<%BANNER%>

Evaluation of Modeling Approaches to Estimate Bahiagrass (Paspalum notatum) Yield Affected by Climate Variability in Florida

Permanent Link: http://ufdc.ufl.edu/UFE0044760/00001

Material Information

Title: Evaluation of Modeling Approaches to Estimate Bahiagrass (Paspalum notatum) Yield Affected by Climate Variability in Florida
Physical Description: 1 online resource (131 p.)
Language: english
Creator: Gelcer, Eduardo Monteiro
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: arid -- bahiagrass -- crop -- drought -- evapotranspiration -- fertilization -- florida -- grass -- grassland -- hay -- index -- irrigation -- model -- modeling -- monteith -- nitrogen -- notatum -- paspalum -- pasture -- penman -- priestley -- rain -- stress -- taylor -- water
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Bahiagrass (Paspalum notatum) is a perennial grass widely used for grazing and hay in Florida. In the past few years, exceptional drought has been reported as impacting the forage and causing losses to the grazing herd. As seasonal climate variability has a large importance on the risks faced by farmers, seasonal climate forecast and tools to assist forage-livestock managers can have a positive impact on forage production systems. The objectives of this study were to adapt a simple crop model that uses the Agricultural Reference Index for Drought (ARID) to predict bahiagrass losses and to understand the spatial and temporal variability of ARID in Florida. To verify the influence of drought and N stress on bahiagrass, a field trial was conducted using two irrigation and two fertilization levels. Simulations based on this trial were conducted to validate the bermudagrass CROPGRO-Forage model, and to use the combination of CROPGRO-Forage and ARID to predict bahiagrass yield. Monthly ARID maps in Florida were created to understand ARID’s spatial and temporal variability and to identify potential anomalies caused by the El Niño Southern Oscillation (ENSO). The results of the field trial showed that yield and crude protein concentration are influenced by irrigation and fertilization, with last one being the main factor. The average yield for fertilized treatments was 3.5 higher that non-fertilized ones. The average crude protein content for fertilized treatments was15.7% and for non-fertilized ones it was 7.4%. The result of the simulations using the CROPGRO-Forage showed better predictions in case of the irrigated and fertilized treatment (D-Stat=0.79). The combination of CROPGRO-Forage model and ARID showed better results (D-Stat=0.91). For all simulations, better results were obtained when low water and N stress was observed. The ARID maps showed that typical ARID values for Florida vary throughout the year. During cold months, south Florida had higher values of ARID ENSO showed strong influence during cold months causing reduction of ARID during El Niño and increase during La Niña. The results of this study show that, the spatial and temporal variability of yield as well as ENSO influence on it can be estimated using ARID.
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 Eduardo Monteiro Gelcer.
Thesis: Thesis (M.S.)--University of Florida, 2012.
Local: Adviser: Fraisse, Clyde W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044760:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044760/00001

Material Information

Title: Evaluation of Modeling Approaches to Estimate Bahiagrass (Paspalum notatum) Yield Affected by Climate Variability in Florida
Physical Description: 1 online resource (131 p.)
Language: english
Creator: Gelcer, Eduardo Monteiro
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: arid -- bahiagrass -- crop -- drought -- evapotranspiration -- fertilization -- florida -- grass -- grassland -- hay -- index -- irrigation -- model -- modeling -- monteith -- nitrogen -- notatum -- paspalum -- pasture -- penman -- priestley -- rain -- stress -- taylor -- water
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Bahiagrass (Paspalum notatum) is a perennial grass widely used for grazing and hay in Florida. In the past few years, exceptional drought has been reported as impacting the forage and causing losses to the grazing herd. As seasonal climate variability has a large importance on the risks faced by farmers, seasonal climate forecast and tools to assist forage-livestock managers can have a positive impact on forage production systems. The objectives of this study were to adapt a simple crop model that uses the Agricultural Reference Index for Drought (ARID) to predict bahiagrass losses and to understand the spatial and temporal variability of ARID in Florida. To verify the influence of drought and N stress on bahiagrass, a field trial was conducted using two irrigation and two fertilization levels. Simulations based on this trial were conducted to validate the bermudagrass CROPGRO-Forage model, and to use the combination of CROPGRO-Forage and ARID to predict bahiagrass yield. Monthly ARID maps in Florida were created to understand ARID’s spatial and temporal variability and to identify potential anomalies caused by the El Niño Southern Oscillation (ENSO). The results of the field trial showed that yield and crude protein concentration are influenced by irrigation and fertilization, with last one being the main factor. The average yield for fertilized treatments was 3.5 higher that non-fertilized ones. The average crude protein content for fertilized treatments was15.7% and for non-fertilized ones it was 7.4%. The result of the simulations using the CROPGRO-Forage showed better predictions in case of the irrigated and fertilized treatment (D-Stat=0.79). The combination of CROPGRO-Forage model and ARID showed better results (D-Stat=0.91). For all simulations, better results were obtained when low water and N stress was observed. The ARID maps showed that typical ARID values for Florida vary throughout the year. During cold months, south Florida had higher values of ARID ENSO showed strong influence during cold months causing reduction of ARID during El Niño and increase during La Niña. The results of this study show that, the spatial and temporal variability of yield as well as ENSO influence on it can be estimated using ARID.
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 Eduardo Monteiro Gelcer.
Thesis: Thesis (M.S.)--University of Florida, 2012.
Local: Adviser: Fraisse, Clyde W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044760:00001


This item has the following downloads:


Full Text

PAGE 1

1 EVALUATION OF MODELING APPROACH E S TO ESTIMATE BAHIAGRASS ( Paspalum n ot atum ) YIELD AFFECTED BY CLIMATE VARIABILITY IN FLORIDA By EDUARDO MONTEIRO GELCER A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Eduardo Monteiro Gelcer

PAGE 3

3 To my parents, Jaime and Keky ; my brothers, Arthur and Daniel; my grandmothers, Anna Theresa and L ia; and my aunt Nina -w ithout their support I would not be able to c omplete this and other projects

PAGE 4

4 ACKNOWLEDGMENTS First, I would like to thank God. I have had an amazing life and all small issues have made me grow as a person. Thanks to Dr. Clyde Fraisse, my committee chair, who gave me the opportunity to study in the United States of America which was one of my dreams. Also, I want to thank Dr. Paulo Sentelhas since he guided me during my degree. He provided me with so many opportunitie s including the one of coming to the USA. I admire these two professors for everything that they have done for me and other students, and for their passion for their work. The endless conversation with them helped me to be a better student, person and pro fessional. I will always be grateful for everything that they have done for me. I also give thanks to the other committee members, Dr. Lincoln Zotarelli, Dr. Senthold Asseng and Dr. Yoana Newman since they were great advisors and effectively participated i n the project. I am grateful to Dr. Jim Jones and Dr. Kenneth Boote for all the help and advisement regarding crop models All were very important for a successful outcome of this research project. I give special thanks to Juliana Muoz. She has been part of my family since 2009 and has been very important for me during these past three years. I do not know if without her, I would have been able start and finish my m degree. Myriane Scalco and her fa mily (Wellington and Lilian) were like family during my first semester at college. Gabriel Bernardes whom I grew up with and our farm experiences together directed me towards the agricultural field. Eliane Maffi Yamada whom was a great friend and helped me to solve many of my personal and professional issues. Daniel and Barbara encouraged me to go to graduate school. When I started the program they gave me housing and then became good friends. I give thanks to Wayne

PAGE 5

5 Williams for his unselfishness with every one. He was always present in the computer lab doing everything that was necessary at the moment. Also, he efficiently solved all the problems that I had in the field and in the office. Thanks go out to Tiago Zortea for his friendship, help and patience wh ile teaching me the R programming language and helping me solve problems with it. Kofikuma Dzotsi for the conversations about GIS and for the help to create the maps. Phillip Alderman for teaching me how to use the CROPGRO Forage model. Andr Giordano whom did a lot of data collection and analysis, and was a great friend for a short period of time. Eduardo Alava for his friendship and help when grinding the samples. The staff from the experimental station in Citra especially Mr. Jim Boyer were very import ant for the success of field trial s Thales Barreto that took care of the field trial for the first year. Special thanks go out to Marta Kohmann, Alisson Kovalesky, Flavio Hazan, Fabio Marin, Renan Mendes, Rafael Rodrigues, Bruna Forcelini, Gi acomo DiSanct is, Verona Montone, Roberto Takahashi and Jos Henrique Andreis who m were my acquaintances, friends, support and help. They opened my mind about modeling, computer programming and life Howard H u was always present in the McNair Simulation Lab and also eff ectively participated in this project. He always taught me lessons about life by how we should work hard, live simple lives and seek our life goals. Boi Babo, my fraternity which was my home while I was school and still remains. Special thanks go out to its older members, since they were like a mirror for me, a glance into my future. Because of them, I saw the path through my degrees. Also, I give thanks to The Cooperative Living Organization for giving me shelter during my ti me in the USA, as well as providing me with friends.

