy GIR range, reducing irrigation by 34% 40% compared to MO. The ET+Edu treatment also followed the GIR at 60% efficiency, but declined in irrigation application by a smaller amount compared to the SMS treatments during the second part of the season, red ucing irrigation by 21% compared to MO. Since ET applied the most irrigation during this season, irrigation application was 4% more than MO and outside of the achievable GIR range. Turfgrass Quality Turfgrass quality ratings by treatment were not significantly different prior to the start of the study, ranging from 6.3 to 6.8 (Table 3 4). Additionally, treatments were not significantly different within each of the seasons that occurred during the stu dy indicating that quality did not suffer as a result of treatments. However, there were

PAGE 104

104 differences across seasons, ranging from 6.4 occurring during dry 2012 to 7.1 occurring during both wet seasons. It is likely that quality increased during the wet s easons due to ideal temperature conditions and lack of water stress that promoted better growth performance compared to the dry seasons. Other unmeasured factors could have affected turfgrass quality such as disease, fertilizer application, mowing practic es, and irrigation system maintenance. Discussion Overall, there was a trend of water savings due to installing a smart technology with additional savings from education and detailed programming. This was evidenced by the Edu treatment s consistently averag ing similar or less irrigation than their non education counterparts ( Table 3 3 ) However, only a few treatments were able to maintain irrigation totals within acceptable GIR ranges and none of the treatments performed with maximum efficiency throughout a ll seasons. The SMS+Edu treatment was the only treatment to irrigate efficiently on both soil types, falling within the high efficiency GIR range, but this occurred during the wet 2013 season only (Figure 3 8 ; Figure 3 9). The SMS treatment also fell wit hin the high efficiency GIR range during the wet 2013 season, but only for the sand soil type. The ET+Edu generally performed better in the sand locations compared to the flatwoods locations, falling within the achievable efficiency GIR range during most seasons. There were instances when weekly irrigation averages did not match the multiplication of irrigation per event by events per week with a maximum difference of almost 8 mm. T here would likely be 1 3 mm attributable to rounding thus further explora tion into the data was required Some treatments during some seasons had cooperators that irrigated unusually large amounts resulting in inflated weekly averages

PAGE 105

105 compared to the median. These instances skew the distribution from normal to non normal, cau sing a discrepancy in the two totals. These outliers are likely due to circumstances outside of the treatment, but were not removed because they are part of the variability in performing irrigation studies in real world, uncontrolled environments. A tran sformation of the data was not performed since this problem affected only 38% of the averages and the minimum sample size of any one combination of treatment, season, and soil type was 289. Sometime between the wet 2012 and dry 2013 seasons, there was a de crease in average weekly irrigation application by MO in the sand locations (Table 3 3). Interaction between UF IFAS and the cooperators in MO had not occurred since the initial on site irrigation evaluation, prior to the beginning of the dry 2012 season, indicating that the change in behavior was due to an external factor unrelated to the study. The most likely factor affecting irrigation totals was a water rate increase that became effective in Oct. 2012. Though the water rate increase was small, maxim um increase of $0.33 per 3.78 m 3 for the highest tier, the perceived increase can affect behavior. This was shown in Whitcomb (2005) where survey respondents were found to be concerned about the cost of water, especially irrigation, but were not knowledge able about the details of true water rates such as price size, and number of tiers. There was a general trend of higher weekly irrigation application for all treatments located in sand soils compared to flatwoods soils. Sandy soils generally have smaller soil water holding capacities compared to the flatwoods soils resulting in the need for more irrigation. This was seen in the GIR ranges where the irrigation

PAGE 106

106 applied assuming 100% efficiency, representing IWR net was 28% greater for the sand locations (1 ,099 mm) than the flatwoods locations (858 mm) (Figure 3 2; Figure 3 3). However, the ET treatment in sand locations was not different from MO, a group shown to be excessive irrigators (Davis and Dukes, in review). Since the only major difference between water holding capacity may have been inappropriate for the actual soil characteristics, causing ov er irrigation despite the default setting description matching the soil type. Though none of the treatments achieved GIR with good efficiency throughout all seasons, the SMS+Edu consistently had better results compared to the other treatments. In additio n to site specific zone runtimes, this technology required a proper burial location in a representative portion of the landscape and an estimation of the threshold based on field capacity of the soil that could be selected through automatic calibration as opposed to user selection. Once these requirements were met, the SMS directly measured soil moisture resulting in bypassing unnecessary irrigation events. The ET+Edu had acceptable results in some seasons, but generally applied more irrigation than SMS+E du. The success of an ET controller required more complex programming and detailed knowledge of both the technology and soil water characteristics of the landscape. The ET controller use s a model as an estimation of soil moisture which is only as good as the model inputs. The complexity of accurate programming in one on site evaluation was an obstacle for maximizing water savings when using ET controllers. It is likely that the amount of water savings achieved from determining more accurate model inputs would not justify the increased costs to

PAGE 107

107 obtaining the information due to other limitations on the technology, such as the accuracy of the weather monitor. In addition to the importance of program settings for ET controllers, there should have been higher accuracy in rainfall accountability in efforts to achieve water savings. Since most of the irrigation scheduling relies on modeling, rainfall was one of the only on site measurements taken by the ET controller. However, there was a lack of response to ra infall during wet seasons, where frequency of irrigation events remained constant or increased in wet months compared to dry months (Table 3 3), resulting in increased application during wet seasons when ET O was higher. Some increase would be expected, as was seen in the higher GIR totals from dry to wet seasons. However, the ET+Edu and SMS+Edu applied similar irrigation depths per event, but ET+Edu always applied more events per week than SMS+Edu in the flatwoods locations (Table 3 3) despite no differen ces in turfgrass quality (Table 3 4). The direct measurement of effective rainfall by the SMS technology allowed the technology to regulate the timer schedule resulting in more efficient irrigation application than the ET controller. Advances to the ET c ontroller technology must occur to address this issue. The benefits received as a part of one of the technology with education treatments included adjusted program settings and a single, on site cooperator education session that lasted no more than five mi nutes. Education is extremely important to the success of smart controllers so that they become a trusted device with proper maintenance to achieve long term water conservation. In the case of this study, it is likely that the single educational on site visit was not as influential to water savings as was the detailed program settings. One setting specific to the education treatments

PAGE 108

108 was restricting irrigation opportunities to three events per week. This was an effective limitation for the ET controller where irrigation savings were not as substantial for the ET treatment due to averaging more than 3 events per week. Though SMS did not average above two events per week, thus regularly bypassing unnecessary irrigation events, the SMS+Edu frequently avera ged fewer events per week than SMS. As a result of restricting irrigation to only three specific days, SMS+Edu was delayed until the next available irrigation day when falling below the soil moisture threshold, allowing either effective rainfall or short periods of deficit conditions between irrigation events. C onclusions The overall trend of water savings occurred due to installing a smart technology with additional savings from education and detailed programming. Results suggest that water savings were greatest for the SMS with education on the flatwoods soils, applying 11 13 mm per week. However, both SMS were relatively as effective as the ET controller with education on the sand soils. Despite the uncontrolled environment, cooperator interactions di d not increase irrigation application to previously excessive levels thus water savings were achieved by introducing a smart technology. Despite positive results based on trends in irrigation application, only a few treatments were able to maintain irrig ation totals within the defined GIR ranges resulting in 21% 52 % water savings by those treatments compared to MO, and none of the treatments performed efficiently in all seasons. A majority of the savings occurred by the cooperators within the sand soils by the SMS, ET+Edu, and SMS+Edu treatments. In the flatwoods soils, only the SMS+Edu fell within the GIR ranges during the wet seasons resulting in water savings of 52% in 2012 and 42% in 2013. None of

PAGE 109

109 the treatments fell within either achievable or hig h efficiency GIR ranges during the dry 2013 season. The SMS+Edu treatment performed better than the other treatments, falling within the achievable efficiency GIR range during dry 2012 and the high efficiency GIR range for both wet seasons regardless of soil type. The performance of SMS+Edu can be attributed to bypassing unnecessary irrigation events based on measured soil moisture and site specific runtimes as a part of education that limited over irrigation on an irrigation day The ET+Edu treatment a lso had promising performance results in the sand locations, falling within the achievable efficiency range during dry 2012 and wet 2013 and within the high efficiency range during wet 2012. I rrigation was greater than the achievable efficiency range in t he flatwoods locations during both 2012 seasons due to more events occurring per week. Both technologies were able to reduce irrigatio n application when implemented on sites with historical excessive irrigation. In general, there was higher consistency in efficient irrigation application by the SMS technology than the ET technology. The SMS was able to produce water savings in both soil types and during most seasons whereas the ET controller was less predictable. The ET controllers may require technolo gical advancement combined with additional educational opportunities to achieve the same amount of water savings as the SMS.

PAGE 110

110 Table 3 1. Number of cooperators by location and treatment. Group Name Soil type ET a ET+Edu b SMS c SMS+Edu MO d Total Hunters Creek A Flatwoods 4 4 4 4 4 20 Hunters Creek B Flatwoods 4 4 4 4 4 20 Keenes Pointe Area Sand 4 4 4 4 3 19 North Tanner Road Area Sand 0 5 0 5 5 15 Turtle Creek Area Sand 4 4 4 4 4 20 Waterford Lakes East Flatwoods 4 4 4 4 4 20 Waterford Lakes South Flatwoods 4 4 4 4 4 20 Waterford Lakes West Flatwoods 4 4 4 4 4 20 Sweetwater Apopka Area Sand 0 5 0 5 3 13 Total 28 38 28 38 35 167 a ET designates cooperators that received a Rain Bird ESP SMT ET controller b Edu designates cooperators that received an on site educational training and optimized programming c SMS designates cooperators that received a Baseline WaterTec S100 SMS d MO designates cooperators that did not receive a technology Table 3 2. Description of seasonal periods used to account for various dry/wet periods occurring throughout the study period. Season Start Date End Date Number of Weeks Cumulative Rainfall (mm) Number of Rainfall Events Study Period Average a Historical Average b (s.d.) Study Period Average Historical Average (s.d ) Dry 2012 Nov. 2011 Apr. 2012 25 123 428 (169) 30 46 (9) Wet 2012 May 2012 Oct. 2012 26 705 895 (170) 75 85 (11) Dry 2013 Nov. 2012 Mar. 2013 22 76 354 (168) 22 39 (9) Wet 2013 Apr. 2013 Aug. 2013 20 782 730 (164) 79 68 (10) a Averages during the study period were calculated using rain gauges installed for the study across all locations. b Historical averages were calculated using a rain gauge installed at the Orlando International Airport ( NCDC 2009).

PAGE 111

111 Table 3 3. Statistics for irrigation application by each treatment over the study period. Flatwoods Sand Treatment Irrigation Per Week (mm) Irrigation Per Event (mm) Events Per Week (#) Irrigation Per Week (mm) Irrigation Per Event (mm) Events Per Week (#) Dry 2012 MO 25 a 12 a 1.8 c 32 a 16 a 1.9 c ET 20 b 5 c 3.6 a 22 b 7 c 3.1 a ET+Edu 16 c 6 c 2.6 b 21 b 9 b 2.3 b SMS 19 b 8 b 1.4 d 21 b 9 b 2.0 c SMS+Edu 12 d 6 c 1.2 e 20 b 9 b 2.0 c Wet 2012 MO 27 a 1 3 a 1.9 c 3 2 a 16 a 1. 8 c ET 23 b 5 d 3.8 a 26 b 6 c 4.1 a ET+Edu 17 d 7 c 2.6 b 19 cd 8 b 2.2 b SMS 20 c 9 b 1.4 d 17 d 8 b 1.9 c SMS+Edu 11 e 5 d 1.3 e 21 c 9 b 1.9 c Dry 2013 MO 26 a 15 a 1.7 c 23 a 15 a 1. 4 e ET 15 c 5 c 3.0 a 17 b 5 d 3.1 a ET+Edu 15 c 6 c 2.5 b 18 b 8 c 2.4 b SMS 18 b 10 b 1. 5 d 21 a 9 bc 2.0 c SMS+Edu 11 d 6 c 0.97 e 21 a 10 b 1.6 d Wet 2013 MO 22 a 1 0 a 1. 6 c 29 a 1 4 a 1.6 c ET 21 a 5 e 3. 6 a 28 a 6 c 4.7 a ET+Edu 22 a 8 c 3.0 b 22 b 9 b 2.6 b SMS 21 a 9 b 1. 8 c 17 c 9 b 1.6 c SMS+Edu 13 b 6 d 1. 2 d 16 c 8 b 1.3 d Note: Values within each column and season followed by different lowercase letters are different at the 95% confidence level.

PAGE 112

112 Table 3 4. Turfgrass quality ratings for each treatment a were evaluated seasonally b Treatment Evaluation Period Turfgrass Quality Dr y 2012 Turfgrass Quality Wet 2012 Turfgrass Quality Dry 2013 Turfgrass Quality Wet 2013 Turfgrass Quality MO 6.3 a 6.4 a 7.0 a 6.6 a 6.9 a ET 6.8 a 6.5 a 7.3 a 6.9 a 7.2 a ET+Edu 6.6 a 6.3 a 7.0 a 6.8 a 7.2 a SMS 6.7 a 6.3 a 7.0 a 6.9 a 7.1 a SMS+Edu 6.7 a 6.4 a 7.0 a 6.6 a 7.0 a Mean 6.6 C 6.4 D 7.1 A 6.7 B 7.1 A a Ratings within each season followed by different lowercase letters are different at the 95% confidence level. b Mean seasonal ratings followed by different uppercase letters are different at the 95% confidence level.

PAGE 113

113 Figure 3 1. Monthly ET O and rainfall for all locations during the study period with error bars representing the standard error of measurements across all study locations.

PAGE 114

114 Figure 3 2. Cumulative irrigation application for all five treatments occurring on the flatwoods locations during the dry 2012 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges. Figure 3 3. Cumulati ve irrigation application for all five treatments occurring on the sand locations during the dry 2012 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges.

PAGE 115

115 Figure 3 4. Cumulative irrigation application f or all five treatments occurring on the flatwoods locations during the wet 2012 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges. Figure 3 5. Cumulative irrigation application for all five treatments occurring on the sand locations during the wet 2012 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges.

PAGE 116

116 Figure 3 6. Cumulative irr igation application for all five treatments occurring on the flatwoods locations during the dry 2013 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges. Figure 3 7. Cumulative irrigation application f or all five treatments occurring on the sand locations during the dry 2013 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges.

PAGE 117

117 Figure 3 8. Cumulative irrigation application for all five treatments occu rring on the flatwoods locations during the wet 2013 season compared to the estimated irrigation requirement assuming achievable and high efficiency ranges. Figure 3 9. Cumulative irrigation application for all five treatments occurring on the sand loc ations during the wet 2013 season compared to the estimated irrigation requirement assuming a and high efficiency ranges.

PAGE 118

118 CHAPTER 4 COMPARISON OF MODELING METHODS TO ESTIMATE SOIL WATER CONTENT IN TURFGRASS BASED LANDSCAPES I ntroduction The American lawn became a fundamental part of the home landscape after the Second World War when suburbanization exploded (Robbins and Birkenholtz 2003). Turfgrass can be a functional part of the landscape by reducing soil erosion, preventing particulate inhalation from excess dust, dissipating heat at the soil surface level, and providing a wildlife habitat. There are also recreational and aesthetic reasons for maintaining turfgrass landscapes such as providing cushioned support for sports and playground activities and improving mental health and quality of life (Beard and Green 1994). Florida was ranked fourth in domestic water use in 2005 ; the only states having higher consumption were California, Texas, and Michigan (Kenny et al. 2009). Using the li mited water supply to irrigate landscapes, which are typically considered ornamental in nature and not purposeful to sustain life such as in agriculture, may not be socially responsible. Finding better ways to preserve turfgrass based landscapes through m odernized irrigation techniques may be a good compromise. One modern technology currently promoted for water conservation is a weather based irrigation controller that use s reference evapotranspiration (ET O ) to estimate plant water needs. Though they vary in functionality by brand and model, most ET controllers use a soil water balance (SWB) to determine how much irrigation is necessary and when to apply it. Research has shown that ET controllers have the potential to produce water savings when implemente d in scenarios with frequent excessive irrigation (Mayer et al. 2009; Davis and Dukes, in review). They have also been shown to increase

PAGE 119

119 irrigation at homes that historically irrigate at or below the theoretical irrigation requirement (Mayer et al. 2009; Davis and Dukes 2014 ). The Irrigation Association developed the Smart Water Application Technologies (SWAT) Testing Protocol for Weather Based Irrigation Controllers as a method to evaluate the ability of ET controllers to schedule irrigation (IA 2008). This testing protocol established in 2002 and now in its 8 th draft specifies a daily soil water balance occurring over a 30 day period to determine if ET controllers can maint ain theoretical soil moisture The protocol specifies that testing results ar e valid when 30 day totals of rainfall and ET O reached 10.2 mm and 63.5 mm, respectively. Publishing SWAT test results are at the discretion of the manufacturer and can be continually tested until achieving acceptable scores during a valid testing period. The USEPA WaterSense program was developed to encourage the adoption of efficient water related products into the consumer market thus ultimately increasing water conservation practices. Through this program, ET controllers are identified to consumers th rough product labeling after undergoing the SWAT test conducted by an independent testing facility (USEPA 2012). Prior to adoption of the SWAT test by USEPA, it was evaluated for transferability, or applicability of scores to different areas of the countr y, and reproducibility, or the ability to obtain identical results using the same input information by various testing labs (Davis and Dukes 2010). During these evaluation efforts, the primary deficiency of the SWAT test was identified as declining schedu ling efficiency scores during periods of frequent rainfall resulting and the lack of transferability (Davis and Dukes 2012). Other significant deficiencies included the low rainfall and ET O requirements for a valid test, penalized scores due to the mandat e that

PAGE 120

120 rainfall must be taken into account before irrigation despite when irrigation occurred prior to rainfall on the same day, and inconsistencies in controller programming compared to theoretical landscape characteristics to achieve high scores. Based o n these deficiencies, it was hypothesized that an hourly soil water balance would better match deterministic soil water models and improves estimates of soil water content compared to a daily soil water balance The objective of this study was to evaluate an hourly soil water balance to determine its ability to estimate soil wa ter content compared to Richard s equation, a deterministic model, and the current SWAT test using a daily soil water balance. Materials and Methods Soil Water Balance The volumetric volume of water to total volume (water, air, and soil particles) of the root zone. In a r ), where additional energy is requi s ), where FC ) is achieved when gravitational forces equal surface tension forces (cohesive and adhesive) between water molecules and soil particl es resulting in no drainage due to gravity (Richards 1931). Water storage FC s FC over time based on soil hydraulic properties. To reduce water stress on the plant material, the water content sh r MAD ) r FC (IA 2013). The soil water balance is a simple model that uses a tipping bucket approach to soil water movement in a homogeneous soil profile (i.e. depth per unit area of root zone)

PAGE 121

121 over a specified timestep (Ranatunga et al. 2008). The model is based on the principle that soil water at the current timestep (i) was affected by the soil water at the previous timestep (i 1) and inputs and output s occurring at i. Z i = Z i 1 ET C,i + R i + I i RO i D i (4 1) The Z (mm) is the depth of water within the soil profile measured from the bottom of the C (mm) is the plant specific ET, R (mm) is rainfall, I (mm) is irrigation RO (mm) is runoff from excess rainfall or irrigation, and D (mm) is deep percolation from excess rainfall or irrigation Plant specific evapotranspiration (ET C ) is calculated for any given plant material by applying a crop coefficient (K C ) u sing Equation 1 28 (Allen et al. 1998) In this study, K C values were selected monthly for warm season turfgrass located in central and southwest Florida (Jia et al. 2009), depending on the origin of the measured soil moisture data (Table 4 1) R ainfall and irrigation are utilized similarly in the model where the depth equals the total that fell (R) or was applied (I) over the unit area during the timestep of interest. s result in separating total depth into effective dept h, which is considered available for plant use and a combination of runoff and deep percolation losses s total depth equals effective depth. Runoff and deep percolation are two different physical proces ses that both FC resulting in water losses from the soil profile. Generally, runoff occurs due to improper cycle/soak scheduling of irrigation and landscape inclines whereas deep percolation occurs due to excess rainfall or irrigation.

PAGE 122

122 It is difficult and unnecessary to separate these losses because they are quantified outside of the soil profile and the variable of interest is the change in depth of water within the soil profile. Daily SWAT Test The SWAT test, specified in the SWAT Climatologically Based C ontrollers 8th Testing Protocol (IA 2008) and used by USEPA WaterSense as a test specification for labeling (USEPA 2012 ), uses a soil water balance to simulate water movement in a landscape plant system using a daily timestep It was developed as a tool to evaluate the ability of ET controllers to adequately and efficiently schedule irrigation. The SWAT test specifies six diverse virtual landscapes tested over a thirty day period during valid weather c onditions that met minimums of 63.5 mm of ET O and 10.2 mm of R over the testing period. The soil water balance specified by the SWAT test was modified to restrict water MAD FC The range is referred to as the root zone working water storage (RZWWS) (IA 2008). In addition to new terminology, water level fluctuations were limited to zero and RZWWS, or the maximum value able to be stored depending on the tested zone. A water level of zero does not equate to the absence of water i n the soil column. Therefore, the soil water balance used for the SWAT test does not represent the actual water level in the soil profile, but the level under which the plant material will no longer be considered well watered. An additional modification included an arbitrary reduction of total daily rainfall by 20% citing losses to non uniformity and runoff (IA 2008). In the SWAT test, runoff and deep percolation were defined as scheduling losses consisting of direct runoff when irrigation cycles exceed maximum runtimes based on

PAGE 123

123 infiltration rates, soak runoff when irrigation cycles occur prior to minimum soak times FC Hourly SWB Model An hourly soil wa ter balance was developed in efforts to address some deficiencies in the SWAT test such as lack of transferability and penalized scores due to mandating that rainfall occurred before irrigation despite actual order of events. All calculations specified in the daily version were maintained so that soil water movement within the RZWWS occurred in the same way. However, the soil water balance was FC MAD These simplifications may benefit the contr ollers during the test by not continually penalizing for deficit or surplus conditions. However, soil water levels in real systems can FC until reaching saturation with deep percolation naturally occurring during each timestep until ret FC Additionally, water levels MAD resulting in deficit conditions can occur in real systems and should be observed in the model. To account for soil water levels higher than field capacity, an exponential decay function was incorporated into each timestep using a stepwise equation. (4 2 ) The Z 0 (mm) is the soil water level at the first instance it exceeds Z FC (mm) as field 4 3.