PAGE 6

6 TABLE OF CONTENTS ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 2 LITERATURE RE VIEW ................................ ................................ .......................... 19 Bahiagrass ................................ ................................ ................................ .............. 19 Drought ................................ ................................ ................................ ................... 27 El Nio Southern Oscillation (ENSO) ................................ ................................ ...... 31 Crop Models ................................ ................................ ................................ ........... 33 3 EFFECTS OF FERTILIZATION AND IRRIGATION ON PASPALUM NOTATUM GROWTH AND CRUDE PROTEIN CONTENT ................................ ...................... 40 Material and Methods ................................ ................................ ............................. 41 Experimental Site ................................ ................................ ............................. 41 Yield and Nutritive Value ................................ ................................ .................. 42 Soil Moisture and Weather Data ................................ ................................ ....... 43 Stolons and Roots ................................ ................................ ............................ 44 Statistical Analysis ................................ ................................ ............................ 44 Results and Discussion ................................ ................................ ........................... 45 Yield ................................ ................................ ................................ ................. 45 Nutritive Value ................................ ................................ ................................ .. 48 Stolon and Roots ................................ ................................ .............................. 49 Soil Water Content ................................ ................................ ........................... 51 Conclusions ................................ ................................ ................................ ............ 52 4 MODELING OF BAHIAGRASS ( PASPALUM NOTATUM ) GROWTH USING CROPGRO FORAGE MODEL AND THE AGRICULTURAL R EFERENCE INDEX FOR DROUGHT ................................ ................................ ......................... 63 Material and Methods ................................ ................................ ............................. 65 Validation of the CROPGRO Forage Model ................................ ..................... 65 Estimation of Yield under Non Nitrogen Stress Conditions .............................. 69 Statistical Analysis ................................ ................................ ............................ 74 Results and Discussion ................................ ................................ ........................... 75 Validation of the CROPGRO Forag e Model ................................ ..................... 75

PAGE 7

7 Estimation of Yield under Non Nitrogen Stress Conditions .............................. 78 Conclusions ................................ ................................ ................................ ............ 80 5 EFFECTS OF EL NIO SOUTHERN OSCILLATION ON THE AGRICULTURAL REFERENCE INDEX FOR DROUGHT (ARID) IN FLORIDA ................................ 92 Material and Methods ................................ ................................ ............................. 93 Study Area and Data ................................ ................................ ........................ 93 Calculation of ARID Values ................................ ................................ .............. 93 ENSO Deviation ................................ ................................ ............................... 97 Spatial Interpolation ................................ ................................ .......................... 97 Results and Discussion ................................ ................................ ........................... 98 Historical Maps ................................ ................................ ................................ 98 ENSO Effects on ARID ................................ ................................ ................... 100 Conclusions ................................ ................................ ................................ .......... 101 6 SUMMARY ................................ ................................ ................................ ........... 114 APPENDIX FIELD DATA ................................ ................................ ................................ ............... 118 LIST OF REFERENCES ................................ ................................ ............................. 121 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 131

PAGE 8

8 LIST OF TABLES Table page 3 1 Monthly averages of minimum, maximum and average temperature, solar radiation and total rainfall and total rainfall + irrigation in Citra, FL ..................... 55 3 2 Analysis o f variance summary for dry matter (kg of DM ha 1 ) as affected by fertilization, irrigation week of the year, and repetition. ................................ ...... 56 3 3 Average dry matter (kg of DM ha 1 ) as affected by fertilization and irrigation ..... 56 3 4 Average dry matter (kg of DM ha 1 ) as affected by fertilization and week of the year ................................ ................................ ................................ .............. 56 3 5 Analysis of variance summary for crude protein concentration (%) as affected by fertilization, irrigation, week of the year and repetition ................................ .. 57 3 6 Average crude protein concentration (%) as affected by fertilization and irrigation ................................ ................................ ................................ .............. 57 3 7 Average crude protein concentration (%) as affected by fertilization and week of the year ................................ ................................ ................................ .......... 57 3 8 Average crude protein concentration (%) as affected by irrigation and week of the year ................................ ................................ ................................ .......... 58 3 9 Bahiagrass stolon root length (cm cm 3 ) in October 2009 and 2011 in different soil depths for four treatments ................................ .............................. 58 3 10 Bahiagrass stolon root distribution (%) in October 2009 and 2011 in different soil depths for four treatments ................................ ................................ ............ 58 4 1 Soil information used to create the soil file ................................ ......................... 83 4 2 Layer of simulated soil water and the corresponding TDR probe used for comparison ................................ ................................ ................................ ......... 83 4 3 Statatistical analysis for biomass (kg of DM ha 1 cut 1 ) predictions based on the bermudagrass CROPGRO Forage model ................................ .................... 83 4 4 Statatistical analys is for crude protein content (%) predictions based on the bermudagrass CROPGRO Forage model ................................ .......................... 84 4 5 Statatistical analysis for soi l moisture predictions (cm 3 cm 3 ) based on the ber mudagrass CROPGRO Forage model ................................ .......................... 84 4 6 Statatistical analysis for biomass predictions based bermudagrass CROPGRO Forage and ARID and bermudagrass CROPGRO Forage N off .... 84

PAGE 9

9 5 1 County, latitude longitude and number of years of data of each of the COOP stations used to calculated historical ARID ................................ ....................... 104 5 2 Monthly ENSO phase classification according to t he Multivariate ENSO Index (MEI) ................................ ................................ ................................ ................. 10 6 A 1 Dry matter (kg of DM ha 1 ) for four treatments ................................ .................. 118 A 2 Crude protein concentration (%) for four treatments ................................ ......... 119 A 3 Average soil water content (cm 3 cm 3 ) for four treatments ................................ 120

PAGE 10

10 LIST OF FIGURES Figure page 3 1 Split plot design of field trial ................................ ................................ ................ 59 3 2 Linear move syst em used ................................ ................................ ................... 59 3 3 Impact sprinkler gun used to irrigate the field after problems with the linear move system ................................ ................................ ................................ ...... 60 3 4 Time domain reflectometry (TDR) probes ................................ .......................... 60 3 5 Dry matter (kg of DM ha 1 ) for four treatments ................................ .................... 60 3 6 Crude protein concentration (%) for four treatments ................................ ........... 61 3 7 Average soil moisture content (cm 3 cm 3 ) in the first 37.5 cm of the soil profile for fertilized treatments ................................ ................................ ....................... 61 3 8 Average soil moisture content (cm 3 cm 3 ) in the first 37.5 cm of the soil profile for non fertilized treatments ................................ ................................ ................ 62 3 9 Observed yield (kg of DM ha 1 ) and daily average soil water content (cm 3 cm 3 ) in the first 67.5 cm of the soil profile ................................ ................................ 62 4 1 Ground covered to avoid evapotranspiration after adding water ........................ 85 4 2 Diagram for the soil water balance of a reference crop (grass) .......................... 85 4 3 Observed vs simulated herbage mass (kg of DM ha 1 ) using bermudagrass CROPGRO Forage model ................................ ................................ .................. 86 4 4 Simulated vs observed crude protein content (%) using bermudagrass CROPGRO Forage model ................................ ................................ .................. 87 4 5 Simulated vs observed soil moisture content (mm 3 mm 3 ) using bermudagrass CROP GRO Forage model ................................ .......................... 88 4 6 Rainfall + irrigation from a 30 day period sample and the respectively soil moisture content (mm 3 mm 3 ) for the I+F treatment ................................ ............ 89 4 7 Observed vs simulate yield (kg of DM ha 1 ) using ARID as the water stress factor ................................ ................................ ................................ .................. 90 4 8 Observed vs simulate herbage mass (kg of DM ha 1 ) using bermudagrass CROPGRO Forage model with no nitrogen N stress ................................ ......... 91

PAGE 11

11 5 1 NWS Cooperative Observer Program (COOP) stations used to calculate historical ARID ................................ ................................ ................................ .. 107 5 2 Spatial distribution of historical monthly average ARID values for Florida ........ 108 5 3 Spatial distribution of monthly average ARID values for Florida during Neutral phase ................................ ................................ ................................ ................ 109 5 4 Spatial distribution of monthly average ARID values for Florida during El Nio phase ................................ ................................ ................................ ................ 110 5 5 Spatial distribution of monthly average ARID values for Florida during La Nia phase ................................ ................................ ................................ ....... 111 5 6 Spatial distribution of monthly deviation in ARID values during El Nio events from All Years ................................ ................................ ................................ ... 112 5 7 Spatial distribution of monthly deviation in ARID values during La Nia events from All Years ................................ ................................ ....................... 113

PAGE 12

12 ABSTRACT Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science EVALUATION OF MODELI NG APPROACH E S TO ESTIMATE BAHIAG RASS ( Paspalum n ot atum ) YIELD AFFECTED BY CLIMATE VARIABILI TY IN FLORIDA By Eduardo Monteiro Gelcer August 2012 Chair: Clyde William Fraisse Major: Agricultural and Biological Engineering Bahiagrass ( Paspalum notatum ) is a perennial grass widely used for grazing and hay in Florida. I n the past few years, exceptional drought has been reported as impacting the forage and causing losses to the grazing herd. As seasonal climate variability has a large importance on the risks faced by farmers, seasonal climate forecast and tools to assist forage livestock managers can have a positive impact on forage production systems. The objectives of this study were to adapt a simple crop model that uses the Agricultural Reference Index for Drought (ARID) to predict bahiagrass losses and to understand t he spatial and temporal variability of ARID in Florida. To verify the influence of drought and N stress on bahiagrass, a field trial was conducted using two irrigation and two fertilization levels. Simulations based on this trial were conducted to validate the bermudagrass CROPGRO Forage model, and to use the combination of CROPGRO Forage and ARID to predict bahiagrass yield. M onthly variability and to identify potential anomalies c aused by the El Nio Southern Oscillation (ENSO). The results of the field trial showed that yield and crude protein concentration are influenced by irrigation and fertilization, with last one being the main factor The

PAGE 13

13 average yield for fertilized treatme nts was 3.5 higher that non fertilized ones T he average crude protein content for fertilized treatments was 15. 3 % and for non fertilized ones it was 7. 3 % The result of the simulations using the CROPGRO Forage showed better predictions in case of the irrigated and fertilized treatment (D Stat=0.79). The combination of CROPGRO Forage model and ARID showed better results (D Stat=0.91). For all simulations, better results were obtained when low water and N stress was observed The ARID maps showed that ty pical ARID values for Florida vary throughout the year. During cold months, south Florida had higher values of ARID ENSO showed strong influence during cold months causing reduction of ARID during El Nio and increase during La Nia. The results of this st udy show that, the spatial and temporal variability of yield as well as ENSO influence on it can be estimated using ARID.

PAGE 14

14 CHAPTER 1 INTRODUCTIO N The land area dedicated to agriculture in Florida in 2007 was estimated at 1.2 million hectares of which 31% was covered by pasture mostly used for raising cattle and calves. In 2010, the estimated herd was more than 1.7 million animals (cattle and calve s), resulting in an income of billions of dollars (USDA NASS, 2007) In Florida, the most planted perennial warm season grass is the Paspalum notatum Fluegg, commonly known as bahiagrass. This tropical grass, native from South America is sprea d by seed or vegetative stolons (Mannetje and Jones, 1992) Due to its high tolerance to severe defoliation, low f ertility, disease and nematodes, bahiagrass has an important role as forage being used for gr azing and hay and it is a key component for the beef industry in Florida (Burson and Watson, 1995; Newman et al., 2010) Stocking rate and amount of hay purchased are important decisions that a pasture manager makes (Breuer et al., 2005) Among all factors (forage, weathe r conditions, fertilization, amount of concentrate being fed, and production goals) influencing the ideal stocking rate in a given situation, the only factor not controlled by the decision maker is weather conditions. For this reason, decision makers need to understand the influence of weather conditions on grass production to minimize production risks and financial losses (Inyang et al., 2010) A severe and persistent drought may harm forage crops causing loss in the grazing herd and consequently being a stressful situat ion for the herd manager (Peel, 2002) In the case of a dry period, forage based livestock producers can m ake some decisions to reduce the problems. They can either i) purchase feed and/or hay or ii) sell some animals. The first option might be the best if short term drought is expected.

PAGE 15

15 However, in case of a long period with low water availability, the investment in feedstock c an be more detrimental if the return is too small. During the prolonged extreme and exceptional drought of 2007, beef cattle producers had to make some risky decisions to overcome drought problems. Many farmers had to over graze their pastures, buy and feed expensive and low quality hay, allow their brood animals to lose weight and condition or delay the decision to sell their animals until mid September, when is a poor timing to sell cows relative to the annual cattle cycle, since prices on brood cows d ecline during this period (Peel and Meyer, 2002) The same was observed in 2011, when the Florida Climate Center 1 1 reported extreme drought during spring and early summ er and th e U.S. Drought M onitor 1 2 showed that in the panhandle, the water shortage lasted until September. Due to these conditions, higher than normal supplemental feed was necessary to compensate the overall dryness. Also, farmers had to send higher numb er of animals to the feed lot due to poor pasture condition s Since climate variability is a key source of production risk and almost 80% of (USDA NASS, 2007) farmers cannot avoid drought conditions. They can only make decisions to adapt and reduce its impact. As seasonal climate variability has a large importance on the risks faced by farmers, seasonal climate forecast and tools to assist forage livestock manager s can have a positive impact in forage production systems reducing risks associated with it (Fraisse et al., 2006; Lara e t al., 2012) Systems models describe plant growth and development, as well its interaction with soil. Therefore, they can be used as a tool to understand the 1 1 http://coaps.fsu.edu/climate_center/index.shtml 1 2 http://droughtmonitor.unl.edu/

PAGE 16

16 relationships between environmental factors and plants in agricultural systems and also the results of this relationship over time (Pedreira et al., 2011) Th ese models are useful tools to guide decision makers about management, yield variability and crop yield risk. There are several crop models available for use e.g. DSSAT (Decision Support System for Agrotechnology Transfer ) (Jones et al., 2003; Hoogenboom e t al., 2004) EPIC (Erosion Productivity Impact Calculator) (Williams et al., 1989) SWAT ( Soil and Water Assessment Tool) (Neitsch et al., 2005) and others. These models use air temperature and solar radiation as environmental factors controlling growth at maximum rate (under non stress conditions) and water and nutrients as limiting factors to growth under stressful conditions. DSSAT simulate s growth, development and yield of a crop. It also simulates changes in soil water, carbon and N that occur during the crop development. It has been used in research around the world to study crop, fertilizer, irrigation, tillage and pest management, clima te change and climate variability, yield forecast, sustainability, variety evaluation and others. In DSSAT, bahiagrass and other forage species growth can be simulated using the CROPGRO which is a mechanistic model that integrates plant, soil, management, and weather inputs to predict crop growth and composition (Boote et al., 1998a; b). A CROPGRO Forage version was created to simulate perennial tropical forage growth. This version was calibrated and used to simulate growth of bahiagrass ( Paspalum notatum F luegg) (Rymph, 2004), Tifton 85 bermudagrass ( Cynodondactylon (L.) Pers.) (Alderman, 2008), palisadegrass ( Brachiaria brizantha ) (Pedreira et al., 2011), and guineagrass ( Panicum maximum ) (Lara et al., 2012).

PAGE 17

17 As CROPGRO simulates a high number of processe s and the soil plant atmosphere interaction, it is then very complex and, thus, requires a high number of inputs to make more accurate predictions. Therefore, its use is restricted to locations where these data are available. Simpler models that take into account less processes and need fewer inputs can be useful tools for decision makers. The Agricultural Reference Index for Drought (ARID) is an example of a simple agricultural drought index based on a reference crop (grass) that is calculated by subtracti ng the ratio of actual to potential transpiration from 1 (Woli et al., 2012). As ARID is based on physiological principles frequently used in crop models to reduce growth when root imilarities with soil water balance used in the DSSAT. The main differences are that ARID does not consider the division of soil into layers nor root growth and distribution. ARID was successfully used to simulate yield losses caused by drought (Woli, 201 0), since there is a straight relationship between crop transpiration and crop dry matter accumulation (Jensen, 1968; Hanks, 1974; Doorenbos and Kassam, 1979; Jones et al., 2003; Woli, 2010). It relates the amount of total dry matter produced per unit of t ranspired water, which is the water use efficiency (Doorenbos and Kassam, 1979). As the water use efficiency is assumed to be constant for a given crop in a given year, actual yield can be expressed as potential yield times the ratio of actual to potential evapotranspiration (Hanks, 1974). This approach using the ratio of actual to potential evapotranspiration is commonly used in crop models. If this ratio is lower than 1, it indicates that stomatal conductance was decreased at some time of the day to avoid plant desiccation thus water stress occurred (Jones et al., 2003).

PAGE 18

18 The hypothesis of this study were: i) The Agricultural Reference Index for Drought (ARID) can be used as an indicator of crop losses due to water stress; and ii) El Nio Southern Oscillati on (ENSO) influences the variability of soil moisture in Florida causing lower water stress during El Nio and higher stress during La Nia. The objectives of this study were adaptation of a crop model that uses ARID to predict bahiagrass losses caused by drought and identify the influence of ENSO phenomenon on the temporal and spatial variability of ARID in Florida.

PAGE 19

19 CHAPTER 2 LITERATURE REVIEW Bahiagrass Bahiagrass ( Paspalum notatum Fluegg) is rhizomatous warm season perennial grass, largely used for grazing with some production of hay and sod in Florida (Newman, 2007) This species is originally from South America but currentl y it is largely distributed in the southern United States of America Central and South America and some regions in Australia, Africa and Asia (Mannetje and Jones, 1992; Burson and Watson, 1995) In the USA, it can be found from east Texas to the Carolinas including Florida, Georgia, Alabama, northern Arkansas and central Tennessee coverin g more than 1.6 million hectares in southeastern USA and about 1 million hectare in Florida (Burson and Watson, 1995; Inyang et al., 2010; Newman et al., 2010) The most commonly used bahiagrass cultivar is Pensacola (Inyang et al., 201 0; Newman et al., 2010) This cultivar has long and narrow leaves, long stems and higher persistence under grazing. It also has a higher tolerance to colder temperatures than others cultivars. Bahiagrass is widely used in Florida and other areas of the southern US A because of its adaptation to wide range of soil conditions, its establishment by seed, high tolerance to insects, diseases and nematodes and relatively high yields in low fertility soils (Burson and Watson, 1995) Bahiagrass has better development in sub humid and humid subtropical climates with optimum temperature between 25 and 30 o C. Although it tolerates temperatures as low as 10 o C, the top growth is killed by frost This species is well adapted to sandy loams soils and tolerates low fertility and pH. The tolerable soil