PAGE 124

124 (4 3 ) depth of water exceeding field capacity thus affecting the amount of time it should take to drain to equilibrium. It was assumed that calculated by substituting field capacit 0 and 24 h for t. The depth of water removed from the soil profile using the exponential decay function, calculated as the difference between Z i and Z i 1 when Z 0 >Z FC was considered lost through deep percolation. The hourly SWB was adapted to incorporate layers within the root zone to evaluate the model at various depths. When more than one layer was evaluated, only the top most layer received rainfall and irrigation as inputs to the soil profile. Each additional layer rece ived only the deep percolation losses calculated using the exponential decay function from the upper adjacent layer. The ET C was separated into evaporation and transpiration. Evaporation was assumed to be 10% of the ET C and was removed from the top layer only (Hanks and Ritchie 1991). Transpiration was linearly proportioned into each layer so that a majority of ET C occurred from the top layers in relation to root density. MAD and deficit conditions occur, transpir ation decreases due to the increases in energy required to remove water from the soil by the plant material. To account for this phenomenon, a stress coefficient (K S ) was introduced to the ET C calculation using a linear relationship (Allen 2000): (4 4)

PAGE 125

125 The Z PWP MAD (mm) is the soil water level at maximum allowable depletion. H YDRUS 1D Richards equation (1931) was developed to estimate t he flow of water through capillary tubes. equation can be expressed by pressure head (h) (Equation 1 33) (Equation 1 34) or a co mbination of both (Equation 1 35) (Celia et al. 1990) Richards equation is a nonlinear partial differential equation that cannot be solved analytically. Instead, numerical approximations such as the finite difference or finite element methods combined with iteration techniques such as Picard or Newton methods are required for d iscrete solutions (Celia et al. 1990). HYDRUS 1D Versio n 4.0 (PC Progress, Prague, Czech Republic) was selected has been established as a stable and reliable mechanism for predicting water movement in variably saturated po rous media such a s a soil column ( et al. 2012). The Windows based program was designed to simulate one dimensional movement of soil water, heat, and solutes in a variably saturat ed soil profile ( et al. 2008). For the purposes of this rese arch, only soil water movement was considered. HYDRUS 1 35 by using Galerkin type linear finite element schemes where integration occurs using an implicit backwards finite difference scheme for both unsaturated and saturated conditions

PAGE 126

126 Field M easured Data PSREU Volumetric water content was collected from a bahiagrass pasture located in Block 4 at the Plant Science Research and Education Unit (PSREU) in Citra, FL. Time domain reflectometry (TDR) se nsors (Campbell Scientific, Logan, UT) were buried in profile at the depths of 15 0 m m, 30 0 m m, 45 0 m m, 60 0 m m, 75 0 m m, and 90 0 m m. Volumetric water content was collected at 15 minute intervals from Jan. 1 2004 to Jun 23, 2004 totaling 174 d ays with data. Irrigation of the turfgrass pasture was accomplished using a linear move system with an irrigation schedule determined by farm management. Irrigation and rainfall were both measured using an Onset Computer Corporation (Bourne, MA) tipping buck et rain gauge. The gauge utilized a Hoboware data logger that recorded each instance when 2.54 mm was collected in the tipping bucket. The rain gauge was mounted at a low height to catch both rainfall and irrigation thus these events were not distinguish able in this dataset. Weather data during this time period was collected from the Florida Automated Weather Network (FAWN) weather station located at the PSREU. Weather parameters included temperature, relative humidity, solar radiation, wind speed, and rainfall collected at 15 minute intervals. Daily and hourly reference evapotranspiration (ET O ) was calculated using the ASCE standardized ET O equation (ASCE 2005). The soil type located in Block 4 at the PSREU was determined from soil survey maps to be Arredondo Fine Sand at 5 to 8 percent slopes (ArC), which is a well drained FC PWP ) of 0.05 mm 3 mm 3 to 0.17 mm 3 mm 3 and a permeability rate of 152 508 mm/h (SCS 1979). On site soil testing occurring on

PAGE 127

127 Aug. 11, 2006 showed that the soil consisted of 94% sand, 3% silt, and 3% clay with surface saturated hydraulic conductivity, related to permeability, averaging 317 mm/h. Soil testing also provided additional estimations required for Richards Equation that in r (0.062 mm 3 mm 3 s (0.329 mm 3 mm 3 1 ). GCREC Twenty St. Augustinegrass plots measuring 7.6 m by 7.9 m were established at the Gulf Coast Research and Education Center (GCREC) in Balm, FL. These plots were created t o evaluate the ability of ET controllers to schedule irrigation in southwest Florida. Results concerning these treatments can be found in Davis et al. (2009), Davis and Dukes (2010), and Davis and Dukes (2012). Each plot was maintained using an individua lized irrigation system with dedicated 114 m m V100 w/ Pulse Output flow meter (AMCO Water Metering Systems, Ocala, FL) to measure volumes of irrigation application per plot. The flow meters were connected to a CR 10X data logger (Campbell Scientific, Loga n, UT) that recorded sub daily irrigation to easily separate irrigation system maintenance activities and treatment related irrigation. The twenty plots were separated into five treatments resulting in four replications of each treatment. There were har dware issues with one of the technologies being tested, so that treatment was removed from this analysis. Therefore, the treatments were: A) Weathermatic SL1600, B) Toro Intelli Sense, C) TIME treatment based on 100% replacement of the net irrigation requ irement (Dukes and Haman 2002), and D) RTIME treatment that was 60% of the TIME treatment. The variable of interest for these treatments was irrigation application only; thus, all plot maintenance and nutrient management remained constant across all treat ments based on University of Florida

PAGE 128

128 Institute of Food and Agricultural Sciences (UF IFAS) recommendations ( S artain 1991; Black and Ruppert 1998 ). Soil moisture was measured using a TDR sensor buried diagonally at a 76 mm to 178 mm depth Each plot rece ived a sensor located in the middle of the southeast quadrant of the plot so that it was equidistant from all sprinkler heads ensuring good sprinkler application at the measurement site. Volumetric water content was collected at 15 minute intervals from Aug. 13, 200 6 to Nov. 12, 200 7 The soil type for this study was Zolfo fine sand according to the soil survey (SCS 1989). The Zolfo series is described as a sandy, siliceous, hyperthermic Grossarenic Entic Haplohumods that is somewhat poorly drained. Th e FC and PWP for Zolfo fine sand was determined from laboratory samples to be 0. 13 mm 3 mm 3 and 0.0 3 mm 3 mm 3 respectively (Carlisle et al. 1985). Model Parameterization, Calibration, and Validation Each model was parameterized to best represent the con ditions of the actual soil profile. HYDRUS 1D was used to determine the best fit of hydraulic characteristics for both studies and all layers using an inverse solution. At the PSREU, initial values and ranges of hydraulic parameters for each layer were s elected based on soil core sampling conducted at the site. The inverse solution was repeated for each layer until volumetric water content of all layers generally approximated the measured data (Table 4 2; Table 4 3; Table 4 4; Table 4 5; Table 4 6; Table 4 7). Detailed soil sampling was not conducted at the GCREC; the hydraulic characteristics were initially selected based on soil survey information and educated guesses that were also fitted using an inverse solution in HYDRUS 1D (Table 4 8; Table 4 9; T able 4 10; Table 4 11).

PAGE 129

129 The accuracy of the model s was evaluated through validation procedures using a subset of the collected data from each study V alidation of the model was completed to quantify its performance compared to the physical system being modeled. The combination of goodness of fit measures, absolute error measures, and graphical assessments provide adequate evaluation s of model perf ormance (Legates and McCabe Jr. 1999; Willmott et al. 1985; McCuen et al. 2006; Krause et al. 2005). Combini ng these three methods will result in overcoming the limitations inherent in each individual measure. The Nash Sutcliffe coefficient of eff iciency (NS E) was selected to determine the goodness of fit (Equation 1 47) The numerator of the equation measures the variation in the data that was not explained by the model and the denominator measures the total variation in the observed data (Mc Cuen et al. 2006). Dimensionless v alues of the coefficient can range from te that the model, if linear, is unbiased. According to Jeong et al. (2010), NS E values larger than 0 indicate minimally acceptable performance and greater than 0.65 on a daily basis is preferred. Ritter and Mu oz Carpena (2013) suggested NSE ranges of u nsatisfactory (<0.650), acceptable (0.650 0.799), good (0.800 0.899), and very good (0.900 1.000). However, the NSE is subjective and may be lower for sub daily results and higher for monthly or annual outputs. L imitations to the NS E include sensit ivity to sample size, outliers, magnitude bias, and time offset bias. The root mean square error (RMSE) was selected to measure the absolute error (Equation 1 49) The root mean square error, also a dimensioned model prediction error, is a function of average error, squared errors, and n 1/2 Though frequently used

PAGE 130

130 as a measure for environmental models, its meaning is often misinterpreted by res earchers (Willmott and Matsuura 2005). Squaring the residual errors allows the magnitude to affect the metric and not the positive or negative difference. However, this exacerbates the influence of large errors allowing a few large errors to increase the RMSE when there may be few small errors. T he FITEVAL software was used to determine the strength and approp riateness of the models evaluated in this paper (Ritter and Muoz Carpena 2013). This software uses a block bootstrap method for stationary dependent data (Politis and Romano 1994) resulting in a probability distribution of NSE and RMSE scores. FITEVAL w as an important tool in determining whether the single values calculated from Equation 1 47 and Equation 1 49 using the entire dataset is representative of all parts of the dataset. Additionally, this software had the ability to incorporate uncertainty ba sed on error inherent in the measured soil moisture estimated by the manufacturer of the TDR sensors as within +/ 2 .5 % by volume of true soil moisture ( Campbell Scientific 2014 ) The models were evaluated against measured soil moisture based on absolute m easurements and uncertainty ranges resulting in two sets of NSE results. Results PSREU There were a total of 70 days that received rainfall or irrigation during the 175 day study period located at the PSREU. Out of the daily totals, 38 days had depths gre ater than 2.54 mm with a maximum depth of 88 mm occurring on Feb. 24, 2004. These events translated to 167 h with water inputs across the 4,175 h period with 61 h that had water inputs greater than 2.54 mm. The maximum intensity of 3.63 mm h 1 occurred a s a single event on May 18, 2004.

PAGE 131

131 This study was separated into two sets of data for calibration and validation purposes. The calibration period was selected as the first 1,000 h of the study period, occurring from Jan. 1, 2004 through Feb. 11, 2004. There were 15 h of rainfall or irriga tion with only 4 h where rainfall was greater than 2.54 mm. The largest event of 35.1 mm occurred in one hour on Feb. 1, 2004. The validation period occurred from Feb. 12, 2004 through Jun. 22, 2004 that included the remaining 3,175 h of the data and 152 h of rainfall or irrigation. Calibration p eriod Out of the six layers representing observation points that occurred every 15 cm until the final depth of 90 cm, the best agreement occurred by HYDRUS 1D with NSE ranging from 0.629 to 0.855 across layers ( Table 4 12). Results were also in an acceptable range for hourly SWB in the top two layer s, with NSE calculated as 0.659 for the first layer and 0.663 for the second layer However, NSE declined for each additional layer occurring below the second layer with bias occurring in the sixth layer for hourly SWB. When matching results by layer, the RMSE for HYDRUS 1D (0.004 mm 3 mm 3 0.009 mm 3 mm 3 ) was either the same or lower than RMSE for hourly SWB (0.007 mm 3 mm 3 0.011 mm 3 mm 3 ) indicating slightly les s error when using HYDRUS 1D. Taking the measurement error of the sensor into consideration using the block bootstrapping method average NSE increased to minimums of 0.9 89 for hourly SWB and 0.9 94 for HYDRUS 1D and remained above 0.900 for all iterations of all layers indicating a cceptable model results with no bias. The daily SWAT is a specific version of a daily soil water balance that does not accommodate layers thus only the surface layer was evaluated for this dataset. The NSE for the daily SWAT wa s 0.514 for this top layer, which was below the limit of

PAGE 132

132 acceptable performance (Ritter and Muoz Carpena 2013) The RMSE was also highest of the three models, estimated as 0.008 mm 3 mm 3 indicating that the daily SWAT had the highest error. In general, both multi layer models best approximated the measured data during periods of declining VWC within the first layer while responding to rainfall or irrigation events by over predicting the increased VWC (Figure 4 1). There were two instances wh en both models responded to an increase in VWC due to a rainfall or irrigation event that was not reflected in the observed data. An over estimated response in VWC occurred by both models on Jan. 10. After this event, the hourly SWB declined at a faster rate than the measured data resulting in slight unde r estimation before the next peak occurring on Jan. 19. However, it appeared that HYDRUS 1D declined at an appropriate rate resulting in continued over estimation of VWC over these ten days. Results for the hourly SWB in the second layer (NSE = 0.663) were similar to the results in the first layer ( NSE = 0.659) whereas NSE r esults declined for the second layer from 0.756 to 0.629 for HYDRUS 1D thus falling just below the threshold for acceptable perform ance of 0.65 (Table 4 12) In general, VWC was over estimated by HYDRUS 1D over the calibration period compared to the hourly SWB that under estimated VWC at the beginning of the period but slightly overestimated VWC for the remainder of the period (Figur e 4 2). Both models responded appropriately to the three sharp increases observed in the measured VWC data with the exception of the last instance where the hourly SWB had a small change in soil moisture. Due to this discrepancy, the hourly SWB did not e xhibit the curved decline to field capacity and overestimated VWC for the remainder of the calibration period.

PAGE 133

133 Both models best approximated the measured VWC during the first 20 days of the calibration period when there was no rainfall or irrigation events indicating that ET C losses were accurately predicted for th e third layer (Figure 4 3). However, the volume of water entering the layer from the upper adjacent layer during rainfall or irrigation was under estimated for two of the events by the hourly SWB resulting in an NSE of 0.454 The HYDRUS 1D better appro ximated the infiltration of water entering the layer resulting in an NSE of 0.855. In the remaining layers, the hourly SWB maintained RMSE ranging from 0.009 mm 3 mm 3 (Layer 5) to 0. 01 1 mm 3 mm 3 (Layer 4) and NSE ranging from 0. 299 (Layer 6) to 0. 224 (Layer 5) thus performance was unacceptable (Table 4 12). There was little difference in RMSE and NSE results between the third, fourth, and fifth layers for the hourly SWB model. Similar to the r esults of the third layer, the hourly SWB best approximated the measured VWC during the first portion of the calibration period for the fourth (Figure 4 4) and fifth (Figure 4 5) layers, but increased only slightly in response to rainfall or irrigation eve nts compared to the measured VWC. The sixth layer was the only layer to exhibit signs of model bias as indicated by a negative NSE score where the error between the observed and predicted values was greater than the error in the observed values and averag e of observed values. However, the NSE results were always acceptable when taking into account measurement error with an average NSE score of 0. 9 89 for the final layer. The VWC was not influenced by rainfall and irrigation events for the hourly SWB in th e sixth layer (Figure 4 6). The model results obtained by using HYDRUS 1D ranged from 0. 0 04 mm 3 mm 3 (Layer 6) to 0. 0 06 mm 3 mm 3 (Layer 4) in RMSE and 0. 743 (Layer 5 ) to 0.849 (Layer 6)

PAGE 134

134 in NSE (Table 4 12). These results indicate that the HYDRUS 1D was co nsistent in estimating VWC at all depths evaluated as a part of this study. Considering measurement error, 100% of NSE remained above 0.9 with averages above 0. 998 for the fourth through sixth layers, respectively. Graphical results showed that the HYDRU S 1D continued to under estimate VWC during the last observed rainfall or irrigation event for all three of the lower layers. This resulted in over estimated VWC occurring after the peak in the fourth layer (Figure 4 4). In the fifth and sixth layers, VW C determined by HYDRUS 1D closely approximated measured VWC with the only deviation occurring in the final peak (Figure 4 5 ; Figure 4 6). Validation p eriod Despite good results during the calibration period, HYDRUS 1D failed to adequately predict VWC for all layers during the validation period with RMSE ranging from 0.020 mm 3 mm 3 (Layer 3 and 5) to 0.035 mm 3 mm 3 (Layer 2) and NSE ranging from 0.452 (Layer 1) to 2.703 (Layer 6) (Table 4 12). Over estimation of VWC occurred for all six layers indicatin g that the calibration results were not adequate in representing the val idation period for this model. Alternative periods were evaluated to determine if the soil hydraulic parameters could have been better approximated based on better model agreement, bu t model outcomes remained consistent to the results presented here. The RMSE for the hourly SWB model ranged from 0.012 mm 3 mm 3 (Layers 3 and 6) to 0. 01 8 mm 3 mm 3 (Layers 1 2, and 4) while the NSE ranged from 0. 729 (Layer 5) to 0. 383 (Layer 1 ), respecti vely (Table 4 12). The results of the first two layers were similar for this model, indicating consistent approximations of VWC in the top 37.5 cm. However, the NSE for all layers were below 0.65 thus resulting in unacceptable

PAGE 135

135 approximations of the field measured VWC. In the first layer, there were over estimations in peak VWC that occurr ed in response to rainfall or irrigation events (Figure 4 7). In the second layer VWC was under estimated multiple times, but did not have the large errors in peak VWC observed in the first layer (Figure 4 8). The bottom four layers did not have good agreement with the measured VWC primarily due to no response to at least six rainfall or irrigation events in layer 3 (Figure 4 9), s even events in layer 4 (Figure 4 10), five events in layer 5 (Figure 4 11), and all events in layer 6 (Figure 4 12). These results indicate that there was model error in the assumptions used to determine percolation to lower layers. Model bias occurred for HYDRUS 1D in all six layers result ing in negative values of NSE. However, t he NSE averages assuming measurement error ranged from 0. 907 (Layer 2 ) to 0. 9 54 (Layer 4 ) during this period. Though the averages fell with in Ritter and Muoz C arpena 2013 ) the probability distribution of NSE for the first and second layers had results falling within range of 0. 800 to 0.899 totaling 5.2% and 36.7% for respective layers The fifth and sixth layers had similar results with 0.3 % and 7. 4 % of results falling into the for all six layers The model results for the daily SWAT were 0.019 mm 3 mm 3 for RMSE and 0.331 for NSE. Both of these performance measu res were slightly worse than the results for the hourly SWB for the surface layer. Based on the performance thresholds proposed by Ritter and Mu oz Carpena (2013), the daily SWAT was not an adequate model for estimating VWC.

PAGE 136

136 In th is study, HYDRUS 1D was s uccessful at predicting VWC during the calibration period, but failed to have the same agreement during the validation period (Table 4 12). According to soil testing conducted in 2007 at the site, the saturated hydraulic conductivity averaged 317 mm h 1 at the 15 cm depth, 774 mm h 1 at the 30 cm depth, and 1,298 mm h 1 at the 45 cm depth. However, the inverse solution feature within HYDRUS 1D calibrated the saturated hydraulic conductivities to 5 mm h 1 5 mm h 1 and 6.2 mm h 1 for the respective laye r s with averages of 0.06 mm 3 mm 3 and 0.33 mm 3 mm 3 r estimated using the inverse solution (0.055 mm 3 mm 3 ) was appropriate for the first layer, estimations of 0.09 mm 3 mm 3 for each of the second and third layers were likely s was estimated as 0.22 mm 3 mm 3 using the inverse solution, which was likely too low. The inverse solution function in HYDRUS 1D was only able to use the information given during the calibrati on period where measured VWC had a minimum of 0.12 mm 3 mm 3 and a maximum of 0.21 mm 3 mm 3 Approximating the soil r s during the calibration period was extremely important to the prediction of VW C by HYDRUS 1D during the validation period but was unavailable for this study However, using the same calibrated parameters in the hourly SWB had less effect on predicting VWC. GCREC Over the 10,968 hours processed for this dataset, there were 447 h wi th rainfall and 85 h where rainfall was greater than 2.54 mm. The maximum rainfall intensity observed was 64 mm/h with a maximum daily rainfall of 125 mm. According to an analysis completed by Davis et al. (2009), rainfall was significantly less frequent and

PAGE 137

137 lower in magnitude for this study period resulting in 33% less cumulative rainfall compared to the historical average. There were 102 (RTIME) to 233 (Weathermatic) irrigation events logged by timestamp for all four treatments evaluated in this study The ET controller treatments tended to irrigate more than twice per week when water restrictions were removed in the middle of the study period, beginning on Feb. 26, 2007. Due to the differences in the ways that the Weathermatic and Toro controllers s chedule irrigation, the Weathermatic had 65 more irrigation events than the Toro. The TIME treatment had one additional event compared to the RTIME treatment due to rainfall occurring between the treatment runtimes that triggered the rain sensor to bypass mode. The data set was separated into two periods for calibration and validation purposes resulting in a calibration period of Aug. 25, 2007 through Nov. 12, 2007 and validation period of Aug. 13, 2006 through Aug. 24, 2007. There were 112 rainfall events and 5 4 irrigation events during the 1,920 h calibration period with one rainfall event lasting 10 h that totaled 85 mm on Oct. 5, 2007. The remaining rainfall (335 h) and irrigation (84 179) events occurred during the validation period. Additionally, two 1 mo nth periods within the validation period were selected as having frequent rainfall (Sep. 1, 2006 Oct. 1, 2006) and infrequent rainfall (Apr. 1, 2007 May 1, 2007) for further graphical evaluation of the models. Calibration p eriod Results from the cali bration period indicate acceptable agreement with measured VWC for all four treatments using all three models. The RMSE ranged from 0.020 mm 3 mm 3 (RTIME) to 0.024 mm 3 mm 3 (Toro, TIME) for the daily SWAT model, 0.017 mm 3 mm 3 (Weathermatic) to 0.022 mm 3 mm 3 (TIME) for the hourly SWB model, and 0.015