PAGE 20

20 pH ranges from 4.5 to 6.5 being 5.5 the ideal (Burson and Watson, 1995; Newman, 2007) I n Florida, bahiagrass annual dry matter (DM) yield varies depending on the soil mois ture and fertility (Newman et al., 2010) In high fertility conditions and adequate water supply, DM yield can be higher, ranging from 13 000 to 16 000 kg of DM ha 1 and daily accumulation varies between 44 to 67 kg of DM ha 1 day 1 (Newman et al., 2010). D ue to warmer temperature and mild winters, production in South Florida is usually higher than in central and north ern areas of the state. From March to May temperatures, solar radiation and daylength might be adequate for bahiagrass growth; however, during this period growth is usually limited by water availability. Inyang et al. (2010) observed yields as high as 25000 kg of DM ha 1 in Ona, FL In this experiment the daily accumulation varied from 38 kg of DM ha 1 day 1 in May 2007 to 158 kg of DM ha 1 day 1 in June 2008. It is known that even with warm temperatures biomass accumulation of bahiagrass and other tropical grasses can be reduced in late summer and fall in the southeast USA (Rymph, 2004) M ore than 85% of yearly bahiagrass production occurs between April and September due to higher temperature and longer days during this period (Mislevy and Everett, 1981; Newman et al., 2010) The limited availability of forage during short daylength months has been one of the most important factor influencing animal production and pasture ma nagement (Sinclair et al., 2001, 2003) To study the environmental factors that may cause reduc tion of biomass accumulation in forages, (Burton et al., 1988) c ompared the biomass accumulation of fertilized, irrigated and rainfed Costal bermudagrass in Tifton, GA. In this experiment the

PAGE 21

21 harvest s were every 24 days from 1 April until 27 October from 1966 to 1968 (23 harvests during this period). For irrigated treatment, the average daily biomass accumulation rate for the period between September 8 and October 2 was 34 kg of DM ha 1 da y 1 which is approximately one third of the accumulation rate of the period between May 9 and June 1 ( 100 kg of DM ha 1 day 1 ) even if the average temperature for the perio d with lower production was 1.1 o C higher. As the treatment was irrigated and well fe rtilized, the authors concluded that some other factor besides nutrition, temperature and water availability would be limiting the growth. They found moderate relationship between yield and temperature as well as between yield and growing degree days. How ever, in this study, yield was highly correlated with daylength and solar radiation. Even if in this study the effects of daylength could not be separated from solar radiation, the authors concluded that daylength strongly affects biomass accumulation. Newman et al. ( 2007) studying the effect of short daylength on grow th of bahiagrass and two other tropical grasses (Common guineagrass, Panicum maximum Jacq; and Tifton 85 bermudagrass, Cynodon spp. Pers.) in three tropical and subtropical locations (South Florida, USA; Gurabo, Puerto Rico; and St. Coix, Virgin Islands) exp osed these forages to an artificially extended photoperiod (15 h) and to natural daylength. In this experiment all treatments were irrigated and fertilized according to its needs. The harvests occurred between October 7, 2002 and April 21, 2003 and between October 6, 2003 and April 19, 2004. Plots were harvest in a 5 week interval from October to February and 4 week interval in March and April. In Florida bahiagrass produced 140 kg of DM ha 1 for natural daylength and 490 kg of DM ha 1 for

PAGE 22

22 extended daylengt h, showing significant difference and yield 3.5 times higher in case of extended photoperiod. In Puerto Rico, the yield was 850 kg of DM ha 1 for natural daylength and 1640 kg of DM ha 1 for extended daylength, which also was a significant difference and t hen corresponds to almost two times more biomass in case of extended photoperiod. Although the difference St. Croix was not significant (560 kg of DM ha 1 for natural daylength and 800 kg of DM ha 1 for extended daylength), the extended daylength treatment resulted in 40% higher yield. The other two species only showed significant increase of biomass for the extended daylength in Florida. In Puerto Rico and St. Croix the difference of biomass between treatments was not significant. For all locations, the cr ude protein concentration in bahiagrass was higher for plants exposed to natural daylength. This difference might be explained by the fact that the crude protein content should be similar for treatments; however, as the dry matter is higher for the plants exposed to extended daylength, the percentage of crude protein is lower. The other species did not show significant difference between the treatments. Also studying the effect of short daylength in forage production, Sinclair et al. (2001, 2003) compared t he yield and crude protein concentration of Pensacola bahiagrass and three other commonly used forage grasses (Tifton 85 bermudagrass; Florakirk bermudagrass, Cynodon dactylon L.; and Florona stargrass, Cynodon nlemfuensis Vanderyst) exposed to natural and extended (15h) daylength. The halogen lamps used to increase the daylength provided less than 2% of photosynthetic active radiation of full sunlight; therefore, the results can be attributed only to variation in daylength and not to the increase of radiat ion. The harvests occurred in a 4 week interval during early summer and a 5 week during the rest of the year. The extended

PAGE 23

23 daylength was imposed from August 19, 1998 to April 21, 1999 and from August 18, 1999 to April 19, 2000. Sinclair et al. (2001) found out that the extended daylength resulted in increase of biomass for the four grasses. However bahiagrass had the biggest increase of 3.59 times for the first year and of 2.02 for the second due to extended daylength treatment, while for the other grasses the maximum value was of 1.84. Sinclair et al. (2003) showed that percentage of crude protein decreased for bahiagrass exposed to extended daylength. T he authors argued that two factors could have influenced the decrease of crude protein: i) decrease of re lative amount of leaves (leaves:total weight ratio) since stems have lower concentration of crude protein than the leaves; and ii) the extended daylength reduced the relative ability to accumulate N during cold months. For Hirata and Pakiding ( 2002) bahiagrass yield decreases during fall due to translocation of reserve substances from the stems to the roots and stolons. Therefore, daylength greatly influences bahiagrass growth and thus being a critical factor in subtropical regions. Bahiagrass is co mmonly known as not being responsive to N and to other mineral nutrients (Twidwell, 2011) However, it has been demonstrated that th e use of balanced fertilization increases yield and bring economic benefits for farmers (Blue and Graetz, 1977; Overman and Blue, 1991; Sumner et al., 1992) Mineral nutrients are part of many components of the plant such as amino acids, nucleic acids, membranes, chlorophyll, cell wall being essential for synthesis of enzymatic and structural pr oteins, osmoregulation, charge balance, energy storage and transfer and others. Nitrogen is the most limiting nutrient for bahiagrass production. It is the one with higher requirement by the plant and when it is used in a balanced fertilization, it usually results in increase

PAGE 24

24 in yield and forage quality (Blue, 1973; Blue and Graetz, 1977; Impithuksa et al., 1984) Potassium is absorbed in large amounts by the plant and due to its mobilization in the soil, it may be deficient in an intensive system, thus regular applications are necessary (Newman et al., 2008) In a study to verify how the amount of N fertilizer applied and its distribution along the growing season affect bahiagrass growth, Blue and Graetz (1977) conducted an experiment near Gainesville applying three different N fertilizer rates (0, 112 and 224 kg of N ha 1 year 1 ). The total amount of N fertilizer was split in 1, 2, 4, 8 or 16 applications through the year. The forage was harvested at 3 c m height four times a year (mid May, early July, mid August and early October). For the treatments with 0, 112 and 224 kg ha 1 of N year 1 the average yield s for the three years were respectively 2780, 8320 and 13210 kg of DM ha 1 year 1 For the tr eatments with 112 kg of N ha 1 year 1 average yields varied from 7880 kg of DM ha 1 year 1 for 8 applications to 8800 for 16 applications. For the treatment with 224 kg of N ha 1 year 1 the yields varied from 12620 kg of DM ha 1 year 1 for 1 application to 13500 for 4 applications. The forage N content increased significantly with N fertilizer rates. In this study, forage production showed high dependency of amount of N fertilizer applied and low correlation with N fertilizer applicatio n splitting. The authors observed that forage yield and N content were influenced by a number of N fertilizer applications in an expected way. For low number of applications, the yield and N content were higher early in the growing season. It was result of higher N availability during spring and early summer when 1 or 2 applications were used.

PAGE 25

25 Sumner et al. (1992) reported a series of field trials in nine counties in South Florida. At each location, five 465 m 2 areas were selected and assigned one of five t reatments: 1) no fertilizer; 2) 67 kg of N ha 1 applied in March; 3) 67 kg of N ha 1 and 50 kg of K and P ha 1 applied in March; 4) 134 kg of N ha 1 split in two applications of 67 kg of N ha 1 in March and September; and 5) 67 kg of N ha 1 100 kg of P ha 1 and 50 kg of K ha 1 applied in March and 67 kg of N ha 1 and 50 kg of K ha 1 applied in September. For all sites, the forage was harvested at 5 cm height every 30 to 60 days from April to December. The authors observed an average increase of 1450 kg of DM ha 1 year 1 for the treatments receiving 67 kg of N ha 1 in co mparison with treatment with no N fertilizer The increase of N fertilizer from 67 to 134 kg of N ha 1 split in two applications resulted in increase of more than 500 kg of DM ha 1 year 1 T he comparison between treatment 4 and 5 resulted in an increase of almost 800 kg of DM ha 1 year 1 indicating a positive response in yield when phosphorus and potassium were applied with the N fertilizer The difference in percentage of crude protein was relatively small between treatments. For all treatments the percentage of crude protein immediately increased after fertilization. The results of these studies show that bahiagrass is benefited by the use of mineral fertilizer resulting in higher yield an d crude protein content. In Florida, in regards to grazing fields has been recommended application of 56 kg ha 1 of N year 1 for low input systems and 180 kg ha 1 of N year 1 ( 90 kg ha 1 in spring and 90 kg ha 1 in fall) for high input systems. In case of hay production, the recommendation is 80 kg of N ha 1 cut 1 until middle of August (Beaty et al., 1974; Newman, 2007; Newman et al., 2010)

PAGE 26

26 Bahiagrass has an excellent drought resistance when grow ing in an appropriate climatic region. This resistance is due to dehydration avoidance and dehydration tolerance (Beard, 1989). The dehydration avoidance is mainly because of the capability of the plant to close their stomata quickly in case of significant water st ress, while dehydration tolerance is due to a dormancy capability during winter (Kim, 1987). To verify the influence of irrigation in bahiagrass growth, Beaty et al. (1974) conducted an experiment for two years in Americus, GA using irrigation and rainfed treatments combined with four N fertilizer rates (0, 84, 168, and 336 kg of N ha 1 ) in six combinations of split applications during the season. The authors observed higher yields in case of irrigation combined with high levels of N fertilizer. Moreover, i n a dry year, the increase of yield due to irrigation was 2165 kg of DM ha 1 while in a wet year, the increase due to irrigation was 1303 kg DM ha 1 indicating that irrigation as well N fertilizer contributed to increase bahiagrass yield. Overman et al. ( 1990) reported two experiments conducted with irrigated and rainfed Pensacola bahiagrass combined with several N fertilizer application rates. The first one in Americus, GA from 1959 to 1962 where the irrigation was added so rainfall plus irrigation summed at least 38 mm per week, and the second in Thorsby, AL from 1956 to 1959 where irrigation was applied when soil moisture tension reached 0.55 atm. In Americus, it was observed an increase of the difference between irrigated and rainfed with the increase o f N fertilizer applied. In this experiment, N fertilizer rates higher than 350 kg of N ha 1 did not contributed to increase yield. In Thorsby, the use of irrigation resulted in increase of yield especially for high N fertilizer application rates. For all N fertilizer rates, the use of irrigation reduced the yield variability between years. For the

PAGE 27

27 same field trial in Thorsby, Ashley et al. (1965) reported that for four years irrigation affected yield and in one year, the difference between irrigated and rai nfed treatments was not significant. For all years, fertilization resulted in significant increase in yield. Furthermore, interaction between irrigation and fertilization was observed for most part of the time. These studies indicate that even though bahia grass is well adapted to the southeast USA and it is a species with excellent tolerance to drought, the use of irrigation resulted in an increase of bahiagrass growth and yield, especially in the case of high N fertilizer application rates. In the case of low N fertilizer application rates, N is the limiting factor for growth; therefore, the use of irrigation has low impact on yield. Drought Agriculture is an economic activity directly related to drought and it is the first one to be affected by it. Drought is responsible for the greatest loss in agriculture production (Song et al., 2004; Wilhite, 2007; Farooq et al., 2009) The constant increase in food demand and the need of hi gh efficiency in food production has created more attention on drought effects on agriculture. Consequently there has been much interest and ongoing research directed towards drought monitoring and forecasting, as well as minimizing drought effects on agri cultural output. Drought is a pervasive natural hazard and a normal part of the climate. Depending upon the perspective, drought can be categorized into four types: meteorological, agricultural, hydrological, and socio economic (Wilhite and Glantz, 1985) For all cases, drought is c aused by insufficient moisture due to precipitation deficit through a certain period of time (Mckee et al., 1993; Piechota and Dracup 1996; Mo and Schemm, 2008) An agricultural drought occurs when the water available in soil is not sufficient to

PAGE 28

28 support a crop or forage growth at a particular time (Wilhite and Buchanan Smith, 2005; Woli et al., 2012) The onset and termination of a drought period is difficult to determine (Dagel, 1997; Wilhite and Buchanan Smith, 2005) Drought indices are used to quantify drought intensity, compare curre nt conditions to previous droughts, and provide a regional overview of potential impacts of droughts (Woli et al., 2012). A drought index is a scientifically based numerical index associated to some cumulative effects of an extended and anomalous moisture deficiency and provides quantification of a drought situation (Wo rld Meteorological Organization, 1984; Lourens and Jager, 1997) Several drought indices have been described and each one has a specific use. A commonly used index to express drought is the Palmer Drought Severity Index (PDSI) (Palmer, 1965) This index is used to determine when a drought or a wet spell begins and ends (Heim Jr., 2002) Since PSDI is based on precipitation departure from normal and soil water ano maly and it is more influenced by weather variables than by soil and plant process, it is more useful as a meteorological index than as an agricultural index (Woli et al., 2012). Also, this index estimates reference evapotranspiration (ETo) using Thornthwa ite methodology, which is not a reliable method to estimate ETo in Florida (Gelcer et al., 2010b) and underestimates runoff (Hayes, 2006) The Standardized Precipitation Index (SPI) presented by (Mckee et al., 1993) is a meteorological index used to express drought in terms of precipitation deficit, percent of normal, and probability of non exceedance (H eim Jr., 2002). Since this index is based on precipitation anomalies and is not soil plant atmosphere relationship based, it does

PAGE 29

29 not account for water losses from evapotranspiration and runoff, therefore, it is weakly related to agricultural drought (Woli et al., 2012). Keetch and Byram ( 1968) presented the Keetch and Byram Drought Index (KBDI), which is based on evapotranspiration and rainfall. This index was primarily created to calculate soil moisture as factor of occurrence and spread of wildfir es. This index takes into account soil plant atmosphere relationship; however, since it was developed to use drought as an indicator of wildfire threat to forests, it is not suitable to indicate drought index (Woli et al., 2012). Since agricultural drought s is the result of the interaction of rainfall, air temperature, solar radiation and other meteorological parameters with soil and plant characteristics (Lourens and Jager, 1997; Sentelhas, 2010; Sivakumar, 2010) an agricultural drought index should take into account these relationships. This type of index should related weather variables with soil moisture availability in relation to crop development and it should be used to monitor water stress in plants (Woli et al., 2012). To better quantify water deficit in crops, Woli et al. (2012) developed the Agricultural Reference Index for Drought (ARID), which is a simple and non crop specific drought index that takes into account the soil plant atmosphere relationship in a daily time step. ARID is based on a reference crop (grass) actively growing in a well drained soil as well as completely covering the soil surface. ARID uses a simple soil water balance for the reference grass having a 400 mm soil layer with evenly distributed roots (Woli et al., 2012). The evaluation of ARID was done in two ways (Woli et al., 2012). The first one using volumetric soil water content in the first 40 cm of soil measured during one year in

PAGE 30

30 ater balance algorithm also computes volumetric soil water content in the root zone. The second one using the water stress index for photosynthesis (WSPD) calculated using the Decision Support Systems for Agrotechnology Transfer (DSSAT) (Jones et al., 2003 ; Hoogenboom et al., 2004) CERES Maize (Jones and Kiniry, 1986) in 16 locations in Alabama, Florida, and Georgia. The WSPD was chosen because the DSSAT CERES Maize has been largely tested and used ( Hodges et al., 1987; Fraisse et al., 2001; 2002 ; Lpez Cedrn et al., 2005). Also, for each one of the 16 locations, seven other drought indices were compared with WSPD. When ARID estimated volumetric soil water content was compared with measured one, the Willmott agreement index (D Stat) (Willmott 1981) was 0.92 for Citra and 0.94 for Bronson and the modeling efficiency was 0.66 for Citra and 0.73 for Bronson, indicating very accurate predictions by the model. When ARID was compared with WSPD, the coefficient of determination (R 2 ) varied between 0 .73 and 0.95 and root mean square error (RMSE) varied between 0.06 and 0.35. Among all drought indices tested, ARID was the one with better predictions of water stress. These results indicate that ARID can be effectively used to calculate soil water moistu re and water stress. Furthermore, ARID can characterize agricultural drought better than several drought indices that are commonly applied to agriculture (Woli et al., 2012). x to calculate yield of four major crops grown under rainfed conditions. Irrigated and rainfed yield data of cotton, maize, peanut, and soybean for several years and locations Variety Testing program and from the database of the DSSAT crop growth model

PAGE 31

31 (Hoogenboom et al., 2004). The irrigated yields were used to estimate the potential yield for a given crop and season while the rainfed yields were used to test ARID. Overall, th e author observed good results using ARID to estimate losses. For maize, the D Stat was 0.87 while for cotton it was 0.55. The best results were observed for maize because it is a crop more sensitive to water stress while for cotton the results were not as good because it is a crop with relative tolerance to water stress. These results indicate that ARID has potential to estimate yield losses due to drought for several crops (Woli, 2010). A RID is calculated from the ratio of actual to potential evapotranspi ration rates; therefore, the only inputs in the calculation are reference evapotranspiration, rainfall data and a few soil characteristics. Due to the simplicity of this index, it can be widely calculated for different regions and time periods. Since ARID is influenced by rainfall and other weather variables, it is greatly impacted by climate variability and weather phenomena (e.g. El Nio Southern Oscillation). El Nio Southern Oscillation (ENSO) The El Nio Southern Oscillation (ENSO) is a coupled ocean a tmosphere phenomenon that is considered the main source of interannual climate variability in the world (Ropelewski and Halpert, 1996; Fraisse et al., 2006) During an El Nio event, the oceanic component is characterized by an abnormal increase of sea surface temperature in the central and/or eastern equatorial Pacific Oce an while the atmospheric component (Southern Oscillation) corresponds to a rise in sea level atmospheric pressure in the western Pacific (Trenberth, 1997) La Nia is the opposite phenomenon, which is the decrease of sea surface temperature in the western South America. During this phase, the trade winds strengthen causing lower atmospheric