PAGE 138

138 mm 3 mm 3 (Weathermatic) to 0.019 mm 3 mm 3 (TIME) using HYDRUS 1D (Table 4 13). The NSE ranged from 0.124 (TIME) to 0.413 (RTIME) for the daily SWAT model, 0.182 (TIME) to 0.310 (Toro) for the hourly SWB model, and 0.384 (Weathermatic) to 0.537 (RTIME) for HYDRUS 1D. With few exceptions, results improved by using the hourly SWB over the daily SWAT model, but the best results were achieved by HYDRUS 1D. Taking into account measurement error of the sensors, average NSE increased for all three models. In general, the daily SWAT model had a lower av erage, ranging from 0.965 (TIME) to 0.978 (Weathermatic, RTIME), than the hourly SWB model, ranging from 0.971 (TIME) to 0.986 (Weathermatic). The highest scores within each treatment occurred by the HYDRUS 1D model, ranging from 0.979 (TIME) to 0.989 (We athermatic). Maximum and minimum scores were obtained by the same treatments for each model, indicating that there was better calibration of the Weathermatic treatment than the TIME treatment likely due to soil profile characteristics that were not accura tely represented in the models. The results showed that the hourly SWB and HYDRUS 1D generally followed the fluctuations in VWC over time while the daily SWAT model maintained the average of the VWC curve without responding to sub daily fluctuations (Figu re 4 13, Figure 4 14, Figure 4 15, Figure 4 16). Additionally, the best agreement by the hourly SWB and HYDRUS 1D occurred when there were small increases in VWC. Large rainfall and irrigation events drained slower than predicted by both models with more error in the hourly SWB than HYDRUS 1D, which was true for all four treatments and clearly observed in the rainfall event on Oct. 5. The VWC frequently exceeded field capacity

PAGE 139

139 resulting in increased error for the daily SWAT model compared to the sub dail y models. Validation p eriod The RMSE for all four treatments ranged from 0.017 mm 3 mm 3 (Weathermatic, RTIME) to 0.021 mm 3 mm 3 (Toro) for the daily SWAT model, 0.016 mm 3 mm 3 (TIME) to 0.022 mm 3 mm 3 (Toro) for the hourly SWAT model, and 0.013 mm 3 mm 3 ( Weathermatic) to 0.018 mm 3 mm 3 (Toro) for HYDRUS 1D (Table 4 13). The NSE ranged from 0.011 (Weathermatic) to 0.490 (RTIME) for the daily SWAT model, 0.224 (Weathermatic) to 0.569 (TIME) for the hourly SWB model, and 0.504 (Toro) to 0.655 (TIME) for HYDR US 1D. The NSE within each treatment improved with the increasing complexity of the model with lowest scores obtained by the daily SWAT model and highest sc ores obtained by HYDRUS 1D. However, the only treatments with acceptable performance occurred for HYDRUS 1D with the TIME and RTIME treatments only, barely exceeding the threshold of 0.65. The average NSE increased for all treatments and models when the measurement error was taken into account (Table 4 13). The NSE ranged from 0.971 (Toro) to 0.982 (Weathermatic) for the daily SWAT model, 0.964 (RTIME) to 0.984 (Weathermatic, TIME) for the hourly SWB, and 0.981 (Toro) to 0.991 (Weathermatic) for HYDRUS 1D. The probability distribution of generated scores using the block bootstrapping technique fell indicating that the performance was stable and not bias. During the month selected as the frequent rainfall period, there were a total of 15 rainfall events with two large rainfall events occurring on Sep. 3, totaling 34 mm, and Sep. 20, totaling 26 mm. In general, good approximation of the VWC draw down curve

PAGE 140

140 occurred by the hourly SWB for the first large rainfall event for all four treatments, but there was significantly more error in the second rai nfall event (Figure 4 17A, Figure 4 18A, Figure 4 19A, Figure 4 20A). An exception occurred for the TIME treatment where the drawdown occurred more quickly in the model compared to the measured VWC during the first rainfall event with general underestimat ion of VWC over the month. The HYDRUS 1D model also underestimated measured VWC over the month, including both large rainfall events, for the TIME treatment. Overestimation of measured VWC occurred for the Weathermatic by both sub daily models with the best approximations occurring for the Toro and RTIME treatments. The daily SWAT model responded to the frequent rainfall events, but failed to increase above field capacity resulting in a moving average of measured VWC. Irrigation was the primary water input to the soil profile during the infrequent rainfall period. The Weathermatic was designed to accumulate deficits until the next allowable watering day, which was set to every day, thus resulting in daily irrigation events for this controller (Figure 4 17B). Due to daily irrigation, runtimes were short resulting in small increases in measured VWC. The daily SWAT model rarely fluctuated at all except during periods of drawdown occurring after a rainfall event whereas the hourly SWB and HYDRUS 1D model s better approximated the sub daily fluctuations in VWC. The Toro also irrigated frequently with shorter runtimes, but irrigated every other day (Figure 4 18B). Irrigation was best approximated by the HYDRUS 1D model due to periods of large underestimati ons by the hourly SWB. The time based treatments, TIME (Figure 4 19B) and RTIME (Figure 4 20B), applied the net irrigation requirement based on historical ET C and rainfall in twice per

PAGE 141

141 week irrigation events to represent local water restrictions. The ra in sensor associated with the TIME and RTIME treatments stopped working on Apr. 19, 2007 causing all irrigation to be bypassed despite many days without rainfall. During this time, measured VWC decreased significantly for both treatments, almost approxima PWP The hourly SWB model also declined with lower predictions than were measured by a maximum difference of 2.3% (RTIME) whereas HYDRUS 1D over predicted VWC by a difference of 2.9% (TIME). This indicates that the stress coefficient added to the E T C MAD was an appropriate estimation of the increasing difficulty of removing water from the soil system as VWC decreases. The daily SWAT model failed MAD that occurred between irrigation events. The hourly SWB showed promising modeling outcomes compared to the daily SWAT model, but there were sources of error where VWC was either under or over predicted in all evaluated treatments resulting in neither model meeting acceptable performance standards This indic ates that some of the soil water interactions were not fully accounted for in the hourly SWB model. Some interactions could include lateral water movement, hysteresis, or a shallow water table. The HYDRUS 1D model also did not account for all soil water interactions, but had better results during the GCREC field study due to better approximations of water movement throughout the soil profile By design, the daily SWAT model was unable to simulate changes in VWC at the sub daily scale or account for VWC ab FC PWP These assumptions may be acceptable for current ET controller testing because ET controllers apply irrigation in short, frequent irrigation events to maintain well watered conditions. However, maintaining well watered conditions i s unnecessary for residential landscapes

PAGE 142

142 where maximum growth is not required. Thus, moving to a similar, more accurate model for ET controller testing that can account for the full range of water movement in the soil profile may encourage ET controller d esigns to allow less than well watered conditions that promotes water conservation. The NSE results during the calibration periods were higher for HYDRUS 1D compared to the hourly SWB for both studies. The major drawback to the soil water balance approach included significantly increased errors in layers that were deeper than 37.5 cm due to inaccurate estimations of infiltration causing compounded error in runoff and deep percolation estimates. Richards equation (used by HYDRUS 1D) or modified Green Ampt with redistribution would have been more appropriate models that have more accurate estimations of VWC (Gowdish and Muoz Carpena 2009, Smith et al. 1993, and Hopmans 2009, Ogden and Saghafian 1997). However, the benefits to the hourly SWB such as less parameterization, shorter processing times, and stable, well known software combined with NSE from the hourly SWB that was better than the current model outweigh the drawbacks to NSE that was lower than HYDRUS 1D. Additionally, s to evaluate performance over a minimum of thirty days thus VWC estimations over longer time periods were more important to this analysis than VWC at specific points in time. Thus, the hourly SWB would be preferable to HYDRUS 1D in association with longe r time periods of relative soil moisture whereas HYDRUS 1D would be more appropriate for short time periods when absolute soil moisture is important. The purpose of the SWAT test was to generate performance scores based on scheduling efficiency and irrigat ion adequacy where 100% scores are achieved when

PAGE 143

143 there is no over and under irrigation, respectively. The performance scores achieved when using the daily SWAT test fail to translate to real world performance due to the inability to represent the real sy stem. The results from the analyses for the hourly SWB model indicate that the performance scores would likely be more accurate in the translation of performance scores to water savings. This relationship will be evaluated in Chapter 5. C onclusions The h ourly soil water balance had better agreement with the deterministic soil water model compared to a daily soil water balance This was shown in the first layer of the PSREU study during the calibration period where the NSE was 0.659 for the hourly SWB but only 0.514 for the daily SWAT, falling below the acceptable threshold for hydrological models of 0.65. Both models failed to meet the threshold during the validation period or during either period of the GCREC study. The hourly SWB had higher NSE result s for the ET controller treatments, ranging from 0.224 to 0.266, during the validation period of the GCREC study compared to the daily SWAT, ranging from 0.011 to 0.181. Based on the results, the hourly SWB improv ed estimates of volumetric water content i n a shallow root zone compared to the current testing standard T he deterministic model required extensive parameterization of soil water characteristics that did not always translate to outside of the calibration period whereas the hourly soil water balan ce was more consistent in estimating volumetric water content using the same estimated parameters This was clearly shown in the PSREU study where NSE results dropped for both models from the calibration period to the validation period, but NSE fell from 0.659 to 0.383 for the first layer whereas NSE fell from 0.756 to 0.452 for HYDRUS 1D in the same layer. Though the hourly SWB was

PAGE 144

144 not acceptable as a hydrological model, there was bias in using HYDRUS 1D during the validation period resulting in extreme ly poor agreement with field measured VWC. T he hourly soil water balance had its best agreement with field measured data within the top 37.5 cm of the soil profile but results declined after the second layer due to accumulating errors as the layer depth increased. This was clear in the PSREU study during the calibration period where acceptable NSE results were 0.659 (layer 1) and 0.663 (layer 2), but increasingly declined in the remaining four layers. The same trend occurred during the validation period where the NSE was 0.383 (layer 1) and 0.358 (layer 2), but was negative for layers 3 5 and was only slightly above zero for layer 6. Thus, using the hourly SWB in layered scenarios greater than two layers is unadvisable without improvements to the model. Though the daily soil water balance is an accepted method for estimating irrigation requirements, it does not accurately reflect the actual water movement in the s oil profile that occurs on a sub daily basis Based on these results, there is confidence in implementing the hourly soil water balance as an alternative method for testing ET controllers that will more accurately reflect true soil water relationships thu s allowing for the estimation of water conservation potential.

PAGE 145

145 Table 4 1. Crop coefficients used for each study location adapted from Jia et al. (2009). Month North Florida Crop Coefficient Southwest Florida Crop Coefficient January 0.35 0.45 February 0.35 0.45 March 0.55 0.55 April 0.80 0.80 May 0.90 0.90 June 0.75 0.75 July 0.70 0.70 August 0.70 0.70 September 0.75 0.75 October 0.70 0.70 November 0.60 0.60 December 0.45 0.45 Table 4 2. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 1 (0 22.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 55 0.0 55 Field Capacity FC (mm 3 mm 3 ) 0.15 NA Saturation S (mm 3 mm 3 ) 0.22 0.22 Root depth RZ (mm) 225 220 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 1.93 Van Genuchten parameter 1 ) NA 0.0 0 1 4 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 5 Note: NA indicates that the parameter was not applicable to the model.

PAGE 146

146 Table 4 3. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 2 (22.5 37.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 9 0.0 9 Field Capacity FC (mm 3 mm 3 ) 0.1 5 NA Saturation S (mm 3 mm 3 ) 0.2 2 0.2 2 Root depth RZ (mm) 150 150 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 2.99 Van Genuchten parameter 1 ) NA 0.0 0 1 1 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 5 Note: NA indicates that the parameter was not applicable to the model. Table 4 4. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 3 (37.5 52.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 9 0.0 9 Field Capacity FC (mm 3 mm 3 ) 0.12 NA Saturation S (mm 3 mm 3 ) 0.2 1 0.2 1 Root depth RZ (mm) 150 150 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 3.67 Van Genuchten parameter 1 ) NA 0.0 0 1 6 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 6.2 Note: NA indicates that the parameter was not applicable to the model.

PAGE 147

147 Table 4 5. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 4 (52.5 67.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 7 8 0.0 7 8 Field Capacity FC (mm 3 mm 3 ) 0.1 6 NA Saturation S (mm 3 mm 3 ) 0.2 4 0.2 4 Root depth RZ (mm) 150 150 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 1. 58 Van Genuchten parameter 1 ) NA 0.0 021 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 11.2 Note: NA indicates that the parameter was not applicable to the model. Table 4 6. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 5 (67.5 82.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 9 0.0 9 Field Capacity FC (mm 3 mm 3 ) 0. 20 NA Saturation S (mm 3 mm 3 ) 0. 23 0. 23 Root depth RZ (mm) 150 150 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 2. 48 Van Genuchten parameter 1 ) NA 0.0 0 1 3 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 16 Note: NA indicates that the parameter was not applicable to the model Table 4 7. Hydraulic parameters required by the hourly SWB and HYDRUS 1D models for Layer 6 (82.5 97.5 cm) at the PSREU location. Characteristic Abbreviation (units) Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) 0.0 9 0.0 9 Field Capacity FC (mm 3 mm 3 ) 0. 20 NA Saturation S (mm 3 mm 3 ) 0.2 2 0.2 2 Root depth RZ (mm) 150 150 Allowable Surface Accumulation ASA (mm) 0 0 Decay Time t (h) 48 NA Van Genuchten parameter n NA 2. 56 Van Genuchten parameter 1 ) NA 0.0 0 1 Saturated Hydraulic Conductivity K s (mm h 1 ) NA 17.8 Note: NA indicates that the parameter was not applicable to the model.

PAGE 148

148 Table 4 8. Hydraulic parameters required for each model for the Weathermatic at the GCREC location. Characteristic Abbreviation (units) Daily SWAT Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) NA 0.0665 0.0665 Field Capacity FC (mm 3 mm 3 ) 0.17 0.165 NA Saturation S (mm 3 mm 3 ) NA 0.480 0.480 Root depth RZ (mm) 305 305 305 Allowable Surface Accumulation ASA (mm) 0 0 0 Decay Time t (h) NA 48 NA Van Genuchten parameter n NA NA 1.925 Van Genuchten parameter 1 ) NA NA 0.0192 Saturated Hydraulic Conductivity K s (mm h 1 ) NA NA 91.2 Note: NA indicates that the parameter was not applicable to the model. Table 4 9. Hydraulic parameters required for each model for the Toro at the GCREC location. Characteristic Abbreviation (units) Daily SWAT Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) NA 0.0414 0.0414 Field Capacity FC (mm 3 mm 3 ) 0.17 0.164 NA Saturation S (mm 3 mm 3 ) NA 0.443 0.443 Root depth RZ (mm) 305 305 305 Allowable Surface Accumulation ASA (mm) 0 0 0 Decay Time t (h) NA 48 NA Van Genuchten parameter n NA NA 2.161 Van Genuchten parameter 1 ) NA NA 0.0105 Saturated Hydraulic Conductivity K s (mm h 1 ) NA NA 12.9 Note: NA indicates that the parameter was not applicable to the model.

PAGE 149

149 Table 4 10. Hydraulic parameters required for each model for TIME at the GCREC location. Characteristic Abbreviation (units) Daily SWAT Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) NA 0.0613 0.0613 Field Capacity FC (mm 3 mm 3 ) 0.17 0.161 NA Saturation S (mm 3 mm 3 ) NA 0.376 0.376 Root depth RZ (mm) 305 305 305 Allowable Surface Accumulation ASA (mm) 0 0 0 Decay Time t (h) NA 48 NA Van Genuchten parameter n NA NA 1.776 Van Genuchten parameter 1 ) NA NA 0.0101 Saturated Hydraulic Conductivity K s (mm h 1 ) NA NA 70.4 Note: NA indicates that the parameter was not applicable to the model. Table 4 11. Hydraulic parameters required for each model for RTIME at the GCREC location Characteristic Abbreviation (units) Daily SWAT Hourly SWB HYDRUS 1D Permanent Wilting Point PWP (mm 3 mm 3 ) NA 0.0491 0.0491 Field Capacity FC (mm 3 mm 3 ) 0.17 0.167 NA Saturation S (mm 3 mm 3 ) NA 0.493 0. 493 Root depth RZ (mm) 305 305 305 Allowable Surface Accumulation ASA (mm) 0 0 0 Decay Time t (h) NA 48 NA Van Genuchten parameter n NA NA 2.319 Van Genuchten parameter 1 ) NA NA 0.0100 Saturated Hydraulic Conductivity K s (mm h 1 ) NA NA 17.6 Note: NA indicates that the parameter was not applicable to the model.

PAGE 150

150 Table 4 12. Evaluation measures of RMSE and NSE used to evaluate the model results during the calibration and validation periods for each layer in the PSREU study. Layer Depth (cm) Sensor Dep th (cm) Daily SWAT a Hourly SWB HYDRUS 1D Calibration Period RMSE (mm 3 mm 3 ) 0 22.5 15 0.008 0.007 0. 006 22.5 37.5 30 NA 0.009 0. 009 37.5 52.5 45 NA 0. 007 0. 004 52.5 67.5 60 NA 0. 011 0. 006 67.5 82.5 75 NA 0.009 0. 005 82.5 97.5 90 NA 0. 010 0. 004 Calibration Period NSE b 0 22.5 15 0.514 ( 0.995) 0. 659 (0.996) 0. 756 (0. 997 ) 22.5 37.5 30 NA 0. 663 (0. 995 ) 0. 629 (0. 994 ) 37.5 52.5 45 NA 0. 454 (0. 994 ) 0. 855 (0. 998 ) 52.5 67.5 60 NA 0. 224 (0. 992 ) 0. 811 (0.998 ) 67.5 82.5 75 NA 0. 171 (0. 993 ) 0. 743 (0.998 ) 82.5 97.5 90 NA 0. 229 (0. 989 ) 0. 849 (0. 999 ) Validation Period RMSE (mm 3 mm 3 ) 0 22.5 15 0.019 0. 018 0. 027 22.5 37.5 30 NA 0. 018 0. 035 37.5 52.5 45 NA 0.012 0. 020 52.5 67.5 60 NA 0. 018 0. 025 67.5 82.5 75 NA 0. 017 0. 020 82.5 97.5 90 NA 0.012 0. 024 Validation Period NSE 0 22.5 15 0.331 (0.969) 0. 383 (0. 972 ) 0. 452 (0.933 ) 22.5 37.5 30 NA 0. 358 (0. 975 ) 1. 377 (0. 907 ) 37.5 52.5 45 NA 0. 040 (0. 981 ) 2.108 (0. 944 ) 52.5 67.5 60 NA 0. 369 (0. 977 ) 1.730 (0. 954 ) 67.5 82.5 75 NA 0. 729 (0. 967 ) 1.509 (0. 952 ) 82.5 97.5 90 NA 0. 051 (0. 983 ) 2.703 (0. 933 ) a The daily SWAT is not capable of calculating layers thus all layers after the initial layer were not applicable (NA). b Values in parentheses were averages determined using FITEVAL software that calculated the probability distribution of scores while incorporating the +/ 2% measurement error inherent in the soil moisture sensors.

PAGE 151

151 Table 4 13. Evaluation measures of RMSE and NSE used to evalua te the model results during the calibration and validation periods for each treatment in the GCREC field study. Treatment Daily SWAT Hourly SWB HYDRUS 1D Calibration Period RMSE (mm 3 mm 3 ) Weathermatic 0.021 0.017 0.015 Toro 0.024 0.021 0.017 TIME 0.024 0.022 0.019 RTIME 0.020 0.020 0.016 Calibration Period NSE Weathermatic 0.165 (0.978) 0.221 (0.986) 0.384 (0.989) Toro 0.294 (0.967) 0.310 (0.976) 0.520 (0.984) TIME 0.124 (0.965) 0.182 (0.971) 0.403 (0.979) RTIME 0.413 (0.978) 0.299 (0.978) 0.537 (0.986) Validation Period RMSE (mm 3 mm 3 ) Weathermatic 0.017 0.017 0.013 Toro 0.021 0.022 0.018 TIME 0.018 0.016 0.014 RTIME 0.017 0.023 0.014 Validation Period NSE Weathermatic 0.011 (0.982) 0.224 (0.984) 0.553 (0.991) Toro 0.181 (0.971) 0.266 (0.972) 0.504 (0.981) TIME 0.361 (0.978) 0.569 (0.984) 0.655 (0.987) RTIME 0.490 (0.981) 0.077 (0.964) 0.651 (0.986) Note: Values in parentheses were averages determined using FITEVAL software that calculated the probability distribution of scores while incorporating the +/ 2% measurement error inherent in the soil moisture sensors.

PAGE 152

152 Figure 4 1. Time series comparison of volumetric water content measured at the 15 cm depth to three models in Citra, FL during the calibration period. Figure 4 2. Time series comparison of volumetric water content measured at the 30 cm depth to two models in Citra, FL during the calibration period.

PAGE 153

153 Figure 4 3. Time series comparison of volumetric water content measured at the 45 cm depth to two models in Citra, FL during the calibration period. Figure 4 4. Time series comparison of volumetric water content measured at the 60 cm depth to two models in Citra, FL during the calibration period.

PAGE 154

154 Figure 4 5. Time series comparison of volumetric water co ntent measured at the 75 cm depth to two models in Citra, FL during the calibration period. Figure 4 6. Time series comparison of volumetric water content measured at the 90 cm depth to two models in Citra, FL during the calibration period.

PAGE 155

155 Figure 4 7. Time series comparison of volumetric water content measured at the 15 cm depth to the three models in Citra, FL during the validation period. Figure 4 8. Time series comparison of volumetric water content measured at the 30 cm depth to the two mo dels in Citra, FL during the validation period.

PAGE 156

156 Figure 4 9. Time series comparison of volumetric water content measured at the 45 cm depth to the two models in Citra, FL during the validation period. Figure 4 10. Time series comparison of volumetr ic water content measured at the 60 cm depth to the two models in Citra, FL during the validation period.

PAGE 157

157 Figure 4 11. Time series comparison of volumetric water content meas ured at the 75 cm depth to the two models in Citra, FL during the validation period. Figure 4 12. Time series comparison of volumetric water content measured at the 90 cm depth to the two models in Citra, FL during the validation period.

PAGE 158

158 Figure 4 13. Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for the Weathermatic during the calibration period. Figure 4 14. Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for the Toro during the calibration period.

PAGE 159

159 Figure 4 15. Time seri es results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for TIME during the calibration period. Figure 4 16. Time series results using the hourly SWB, daily SWAT, and HYDRUS 1 D models for RTIME during the calibration period.

PAGE 160

160 Figure 4 17. The volumetric water content was estimated using all three models for the Weathermatic during the validation period at the GCREC during a month with A) frequent rainfall and B) infrequent rainfall.

PAGE 161

161 Figure 4 18. The volumetric water content was estimated using all three models for the Toro during the validation period at the GCREC during a month with A) frequent rainfall and B) infrequent rainfall.

PAGE 162

162 Figure 4 19. The volumetric water content was est imated using all three models for TIME during the validation period at the GCREC during a month with A) frequent rainfall and B) infrequent rainfall

PAGE 163

163 Figure 4 20. The volumetric water content was estimated using all three models for RTIME during the validation period at the GCREC during a month with A) frequent rainfall and B) infrequent rainfall.