PAGE 32

32 pressure and stronger rainfall in western Pacific. Changes in the atmospheric circulation cause changes in sea surface temperature as wind stress affect the movement and surface temperature of the equatorial Pacific ocean, which in turn influence rainfall patterns due to latent heat exchange feedbacks between the waters and the atmosphere (Oliver and Hidore, 2002; Fraisse et al., 2006) El Nio and La Nia events return every two to seven years (Winsberg, 2003) and its peak is between November and January (Trenberth, 1997). If neither El Nio nor La Nia is present in the Pacific Ocean, the condition is called Neutral which represents more than 50% of the years (W insberg, 2003) ENSO influences climate pattern in different regions around the world due to global ocean and atmospheric flow patterns. There is evidence linking El Nio to drought events in Asia, Africa, Australia and Central America (Oliver and Hidore, 2002; Fraisse et al., 2006) In southeast USA ENSO has a strong influence during cold months while this influence is much lower during warm months (Winsberg, 2003). During El Nio, the jet stream is fortified and pulled southern driving the storms from California to Florida (Hildebrand, 2006) Also, during this phase, moisture is advected by the subtropical jet from the tropical Pacif ic into southeast USA (Ropelewski and Ha lpert, 1996) La Nia phenomenon has the opposite effect in Florida, weakening and moving the jet streams northwards. The shifting of jet streams caused by ENSO phenomenon results in, on average during winter, increase of 30 to 40% of rainfall during El Nio and decrease of 10 to 30% during La Nia The variation of clouds and rainfall among Neutral, El Nio and La Nia phases influences the incoming solar radiation and then the air temperature. During winter

PAGE 33

33 months, the average temperature is 1 to 1.5 C below normal in El Nio years and 1 to 2.2 C above normal in La Nia years. La Nia influence in temperatures during this Winsberg, 2003). For the months of January and February temperatures exhibit greater departure from normal for both ENSO Due to its strong influence in climate, ENSO has been associated with drought. In southeast USA, La Nia is related to drought while El Nio to higher water availability Schemm, 2008) As drought indices are used to quantify drought, they are i nfluenced by ENSO and several authors have described this relationship in the USA. Piechota and Dracup (1996) studying the association between the Palmer Drought Severity Index (PDSI) and ENSO found relationship between drought and ENSO for several locatio ns in the USA. In the s outheast USA, the authors found that drought episodes occur regularly during La Nia. Karnauskas et al. (2008) asserted that in the s outheast USA PDSI is significantly correlated with ENSO. In a study to verify the relationships betw een ENSO and the 6 month Standardized Precipitation Index (SPI 6) over the southeastern USA, Mo and Schemm (2008) found that wet spell is more likely to start during La Nia while the opposite is true for El Nio. Can et al. ( 2007) used the Standardized Precipitation Index (SPI) to analyze abnormal moisture conditions in the Colorado River Basin ( s outhwest USA) and demonstrated that La Nia is related to extreme dry events in the region. Crop Models Crop models are used to understand the relationship among weather, crop growth and soil in agricultural systems over time. These models describe crop growth and

PAGE 34

34 development using mathemati cal models (Wallach, 2006a; Pedreira et al., 2011) thus it can be an important tool in agriculture to simulate growth and yield. When validate for certain conditions, a mod el allows experimentation and provides support for planning decisions in research, technology transfer, and agricultural development (Jones et al., 2003) These models are useful as a t ool to guide decision makers regarding management, yield variability and crop risk. A well known and tested simulation model is the Decision Support System for Agrotechnology Transfer (DSSAT) (Jones et al., 2003; Hoogenboom et al., 2004) which is a worldw ide used software that integrates knowledge about soil climate, crops and management. This software simulates growth, development and yield of a crop and changes in soil water, carbon and N that occur during the crop development. DSSAT incorporates models of 16 crops and offers a user friendly interface that allows users to simulate options for crop management over a number of years to assess the risks associated with each option (Jones et al 2003). In DSSAT, bahiagrass and other forage species growth can be simulate using the CROPGRO which is a mechanistic model that integrates plant, soil, management and weather inputs to predict crop growth and composition (Boote et al., 1998a; b) The CROPGRO has been tested and used for different crops such as soybean ( Glycine max (L.) Merr.), peanut ( Arachis hypogaea L.), dry bean ( Phaseolus vulgaris L.) (Boote et al., 1998b), faba bean ( Vicia faba L.) (Boote et al., 2002) cowpea ( Vigna unguiculata L. Walp.) (Bastos et al., 2002) velvet bean ( Mucuna pruriens ) (Hartkamp et al., 2002) bahiagrass (Rymph, 2004) bermudagrass ( Cynodondactylon (L.) Pers.) (Alderman, 2008) canola ( Brassica napus L.)

PAGE 35

35 (Sassendran et al., 2010) palisadegrass ( Brachiaria brizantha (A. Rich.) Stapf) (Pedreira et al., 2011) and guinegrass ( Panicum maximum Jacq.) (Lara et al., 2012) During the first half of the 1990s, Kelly ( 1995) adapted the CROPGRO to model bahiagrass growth as a component of a crop rotation system for simulating peanut cropping systems in Florida. The author changed some of the species paramet ers and calibrated the model using published experimental data from three locations. The first one in Jay, Florida from 1952 to 1959; the second one in Thorsby, Alabama from 1956 to 1959; and the third one in Eagle Lake, Texas from 1978 to 1980. In Thorsby irrigated and rainfed treatments were combined with four N fertilizer application rates. In the other two experiments, five N fertilizer rates were compared under rainfed conditions. To simulate the cutting or grazing, the DSSAT pest damage routine was used in which the mower or cows were the pest. The annual version of the model showed relationship with N fertilizer application rate and responded to interannual variability, however, for all situations, the simulations overestimated the biomass production. Based on CROPGRO used to simulate bahiagrass growth, Giraldo et al. ( 2001) parameterized the model for Brachiaria decum bens The author adjusted the input parameters that define crop species, cultivar and ecotype. The B. decumbens model was calibrated using experimental data from four field trials in Colombia and the validation using two data sets also from the same countr y. In the validation process was observed that the simulated values ranged from 81 to 101 % of observed values. With these results, the authors attested that the calibration of CROPGRO for B. decumbens resulted in better results than previous ones obtained for bahiagrass. However, Rymph (2004) demonstrate that this code structure resulted in good biomass predictions during

PAGE 36

36 warm months but overpredicted during winter and spring. It was mainly due to the fact that this model is annual in opposite to the grass es that are perennial. Annual versions of the model do not allow winter dormancy and regrowth after complete harvest of leaves and stems or after freeze damage (Rymph, 2004). To add characteristics from a tropical perennial grass to the model, Rymph (2004) created a CROPGRO Forage version. In this version, the model was modified to simulate the C4 photosynthetic pathway, to add a perennial storage organ (stolon), a dormancy process to alter partitioning of growth and mobilization of N during short daylengt h periods. Also, the freeze damage routine was modified to allow partial loss of green structures due to freezing temperatures. This new version was based on the version of CROPGRO which was distributed as part of DSSATv4 (Hoogenboom et al., 2003) Data from two experiments in Gainesville, Florida, two in Ona, Florida and one in Eagle Lake, Texas (the same used to create the first bahiagrass model) were used to calibrate the model. In these five field trials was done a total of 303 measurements of forage mass and 227 of forage crude protein. The average herbage mass of all experiments was 3066 kg of DM ha 1 while the predicted one was 3145 kg of DM ha 1 when leaf level photosynthesis option was used and 3150 kg of DM ha 1 when daily canopy photosynthesis option was used. The use of the new version resulted in reduction of the RMSE from 1003 to 857 kg of DM ha 1 when leaf level photosynthesis option was used and from 930 to 782 kg of DM ha 1 using daily c anopy photosynthesis. Analyzing the scatterplot, the author observed that the CROPGRO Forage version overestimated herbage mass at higher levels and underestimated at lower levels of growth. The prediction of herbage N concentration and its seasonal variab ility were also

PAGE 37

37 improved for the forage version. The observed N concentration was 15.6 g N ha 1 whist the use of the leaf level photosynthesis resulted in underestimation of N concentration (14.6 g N ha 1 ) and the daily canopy photosynthesis resulted in ov erestimation (17.6 g N ha 1 ). The CROPGRO Forage model was modified and calibrated by Alderman ( 2 008) to simulate Tifton 85 Bermudagrass. The author made two minor adjustments in the code. The first one was in relation the harvesting algorithm that was modified to better simulate bermudagrass harvest. The change was done to reduce the amount of leaves left in the field after the each harvest. The second modification is related to the temperature sensitivity for soil organic matter decomposition. Parameters were changed to increase the base temperature for decomposition, resulting in lower N mineralizat ion rates during winter and early spring. The aut hor used data from a field trial conducted in 2006 and 2007 near Gainesville, Florida. In this trial, four N fertilizer application rates (0, 45, 90, and 135 kg of N ha 1 cutting 1 ) were used. For each year, harvests started in mid July and occurred every 28 days until mid October resulting in four regrowth cycles for each year. For the second and fourth growing period of each year, weekly plant samples and photosynthesis measurements were done. The results of this study showed good prediction of biomass, leaf area index and leaves mass. The RMSE for shoot mass varied from 190 to 745 kg ha 1 and D Stat vari ed from 0.54 to 0.95. For leaf N concentration the RMSE varied from 1.8 t o 8.4 % and D Stat varied betwe en 0.33 and 0.94. It shows that the calibrated CROPGRO Forage well predicted bermudagrass yield and leaf N concentration.