PAGE 164

164 CHAPTER 5 SIMULATED PERFORMANCE OF E VAPOTRANSPIRATION BASED IRRIGATION CONTROLLERS USING SOIL WATER BALANCE MODEL S I ntroduction Models are used to solve problems in the real world when actual data cannot be collected. In soil water modeling, mathematical representations are used to simulate hydrological processes in the variably saturated shallow soil profile. Models can be descr ibed in a number of ways such as physically based vs. empirically based, temporal vs. spatial, or static vs. dynamic. The appropriate model should be created or selected based on the purpose of the model while factoring the types of available input data an d ability to produce meaningful outcomes. Complex models require extensive input data provided by the user, numerical approximation techniques, and intense computations compared to simple models that are genera lly easier to use (Jeong et al. 2010). Howev er, simple models sometimes lack key biological and hydrological processes making the model l ess realistic (Ranatunga et al. 2008). Thus, it is important to find a good balance between simplistic models that lack computational intensity and complex models that better approximate the real system. One modern technology currently promoted for water conservation is the weather based irrigation controller that uses reference evapotranspiration (ET O ) to estimate plant water needs. For this reason, weather based irrigation controllers are commonly called ET controllers. Though they vary in functionality by brand and model, most ET controllers use a soil water balance to determine when and how much irrigation is needed Research h as shown that ET controllers have the potential to produce water savings when implemented in scenarios with frequent excessive irrigation (Mayer et al. 2009; Davis and Dukes, in review). They have also been shown

PAGE 165

165 to increase irrigation at homes that histo rically irrigate at or below the theoretical irrigation requirement (Mayer et al. 2009; Davis and Dukes 2014 ). The Irrigation Association developed the Smart Water Application Technologies (SWAT) Testing Protocol for Weather Based Irrigation Controllers as a method to evaluate the ability of ET controllers to adequately and efficiently schedule irrigation (IA 2008). This testing protocol established in 2002 and in its 8th draft specifies a 30 day test with baseline totals of 63.5 mm of ET O and 10.2 mm of rainfall to determine if ET controllers irrigate efficiently. Publishing SWAT test results are at the discretion of the manufacturer and can be continually tested until achieving acceptable scores during a valid testing period. In June 2006, the United States Environmental Protection Agency (USEPA) started the WaterSense program to certify products as water saving devices by incorporating a WaterSense label on product packaging By Jan 2007, the first device labeled as part of this program was the hig h efficiency toilet that use d less than 4 9 L per flush with other products following such as faucets, showerheads, and urinals. The first outdoor product eligible for the WaterSense label w as the ET controller with the final specification released on Nov 3 2011 and the first device officially labeled in May 2012 (USEPA 2012). This specification was developed by adopting the Irrigation with minor modifications Instead of publishing test scores as the Irrigation Assoc iation does, the USEPA chose minimums of 80% for irrigation adequacy and 95% for scheduling efficiency as the score requirements for WaterSense labeling and publishes the controller information when they meet or exceed these scores.

PAGE 166

166 In 2009 and 2010, the SWAT test was evaluated for transferability, or applicability of scores to different areas of the country, and reproducibility, or the ability to obtain identical results using the same input information by various testing labs (Davis and Dukes 2010) It was determined that t he results of the SWAT test w ere not transferrable across the United States due to decreas ing scheduling efficiency scores during periods of increased rainfall frequency (Davis and Dukes 2012) Other significant deficiencies included the low rainfall and ET O requirements for a valid test, penalized scores due to the mandate that rainfall must be taken into account before irrigation despite when irrigation occurred prior to rainfall on the same day, and inconsistencies in controller programming compared to the oretical landscape characteristics to achieve high scores. Based on these deficiencies, it was hypothesized that an hourly soil water balance would better match actual soil water content than the daily soil water balance specified by the SWAT test thus t ranslating efficacy results into better predictions of real world performance. The objective of this research was to compare the performance results of ET controllers, quantified as irrigation adequacy and scheduling efficiency, using an hourly soil water balance and the SWAT test. M aterials and Methods Soil Water Balance to total volume (water, air, and soil particles) of the root zone. In a natural landscape r ), where additional energy is s ), where all pore spaces are FC ) is achieved when gravitational forces equal

PAGE 167

167 surface tensi on forces (cohesive and adhesive) between water molecules and soil particles resulting in no drainage due to gravity (Richards 1931). Water storage at FC s FC over time based on soil hydraulic prope rties. To reduce water stress on the plant material, the water r MAD ) r FC (IA 2013). The soil water balance is a simple model that uses a ti pping bucket approach to soil water movement in a homogeneous soil profile (i.e. depth per unit area of root zone) over a specified timestep (Ranatunga et al. 2008). The model is based on the principle that soil water at the current depth (Z) and timestep (i) was affected by the soil water at the previous timestep (i 1) and inputs and outputs of crop evapotranspiration (ET C ), rainfall (R), irrigation (I), runoff (RO), and deep percolation (D) occurring at i (Equation 4 1) The ET C (Equation 1 28) was cal culated from ET O (Equation 1 13) determined from local microclimate variables and crop coefficients specific to the studies (Table 5 1). Rainfall and irrigation we re utilized similarly in the model where the depth equals the total that fell (R) or was app lied (I) over the unit area during the timestep of interest. an upper limit, typically FC or s result in separating total depth into effective depth, which is considered available for plant use, and a combination of runoff less than the upper limit total depth equals effective depth. Runoff and deep percolation FC resulting in water losses from the soil profile. Generally, runoff occurs due to improper

PAGE 168

168 cycle/soak scheduling of irrigation and landscape inclines whereas deep percolation occurs due to excess rainfall or irrigation. It is difficult and unnecessary to separate these losses because they are quantified outside of the soil profile and the variable of interest is the change in depth of water within the soil profile. Daily SWAT Model The SWAT test (IA 2008) uses a modified daily soil water balance to simul ate the soil water dynamics of a landscape plant system. It was developed as a n industry tool to evaluate the irrigation scheduling capabilities of ET controllers. Test results are considered valid if meeting the 30 day baseline totals of 63.5 mm and 10. 2 mm for ET O and R, respectively. Publishing SWAT test results to the public are at the discretion of the manufacturer and can be continually tested until achieving acceptable scores during a valid testing period. The SWAT protocol specifies six diverse virtual landscapes that are primarily described by plant material, sprinkler type, soil type, and slope (IA 2008). The soil water balance specified by the SWAT test was modified to restrict water MAD FC The range is referred to as the root zone working water storage (RZWWS) (IA 2008). In addition to new terminology, water level fluctuations were limited to zero and RZWWS, or the maximum value able to be stored depending on the tested zone. A water level of zero does not equate to the absence of water in the soil column. Therefore, the soil water balance used for the SWAT test does not represent the actual water level in the soil profile, but the level under which the plant material will no longer be considered well watered. An additional modification included reducing total daily rainfall by 20% citing losses to non uniformity and runoff (IA 2008). In the SWAT test, runoff and deep percolation were defined as scheduling losses consisting of direct runoff when irrigation cycle s exceed maximum runtimes based on

PAGE 169

169 infiltration rates, soak runoff when irrigation cycles occur prior to minimum soak times FC The performance results of ET controllers evaluated by the SWAT test are presented as scheduling efficiency (E) and irrigation adequacy (A). Scheduling efficiency is a measure of over irrigation quantified by estimating scheduling losses over a 30 day period combined from three so urces: soil moisture rising above field capacity, too long cycle times, and too short soak times (Equation 1 52) The I N (mm) is net irrigation and SL (mm) is scheduling losses, both summed over thirty days, resulting in a single E (%) for each 30 day per iod. Similarly, irrigation adequacy is a measure of under irrigation and can be quantified by depths when conditions were not considered well watered over the 30 day period (Equation 1 53) The ET C (mm) and D (mm) as the amount of deficit were summed ove r thirty days, resulting in a single A (%) for each 30 day period. Deficit is calculated as the cumulative depth below well watered conditions, MAD Results of 100% for either score would indicate that the irrigation scheduling performanc e by the ET controller was effectively perfect. In addition to evaluating irrigation schedules, the model can be used to predict the gross irrigation requirement (GIR) by implementing a minor change in the irrigation reference. Instead of inputting I N into the model directly, it is calculated as the depth of FC MAD is reached. Hourly SWB Model An hourly soil water balance was developed in efforts to address some deficiencies in the SWAT test such as lack of transferabil ity and penalized scores due to mandating that rainfall occurred before irrigation despite actual order of events. All

PAGE 170

170 calculations specified in the daily version were maintained so that soil water movement within the RZWWS occurred in the same way. Howe ver, the soil water balance was FC MAD These simplifications may benefit the controllers during the test by not continually penalizing for deficit or surplus conditions. However, soil water le vels in real systems can FC until reaching saturation with deep percolation naturally occurring FC Additionally, water levels MAD resulting in deficit conditions ca n occur in real systems and should be observed in the model. To account for soil water levels higher than field capacity, an exponential decay function (Equation 4 2 ) was incorporated into each timestep used to represent drainage to equilibrium. When d eficit conditions occur and soil moisture falls below MAD the ability for ET C to be removed from the soil profile by the plant material becomes more difficult. To account for this phenomenon, a stress coefficient ( Equation 4 4 ) was introduced to the ET C calculation Data Description Bench t esting s tudy The Weathermatic SL1600 (Garland, TX), Toro Intelli Sense (Riverside, CA), and ET Water Smart Controller 100 (Novato, CA) were installed at the University of Florida Agricultural and Biological Engineering turfgrass research facility in Gainesville, FL. The Toro and ET Water controllers were signal based where they received daily ET O information through cellular technology. These two controllers were d uplicate d so th at one controller of each brand utilized an additional Mini Clik rain sensor (Hunter Industries, Inc., San Marcos, CA) set at a 6 mm threshold denoted as with rain sensor,

PAGE 171

171 while the remaining controllers relied on product abilities alone, denoted as witho ut rain sensor The Weathermatic was a standalone controller that calculated ET O using temperature information collected on site These controller models were selected based on their previous SWAT testing by the Center for Irrigation Technology (CIT) in Fresno, California as well as previous testing conducted by our research group (Davis et al. 2009; Davis and Dukes 2010) Results from the 2010 Florida SWAT Test using three controllers of each brand showed that there was little variability between repli cations (Davis et al. 2009). Thus, performance results of these brands are likely to be similar to controller s of the same brand being utilized in the real world but may not be representative of all controller brands. Based on results from Davis and Dukes (2011), the duplicate controllers performed almost identically despite the addition of a rain sensor. The three controllers with rain sensors were selected for evaluation in this study. The treatments with rain sensors were chosen over the treatmen ts without rain sensors based on results by Rutland and Dukes (2013) that suggested the addition of a rain sensor was beneficial to the functionality of an ET controller in Florida. Each controller was connected to a CR 10X datalogger (Campbell Scientific, Logan, UT) via a set of relays to record time and date at the beginning and end of each irrigation event for all active zones. Weather data were collected from a weather station located on site and managed by research personnel. Weather parameters include d temperature, relative humidity, solar radiation, and wind speed collected at 15 minute

PAGE 172

172 intervals. Values for reference evapotranspiration (ET O ) were calculated using Equation 1 13 as specified in the SWAT protocol (IA 2008). Each controller was program supervision. It was intended that the settings would duplicate the original SWAT test performed by CIT. However, since controller settings are not reported in the current SWAT protocol, we were unable t o verify identical settings to the original SWAT test for the controllers. The official test began when the manufacturer determined that the controller(s) were programmed correctly. Four 60 day datasets were selected to represent a variety of weather con ditions specified as f requent rainfall i nfrequent rainfall h igh ET O and l ow ET O ( Table 5 2) Each period was selected by comparing the rainfall and ET O amount over the 60 day period with the historical averages over the same sixty days determined from 37 years of Gainesville Regional Airport weather data from 1970 through 2006 (NCDC 2007). Field p lot s tudy Sixteen St. Augustinegrass plots measuring 7.6 m by 7.9 m were established at the University of Florida Gulf Coast Research and Education Center (GCR EC) in Balm, FL. These plots were created to evaluate the ability of ET controllers to apply irrigation efficiently in southwest Florida. The treatments, replicated four time s were: A) Weathermatic SL1600, B) Toro Intelli Sense, C) TIME treatment based on 100% replacement of the net irrigation requirement, and D) RTIME treatment that was 60% of the TIME treatment. The variable of interest for these treatments was irrigation application only; thus, all plot maintenance and nutrient management remained co nstant across all treatments based on University of Florida Institute of Food and Agricultural Sciences recommendations (Sartain 1991; Black and Ruppert 1998).

PAGE 173

173 Detailed methods and results concerning these treatments can be found in Davis et al. (2009), Davis and Dukes (2010), and Davis and Dukes (2012). Each plot was maintained using an irrigation system with dedicated flow meter to measure volumes of irrigation application per plot. The flow meters were connected to a data logger that recorded sub dai ly irrigation to easily separate irrigation system maintenance activities and treatment related irrigation. Weather data during the study period ranging from Aug. 13, 2006 to Nov. 27, 2007, was collected from the Florida Automated Weather Network (FAWN) weather station located at the GCREC Weather parameters included temperature, relative humidity, solar radiation, and wind speed collected at 15 minute intervals. The ET O was calculated using the ASCE ET equation (ASCE 2005). Rainfall was also collect ed from this weather station. Statistical a nalysis Statistical trends in performance results of irrigation adequacy and scheduling efficiency were analyzed using Statistical Analytical Systems (SAS) software (Cary, NC). The general linear model procedure was used to determine significant differences in means of performance scores based on controller brand, model type, and weather conditions. Significance was determined at a 95% confidence level. The correlation procedure was used to determine if there wa s a significant dependence between irrigation application and performance results. Correlation coefficients can range from 1 (strong negative correlation) to 1 (strong positive correlation) where 0 indicates no correlation between the variables. Irrigat ion adequacy and scheduling efficiency scores are measures of under and over irrigation, respectively, and should not be affected by the opposite circumstance. As a result,

PAGE 174

174 correlations were limited to 30 day scores when irrigation was less than GIR for i rrigation adequacy and greater than GIR for scheduling efficiency. R esults Bench Testing Study Frequent r ainfall p eriod The cumulative rainfall during the frequent rainfall period total ed 269 mm occurring over 60 days (Table 5 2 ). This amount of rainfall was less than the historical average by 17%; h owever, the number of rain events during the period was similar with 28 events compared to an average of 27 events. The 30 day ET O values calculated on a daily basis ranged from 105 mm to 133 mm and on an hourly basis ranged from 102 mm to 115 mm (Figure 5 1). The ET O totals were above the minimum requirement of 63.5 mm for all 30 day periods. Rainfall was also above the minimum of 10.2 mm, ranging from 123 mm to 183 mm. Using the GIR 107 mm and 142 mm of irrigation were required according to the daily SWAT model and hourly SWAT model, respectively (Table 5 3 ) despite frequent rainfall over this period due to the small soil water holding capacity Irrigation application over the 60 day period ranged from 74 mm by Weathermatic, 31% savings from daily GIR and 48% savings from hourly GIR, to 150 mm by Toro, 40% more than daily GIR and 6% more than hourly GIR. Irrigation application by t he ET Water fell within this range, applying 13 2 mm. This controller increased irrigation by 23% more than daily GIR, but had water savings of 7% compared to the hourly GIR. In general, irrigation adequacy results reflected irrigation application by the controllers compared to the GIR. Irrigation ade quacy for the ET Water were 100% for all thirty scores using the daily SWAT model where irrigation application was higher

PAGE 175

175 than GIR, but were slightly lower and declined over time using the hourly SWB where irrigation was slightly less than GIR, ranging fro m 73% to 93% (Figure 5 2). The Toro controller maintained scores of 100% using the daily SWAT model and above 96% using the hourly SWB while consistently irrigating more than GIR using both models (Figure 5 3). The Weathermatic applied less than the GIR for both models resulting in scores ranging from 85% to 100% for the daily SWAT model and 59% to 71% for the hourly SWB model (Figure 5 4). Scheduling efficiency results were maintained at 100% for both models by the Weathermatic only. Results ranged fr om 92% to 97% for ET Water and 97% to 100% for Toro using the hourly SWB. Decreased scores for ET Water and Toro occurred when using the daily SWAT model, reaching a minimum of 87% and 92%, respectively. A decline in scheduling efficiency scores by the E T Water and Toro controllers occurred due to over irrigation of 23% and 40%, respectively, compared to GIR. Mean irrigation adequacy and scheduling efficiency scores were significantly different between the daily SWAT and hourly SWB models for each contr oller if the controller had scores less than 100% (Table 5 3). Irrigation adequacy scores decreased from 100% to 91% for the ET Water and from 88% to 65% for the Weathermatic. The opposite effect occurred for scheduling efficiency scores, showing increas es from 90% to 94% for ET Water and 95% to 98% for the Toro. As a result, there were differences in performance results attributed to the model. Infrequent r ainfall p eriod The rainfall over the infrequent rainfall period totaled 46 mm in 10 events resulti ng in 64% less rainfall than the historical average (Table 5 2) There were three 30 day periods at the end of the season that did not meet the minimum ET O

PAGE 176

176 requirement of 63.5 mm when using the daily values, ranging from 58 mm to 100 mm (Figure 5 5) How ever, the 30 day ET O using the hourly calculations met the minimum requirement during all 30 day periods ranging from 64 mm to 90 mm. Additionally, all 30 day periods met the minimum rainfall requirement of 10.2 mm, ranging from 14 mm to 27 mm. Irrigation application by the treatments ranged from 94 mm to 137 mm by the ET Water and Toro controllers, respectively (Table 5 4). The ET Water had reduced irrigation application compared to GIR using both models, resulting in 3% reduction from daily SWAT and 33% reduction from hourly SWB. The Toro and Weathermatic controllers had reductions in irrigation application compared to hourly GIR by 3% and 10%, but increased irrigation compared to daily GIR by 41% and 31%, respectively. Irrigation adequacy for ET Wat er ranged from 84% to 96%, averaging 89%, for the daily SWAT model and 34% to 79%, averaging 64%, for the hourly SWB model (Table 5 4; Figure 5 6). Though cumulative irrigation by this controller was similar to GIR calculated using the daily SWAT model, t here were many times where deficit conditions occurred resulting in lower irrigation adequacy scores across 30 day periods within the 60 day period Irrigation adequacy results for the Toro and Weathermatic controllers scored 100% for all 30 day periods using both models due to frequent irrigation (Figure 5 7; Figure 5 8). Mean scheduling efficiency scores were not different across controller brands for the daily SWAT model, but scores from the hourly SWB model were different for the ET Water and Toro controllers, averaging 92% and 94%, respectively. The ET Water controller had perfect scores using the daily SWAT model, but scores began at 87%

PAGE 177

177 and steadily increased to 100% by the end of the period using the hourly SWB model. Results for the Toro ranged from 91% to 100% whereas the Weathermatic had repeatedly perfect scores Short and frequent irrigation events resulted in relatively little over irrigation. Low ET O p eriod During t he low ET O period ET O totaled 163 mm with a daily average of 3 mm, 22% below the historical average (Table 5 2) Thirty day ET O increased over the period from 72 mm to 96 mm using the daily values and 57 mm to 81 mm for the hourly values (Figure 5 9). Due to the lower values using the hourly ET O there were seven 30 day periods that were below the minimum ET O requirement of 63.5 mm. Rainfall w as consistently greater than the minimum requirement of 10.2 mm for this period with a total of 63 mm and a range of 63 mm to 115 mm It was estimated that 76 mm and 94 mm of irrigation were required according to the daily GIR and hourly GIR, respectivel y, over this 60 day period to supplement rainfall (Table 5 5). The ET Water irrigated 11% more than daily GIR and 11% less than hourly GIR, totaling 84 mm. The Toro irrigated less than the GIR by 9% on a daily basis and 27% on an hourly basis by applying 69 mm. Results were not reported for the Weathermatic due to equipment malfunction. Irrigation adequacy results for the daily SWAT model ranged from 93% to 100% for the ET Water (Figure 5 10) and 90% to 98% for the Toro (Figure 5 11). There was no diffe rence in scores using the daily SWAT model for these two controllers (Table 5 5). Scores decreased when using the hourly SWB model with ranges of 48% to 94% for ET Water and 12% to 72% for Toro. Mean scores using the hourly SWB were different from daily S WAT model, decreasing to 75% for ET Water and 43% for Toro.

PAGE 178

178 Scheduling efficiency results for both controllers using the hourly SWB model were 100% across all 30 day periods. This was also true for Toro when using the daily SWAT model. However, ET Water had mostly less than perfect scheduling efficiency scores using the daily SWAT model with results ranging from 70% to 100% and averaging 82%, which was different from the hourly SWB model. High ET O p eriod T he highest 60 day ET O totaling 287 mm, occurred during t he high ET O period with a 5 mm daily average (Table 5 2) The total ET O was 14% less than the historical average and r ainfall was also less than its historical average by 22%, totaling 124 mm The 30 day ET O values ranged from 122 mm to 170 mm f or the daily calculations and 124 mm to 143 mm for the hourly calculations (Figure 5 12). The 30 day rainfall values ranged from 55 mm to 100 mm for this time period. Minimum requirements for the SWAT test were met for all time periods using both methods There was a high demand for irrigation during this time period due to the combination of high ET O and less than normal rainfall totals. As with all other periods, the GIR was higher for the hourly SWB model than the daily SWAT model with a total irrigat ion requirement of 231 mm for the daily GIR and 283 mm for the hourly GIR (Table 5 6). The ET Water applied more than daily GIR by 13% and less than hourly GIR by 7%, applying 262 mm. The Toro applied the most similar amount of irrigation to daily GIR, t otaling 236 mm, but was 17% less than the hourly GIR. The Weathermatic controller applied less than what was required for both models, totaling 185 mm, with results of 20% to 35% less than the daily and hourly GIR, respectively. The irrigation adequacy re sults for the ET Water and Toro controllers using the daily SWAT ranged from 91% to 99% (Figure 5 13) and 87% to 100% (Figure 5 14),

PAGE 179

179 respectively, with no difference between mean scores (Table 5 6). Results were much lower for these controllers using the hourly SWB with ranges of 18% to 36% for ET Water and 15% to 32% for Toro. These results were significantly different from their respective daily SWAT scores, but were not different from each other. Irrigation adequacy for the Weathermatic using the dail y model had a decreasing trend in scores from 100% to 75% (Figure 5 15). However, irrigation adequacy results for the hourly SWB ranged from 19% to 80% with a general upward trend. Mean scores were different for this controller, averaging 82% using the d aily SWAT model and 47% using the hourly SWB model. Scheduling efficiency results were consistently 100% for ET Water and Toro using the hourly SWB, but had respective minimums of 87% and 95% at the beginning of the period for the daily SWAT model. The me an scores for the ET Water were different due to lower scores at the beginning of the period. The Weathermatic had the opposite trend where daily SWAT scores were 100% whereas the hourly SWB scores decreased after the fourth 30 day period to 70%, also resu lting in significant differences between model types. Correlation c oefficients Correlation coefficients for irrigation adequacy scores ranged from 0.280 (Weathermatic) to 0.125 (ET Water) for the daily SWAT model and from 0.324 (Weathermatic) to 0.783 (ET Water) for the hourly SWB model (Table 5 7). For the daily SWAT model, the coefficient for the Weathermatic was significant (P value = 0.0250) indicating that irrigation adequacy scores were slightly dependent on irrigation application. Also, correlation coefficients for the hourly SWB were significant for all three controllers thus also having dependence. However, coefficients were not

PAGE 180

180 significant for the ET Water (P value = 0.4185) and Toro (P value = 0.3351) controllers using the daily SWAT model. These results indicate that there was a stronger relationship between irrigatio n application and irrigation adequacy scores for the hourly SWB model than for the daily SWAT model. For scheduling efficiency scores associated with the daily SWAT model, correlation coefficients ranged from 0.304 (Weathermatic) to 0.224 (ET Water), but none of the coefficients were significant thus there was not enough information to show a correlation. However, correlation coefficients were significant (P value < 0.0001) for the ET Water and Toro, equaling 0.518 and 0.483, respectively. These coeffici ents indicate that there was a relationship between irrigation application and scheduling efficiency scores for the hourly SWB model, but not the daily SWAT model. Since the Weathermatic did not irrigate more than the GIR, there was a lack of data to calc ulate a coefficient for this controller. Field Plot Study This study was conducted during a drought period (Davis et al. 2009) resulting in high ET O (2,475 mm) and low rainfall (1,293 mm) compared to historical averages of 1,845 mm and 1,738 mm, respecti vely (Table 5 2). There were similar numbers of rainfall events with 145 events occurring during the study period and 153 events as the historical average. The 30 day ET O ranged from 91 mm to 222 mm when calculated using daily climatological data and fro m 63 mm to 184 mm when calculated using hourly data (Figure 5 16). The 30 day rainfall totals ranged from 0 mm to 210 mm. The irrigation adequacy scores for the Weathermatic showed that the daily SWAT model predicted very little periods of deficit with two instances where the scores reached lows of 93% (mid December 2006) and 84% (mid June 2007) (Figure 5 17).

PAGE 181

181 Perfect irrigation adequacy scores were achieved by the hourly SWB model. Scheduling efficiency scores began as 100% values during the beginning of the study with fluctuating scores occurring during beginning periods of frequent rainfall occurring over winter 2006, minimum score of 81%, and summer 2007, minimum score of 48%, for the daily SWAT model. The same trends occurred for the hourly SWB mo del with declining scores during initial periods of frequent rainfall. However, the hourly SWB had lower scores overall with minimums of 60% in mid December 2006 and 28% in mid June 2007. For the Toro controller, irrigation adequacy fluctuated with sh ort periods of scores just greater than 80% according to the daily SWAT model whereas the scores for the hourly SWB approximated 100% for most of the study period (Figure 5 18). There was a period in the fall 2006 season where irrigation adequacy declined for both models, reaching a minimum score of 57% for daily SWAT and 30% for hourly SWB. Scheduling efficiency was near perfect for both models during the period when irrigation adequacy scores were low, however, there were fluctuating results during the rest of the study period with minimums of 73% and 46% for the daily SWAT and hourly SWB, respectively. These periods of low values correspond to periods of frequent rainfall. The TIME and RTIME treatments were developed to replace the historical net irrig ation requirement at rates of 100% and 60%. There was a calculation error resulting in an irrigation runtime that was much lower than the net irrigation requirement during the month of Oct. 2006. Additionally, there was a period of a few weeks were irrig ation was shut off to these treatments resulting in no irrigation application despite the schedule in Apr. 2007.

PAGE 182

182 The irrigation adequacy results remained above 95% for both models and both time based treatments during most of the study period with the on ly exceptions occurring during Oct. 2006 and Apr. 2007 corresponding to errors in the irrigation schedule pr eviously described (Figure 5 19; Figure 5 20). The declines in irrigation adequacy scores were much more apparent in RTIME, reaching a score of 0% at one point, compared to TIME where the minimum score was 49%. Though the actual values of the irrigation adequacy scores varied between models, the differences were small during a majority of the 30 day periods. Scheduling efficiency scores mirrored t he irrigation adequacy results with high scores occurring during the periods of reduced runtimes/no irrigation and lower scores occurring the rest of the study period. The daily SWAT model had higher scores than the hourly SWB model for both time based tr eatments. Additionally, there were higher scores by the RTIME compared to TIME due to the shorter runtimes since only 60% of the net irrigation requirement was replaced. Despite the reduced schedule, over irrigation still occurred by RTIME indicating tha t it was not a deficit treatment. D iscussion The GIR was 23% to 45% larger when calculated using the hourly SWB compared to the daily SWAT model using the seasonal periods despite the general trend of lower cumulative ET O when calculated hourly. This may have occurred due to the lack of irrigation restrictions on the hourly SWB where irrigation was allowed to occur during any hour that MB dropped below maximum allowable depletion regardless of time of day or day of week. Due to the mechanics of a daily mo del, irrigation frequency must be in 24 h timesteps thus resulting in fewer opportunities for irrigation and more opportunities to take advantage of rainfall. For example, there were four

PAGE 183

183 irrigation events for the daily GIR and six events for the hourly G IR during the frequent rainfall period, resulting in 35 mm of additional irrigation. With gross irrigation events averaging 23 mm, a few additional irrigation events would impact the GIR. Including an allowable watering window in the hourly SWB may reduc e the GIR, but would introduce sub daily periods of deficit conditions. The results for the bench testing study showed general trends of declining irrigation adequacy scores and high scheduling efficiency scores for the controllers indicating that the cont rollers erred on the side of deficit conditions. However, the results for the field study showed the opposite trend of high irrigation adequacy and fluctuating scheduling efficiency indicating over irrigation. Additionally, there were few instances of co rrelation between irrigation application and SWAT performance scores for the daily SWAT model used during the bench testing study The settings programmed into the controllers for these testing periods were selected by the manufacturers as representative of the landscapes described in the SWAT protocol. However, the settings used were not always identical to the landscape descriptions despite having the option available. The controllers in the field study were programmed with settings that described the landscape such as sand for soil type and warm season turfgrass for plant type (Davis et al. 2009). As a result, predictions of conservative water use based on SWAT test results were not observed in the field study where over irrigation occurred. Based on the settings discrepancy, the results from daily SWAT tests are not representative of actual ET controller performance when implemented in an actual landscape setting.

PAGE 184

184 The USEPA chose minimums of 80% for irrigation adequacy and 95% for scheduling effici ency as the score requirements for WaterSense labeling. In most cases of the bench testing study the scores fell much lower than 80% when using the hourly SWB model in situations where the irrigation adequacy was less than perfect but above 80% using the daily SWAT model. In these instances, scheduling efficiency scores remained above 95% resulting in no change in outcome based on this requirement. Though it would be undesirable to penalize a controller for a small error that dramatically drops the scor e, such large discrepancies with the model are most likely connected to the programming issue previously discussed. As was found in Chapter 4, the hourly SWB more accurately reflects actual soil water movement in a soil profile and scores may benefit from more realistic program settings. This was seen in the field study where non passing scores for irrigation adequacy and scheduling efficiency occurred with both models for mostly the same time periods despite differences in score magnitude. Previously published results from the field study showed that the ET controllers averaged 43% annual water savings with maximum savings of 60% occurring in the fall/winter months and some over irrigation during the hot and dry spring season (Davis et al. 2009). Thes e savings were calculated based on the TIME treatment that replaced 100% of the net irrigation requirement. However, TIME also showed gross over watering during most of the study (Davis and Dukes 2010), which was evidenced by low scheduling efficiency sco res using both models. Additionally, RTIME was just as beneficial in water savings as the ET controllers by replacing 60% of the net irrigation requirement (Davis et al. 2009). Though TIME may reflect actual homeowner practices

PAGE 185

185 in Florida and an ET contr oller would be beneficial for excessive irrigators, it was not accurate as the sole measure of water conservation performance by ET controllers. C onclusion The hourly SWB was preferential to the daily SWAT due to the realistic representation of soil water movement by hourly SWB resulting in a correlation of SWAT scores to irrigation application An additional benefit to hourly SWB was that higher performance scores were achieved when using accurate landscape prog ramming instead of manufacturer selected settings that forced short, frequent irrigation events. In m any cases, t he performance scores declined when using the hourly SWB in the bench testing study where program settings were not realistic for the describ ed landscape resulting in frequent under irrigation As a result, there was an increased failure rate of achieving the WaterSense label using the hourly SWB The irrigation adequacy results were most affected with scores declining as much as 80 percentile points. Thus, the failure rate of irrigation adequacy results frequently went from 0% with the daily SWAT to 100% with the hourly SWB, with the change in the failing rate ranging from 13 percentile points to 100 percentile points The decline in scheduling efficiency results were less drastic, falling no more than 30 percentile points and typically falling less than 10 percentile points at any one 30 day period W hen program settings were related to the landscape characteristics as in the fi eld plot study, a ll treatments tended to have over irrigation As a result, scheduling efficiency was the primary performance measure affected by the chosen model. However, the WaterSense threshold of 95% for scheduling efficiency is already s o high that both models frequently resulted in failing scores with failure rates of 47% to 88%

PAGE 186

186 for the daily SWAT and 75% to 94% for the hourly SWB Due to more irrigation being considered over irrigation with the hourly SWB model, there was an increase in passing irrigation adequacy results at the 80% threshold with a minimum passing rate of 74% for the daily SWAT and 80% for the hourly SWB The time based treatments selected to represent actual homeowner practices in Florida resulted in excessive irrigation during the field plot study. Though the ET controllers were able to decrease irrigation compared to 100% replacement of the net irrigation requirement, there were no additional water savings compared to 60% replacement. Additionally, the performance scores usi ng either model for the ET controllers scheduling efficiency averaging 75 77%, were not very different from the irrigation treatment scheduled for 60% replacement scheduling efficiency averaging 81%, with all treatments experiencing over irrigation. Bas ed on the results of the field study ET controllers reduced irrigation application compared to excessive irrigators, but still over irrigated during the study period thus the ET controllers failed to meet true irrigation efficiency standards.

PAGE 187

187 Table 5 1. Crop coefficients for each study. Month Bench Test Study a Field Plot Study b Jan. 0.52 0.45 Feb. 0.64 0.45 Mar. 0.70 0.55 Apr. 0.73 0.80 May 0.73 0.90 Jun. 0.71 0.75 Jul. 0.69 0.70 Aug. 0.67 0.70 Sep. 0.64 0.75 Oct. 0.60 0.70 Nov. 0.57 0.60 Dec. 0.53 0.45 a Specified for zone 2 of the SWAT protocol (IA 2008) b Specified for southwest Florida (Jia et al. 2009)

PAGE 188

188 Table 5 2. Summary of treatment periods evaluated by the models. Data Period Start Date End Date Cumulative Daily ET O (mm) Cumulative Rainfall (mm) Rainfall Events (#) Obs erved Hist orical Obs erved Hist orical Obs erved Hist orical Freq uent Rain Jul. 28, 2009 Sep. 25, 2009 239 269 269 323 28 27 Infreq uent Rain Sep. 26, 2009 Nov. 24, 2009 157 203 46 127 10 12 Low ET O Jan. 27, 2010 Mar. 26, 2010 163 208 198 185 17 14 High ET O Apr. 7, 2010 Jun. 5, 2010 287 333 124 160 16 13 Field Plots Aug. 13, 2006 Nov. 27, 2007 2,475 1,845 1,293 1,738 145 153

PAGE 189

189 Table 5 3. Cumulative irrigation, water savings compared to GIR and performance scores during the frequent rainfall period, lasting sixty days, for each controller. Controller Irrigation (mm) Water Savings (%) Irrigation Adequacy (%) Scheduling Efficiency (%) Daily SWAT ET Water 132 23 100 a 90 E Toro 150 40 100 a 95 C Weathermatic 74 31 88 c 100 A GIR 107 0 100 100 Hourly SWB ET Water 132 7 91 b 94 D Toro 150 6 100 a 98 B Weathermatic 74 48 65 d 100 A GIR 142 0 100 100 Note: Different letters within the same case indicate significant differences between treatments. Table 5 4. Cumulative irrigation, water savings, and performance scores during the infrequent rainfall period, lasting sixty days, for each controller. Controller Irrigation (mm) Water Savings (%) Irrigation Adequacy (%) Scheduling Efficiency (%) Daily SWAT ET Water 94 3 89 b 100 A Toro 137 41 100 a 99 A Weathermatic 127 31 100 a 99 A GIR 97 0 100 100 Hourly SWB ET Water 94 33 64 c 92 C Toro 137 3 100 a 94 B Weathermatic 127 10 100 a 100 A GIR 141 0 100 100 Note: Different letters within the same case indicate significant differences between treatments.

PAGE 190

190 Table 5 5. Cumulative irrigation, water savings, and performance scores during the low ET O period, lasting sixty days, for each controller. Controller Irrigation (mm) Water Savings (%) Irrigation Adequacy (%) Scheduling Efficiency (%) Daily SWAT ET Water 84 11 95 a 82 B Toro 69 9 92 a 100 A Weathermatic NA NA NA NA GIR 76 0 100 100 Hourly SWB ET Water 84 11 75 b 100 A Toro 69 27 43 c 100 A Weathermatic NA NA NA NA GIR 94 0 100 100 Note: Different letters within the same case indicate significant differences between treatments. Table 5 6. Cumulative irrigation, water savings, and performance scores during the high ET O period, lasting sixty days, for each controller. Controller Irrigation (mm) Water Savings (%) Irrigation Adequacy (%) Scheduling Efficiency (%) Daily SWAT ET Water 262 13 94 a 92 B Toro 236 2 93 a 98 A Weathermatic 185 20 82 b 100 A GIR 231 0 100 100 Hourly SWB ET Water 262 7 23 d 100 A Toro 236 17 19 d 100 A Weathermatic 185 35 47 c 74 C GIR 283 0 100 100 Note: Different letters within the same case indicate significant differences between treatments.

PAGE 191

191 Table 5 7. Correlation coefficients were used to determine dependence between the performance results and irrigation application for each treatment and model. Daily SWAT Hourly SWB Controller Correlation Coefficient P Value Correlation Coefficient P Value Irrigation Adequacy ET Water 0.125 0.4185 0.783 <0.0001 Toro 0.142 0.3351 0.585 <0.0001 Weathermatic 0.280 0.0250 0.324 0.0090 Scheduling Efficiency ET Water 0.224 0.0598 0.518 <0.0001 Toro 0.074 0.5330 0.483 <0.0001 Weathermatic 0.304 0.1306 NR NR Note: The Weathermatic had too few instances of over irrigation for statistical analysis thus results related to scheduling efficiency could not be reported.

PAGE 192

192 Figure 5 1. Thirty day totals of ET O and rainfall for the frequent rainfall period that must meet or exceed the thresholds for a valid testing period

PAGE 193

193 Figure 5 2. Thirty day performance scores for the frequent rainfall period by the ET Water.

PAGE 194

194 Figure 5 3. Thirty day performance scores f or the frequent rainfall period by the Toro.

PAGE 195

195 Figure 5 4. Thirty day performance scores for the frequent rainfall period by the Weathermatic.

PAGE 196

196 Figure 5 5. Thirty day totals of ET O and rainfall for the infrequent rainfall period that must meet or exceed the thresholds for a valid testing period

PAGE 197

197 Figure 5 6. Thirty day performance scores for the infrequent rainfall period by the ET Water.

PAGE 198

198 Figure 5 7. Thirty day performance scores for the infrequent rainfall period by the Tor o.

PAGE 199

199 Figure 5 8. Thirty day performance scores for the infrequent rainfall period by the Weathermatic.

PAGE 200

200 Figure 5 9. Thirty day totals of ET O and rainfall for the low ET O period that must meet or exceed the thresholds for a valid testing period

PAGE 201

201 Figure 5 10. Thirty day performance scores for the low ET O period by the ET Water.

PAGE 202

202 Figure 5 11. Thirty day performance scores for the low ET O period by the Toro.

PAGE 203

203 Figure 5 12. Thirty day totals of ET O and rainfall for the high ET O period that must meet or exceed the thresholds for a valid testing period

PAGE 204

204 Figure 5 13. Thirty day performance scores for the high ET O period by the ET Water.

PAGE 205

205 Figure 5 14. Thirty day performance scores for the high ET O period by the Toro.

PAGE 206

206 Figure 5 15. Thirty day performan ce scores for the high ET O period by the Weathermatic.

PAGE 207

207 Figure 5 16. Thirty day totals of ET O and rainfall for the field plot study at the GCREC that must meet or exceed the thresholds for a valid testing period

PAGE 208

208 Figure 5 17. Thirty day performance scores for the Weathermatic evaluated during the field plot study at the GCREC.

PAGE 209

2 09 Figure 5 18. Thirty day performance scores for the Toro evaluated during the field plot study at the GCREC.

PAGE 210

210 Figure 5 19. Thirty day performance scores for the TIME treatment evaluated during the field plot study at the GCREC.

PAGE 211

211 Figure 5 20. Thirty day performance scores for the RTIME treatment evaluated during the field plot study at the GCREC.

PAGE 212

212 CHAPTER 6 REGRESSION ANALYSIS FOR SENSITIVITY OF AN HOURLY SOI L WATER BALANCE MODEL Introduction In efforts to conserve our limited fresh water resources, smart controller technologies were developed as devices designed to improve the efficiency of residential irrigation application. One smart technology is the weat her based irrigation controller that uses reference evapotranspiration (ET O ) to estimate plant water needs, commonly referred to as the ET controller. Though they vary in functionality by brand and model, some ET controllers use a soil water balance (SWB) to determine how much irrigation is necessary and when to apply it. Research has shown that ET controllers have the potential to produce water savings when implemented in scenarios with frequent excessive irrigation (Mayer et al. 2009; Davis and Dukes, i n review). They have also been shown to increase irrigation at homes that historically irrigate at or below the theoretical irrigation requirement (Mayer et al. 2009; Davis and Dukes 2014). The Irrigation Association developed the Smart Water Applicatio n Technologies (SWAT) Testing Protocol for Weather Based Irrigation Controllers as a method to evaluate the irrigation scheduling capabilities of ET controllers (IA 2008). This testing protocol, established in 2002 and now in its 8th draft, specifies a da ily soil water balance occurring over a 30 day period to determine if ET controllers can maintain theoretical soil moisture. Results are quantified as irrigation adequacy, a measure of under irrigation, and scheduling efficiency, a measure of over irrigat ion. The protocol specifies that testing results are valid when 30 day totals of rainfall and ET O meet or exceed 10.2 mm and 63.5 mm, respectively. Publishing SWAT test results are at the discretion of

PAGE 213

213 the manufacturer and can be continually tested until achieving acceptable scores during a valid testing period. The United States Environmental Protection Agency (USEPA) adopted the daily SWAT test as part of their specification for certifying weather based irrigation controllers as water saving devices. I nstead of publishing performance results as the Irrigation Association does, the USEPA chose minimums of 80% for irrigation adequacy and 95% for scheduling efficiency as the minimum requirements to achieve certification. The results after the first valid 30 day period are compared to the thresholds to determine pass/fail status. Only the controller models and not the quantitative results are published if meeting the minimum requirements. In chapter 4, an hourly SWB was developed to address the deficiencie s in the daily SWAT test that were apparent when evaluated against measured soil moisture data and negatively impacted the transferability of the performance results. The primary deficiencies include inability to store water above field capacity, inabilit y to allow deficit conditions, and the adaptation of transpiration rates based on water storage. The hourly SWB better represented the sub daily fluctuations in soil moisture resulting in a more accurate estimation of soil water content. The results of c hapter 5 indicated that controllers tested using settings meant to pass the daily SWAT test resulted in under irrigation using the hourly SWB, thus compromising whether the controller achieved passing performance scores for WaterSense certification. As a result, water use efficiency did not translate to controllers implemented in field plot studies that tended to over irrigate when programmed with parameters specific to landscape descriptions.

PAGE 214

214 Though the results from these chapters were positive, the limit ations to the application of the hourly SWB model are unknown. The method for determining the value of the model and appropriateness in future applications is a sensitivity and uncertainty analysis (Beven 2006). A sensitivity analysis evaluates the cont ribution of uncertainty in the input factors, such as hydraulic parameters in a hydrological model, to the outcomes of the model and an uncertainty analysis quantifies the uncertainty due to the input factors (Helton 1993). Typically, the testing required in quantifying uncertainty is extensive and costly making it unfeasible for some modeling efforts, but determining the sensitivity of each input factor can be completed without additional testing. The objective of this study was to evaluate the uncertain ty of the input factors on the performance results obtained using the hourly SWB by performing a regression based sensitivity analysis. Materials and Methods Hourly SWB S oil volume of water to total volume (water, air, and soil particles ). In a natural landscape PWP ), where additional energy is s ), where all pore spaces are FC ) is achieved when gravitational forces equal surface tension forces (cohesive and adhesive) between water molecules and soil particles resulting in no drainage due to gravity (Richards 1931). Water storage at water FC s FC over time based on soil hydraulic properties. To reduce water stress on the plant material, the water

PAGE 215

215 PWP MAD ) selected as a PWP FC (IA 2013). The soil water balance is a simple model that uses a tipping bucket approach to soil water movement in a homogeneous soil profile (i.e. depth per unit area of root zone) over a specified timestep (Ranatunga et a l. 2008). The model is based on the principle that soil water at the current depth (Z) and timestep (i) was affected by the soil water at the previous timestep (i 1) and inputs and outputs of crop evapotranspiration (ET C ), rainfall (R), irrigation (I), ru noff (RO), and deep percolation (D) occurring at i (Equation 4 1). The ET C was calculated using Equation 4 2 from ET O determined from local microclimate variables and crop coefficients specific to the studies (Table 5 1). Rainfall and irrigation were utilized similarly in the model where the depth equals the total that fell (R) or was applied (I) over the unit area duri ng the timestep of interest. Events that s resulted in separating total depth into effective depth, which is considered available for plant use, and a combination of runoff and deep percolation losses. When total dept depth equals effective depth. Runoff and deep percolation are two different physical FC resulting in water losses from the soil profile. Generall y, runoff occurs due to improper cycle/soak scheduling of irrigation and landscape inclines whereas deep percolation occurs due to excess rainfall or irrigation. It is difficult and unnecessary to separate these losses because they are quantified outside o f the soil profile and the variable of interest is the change in depth of water within the soil profile.