PAGE 38

38 To predict palisadegrass ( Brachiaria brizantha (A. Rich.) Stapf. cv. Xaraes) growth, Pedreira et al. (2011) modified and calibrated th e CROPGRO Forage version. For this new perennial version, an improved method and code for defining forage harvest conditions was used. A new input file (MOW) was created defining the harvest dates, the MOW and RSPLF parameters which correspond respectively to stubble mass and leaf fraction in the stubble corresponding to the time when top growth harvest is simulated. The parameters were developed using values and relationships reported in the literature and comparing simulated with observed values from a fi eld experiment conducted in Piracicaba, Brazil during two seasons (2005/06 and 2007/08) under irrigation and high N conditions. In this field experiment, the average observed biomass was 3358 kg ha while the predicted one was 3573 kg of DM ha showing a small overestimation by the simulation. The RMSE observed for palisadegrass simulations (538 kg of DM ha ) was lower than the previously observed indicating better prediction through the year for the palisadegrass. The simulated and observed leaf area index also showed great similarity during the growing season. Using the palisadegrass model as a starting point, Lara et al. (2012) calibrated the CROPGRO Forage model to predict guineagrass ( Panicum maximum Jacq. cv. se these two species are C4 and have similar morphology (Lara et al., 2012) Data from 17 months (December 2002 to April 2004) and 11 regrowth cycles of an irrigated and well fertilized treatment localized in Piracicaba, Brazil were used to calibrate the model. The calibration proces s was done similarly to what was described above for the palisadegrass model. To validate the model, data from an irrigated and fertilized treatment at the same location for the

PAGE 39

39 growing season 2005/06 were used. The results of this study showed that the le af area index was well predicted during summer and poorer relationship was found during winter. The authors attributed this to the fact that during winter, there was a lot dead material in the pasture that might be registered by the scanner as leaf area. D uring the calibration process, the average predicted biomass was 6576 kg of DM ha while the observed was 6535 kg of DM ha and the RMSE was 494.2 kg of DM ha For the validation, the predicted biomass was 6678 kg of DM ha whereas the observed was 65 48 kg ha and the RMSE 478.3 of DM kg ha The results of these studies show that CROPGRO Forage model is an efficient tool to integrate physiological aspects of different species of forage. It can be used as tool to simulate plant growth, to understand how management affects crop development, to make decisions regarding management and to understand the risks associated with forage production.

PAGE 40

40 CHAPTER 3 EFFECTS OF FERTILIZATION AND IRRIGATION ON PASPALUM NOTATUM GROWTH AND CRUDE PROTEIN CONTENT Agriculture is an economic activity directly related to drought and it is the first one to be affected by it. Drought is responsible for the greatest loss in agriculture production (Song et al., 2004; Wilhite, 2007; Farooq et al., 2009) A severe and persistent drought may harm the forage crops causing losses in the grazing herd and consequently being a stressful situation for the herd manager (Peel, 2002) During the prolonged extreme and exceptional drought of 2007, beef cattle producers had to make some risky decisions to overcome drough t problems. Many farmers had to over graze their pastures, buy and feed expensive and low quality hay. They also let their brood animals to lose weight and condition or delay the decision to sell them until mid September. This being a poor time to sell cow s relative to the annual cattle cycle, since prices on brood cows decline during this period (Peel and Meyer, 2002) In 2011, t he Florida Climate Center 3 1 reported extreme drought during spring and early summer and the U.S Drought Monitor 3 2 showed that in the panhandle, the water shortage lasted until September. Due to these conditions, higher than normal supplemental feed was necessary to compensate the overall dryness. Also, farmers had to send higher number of animals to the feed lot due to poor condition of pasture. The objective of this study was to verify the influence of drought, nitrogen stress, time of the year and their combination on bahiagrass dry matter accumulation and nutritive value. 3 1 http://coaps.fsu.edu/climate_center/index.shtml 3 2 http://droughtmonitor.unl.edu/

PAGE 41

41 Material and Methods Experimental Site This study was conducted between March 2009 and November 2011 at the University of Florida at the Plant Science Research and Education Unit (PSREU), Citra, bahiagrass that was established in the field at least five years prior to the experiment. T he soil in the field experiment was Arredondo Gainesvill e association (Thomas et al., 1979) Using Kppen classification, the climate in C itra is humid subtropical (Henry et al., 1994) w here the averag e annual temperature is 21.5 o C and the total annual precipitation is above 1200 mm. Two irrigation levels (irrigated and non irrigated) and two fertilizer levels (fertilized and non fertilized) were established in a strip split plot design r esulting in four treatments: 1) i rrigated and fertilized (I+F); 2) irrigated and non fertilized (I+NoF); 3) non irrigated and fertilized (NoI+F); and 4) non irrigated and non fertilized (NoI+NoF). Each plot had an area of 930 m 2 (Fig. 3 1). From April 2009 to June 2011 irrigation was applied using a line ar move system (Fig. 3 2) at a fixed rate of 17 19 mm per application twice a week (Mondays and Thursdays). No irrigation was applied when a rainfall event greater than 25 mm occurred previously two days to the scheduled irrigation event. Due to mechanical problems in the irrigation system an impact sprinkler gun (Fig. 3 3) was used to apply 12 15 mm of water three times a week (Monday, Wednesday and Friday) between June and September of 2011 and 7 10 mm between September and November of 2011. The irrigation was skipped if the total rainfall was higher than 10 mm on the previous two days to the scheduled irrigation event.

PAGE 42

42 Fertilized treatments received a broadcast application of N and K at a rate of 90 kg of N and K ha 1 28 days before the first harvest of each year and 90 kg of N ha 1 and 45 kg of K ha 1 on the day right after each harvest. If necessary, 3.5 l ha 1 of 2 4 D was applied in the field at the beginning of each year to reduce weeds infestation. Yield and Nutritive Value Four weeks before the first scheduled harvest of each year, all treatments were staged by mowing Fresh biomass harvests happened from May 2009 to December 2009, from June 2010 to November 2010 and from May 201 1 to November 2011 in a 4 week interval except for the last harvest of 2009 that occurred in 6 week interval, totalizing 23 harvests during the experiment period. For all harvests the stubble height was approximately 5. 1 cm. Plant dry matter (DM) accumulat ion for each treatment was evaluated by harvesting five representative areas of 5. 4 m 2 in 2009. Due to change in the equipment used to harvest, for following two years the size of each one of the representative areas was 6. 1 m 2 From each harvested area, s ubsamples were taken to measure dry matter. For that, the fresh subsamples were weighted and then dried at 65 o C for subsequent dry weight determination. In the afternoon after each sample collection, all treatments were staged by mowing. Herbage crude prot ein (CP) determination was on three out of the five dry subsamples for each treatment for each harvest. These dry subsamples were ground in a Wiley mill (Model 4 Thomas Wiley Laboratory Mill, Thomas Scientific, Swedeboro, NJ) to pass a 1 mm stainless steel screen. The dry subsamples from July 29 2009 were not available ; thus it was not possible to do N analysis for this harvest. Nitrogen analyses were carried out at the Forage Evaluation Support Laboratory (FESL). In the lab the samples were digested usin g a modification of the aluminum

PAGE 43

43 block digestion procedure of Gallaher et al. ( 1975) Sample weight was 0.25 g, the catalyst used was 1.5 g of 9:1 K2SO4:CuSO 4 and digestion was conducted for at least 4h at 375C using 6 ml of H 2 SO 4 and 2 ml H 2 O 2 Nitrogen in the digestate was determined by semiautomated colorimetry (Hambleton, 1977) Percentage of CP was estimated as N multiplied by 6.25. Soil Moisture and Weather Data Time domain reflectometry (TDR) probes (CS 616, Campbell Scientific, Inc. Lo gan, UT) connected to a datalogger (CR 10X, Campbell Scien tific, Inc., Logan, UT ) were installed on May 14, 2009 at five depths, 5, 15, 30, 45, and 60 cm (Fig. 3 5a) in each of fertilized plot and at three depths, 5, 15, and 30 cm (Fig. 3 5b) in each of th e non fertilized plots. The soil moisture was logged at 15 min u te interval Since for non fertilized treatments, the deepest TDR probe was at 30 cm depth, the comparison between soil moisture in fertilized and non fertilized treatment was done on the first 37.5 cm of the soil profile. 37.5 cm was used because the sensitive volume extends approximately 7.5 cm from the rods and the deepest probe for non fertilizer treatment was at 30 cm. Two automatic rain gauges (Campbell Scientific, Inc., Logan, Utah) were installed and the rainfall for rainfed treatments and rainfall plus irrigation for irrigated treatments were recorded every 15 min. Additional weather data necessary (minimum and maxim um air temperature, solar radiation, wind speed and relative humidity) were retrieved from the Florida Automated Weather Network (FAWN) weather station located at the PSREU farm, approximately 700 m from the experimental field. Table 3 1 shows a summary of the weather data from FAWN weather station and from the two rain gauges.