PAGE 216

216 To account for soil water levels higher than field capacity, an exponential decay function (Equation 4 3) was incorporated into each timestep used t o represent drainage to equilibrium. When deficit conditions occur and soil moisture falls below MAD the ability for ET C to be removed from the soil profile by the plant material becomes more difficult. To account for this phenomenon, a stress coeffic ient (Equation 4 5) was introduced to the ET C calculation. Additional details on these changes can be found in Chapter 4. The performance results of ET controllers determined from the soil water balance are presented as scheduling efficiency (E) and irrig ation adequacy (A). Scheduling efficiency is a measure of over irrigation quantified by estimating scheduling losses over a 30 day period combined from three sources: soil moisture rising above field capacity, too long cycle times, and too short soak time s (Equation 5 1). Similarly, irrigation adequacy is a measure of under irrigation and can be quantified by depths when conditions were not considered well watered over the 30 day period (Equation 5 2). Results of 100% for either score would indicate that the irrigation scheduling performance by the ET controller was effectively perfect. Case Studies Bench test s tudy The Weathermatic SL1600 (Garland, TX), Toro Intelli Sense (Riverside, CA), and ET Water Smart Controller 100 (Novato, CA) were installed at t he University of Florida Agricultural and Biological Engineering turfgrass research facility in Gainesville, FL. Results from the 2010 Florida SWAT Test using three controllers of each brand showed that there was little variability between replications (D avis et al. 2009). Thus, performance results for these brands are likely to be similar to controllers of the same

PAGE 217

217 brand being utilized in the real world, but may not be representative of all controller brands. Additionally, the Toro and ET Water were pro vided rain sensors and the Weathermatic was equipped with a rain sensor, all set at a 6 mm threshold. Each controller was connected to a CR 10X datalogger (Campbell Scientific, Logan, UT) via a set of relays to record time and date at the beginning and end of each irrigation event for all active zones. Weather data were collected from a weather station located on site and managed by research personnel. Weather parameters included temperature, relative humidity, solar radiation, and wind speed collected a t fifteen minute intervals. Values for ET O were calculated using the American Society of Civil Engineers standardized ET O equation (ASCE 2005) as specified in the SWAT protocol (IA 2008). Each controller was programmed by the manufacturer, or with the ma supervision, with all six zones specified in the SWAT test. For this analysis, only zone 2 description to landscapes in Florida. In Chapter 5, four 60 day datasets were selected to represent a variety of weather conditions specified as frequent rainfall, infrequent rainfall, high ET O and low ET O (Table 5 2). Each period was selected by comparing the rainfall and ET O amount over the 60 day period with the historical averages over the same sixty days determined from 37 years of Gainesville Regional Airport weather data from 1970 through 2006 (NCDC 2007). The high ET O period was chosen as a representative period for this analysis due to the occurrence of high ET O frequent rainfall, and disparity between performance results using the hourly SWB compared to the daily SWAT test (IA 2008) found in Chapter 5.

PAGE 218

218 GCREC field plot study Sixteen St. Augustinegrass plots measuring 7.6 m by 7.9 m were established a t the University of Florida Gulf Coast Research and Education Center (GCREC) in Balm, FL. These plots were created to evaluate the ability of ET controllers to apply irrigation efficiently in southwest Florida. The treatments, replicated four times, were : A) Weathermatic SL1600, B) Toro Intelli Sense, C) TIME treatment based on 100% replacement of the net irrigation requirement, and D) RTIME treatment that was 60% of the TIME treatment. The variable of interest for these treatments was irrigation applica tion only; thus, all plot maintenance and nutrient management remained constant across all treatments based on University of Florida Institute of Food and Agricultural Sciences recommendations (Sartain 1991; Black and Ruppert 1998). Detailed methods and results concerning these treatments can be found in Davis et al. (2009), Davis and Dukes (2010), and Davis and Dukes (2012). Each plot was maintained using an irrigation system with dedicated flow meter to measure volumes of irrigation application per plo t. The flow meters were connected to a data logger that recorded sub daily irrigation to easily separate irrigation system maintenance activities and treatment related irrigation. Weather data during the study period, ranging from Aug. 13, 2006 to Nov. 12, 2007, was collected from the Florida Automated Weather Network (FAWN) weather station located at the GCREC. Weather parameters included temperature, relative humidity, solar radiation, and wind speed collected at fifteen minute intervals. The ET O was calculated using the ASCE EWRI ET equation (ASCE 2005). Rainfall was also collected from this weather station.

PAGE 219

219 Sensitivity Analysis S ensitivity analysis methods are used to determine the sensitivity of model results to input parameters (Saltelli et al. 2 000). This helps to avoid over parameterization and verify the representation of the physical system. In this study, the Monte Carlo simulation was used to perform a regression analysis to evaluate the effect of changing various input factors on the mode l output. This type of evaluation is a version of a one at a time (OAT) sensitivity analysis. The simulation involves three important steps (Helton 1993). Select range and distribution of input factors Generate sample input factors based on the distribut ions Produce model results from each sample input sequence The input factors selected as important to the hourly SWB were PWP FC s and the fraction (MAD) that determines MAD Because these factors were already determined using the inverse solution feature of HYDRUS 1D (Chapter 4), preference was given to the inverse solution value in the probability distribution functions. Thus, a triangular distribution was selected for all four parameters with the maximum and minimum limits of each range being l ess likely to occur in the real system. There were 1,000 randomly selected combinations of input factors chosen using a Microsoft Office Excel add in called EasyFitXL (Mathwave Technologies, Spokane, WA) for the bench test study (Figure 6 1) and the field plot study (Figure 6 2). These combinations were chosen once and applied identically to all treatments. The hourly SWB was performed for each combination of input factors resulting in 1,000 results for the response variables of irrigation adequacy and sc heduling efficiency for each timestep evaluated. Only one 30 day period was evaluated for the high ET O

PAGE 220

220 period for the bench test study, with the 30 day period ending Jun. 5, 2010. In Chapter 5, irrigation adequacy and scheduling efficiency results for th e GCREC field plot study were calculated for every 30 day period occurring in the study period resulting in rolling averages. For this analysis, only completely independent scores were evaluated thus only fifteen 30 day periods were applicable. Statisti cal analyses were completed using Statistical Analytical Systems (SAS) software (Cary, NC). The correlation procedure was used to determine the Pearson correlation coefficient between model results of irrigation adequacy and scheduling efficiency to each of the input factors. The correlation coefficient ranges from 1 to 1 where more positive or more negative values indicate a stronger relationship between the two variables. For the purposes of this analysis, a weak correlation was considered between 0. 2 and 0. 2 when the coefficient was significant at a 95% confidence level. A moderate correlation was considered between 0. 2 and 0.5 or between 0. 5 to 0. 2 A strong correlation was greater than 0.5 or less than 0. 5. Results Irrigation Adequacy Bench te st study The net irrigation requirement can be generalized as the difference between ET O and rainfall and is usually an over estimated benchmark. During the thirty days that represented the high ET O period, the net irrigation requirement approximated 76 m m (Table 6 1). The Toro and ET Water controllers applied less irrigation than the net irrigation requirement, totaling 58 mm and 54 mm. The Weathermatic applied 366 mm, which was 382% more than the net irrigation requirement.

PAGE 221

221 Across all 1,000 iterations o f the Monte Carlo simulation for the ET Water, irrigation adequacy results averaged 3 7.8 % with a CV of 0. 108 (Table 6 2). This controller had only 2% of results that fell above the EPA WaterSense threshold of 80% indicating very little chance of passing t he test. There was no correlation between irrigation adequacy results and the input factors of PWP FC or s However, there was a strong correlation for MAD with a correlation coefficient of 0. 800 Thus, the irrigation adequacy results were highly influenced by MAD The irrigation adequacy for the Toro averaged 24. 7 % with a CV of 0. 262 indicating a large distribution of results (Table 6 2). The Toro had very few results that exceeded the EPA Wat erSense threshold of 80%, totaling 1% of the 1,000 iterations. There was no correlation between results and s but there was a moderate correlation for PWP and FC with a coefficient s of 0. 399 and 0.406, respectively Similar to the results of the ET Water, there was a strong correlation with MAD having a correlation coefficient of 0. 738 Due to the over application of irrigation by the Weathermatic, irrigation adequacy averaged 8 2 8 % across the 1,000 replications using the Monte Carlo simulation wit h a coefficient of variation (CV) of 0. 0 35 (Table 6 2). Due to over irrigation, the Weathermatic was the only controller with considerable number of scores above the WaterSense threshold of 80%, totaling 66%. There was a strong correlation between irrigation adequacy results and PWP with a coefficient of 0.537 FC with a coefficient of 0.582, and MAD, with a coefficient of 0.545 There was no correlation between irrigation adequacy and s

PAGE 222

222 The MAD is the fraction of the available water holding capacity that is considered well watered with higher values indicating a larger fraction of well watered storage. Irrigation adequacy scores for ET Water (Figure 6 3A) and Toro (Figure 6 3B) had similar trends with a steadily incre asing slope across the 0.3 to 0.7 range. O ver irrigation was the primary driving factor for the Weathermatic which limits the range of irrigation adequacy scores, resulting in a less inclined slope across the range of MAD (Figure 6 3C). GCREC field plot study The net irrigation requirement ranged from no irrigation required due to enough rainfall to satisfy the plant water needs to 142 mm required for the 30 day period ending May 12, 2007 due to the combination of 170 mm of ET O but only 28 mm of rainfall (Table 6 3). The previous 30 day period ending Apr. 12, 2007 had similar irrigation requirements, estimated as 129 mm, due to high ET O and low rainfall. As would be expected, months with low or no irrigation requirement occurred during winter periods whe re ET O was low. Periods where the irrigation application did not satisfy the net irrigation requirement occurred for all four treatments, occurring twice for the Weathermatic, five times for the Toro, twice for TIME, and seven times for RTIME. There was very little variation in irrigation adequacy across the 30 day periods for the Weathermatic, with means ranging from 94.2 % to 100 % and the maximum CV calculated as 0. 094 (Table 6 4). During the period with the lowest mean, ending on Nov 11, 200 6 76% of the scores were above the EPA WaterSense threshold of 80%. Likewise, the period s with the highest mean, occurring three times had 98% of scores that met or exceeded the WaterSense threshold. Only one period had a strong correlation between PWP and irrigation adequacy while the remaining 14 periods had a

PAGE 223

223 weak to moderate correlation. Similar results were also true for FC There were ten periods wehre irrigation adequacy and s were not significantly correlated while the remaining periods w ere only weakly correlated. A wide range of correlations occurred for MAD, ranging from not significant (five periods) to 0.564 When evaluating one of the period s with a strong correlation, most of the results fell at 100% with a considerable amount of scatter (Figure 6 4). Average irrigation adequacy ranged from 74.2 % (Nov. 11, 2006), with the maximum CV of 0. 160 occurring on Dec. 11, 2006 to 99.9 % (Aug. 10, 2007), with the minimum CV of 0.0 10 for the Toro controller (Table 6 5). There were only 9% of the results exceeding the EPA WaterSense threshold of 80% during the Nov. 2006 period whereas there were 97% of results exceeding the WaterSense threshold for the period with the highest average. Moderate to no correlations occurred for PWP and s when evaluating irrigation adequacy with two periods of strong correlation for PWP There was a more consistent relationship for FC where the coefficients ranged from 0.156 to 0.614 with six periods above 0.5. A majority of moderate to strong correlations o ccurred for MAD, ranging from 0.215 ( Sep 1 2, 2006 ) to 0. 871 (Nov. 11, 2006). In the period with the strongest correlation, irrigation adequacy increased steadily as MAD increased (Figure 6 5). Irrigation adequacy results for the TIME treatment ranged fro m 5 6 .3% (Nov. 11, 2006) to 100 % (Jan. 12, 2007) with the lowest CV of 0.0 01 (Table 6 6). As a result, only 6% of irrigation adequacy scores exceeded the 80% WaterSense threshold for the period with the lowest average compared to 99% of the scores exceedin g the threshold for the period with the highest average. Similar to the other treatments, the i nput factor

PAGE 224

224 with the strongest correlations was MAD, ranging from 0. 107 ( Mar 1 3 2007) to 0. 993 (May 12, 2007). The lowe r correlation coefficient s occurred during the period s with the highest average s indicating that there were too few deficit conditions to develop a full relationship between irrigation adequacy and MAD. During the period of strongest correlation between MAD and irrigation adequac y, irrigation adequacy increased as MAD increased (Figure 6 6). Irrigation adequacy for RTIME ranged from 38.2 % (Jun. 11, 2007) to 100 % (Jan. 12, 2007) with respective CV values of 0. 17 9 and 0.0 02 (Table 6 7) The period with the highest mean had 99% of scores that exceeded EPA WaterSense criteria whereas the period with the lowest mean had only 3% of passing scores. Similar to the other treatments, there was a weak to moderate correlation between irrigation adequacy and PWP or FC whereas there was li ttle to no correlation with s Also, there was a weak to strong correlation with MAD, ranging from 0. 147 ( Mar 1 3 2007) to 0. 833 (May 12, 2007). The period s with the correlation coefficient s that were not significant for MAD coincided with the period s with the highest average s resulting in little variation in irrigation adequacy due to consistently high scores. The linear relationship increased as MAD increased across the period with the highest correlation coefficient (Figure 6 7), likely due to persi stent deficit conditions by RTIME with an estimated 30 day deficit of 60 mm (Table 6 3). Scheduling Efficiency Bench t est s tudy Scheduling efficiency is a measure of over irrigation and will not fluctuate due during prolonged deficit conditions. Due to under irrigation occurring by the ET Water

PAGE 225

225 and Toro controllers during the high ET O period, scheduling efficiency results remained at 100% despite repetitive variations of all four input factors (Table 6 2). The Weathermatic frequently had over irrigation during this period resulting in an average of 51.7% for scheduling efficiency (Table 6 2). There was very little change in sche duling efficiency from varying the input factors resulting in a low CV of 0.02 4 All four input factors had significant correlation coefficients for this treatment. There was a moderate PWP and MAD with coefficients of 0.274 and 0.256, respectively. Also, there was a weak correlation with FC having a coefficient of 0.1 02 A strong s with a coefficient of 0. 693 Though s was strongly correlated, the magnitude of the change in scheduling efficiency results wa s small s had little impact on final score (Figure 6 8) Compared to the WaterSense threshold of 95% for scheduling efficiency, there were no passing scores for the Weathermatic GCREC field plot study There were four out of fifteen periods that did not require irrigation according to the net irrigation requirement (Table 6 3). Irrigation was applied by all four treatments during all periods thus resulting in over irrigation and declining scheduling efficiency scores. Over irrigation was highest during the summer months with maximum excess totaling between 113 mm (RTIME, Aug. 10, 2007) and 209 mm (Weathermatic, Jul. 11, 2007) across treatments due to a lack of effective means fo r handling rainfall. For the Weathermatic, scheduling efficiency ranged from 20 % (Jul. 11, 2007) to 94.5% (Jun. 11, 2007) across the fifteen time periods (Table 6 4). In most cases, a high frequency of scores occurred around the mean resulting in low CV values, reaching a maximum of 0.181. A majority of periods would not have met the EPA WaterSense

PAGE 226

226 threshold for scheduling efficiency of 95% due to such precision around the mean score. There was a weak or no correlation between scheduling efficiency and PWP or FC However, FC for the period ending Nov. 11, 2006 and a strong correlation for the period ending Dec. 11, 2006. These were the only two periods where the Weathermatic did not apply more than the net irrigation requirement. The majority of moderate to strong correlati ons occurred between scheduling efficiency and s ranging from 0.197 (Dec. 11, 2006) to 0.9 19 (Aug. 10, 2007). The lowest correlation FC was s increased scheduli s was 0.26 mm 3 mm 3 (Figure 6 9). Scheduling efficiency ranged from 17.0% (Sep. 12, 2006) to 99.9% (Dec. 11, 2006) for the Toro controller with the highest CV of 0.152 (Table 6 5). The period with the highest mean a lso had 99% of results above the EPA WaterSense threshold of 95%. Very few instances of moderate correlation occurred between scheduling PWP FC and MAD. However, a majority of moderate to strong s with correl ati on coefficients ranging from 0.078 (Nov. 11, 2006) to 0.8 80 (Apr. 12, 2007). Similar to the Weathermatic, the period with the lowest FC had the highest coefficient due to 36 mm of deficit. Also, the period with the strongest correlation had a linearly increasing trend beginning at 0.28 mm 3 mm 3 (Figure 6 10). Scheduling efficiency for TIME resulted in a range of 16.1% (Sep. 12, 2006) to 8 9 7 % (Jun. 11, 2007) with the largest CV calculated as 0.1 08 indicating very little

PAGE 227

227 change of scores due to variations in the input factors (Table 6 6). Only the period with the highest average had results that exceeded the EPA WaterSense criteria of 95%, totaling 84%. The periods ending May 12, 2007 and Jun. 11, 2007 had moderate PWP FC s due to deficit conditions occurring during parts of the periods. Otherwise, moderate to strong s wit h a maximum coe fficient of 0.950 for the period ending Dec. 11, 2006. The linear, increasing trend of scheduling s occurred over the entire range with no limitations to the relationship (Figure 6 11). Scheduling efficiency ranged from 27. 8 % (Sep. 12, 2006) to 99.9% (Jun. 11, 2007) for RTIME (Table 6 7), thus mirroring the results of the TIME treatment. There were more periods of deficit for this treatment due to having a reduced timer schedule resulting in a few periods with moderate to strong correla tions between scheduling FC PWP and MAD. However, a majority s with a maximum coefficient of 0.8 44 for the period ending Oct 9 200 7 The linear, increasin g trend of scheduling efficiency s occurred for RTIME in the same manner as the trend for the TIME treatment (Figure 6 12). The magnitude of scores were higher for RTIME due to less irrigation application than TIME, but the change in scores across s was the same, 4 5 % to 65% for TIME (Figure 6 11) and 75 % to 95% for RTIME (Figure 6 12), totaling approximately twenty percentile points. Discussion Both SWAT test results of irrigation adequacy and scheduling efficiency were impacted by variations of PWP FC s and MAD. In both studies,

PAGE 228

228 there was a strong correlation between irrigation adequacy and MAD whereas the s during periods of over irrigation and the other three input factors when under irrigation occurredThe relationship s between irrigation adequacy and MAD as well as scheduling s were a linear trend. Despite the high correlation coefficients, there was still scatter in portions o f the scatter plots, indicating that there may be other unexplored input factors affecting the model results. Root depth would be a likely factor to significantly impact the results. However, this parameter was not identified as an input factor in this a nalysis due to the correlation between root depth and all four input factors that would not be addressed. Helton (1993) explained that relationships between the input factors create complexities to a sensitivity analysis that are beyond the scope of this chapter. Though there were strong correlations in both studies, irrigation adequacy and scheduling efficiency were largely unaffected by the variations in all input factors. The maximum difference was approximately thirty percentile points despite using e xtreme values. Consequently, t he primary input that affected the irrigation adequacy and scheduling efficiency scores was irrigation. In 30 day periods where irrigation and rainfall exceeded ET O scheduling efficiency was low. Likewise, when irrigation and rainfall totaled less than ET O there was the possibility of deficit conditions occurring during the 30 day period resulting in low irrigation adequacy. Considering its purpose was to judge irrigation scheduling performance, the model was stable across wide ranges of assumed hydraulic parameters. In most cases, the irrigation schedule was a better predictor of performance than the input factors.

PAGE 229

229 The inputs of ET O and rainfall were also extremely important to the irrigation adequacy and scheduling eff iciency scores. The soil hydraulic parameters were the same for each 30 day period in the GCREC field plot study. Thus differences in average SWAT test results were primarily due to varying ET O and rainfall and how the controller scheduled irrigation in reaction to these weather conditions. Chopart and Vauclin (1990) also found that water related inputs were the most influential parameters to the results of the water balance model and not the soil properties. In general, improper estimation of the soil hydraulic parameters were not an important factor for meeting the EPA WaterSense criteria. The selection of the input factors and their assumed probability distributions were based on educated guesses and may have introduced error into the analysis. Howev er, the maximums and minimums of the ranges were selected to evaluate the stability of the model by using extreme but realistic values and defining the limits of the model through the regression analysis (Jones and Luyten 1998; Saltelli et al. 2000). The model consistently produced results that were reasonable given the inputs thus justifying the introduced error in assumptions. Conclusion Irrigation adequacy was most sensitive to the model inputs with the only clear relationship occurring between irrigati on adequacy and MAD. The correlation between MAD and irrigation adequacy ranged from 0.545 to 0.800 for the bench test study and maximum correlations ranged from 0.564 to 0.933 for the field plot study. Despite such high correlation for all treatments, t here was very little increase in irrigation adequacy results due to increasing values of MAD within a reasonable range.

PAGE 230

230 Scheduling efficiency had a moderate to strong relationship to varying s when over irrigation was prominent, but also had a small effec t on final scores. For the bench test study, the correlation coefficient was 0.693 for the one treatment that did not have a perfect score Likewise, the maximum correlation coefficients ranged from 0.844 to 0.950 for the field plot study. When there we re periods of deficit conditions, the model input of FC had a strong correlation combined with a weak correlation to s resulting in a minimum coefficient of 0.577 for FC Both performance results were highly dependent on irrigation application thus satisfying the purpose of the model with good stability. There was also a seasonal dependence on irrigation adequacy and scheduling efficiency results that was primarily based on ET O and rainfall that was directly inputted into the model. It would be ben eficial to continue with the evaluation of the sensitivity and uncertainty of the model using global sensitivity techniques.

PAGE 231

231 Table 6 1. Summary of data used for the soil water balance for the bench test study during the thirty days ending Jun. 5, 2010, representing the high ET O period. Description Depth (mm) 30 day ET O 143 30 day Rainfall 67 30 day Irrigation for Weathermatic 366 30 day Irrigation for Toro 58 30 day Irrigation for ET Water 54 Table 6 2. Statistical results a of irrigation adequacy and scheduling efficiency for each ET controller in the bench test study after performing the global sensitivity analysis using the Monte Carlo simulation. Statistic Weathermatic Toro ET Water Irrigation Adequacy Mean (%) 8 2 8 24. 7 37 .8 Coefficient of Variation 0. 035 0.262 0. 108 Correlation Coefficient for PWP 0.537 0.399 0.319 Correlation Coefficient for FC 0.582 0.406 0.311 Correlation Coefficient for s NS NS NS Correlation Coefficient for MAD 0.545 0.738* 0.800* Scheduling Efficiency Mean 51.7 100 100 Coefficient of Variation 0.02 4 0 0 Correlation Coefficient for PWP 0.274 NA c NA Correlation Coefficient for FC 0.102 NA NA Correlation Coefficient for s 0.693 NA NA Correlation Coefficient for MAD 0.256 NA NA a Asterisk denotes strong correlation selected as greater than 0. 5 or less than 0. 5 b Correlation coefficients that were not significant at a 95% confidence level were indicated by NS. c Correlation coefficients that could not be calculated are indicated by NA.

PAGE 232

232 Table 6 3. Summary of data used for the soil water balance during each 30 day evaluation period for the GCREC field plot study. End Date of Evaluation Period 30 day ET O (mm) 30 day Rainfall (mm) 30 day Irrigation for Weathermatic (mm) 30 day Irrigation for Toro (mm) 30 day Irrigation for TIME (mm) 30 day Irrigation for RTIME (mm) 9/12/2006 115 199 15 64 72 43 10/12/2006 130 61 85 51 72 45 11/11/2006 110 29 69 45 43 28 12/11/2006 80 18 53 36 71 48 1/12/2007 64 59 31 34 70 44 2/11/2007 77 78 38 25 55 35 3/13/2007 104 33 81 65 81 51 4/12/2007 142 13 132 134 135 82 5/12/2007 170 28 148 125 123 82 6/11/2007 163 103 124 123 124 74 7/11/2007 157 93 273 136 128 63 8/10/2007 145 146 128 147 204 113 9/9/2007 157 137 87 81 118 67 10/9/2007 126 195 116 84 115 68 11/8/2007 105 41 83 68 153 90

PAGE 233

233 Table 6 4. Statistical results a of irrigation adequacy and scheduling efficiency for the Weathermatic in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation. Statistic Mean (%) CV Correlation Coefficient for PWP Correlation Coe fficient for FC Correlation Coefficient for s Correlation Coefficient for MAD Irrigation Adequacy 9/12/2006 96.5 0.043 0.486 0.502 NS 0.564 10/12/2006 97.7 0.056 0.426 0.465 0.097 0.342 11/11/2006 94.2 0.094 0.513 0.333 NS 0.545 12/11/2006 98.5 0.050 0.369 0.359 0.129 0.293 1/12/2007 100.0 0.006 0.145 0.142 NS NS 2/11/2007 99.9 0.008 0.127 0.129 NS NS 3/13/2007 100.0 0.004 0.115 0.125 NS NS 4/12/2007 99.9 0.006 0.172 0.183 NS NS 5/12/2007 99.2 0.021 0.396 0.436 0.078 0.295 6/11/2007 95.2 0.084 0.454 0.547 0.066 0.463 7/11/2007 97.5 0.021 0.442 0.576 NS 0.554 8/10/2007 98.8 0.026 0.426 0.488 NS 0.409 9/9/2007 99.2 0.026 0.377 0.404 NS 0.289 10/9/2007 99.7 0.010 0.350 0.350 0.065 0.246 11/8/2007 100.0 0.002 0.118 0.124 NS NS Scheduling Efficiency 9/12/2006 73.3 0.058 0.139 NS 0.379 0.077 10/12/2006 78.8 0.081 0.124 NS 0.633 NS 11/11/2006 82.6 0.064 0.098 0.454 0.547 NS 12/11/2006 90.7 0.063 0.111 0.870 0.197 NS 1/12/2007 63.2 0.043 NS 0.095 0.609 NS 2/11/2007 62.2 0.052 NS 0.172 0.799 NS 3/13/2007 54.8 0.114 NS 0.195 0.797 NS 4/12/2007 63.9 0.148 NS 0.202 0.750 NS 5/12/2007 83.3 0.130 NS 0.207 0.652 NS 6/11/2007 94.5 0.097 NS 0.167 0.413 NS 7/11/2007 20.0 0.125 0.148 NS 0.870 0.076 8/10/2007 42.3 0.181 NS 0.166 0.919 NS 9/9/2007 93.5 0.038 NS 0.145 0.361 NS 10/9/2007 54.8 0.097 NS 0.183 0.827 NS 11/8/2007 72.2 0.116 NS 0.196 0.693 NS a Asterisk denotes strong correlation selected as greater than 0. 5 or less than 0. 5 b Correlation coefficients that were not significant at a 95% confidence level were indicated by NS.

PAGE 234

234 Table 6 5. Statistical results a of irrigation adequacy and scheduling efficiency for the Toro in the GCREC field plot study after pe rforming the global sensitivity analysis using the Monte Carlo simulation. Statistic Mean (%) CV Correlation Coefficient for PWP Correlation Coefficient for FC Correlation Coefficient for s Correlation Coefficient for MAD Irrigation Adequacy 9/12/2006 99.7 0.013 0.337 0.340 NS 0.215 10/12/2006 74.2 0.146 0.349 0.407 0.106 0.706 11/11/2006 53.3 0.114 0.081 NS 0.094 0.871 12/11/2006 87.9 0.160 0.484 0.556 0.208 0.542 1/12/2007 97.2 0.041 0.414 0.470 0.196 0.554 2/11/2007 99.6 0.021 0.322 0.317 NS 0.180 3/13/2007 99.9 0.010 0.157 0.156 NS NS 4/12/2007 99.8 0.013 0.253 0.240 0.084 0.115 5/12/2007 94.3 0.046 0.476 0.584 0.106 0.616 6/11/2007 90.3 0.103 0.479 0.589 0.062 0.597 7/11/2007 97.1 0.030 0.505 0.593 0.064 0.526 8/10/2007 99.7 0.013 0.318 0.300 NS 0.184 9/9/2007 93.4 0.077 0.511 0.614 NS 0.494 10/9/2007 99.0 0.032 0.377 0.391 0.087 0.264 11/8/2007 96.5 0.067 0.404 0.518 0.065 0.417 Scheduling Efficiency 9/12/2006 17.0 0.080 NS 0.267 0.527 NS 10/12/2006 97.5 0.021 0.249 0.165 0.309 NS 11/11/2006 85.0 0.057 0.424 0.696 0.078 0.246 12/11/2006 99.9 0.008 0.130 0.137 NS NS 1/12/2007 45.3 0.087 0.361 0.438 0.540 0.245 2/11/2007 72.2 0.044 NS 0.168 0.740 NS 3/13/2007 56.2 0.067 NS 0.172 0.876 NS 4/12/2007 53.4 0.153 NS 0.196 0.880 NS 5/12/2007 93.6 0.063 0.238 0.096 0.476 0.114 6/11/2007 92.4 0.055 NS 0.104 0.358 NS 7/11/2007 56.2 0.076 0.163 NS 0.798 NS 8/10/2007 44.4 0.133 NS 0.198 0.656 NS 9/9/2007 85.4 0.044 NS 0.184 0.687 NS 10/9/2007 46.9 0.075 0.075 0.098 0.870 NS 11/8/2007 71.4 0.041 0.253 0.122 0.509 0.094 a Asterisk denotes strong correlation selected as greater than 0. 5 or less than 0. 5 b Correlation coefficients that were not significant at a 95% confidence level were indicated by NS.

PAGE 235

235 Table 6 6. Statistical results a of irrigation adequacy and scheduling efficiency for TIME in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation. Statistic Mean (%) CV Correlation Coefficient for PWP Correlation Coefficient for FC Correlation Coefficient for s Correlation Coefficient for MAD Irrigation Adequacy 9/12/2006 99.7 0.012 0.335 0.343 0.062 0.205 10/12/2006 89.8 0.101 0.475 0.596 0.117 0.534 11/11/2006 56.3 0.072 0.207 0.170 0.102 0.635 12/11/2006 98.6 0.032 0.423 0.488 0.098 0.352 1/12/2007 100.0 0.001 0.064 0.074 NS NS 2/11/2007 99.9 0.009 0.125 0.127 0.067 NS 3/13/2007 99.8 0.016 0.242 0.226 NS 0.106 4/12/2007 99.0 0.036 0.361 0.362 0.073 0.247 5/12/2007 81.4 0.082 0.197 0.181 0.137 0.933 6/11/2007 79.8 0.107 0.409 0.523 0.207 0.730 7/11/2007 95.0 0.066 0.493 0.582 0.074 0.490 8/10/2007 98.8 0.028 0.414 0.478 0.117 0.345 9/9/2007 98.1 0.042 0.443 0.467 0.080 0.377 10/9/2007 99.0 0.033 0.367 0.389 NS 0.261 11/8/2007 98.9 0.028 0.403 0.467 NS 0.329 Scheduling Efficiency 9/12/2006 16.1 0.080 NS NS 0.817 NS 10/12/2006 78.4 0.029 0.089 NS 0.490 NS 11/11/2006 68.7 0.083 NS 0.737 0.332 NS 12/11/2006 58.1 0.075 0.068 0.065 0.950 NS 1/12/2007 32.0 0.072 NS 0.138 0.776 NS 2/11/2007 54.0 0.083 NS 0.139 0.907 NS 3/13/2007 47.1 0.061 NS 0.164 0.833 NS 4/12/2007 45.5 0.067 0.073 0.112 0.853 NS 5/12/2007 72.6 0.108 0.558 0.690 0.079 0.132 6/11/2007 89.7 0.070 0.529 0.577 0.096 0.227 7/11/2007 49.0 0.073 0.391 0.324 0.525 0.191 8/10/2007 27.8 0.087 0.107 0.103 0.799 NS 9/9/2007 45.8 0.062 0.151 0.331 0.579 NS 10/9/2007 29.4 0.057 0.084 0.154 0.725 NS 11/8/2007 34.1 0.045 0.131 NS 0.317 NS a Asterisk denotes strong correlation selected as greater than 0. 5 or less than 0. 5 b Correlation coefficients that were not significant at a 95% confidence level were indicated by NS.

PAGE 236

236 Table 6 7. Statistical results a of irrigation adequacy and scheduling efficiency for RTIME in the GCREC field plot study after performing the global sensitivity analysis using the Monte Carlo simulation. Statistic Mean (%) CV Correlation Coefficient for PWP Correlation Coeffici ent for FC Correlation Coefficient for s Correlation Coefficient for MAD Irrigation Adequacy 9/12/2006 99.7 0.012 0.335 0.340 NS 0.209 10/12/2006 72.4 0.134 0.409 0.516 0.091 0.612 11/11/2006 63.5 0.063 NS 0.274 NS 0.450 12/11/2006 98.0 0.044 0.436 0.489 NS 0.379 1/12/2007 100.0 0.002 0.078 0.087 NS NS 2/11/2007 99.9 0.008 0.152 0.151 NS NS 3/13/2007 99.5 0.028 0.279 0.261 NS 0.147 4/12/2007 96.8 0.074 0.430 0.489 NS 0.375 5/12/2007 63.1 0.125 0.254 0.409 0.112 0.833 6/11/2007 38.2 0.179 0.403 0.571 NS 0.604 7/11/2007 73.0 0.127 0.388 0.455 NS 0.768 8/10/2007 98.6 0.030 0.434 0.499 0.065 0.386 9/9/2007 93.3 0.074 0.506 0.594 0.105 0.532 10/9/2007 98.3 0.044 0.414 0.461 0.069 0.335 11/8/2007 97.6 0.054 0.410 0.499 NS 0.382 Scheduling Efficiency 9/12/2006 27.8 0.103 NS 0.257 0.673 NS 10/12/2006 90.8 0.017 NS 0.183 0.766 NS 11/11/2006 93.4 0.066 NS 0.881 NS NS 12/11/2006 89.1 0.035 NS 0.132 0.832 NS 1/12/2007 52.5 0.059 NS NS 0.803 NS 2/11/2007 84.9 0.035 NS 0.156 0.518 NS 3/13/2007 81.0 0.026 NS 0.086 0.635 NS 4/12/2007 86.6 0.027 0.239 0.133 0.433 0.068 5/12/2007 89.4 0.052 0.406 0.347 0.360 NS 6/11/2007 99.9 0.015 0.142 0.150 NS NS 7/11/2007 94.0 0.063 0.557 0.702 NS 0.275 8/10/2007 52.1 0.052 0.196 NS 0.761 0.074 9/9/2007 74.7 0.039 0.075 0.096 0.744 NS 10/9/2007 54.0 0.053 0.107 0.078 0.844 NS 11/8/2007 66.6 0.033 0.248 0.136 0.145 0.079 a Asterisk denotes strong correlation selected as greater than 0. 5 or less than 0. 5 b Correlation coefficients that were not significant at a 95% confidence level were indicated by NS.

PAGE 237

237 Figure 6 1. Probability densities used in the bench test study for A) permanent wilting point, B) field capacity, C) saturation, and D) maximum allowable depletion. A B C D

PAGE 238

238 Figure 6 2. Probability densities used in the GCREC field plot study for A) permanent wilting point, B) field capacity, C) saturation, and D) maximum allowable dep letion. A B C D

PAGE 239

239 Figure 6 3. Response in ir rigation adequacy to varying the input factor of maximum allowable depletion during the bench test study for the A) ET Water, B) Toro, and C) Weathermatic. A B C

PAGE 240

240 Figure 6 4. Response in irrigation adequacy due to varying the input factor of maximum allowable depletion for the Weathermatic for the period ending Nov. 11, 2006 in the GCREC field study where the correlation coefficient was 0. 545 Figure 6 5 Response in irrigation adequac y due to varying the input factor of maximum allowable depletion for the Toro for the period ending Nov. 11, 2006 in the GCREC field study where the correlation coefficient was 0. 871

PAGE 241

241 Figure 6 6. Response in irrigation adequacy due to varying the input factor of maximum allowable depletion for TIME for the period ending May 12, 2007 in the GCREC field study where the c orrelation coefficient was 0.933 Figure 6 7. Response in irrigation adequacy due to varying the input factor of maximum allowable dep letion for RTIME for the period ending May 12, 2007 in the GCREC field study where the correlation coefficient was 0. 833

PAGE 242

242 Figure 6 8. Response by the Weathermatic during the bench test study between scheduling efficiency and varying the input factors of s Figure 6 9. Response in scheduling efficiency due to varying the input factor of saturation for the Weathermatic for the period ending Aug. 10, 2007 in the GCREC field study where the correlation coefficient was 0.920.

PAGE 243

243 Figure 6 10. Response in scheduling efficiency due to varying the input factor of saturation for the Toro for the period ending Apr. 12, 2007 in the GCREC field study where the correlation coefficient was 0.877. Figure 6 11. Response in scheduling efficiency due to varying the input factor of saturation for TIME for the period ending Dec. 11, 2006 in the GCREC field study where the correlation coefficient was 0.931.

PAGE 244

244 Figure 6 12. Response in scheduling efficiency due to varying the input factor of saturation for RTIME for the period ending Dec. 11, 2006 in the GCREC field study where the correlation coefficient was 0.800.

PAGE 245

245 CHAPTER 7 CONCLUSION S The ET controller has the potential to produce water savings without sacrificing landscape quality in Florida when implemented at residences that habitually irrigate excessively and programmed with settings that are appropriate for the landscape. In Hills borough County, where cooperators were already conservative in their irrigation habits, t otal irrigation reductions ranged from 23% 41% compared to the GIR and 23% 34% compared to the ir historical average irrigation In Orange County, where cooperators re gularly irrigated more than 1.5 times the GIR prior to the study, cooperators continued to irrigate at or above the GIR. However, due to the over irrigation habits of the Orange County cooperators, the ET treatment generally reduced irrigation by 12 % to 39 % compared to the neighboring homes with no added technology This treatment increased irrigation application during the wet seasons on both soil types The education provided as part of the ET+Edu treatment was effective at reducing irrigation application by 1 to 49 percentile points in addition to water savings by ET treatment. The time based treatments selected to represent actual homeowner practices in Florida resulted in excessive irrigation during the field plot study. Though the ET controllers decreased irrigation compared to 100% replacement of the net irrigation requirement, there were no additional water savings compared to 60% replacement. A dditionally, the performance scores of irrigation adequacy and scheduling efficiency for the ET controllers were not different from the 60% replacement with all treatments experiencing over irrigation.

PAGE 246

246 Homeowners who already irrigate less than well watered conditions or accept declines in landscape quality on a regular basis will not benefit from using an ET controller in terms of water savings. This was seen in Riverview where cooperators with an ET controller had historically irrigated 23% less than GIR, but increased their irrigation application by 54% compared to their neighbors during the study. H omeowners who set their time clocks to a peak irrigation schedule without updating the controller settings throughout the year may see significant reductions in irrigation by implementing an ET controller, specifically during the fall and winter seasons. Proper controller programming was essential to achieving maximum water savings while maintaining acceptable landscape quality. This occurred in Orange Count y where the ET treatment frequently irrigated 2% to 68% more than the GIR assuming a 60% efficiency, which is an achievable standard, across all seasons. The ET+Edu, which included site specific programming, reduced irrigation by 9 to 72 percentile points compared to ET. If the time and effort of proper programming is not achievable, o ther smart irrigation technologies that do not increase existing irrigation, such as soil moisture sensors, would be more appropriate. Even though water savings are achievab le when using ET controllers, controllers vary by brand and model thus requiring an adequate method for determining the efficacy of the controllers. The current industry standard for evaluating ET controllers, termed the Smart Water Application Technologi es (SWAT) test (IA 2008), uses a daily soil water balance to determine whether the controller over or under irrigates during a 30 day period for a theoretical landscape. However, performance scores of irrigation adequacy (a measure of under irrigation) a nd scheduling efficiency (a measure of over

PAGE 247

247 irrigation) do not translate to real world performance. The EPA adopted the daily SWAT test as the testing methodology for the WaterSense program where ET controllers are considered water saving devices if achie ving thresholds of 80% for irrigation adequacy and 95% for scheduling efficiency. A variety of deficiencies were identified in the daily SWAT test that hinder s the application of the test results outside of the exact testing conditions. In response to som e of these deficiencies, an hourly soil water balance (SWB) was developed as an alternative to the daily SWAT test and b oth models were evaluated against HYDRUS The hourly SWB better match ed t he deterministic soil water model compared to the daily SWAT test thus improv ing estimates of volumetric water content in a shallow root zone This was shown in the first layer of the PSREU study during the calibration period where the NSE was 0.659 for t he hourly SWB but only 0.514 for the daily SWAT, falling below the acceptable threshold for hydrological models of 0.65. Both models failed to meet the threshold during the validation period or during either period of the GCREC study. The hourly SWB had higher NSE for the ET controller treatments, ranging from 0.224 to 0.266, during the validation period of the GCREC study compared to the daily SWAT, ranging from 0.011 to 0.181. Based on the results, the hourly SWB improv ed estimates of volumetric water content in a shallow root zone compared to the current testing standard T he hourly SWB was preferential to the daily SWAT based on the combination of more accurate landscape programming required to obtain high performance scores and the realistic represen tation of soil water movement by hourly SWB resulting in a more

PAGE 248

248 accurate translation of SWAT scores to controller performance in the landscape. The correlation between irrigation adequacy and under irrigation when using the hourly SWB ranged from 0.324 t o 0.783 for the bench test study. For the same study, the correlation was 0.483 to 0.518 for scheduling efficiency and over irrigation. There was not a significant correlation for either performance measure for the daily SWAT model. The ability of the c ontrollers to surpass the WaterSense thresholds was affected by model type with controllers unable to pass the test using the hourly SWB, when manufacturers used program settings that were not realistic for the described landscape. The irrigation adequac y results for the bench test study were most affected due to frequent under irrigation with scores declining as much as 80 percentile points. Thus, the failure rate of irrigation adequacy results frequently went from 0% with the daily SWAT to 100% with th e hourly SWB, with the change in the failing rate ranging from 13 percentile points to 100 percentile points. The decline in scheduling efficiency results were less drastic, falling no more than 30 percentile points and typically falling less than 10 perc entile points. T he differences in scores by model type were rarely a factor for WaterSense labeling criteria in the field plot study when program settings were accurate for the landscape. There was frequent over irrigation thus scheduling efficiency was the primary performance measure affected by the model choice. Failure rates were higher for the hourly SWB, ranging from 75% to 94% compared to failure rates of 47% to 88% for the daily SWAT. However, the WaterSense threshold for scheduling efficiency w as already so high at 95% that both models failed to meet the criteria.

PAGE 249

249 A regression analysis was used to evaluate the stability of the model results on selecting the soil hydraulic parameters of permanent wilting point, field capacity, saturation, and maximum allowable depletion fraction. These four parameters were varied using a Monte Carlo simulation across a wide range using a triangular distribution Irrigation adequacy was most sensitive to the model inputs with a clear relationship with MAD, having correlation coefficients ranging from 0.545 to 0.800 for the bench test study and m aximum correlations ranged from 0.564 to 0.933 for the field plot study. Scheduling efficiency had a strong relationship to s with a correlation coefficient of 0.693 for the bench test study and maximum correlation coefficients ranging from 0.844 to 0.95 0 for the field plot study. Despite such high correlation for all treatments, there was very little increase in the performance results of irrigation adequacy and scheduling efficiency due to increasing values of MAD and s respectively. Therefore, the v ariations in the performance results were highly dependent on the irrigation scheduling techniques of the ET controller and not the model input factors, thus satisfying the purpose of the model with good stability. T he hourly SWB can be used as an altern ative method for testing ET controllers that will more accurately reflect true soil water relationships thus allowing for the estimation of water conservation potential. The improved relationship between ET controller performance and the performance measu res of irrigation adequacy and scheduling efficiency used by the EPA WaterSense program allows the possibility of translatable scores to scenarios beyond the conditions of the testing period.

PAGE 250

250 LIST OF REFERENCES Alle n, R. G., Pereira, L. S., Raes, D., and Smith, M. e vapotranspiration: Irrig Drain Paper No. 56 Food and Agricultural Organization of the United Nations, Rome. Allen, R. G. 56 dual crop coefficient method over an irrigated J. Hydrol ., 229(2000), 27 41. Aquacraft, Inc. ( 2002 ) Performance evaluation of WeatherTRAK irrigation controll ers in Colorado < www.aquacraft.com > ( Oc t. 21, 2005 ) Aquacraft, Inc. ( 2003 ) Report on performance of ET based irrigation controller: Analysis of operation of WeatherTRAK controller in field conditions during 2 002. < www.aquacraft.com > (Oct. 21, 2005 ) ASCE Task Committee on Standardization of Reference Evapotranspiration (ASCE) (2005). T he ASCE standardized reference evapotranspiration equation, R. G. Allen, ed., Reston, VA. Bamezai, A ( 2004 ) LADWP weather based irrigation controller pilot study. < http://www.cuwcc.org/uploads/product/LADWP Irrigation Controller Pilot Study.pdf > (Oct. 31, 2005 ) Banuelos, J. B., Walworth, J. L. Brown, P. W. and Kopec D. M ( 2011 ) Deficit irrigation of seashore paspalum and bermudagrass. Agron J ., 103(6) 1567 1577. < http://www3.wabash.edu/econometrics/EconometricsBook/Basic% 20Tools/Exce lAddIns/MCSiM.htm > (Jan. 23, 2014). Beard, J. B. and Green, R. L. (1994) J. Environ. Qual ., 23(3), 452 460. Hydrol. Process. 20, 3141 3146. Biran, I., Bravdo, B. Bushkin Harav, I., and Rawitz E ( 1981 ) Water consumption and growth rate of 11 turfgrasses as affected by mowing height, irrigation frequency, and soil moisture. Agron J ., 73(1) 85 90. Black, R. J., and Ruppert, K. (1998). Your Florida Landscape: A Complete Guide to Planting and Maintenance. UF IFAS publication. University Press of Florida. Gainesville, FL. Bonos, S. A. and Murphy J. A ( 1999 ) Growth responses and performance of Kentucky bluegrass under summer stress. Crop Sci 39 770 774.

PAGE 251

251 Brooks, R. H. and Corey A. T ( 1964 ) Hydraulic properties of porous media. Hydrology Papers N o. 3 Colorado State University, Fort Collins, CO. Burdine, N. T. ( 1953 ) Relative permeability calculations from pore size distribution data. J Pet Technol ., 5(3) 71 78. Burt, C. M., Clemmens, A. J. Strelkoff, T. S. Solomon, K. H. Bliesner, R. D. Hardy, L. A. Howell, T. A. and Eisenhauer D. E ( 1997 ) Irrigation performa nce measures: efficiency and uniformity. J Irrig Drain Eng ., 123(6), 423 442. < http://www.campbellsci.com/c s616 l specifications > (Mar. 29, 2014). Cardenas Lailhacar, B., and Dukes, M. D. (2012). irrigation controllers: A review of multi Trans. ASABE 55(2), 581 590. Carlisle, V. W., Collins, M. E., Sodek, III, F., and Hammond, L. C. (1985). 85 1. University of Florida, Institute of Food and Agricultural Sciences, Gainesville, FL. Carrik er, R. ( 2006 ) Overview. Department of Food and Resource Economics Report FE 643 University of Florida, Gainesville, FL. Celia, M. A., Bouloutas, E. T., and Zarba, R. L. c onservative Water Resour Res ., 26(7), 1483 1496. Chang, H. J. ( 2007 ) Comparative streamflow characteristics in urbanizing basins in the Portland Metropolitan Area, Oregon, USA. Hydrol Process 21(2) 211 222. Childs, E. C. ( 1969 ) The physical basis of soil water phenomena. Wiley Interscience New York, NY Childs, E. C. and Collis George N ( 1950 ) The permeability of porous materials. Proceedings of the Royal Society of London Ser. A, 201, 392 405. Soil Sci. Soc. Am. J. 54, 1377 1384. Chu, S. T. ( 1978 ) Infiltration during an unsteady rain. Water Resour Res ., 14(3) 461 466. Davis, S. L. ( 2008 ) Evapotranspiration based irrigation controller under dry conditions in Florida. ME Thesis University of Florida, Gainesville, FL.

PAGE 252

252 Davis, S. L. and Dukes, M. D. Perform ance Analysis of Weather < http://www.epa.gov/watersense/docs/university_of_florida_report.pdf > (Dec. 9, 2013). Davis, S. L., Dukes, M. D., an d Miller, G. L. evapotranspiration based irrigation controllers under dry conditions in Southwest Ag r Water M ana g e ., 96( 12 ), 1828 1836. Davis, S. L. and Dukes, M. D. evapotranspiration Agr. Water Manage ., 98(1), 19 28. Davis, S. L. and Dukes, M. D. (2011). Performance Analysis of Weather Based Irrigation Contr ollers: Update with < http://www.epa.gov/watersense/docs/controller_2010 epa report.pdf > (Dec. 9, 2013). Davis S. L. and Dukes, M. D. igation with evapotranspiration Trans. ASABE 55(2), 571 580. Davis, S. L. and Dukes, M. D. (201 4 ) Toro Intelli J. Irrig. Drain. Eng., 140(3) 04013020. Davis, S. L. and Dukes, M. D. In review. J. Irrig. Drain. Eng., 2013. Devitt, D. A., Morris, R. L. and Bowman D. C ( 1992 ) Evapotranspiration, c rop coefficients, and leaching fractions of irrigated desert turfgrass systems. Agron J ., 84 717 723. Devitt, D. A., Carstensen, K., and Morris, R. L. (2008). Residential water savings associated with satellite J. Irrig. Drain. Eng ., 134(1), 74 82. Dukes, M. D. (2012). Trans. ASABE 55(2), 563 569. Dukes, M. D. and Haman, D. Z. CIR1421, Institute of Food and Agricultural Sciences, Universit y of Florida, Gainesville, FL Dziegielewski, B. and Keifer, J. C. (2010). J. AWWA 102(6), 1 15. Florida Statutes. (2006). < http://www.leg.state.fl.us > (Dec. 4, 2006).

PAGE 253

253 Friedman, K., Heaney, J. P., Morales, M., and Palenchar, J. E. (2013). managing residential potable irrigation using parcel J. AWWA 105(7), E372 E386. Fu, J., Fry, J. and Huang B ( 2004 ) Minimum water requirements of four turfgrasses in the transition zone. HortScience 39(7) 1740 1744. Gardner, W. R. ( 1958 ) Some steady state solutions of the unsaturated moisture flow equation with application to evaporation from a water table. Soil Sci ., 85 228 232. Garen, D. C., and Moore D. S ( 2005 ) Curve number hydrology in wate r quality modeling: uses, abuses, and future directions. J AWRA 41(2) 377 388. Gibeault, V. A., Cocker ham, S. Henry, J. M. and Meyer J ( 1989 ) California California Turfgrass Culture 39( 3) 1 9. Gilman, E. F., Wiese, C. L., Paz, M., Shober, A. L., Scheiber, S. M., Moore, K. A., and Brennan, M. (2009). J. Environ. Hort., 27(3), 149 154. Global Water Intellig ence ( GWI ) ( 2011 ) GWI OECD global water tariff survey. Global Water Intelligence 12(9). Gowdish, L. and Muoz Carpena, R. Ampt infiltration and Vadose Zone J ., 8, 470 479. Grabow, G. L., Ghali, I. E. Huffman, R. L. Miller, G. L. Bowman, D. and Vasanth A ( 2013 ) Water application efficiency and adequacy of ET based and soil moisture based irrigation controllers for turfgrass irrigation. J Irrig Dr ain Eng ., 139(2) 113 123. Green, W. H. and Ampt G. ( 1911 ) Studies of soil physics, part I: the flow of air and water through soils. J Agr Sci ., 4 1 24. Haley, M. B., Dukes, M. D., Miller, G. L. (2007). J. Irrig. Drain. Eng., 133(5), 427 434. Haley, M. B., and Dukes, M. D. (2012). J. Irrig. Drain. Eng ., 138(2), 135 144. Hanks, J. and Ritchie, J. T. Soil Sci. Soc. Am Madison, WI.

PAGE 254

254 performance assessment for radioactive waste d Safety, 42(1993), 327 367. Holtan, H. N. ( 1961 ) A Concept for Infiltration Estimates in Watershed Engineering. U.S. Department of Agriculture Agricultural Research Service, Bulletin 41 51, p. 25. Hsiao, T. C. and Acevedo E ( 1974 ) Plant responses to water deficits, water use efficiency, and drought resistance. Agr Meteorol ., 14 59 84. Irrigation Association (IA) (2005). Irrig. Assoc. Water M ana g e Committee Falls Church, VA. Irrigation Association (IA) ( 2007 ) Definition of a smart controller. < www.irrigation.org/WorkArea/linkit.aspx?L inkIdentifier=id&ItemID=405&libID=42 7 > ( Dec 13, 2010 ) Irrigation Association (IA) and Landscape Irrigation Equipment Testing Protocol for Climatologically Based Controllers: 8 th < http://www.irrigation.org/SWAT/Draft_Protocols/Climate Based.aspx > (Mar. 15, 2007) Irrigation Association (IA) (2013). Landscape irrigation auditor 3rd E d. Irrigation Fou ndation Falls Church, VA. Izady, A., Ghahraman, B. and Davari K. (2009). Hysteresis: Phenomenon and modeling in soil water relationship. Iran Agr Res ., 28(1) 47 64. Jia, X., Dukes, M. D., and Jacobs, J. M. J. Irrig. Sci. 28(1), 5 15. Jenson, M. E., Burman, R. D. and Allen R. G ( 1990 ) Evapotranspiration and irrigation water requirements. ASCE manuals and reports on engineering practices No. 70. American Society of Civil Engineers. New York, New York. Jeong, J., Kannan, N., Arnold, J., Glick, R., Gosselink, L. and Srinivasan, R. (2010). hourly rainfall runoff modeli ng capability Water Resour Manage ., 24, 4505 4527. Jones, J. W. and Luyten J. C ( 1998 ) Simulation of biological processes. Agricultural Systems Modeling and Simulation University of Florida, Gainesville, FL. Kenny, J F., Barber, N. L., Hutson, S. S., Linsey, K. S., Lovelace, J. K., and Maupin, M. A. (2009). U.S. Geological Survey Circular 1344.

PAGE 255

255 King, K. W. ( 2000 ) Response of Green Ampt Mein Larsen simulate d runoff volumes to temporally aggregated precipitation. J AWRA 36(4), 791 797. Knisel, W. G. (Ed.). ( 1980 ) CREAMS, A field scale model for chemicals, runoff, and erosion from agricultural management systems. USDA Conserv. Res. Rep., No. 26, 643 pp. Krause, P., Boyle, D. P., and Base, F. Adv Geosci 5, 89 97. Kroes, J. G., van Dam, J. C. Groenendijk, P. Hendriks, R. F. A. and Jacobs C. M. J ( 2008 ) t heory description and user manual. Alterra report 1649, Alterra, Research Institute, Wageningen, The Netherlands. 262 pp. Legates, D. R. and McCabe Jr., G. J. (1999). of fit measures in h Water Resour Res ., 35(1), 233 241. Marshall, T. J. ( 1958 ) A relation between permeability and size distribution of pores. J Soil Sci ., 9 1 8. Mayer, P. W., DeOreo, W. B., Opitz, E. M., Kiefer, J. C., Davis, W. Y., Dziegielewski, B., AWWA Res. Foundation Denver, CO. Mayer, P., DeOreo, W., Hayden, M., Davis, R., Caldwell, E., Miller, T., and Bickel, P. J. irrigation controller < http://www.aquacraft.com /Download_Reports/Evaluation_of_California_Smart_ Controller_Programs_ _Final_Report.pdf > (Jul. 14, 2010). McCready, M. S., Dukes, M. D. and Miller G. L ( 2009 ) Water conservation potential of smart irrigation controllers on St. Augustinegrass. Agr Water Manage ., 96 1623 1632. McCuen, R. H., Knight, Z., and Cutter, A. G. Sutcliffe J. Hydrol. Eng ., 11(6), 597 602. Mein, R. G., and Larson C. L ( 1973 ) Modeling infiltration during a steady rain. Water Resour Res ., 9(2) 384 394. Mendoza, C. A., Therrien, R. and Sudicky E. A ( 1991 ) Version 1.02. Waterloo Center for Groundwater Research University of Waterloo, Waterloo, Ontario, Canada. Metropolitan Water D i strict of Southern California (MWD) ( 2004 ) Weather based controller bench report. < http://www.cuwcc.org/WorkArea/showcontent.aspx?id=2542 > ( Nov 1, 2012 )

PAGE 256

256 Morgan, M. G., and Henr ion M ( 1992 ) Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge University Press Cambridge, U.K Mualem, Y. ( 1976 ) A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resour Res ., 12(3) 513 522. Municipal W ater District of Orange County (MWDOC) and Irvine Ranch Water District (IRWD) (2004). < http://www.allianceforwaterefficiency.org/Landscape_and_Irrigation_Introductio n.aspx?terms=residential+runoff+reduction+study > (Oct. 10, 2012). Muoz Carpena, R., Zajac, Z. and Kuo Y. M (2007). Global sensitivity and uncertainty analyses of the water quality model VFSMOD W. Trans ASABE 50(5) 1719 1732. National Climatic Data Center (NCDC) Radar Data Inventories for < http://www4.ncdc.noaa.gov/cgi win/wwcgi.dll?wwDI~StnSrch~StnID~20004544 > (Jan. 12, 2008). National Climatic Data Center (NCDC) (200 9 Radar Data Inventories for Orlando International < http://www.ncdc.noaa.gov > ( Feb. 20, 2012 ). National Turfgrass Federation, Inc. ( 2003 ) The national turfgrass research initiative. National Turfgrass Evaluation Program Beltsv ille, Maryland. Nielson, D. R., van Genuchten, M. T. and Biggar J. W ( 1986 ) Water flow and solute transport processes in the unsaturated zone. Water Resour Res ., 22(9) 89S 108S. Nofziger, D. L. and Wu J ( 2003 ) CHEMFLO 2000: Interactive S oftware for Simulating Water and Chemical Movement in Unsaturated Soils. National Risk Management Research Laboratory U.S. EPA, Cincinnati, OH. Ogden, F. L. and Saghafian, B. (1997). J. Irrig. Drain. Eng ., 123(5), 386 393. Parlange, J. Y., Lisle I. G. Prasad, S. N. and Romkens M. J. M ( 1982 ) Wetting front analysis of the nonlinear diffusion equation. Water Resour Res 20(5) 636 638. Peacock, C. H. and Dudek A. E ( 1984 ) Physiological response of St. Augustinegrass to irrigation scheduling. Agron J ., 76 275 279. Philip, J. R. ( 1957 ) The theory of infiltration: 4. Sorptivity and algebraic infiltration equations. Soil Sci ., 84 257 264.

PAGE 257

257 Pittenger, D. R., Shaw, D. A. and Richie W. E ( 2004 ) Evaluation of weather sensing landscape irrigation controllers. < http://ucanr.edu/sites/UrbanHort/files/131533.pdf > (Nov. 1, 2012 ) Politis, D. N., and Romano, J. P. J. Am. Stat. Assoc ., 89(428), 1303 1313. Environ. Model l. Softw 23, 1182 1206. Rawls, W. J., Brakensiek, D. L. and Saxton K. E ( 1982 ) Estimation of soil water properties. Trans ASAE 25(5) 1316 1320, 1328. Rawls, W. J., Brakensiek, D. L. and Miller N ( 1983 ) Green Ampt infiltration paramet ers from soils data. J Hydraul Eng ., 109(1) 62 70. Richards, L. A. Phys ., 1, 318 333. Ritter, A. and Muoz Carpena, R. ( Performance evaluation of hydrological models: Statistical significance for reducing subjectivity in goodness of fit J. Hydrol ., 480(2013), 33 45. Robbins, P. and Birkenholtz, T. of th Land Use Policy 20, 181 194. Romero, C. C., and Dukes, M. D. irrigated in central and southwest Florida? A spatial th National Decennial Irrigation Conference, Phoeni x, AZ. Romero, C. C. and Dukes, M. D. (201 4 single J. Irrig. Drain. Eng. 140(2), 04013011 Rosenberg, D. E., Kopp, K. Kratsch, H. A. Rupp, L. Johnson, P. and Kjelgren R ( 2011 ) Value landscape engineering: identifying costs, water use, labor, and impacts to support landscape choice. J AWRA 47(3) 635 649. signal J. Irrig. Drain. Eng ., 138(11), 972 977. Saltelli, A., Tarantola, S. and Campolongo F ( 2000 ) Sensitivity anal ysis as an ingredient of modeling. Stat Sci ., 15(4) 377 395. Sartain, J. B. SL21, Institute of Food and Agricultural Sciences University of Florida, Gainesville, F L.

PAGE 258

258 Scheiber, S. M., Gilman, E. F., Sandrock, D. R., Paz, M., Wiese, C., and Brennan, M. M. (2008). HortTechnol. 18(1), 59 67. Shearman, R. C., and Morris, K. N. (1998). National Turfgrass Evaluation Program Turfgrass Evaluation Workshop Beltsville, MD. Simmons C. S., Nielsen, D. R. and Biggar J. W ( 1979 ) Scaling of field measured soil water properties Hilgardia 47 77 154. im Ecol. Modell. 220(2009), 505 521. J., van Genuchten, M. T., and Sejna, M. lopment and a pplications of the HYDRUS and STANMOD software packages and related Vadose Zone J ., 7, 587 600. J., van Genuchten, M. T., and Sejna, M. Trans. ASABE 55(4), 1261 1274. Smith R. E., Corradini, C., and Melone, F. Water Resour Res ., 29(1), 133 144. Soil Conservation Service ( SCS ) South Florida Water Management District (SFWMD). (2012). round landscape < http://www.sfwmd.gov/portal/page/portal/xweb%20 %20release%203%20water%20conservation/year%20round%20irrigation%20rul e#countyli st > (Mar. 29, 2014). < https://www.swfwmd.state.fl.us/conservation/restrict ions/swfwmd.php > (Mar. 29, 2014). http://www.sjrwmd.com/wateringrestrictions/ > (Mar. 29 2014). round Lawn < http://www.srwmd.state.fl.us/index.aspx?NID=337 > (Mar. 29, 2014).

PAGE 259

259 Sw anson, C. and Fipps G ( 2009 ) Evaluation of smart irrigation controllers: initial bench testing results. AgriLIFE Extension Service, Report TR 354 Texas A&M University, College Station, TX. Swanson, C. and Fipps G ( 2010 ) Initial evaluation of smart irrigation controllers: year two (2009) results. AgriLIFE Extension Service, Report TR 364 Texas A&M University, College Station, TX. Swanson, C. and Fipps G ( 2011 ) Evaluation of smart irrigation controllers: year 2010 results. AgriLIFE Extension Service, Report TR 401, Texas A&M University, College Station, TX. TexasET Network. ( 2012 ) Landscape Plant Water Requirement Calculator. < http://TexasET.tamu.edu > ( Oct 28, 2012 ) The Toro Company. (2 011). < http://media.toro.com/CatalogDocuments/Product%20Literature/11 1052_tis_fam.pdf > (Apr. 2, 2013). U.S. Census Bureau. (20 < http://quickfacts.census.gov/qfd/states/12/12057.html > (Aug. 19, 2012). U .S. Census Bureau. 2011 American Community Survey: Resu lts < http://factfinder2.census.gov/faces/tableservices/jsf/pages/productview.xhtml?sr c=bkmk > (Apr. 25, 2013). U. S. De partment of Agriculture ( USDA ) < http://soils.usda.gov/survey/online_surveys/florida/ > ( Feb. 9, 2013). U. S. Department of Agriculture ( USDA ) < http://SoilDataMart.nrcs.usda.gov/ > ( Jan. 15, 2011). U S. Department of Agriculture ( USDA ) ( 2008 ) Products and Services: HYDRUS 2D Mod el. U. S. Salinity Laboratory. < http://www.ars.usda.gov/Services/docs.htm?docid=8939 > ( Nov 1, 2012 ) U S Environmental Protection Agency (USEPA) ( 2002 ) The Clean Water and Drinkin g Water Infrastructure Office of Water Washington, D.C. U S < http://www.epa.gov/watersense/ about_us/milestones.html > (Oct. 12, 2012). van Genuchten, M. T. ( 1980 ) A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci Soc Am J ., 44 892 898.

PAGE 260

260 Verdi, R. J., Tomlinson, S. A., and Marella, R. L. (2006). 2002: U.S. Geological Survey, Circular 1295 Reston, VA Whitcomb, J. B. on of single < http://www.swfwmd.state.fl.us/documents/reports/water_rate_report.pdf > (Dec. 2, 2013). Willmott, C. J., Ackleson, S. G., Davis, R. E., Feddema, J. J., Klink, K. M., Legates D. J. Geophys. Res. 90(C5), 8995 9005. (MAE) over the root mean sq uare error (RMSE) in assessing average model Clim. Res ., 30(2005), 79 82. Zhang, X., Lin, H. Xiuju, B. Bingziang, Z. Fahe, C. and Zinzhang S ( 2007 ) The most economical irrigation amount and evapotranspiration of the turfgrasses in Beijing City, China. Agr Water Manage ., 89 98 104.

PAGE 261

261 BIOGRAPHICAL SKETCH Stacia L. Davis has always enjoyed learning while attending Colfax Elementary School in Fairmont, WV Wendover Middle School in Greensburg, PA, and Hempfield Area Senior High Schoo l in Greensburg, PA. These schools were important stepping stone s in determining her academi c path as she concentrate d o n the advanced math and science classes offered. It was clear to everyone, including Stacia, that engineering was her niche. Ms. Davis attend ed the University of Pittsburgh where she stud ied civil and environmental engineering. She also worked in the engineering field for companies such as CDM, Allegheny Energy, and Rhea Engineers She passed the FE exam, adding the title of Engineer I n Training before completing her Bachelor of Science and fully intends on sitting for the professional engineering exam as well. Ms. Davis narrowed her focus when leaving PITT and chose to pursue a Masters in Engineering degree in irrigation engineering in the Agricultural and Biological Engineering department at the University of Florida. Here, she researched water conservation in reside ntial irrigation focusing on ET controllers and turfgrass. She continued her research in areas of water modeling and performance testing of ET controllers during her recently completed PhD program also in irrigation engineering at the University of Florida Ms. Davis accepted a faculty position with Louisiana State University located at the Red River Research Station in Bossier City, LA. Her position is part of a water cluster that will address the water issues of the state of Louisiana with her focus on irrigation. She is excited to begin this new chapter of her life with exciting opportunities in research and extensi on.