PAGE 44

44 Data treatment was done to reduce the errors in relation to rainfall and irrigation measurements. If the difference in measurement of the two rain gauges was 4 mm or more on a non ir rigated day, data from the FAWN station was used to indicate which rain gauge was showing the right value. On 17 November 2010 when the field was being staged, the cables connecting the TDR to the datalogger were broken and they were fixed on March 2011; therefore during that time, soil moisture and rainfall data were not collected. For this reason, rainfall data from the FAWN station were used during this period. Stolons and Roots Bahiagrass stolons roots were collected on October 2009 and October 2011 fo r all treatments. Using a 0.05 m diameter soil auger of know n volume, soil cores were extracted at four different depths: 0 15, 15 30, 30 60, and 60 90 cm in five different points of each treatment plot. After washing away the soil above a fine sieve, stol ons roots and organic debris were stored in plastic bags at 4 o C until further cleaning. Samples were then placed in a glass bowl placed above a light plate and stolons roots were handpicked and placed in Petri dishes. Stolon roots length distribution (SRLD ) for each soil core was determined using a scanner and Winrhizo (Rgent Instrument Inc., Quebec City, Canada) software. Statistical Analysis For yield and nutritive value, statistical analyses were performed with a general linear model as follow: where Y ijkl he residual error

PAGE 45

45 for i=2, j=2, k=5, l=9. Multiple comparisons of mean were performed with Bonferroni test using the pairwise.t.test function in R language (Venables et al., 2011) and significance was accepted at P<0.05. Results and Discussion Yield In general, fertilization irrigation and week of the year that the harvest occurred in and the interaction between fertilization and irrigation as well between fertilization and week of the year significantly affected DM (Table 3 2). Fertilization, irrigation and the joint effec t between them were significant because for non fertilized treatments, N and K were the limiting factors for growth; therefore, the use of irrigation had low effect on yield. However, for the fertilized treatments, low nutrient stress occurred and therefor e water was the limiting factor for growth, thus irrigation resulted in an increase of DM production (Table 3 3) As weeks represent the time of the year that the harvest happened and these factors significantly affected DM, it was noticeable that atmosphe re conditions throughout the year influenced plant growth. It was mainly due to the fact that bahiagrass is a warm season grass, from tropical and sub tropical regions influenced by temperature, solar radiation and especially daylength (Sinclair et al., 2001, 2003; Rymph, 2004; Newman et al., 2007) The joint effect between fertilization and week was also significant, indicating that the interaction between these factors. During summer, temperature, solar radiation a nd daylength were favorable for plant growth; therefore; the use of fertilization greatly affected yield because nutrients were limiting growth. On the other hand in fall, weather conditions limited plant growth resulting in less effect of fertilization d uring this period (Table 3 4)

PAGE 46

46 F ertilization greatly impacted yield ( Table 3 3 ). The average yield for fertilized treatments was 3059 kg of DM ha 1 which is 3.5 times higher than average yield for non fertilized treatments (855 kg of DM ha 1 ). The use of i rrigation resulted in increase in yield ( Table 3 3 ) H owever, it was lower than the impact caused by fertilization. The average yield for irrigated treatment was 2124 kg of DM ha 1 which is 1.2 times higher than the average yield for non irrigated treatmen ts (1790 kg of DM ha 1 ). These results are consistent with previous research. For Beaty et al. ( 1974) the irrigated treatment produced 1.1 times more than rainfed treatment, while the production for the treatment with 336 kg of N ha 1 year 1 was 4 times higher than the treatment with no fertilizer. Blue ( 1973) and B lue and Graetz ( 1977) observed yields 4 times higher for treatment with 448 kg of N ha 1 year 1 than with no fertilizer. For the I+F treatment the peak production occurred from early June to late August (Fi g. 3 5) which corresponds to a growing period from May to August. During this period, temperature and solar radiation were optimum, resulting in a higher pote ntial for the plant to growth. In each year, the highest yield for this treatment was observed on late June/early July what was also observed by Johnson et al. ( 2001) It shows that the combination of high temperature, solar radiation and long daylength in June associated with low or no water and nutrient stress was favorable for plant growth. For the NoI+ F treatment, the peak production varied from year to year due to atmosphere conditions associated with variability on water stress (Fig 3 5) For fertilized treatments, yield start ed to decrease on late August, when daylength is around 12.5 h. Rymph (2004 ) used 12.5 h of light as threshold for dormancy. For the author, if

PAGE 47

47 daylength is lower than 12.5 h, bahiagrass dormancy is progressively induced and partitioning of carbohydrate and N to the stolons is increased. The differences between fertilized treatme nts were more prominent for the harvests corresponding to the May and June growing cycles. It is mainly due to the lack of rainfall associated with high temperatures in May (Gelcer et al., 2010a), causing high water stress on rainfed areas. However, for ir rigated fields, this period has good conditions for plant development, because the average temperature is close to the optimum, solar radiation is abundant and daylength is long. In 2009 and 2011, the late June/early July harvests were also different for I +F and NoI+F treatments even if in June 2009, only 19 mm of water was applied through irrigation and in June 2011 it was only 22.9 mm For the September and October harvests of 2011, the differences between fertilized treatments were large even if during this period only 28.7 mm of water was applied. However, an average of 125.6 mm month 1 of water was applied in July and August 2011 and during this period the difference in yield between these treatments was small It indicates that the difference in yield between fertilized treatments was not only result of water availability on the current growing cycle, it also was results of environmental conditions and interaction of other factors in the past months. As mentioned before, the interaction of irrigation a nd atmosphere conditions affected yield For non fertilized treatments, irrigation had a small effect. Larger difference s between I+NoF and NoI+NoF treatments were observed only for two harvests ( 06/29/10 and 10/19/11). For the rest of the period the diff erence between these treatments was small It was mainly the result of the nutrients stress in the non fertilized plots. As N K

PAGE 48

48 or both were limiting the growth, the addition of water had low impact on yield. It indeed agrees with the values reported in t he literature (Beaty et al., 1974) Nutritive Value The nutritive value was also affected by fertilization, irrigation and week of the year, as well by the interaction between irrigation and week and between fertilization and week (Table 3 5 ). Di fferently from yield, CP concentration was not impacted by the interaction between fertilization and irrigation. It indicates that irrigation and fertilization affected CP concentration, however, the effect of the fertilizer was not influenced by the irrig ation and the effect of irrigation was not affected by the fertilizer. The average CP concentration for fertilized treatments was 15. 3 % which corresponds to more than double of the CP concentration of non fertilized treatments (7. 3 %). This variation is lar ger than the ones previously reported in the literature (Blue, 1973; Blue and Graetz, 1977; Sumner et al., 1992; Johnson et al., 2001) F or few harvests, large differences in CP concentration between the two fertilized treatments were observed, and for all harvests, the difference between the non fertilized treatments was small ( Fig. 3 6 ). The average CP concentration for i rrigated treatment was 11. 1 % and for non irrigated treatment was 11. 3 %, which is not significantly different (Table 3 6) It indicates that CP concentration is strongly affected by fertilization, but in this case, the use of irrigation did not affect yearly average CP concentration. Time of the year that the harvest happened also affected CP concentration. For all treatments, CP concentration during warm period is lower than for cold periods (Table 3 7) These results are consistent with previous resul ts for bahiagrass ( Sumner et al., 1992 ; Sinclair et al., 2003 ; Newman et al., 2007). However, the reason for this effect is unknown. Sinclair et al. (2003) argued that it could be the result of decrease of leaves:

PAGE 49

49 stems ratio due to increase of stems which is material with lower CP concentration or due to a decrease of relative plant ability to accumulate N during cooler periods. For Newman et al. (2007) this response could be result of dilution of total CP because during warm periods DM is higher causing l ower CP concentration. Stolon and Roots For all treatments, most part of the stolon roots was found in the first 15 cm of the soil profile (from 4 1 to 6 5 %). The relative amount of total stolon roots in the first 30 cm varied between 58 and 7 8 %, in the 30 60 cm layer it was between 10 and 20% and in the 60 90 cm layer it was between 6 to 22%. Similar results were observed by Doss et al. ( 1960) that found 50% of roots in the first 10 cm and 76% in the first 25 cm of the soil profile. In average, higher amount of stolon roots were found in the non fertilized treatments. The total of stolon roots length in the soil profile in the I+F treatment was 11 2 cm while in the I+NoF it was 13 7 cm. For N oI+F, the total of stolon roots length was 132 cm while for NoI+NoF it was 154 cm. These results do not agree with the previously reported in the literature Hirata and Pakiding ( 2003) observed an increase of amount of stolon roots when fertilizer was applie d. However, Blue (1973) showed a non linear relationship between the amount stolon roots mass and fertilizer applied. Furthermore, the author obtained a small reduction of stolon roots mass when fertilization was doubled from 224 to 448 kg of N ha 1 It ma y indicate that small and medium N fertilize r application levels cause an increase of stolon and roots mass, but high levels reduce underground biomass accumulation. Also, these two experiments did not account for potassium, which was also applied after each harvest in Citra.

PAGE 50

50 The use of irrigation increased the relative amount stolon roots in the first 30 cm of the soil profile from 2009 to 2011. Even if for the I+F treatment the stolon roots length reduced from 4.4 to 2.5 cm cm 3 in the 0 15 cm layer, th e relative amount of stolon roots in the first 30 cm of the soil profile increased from 58 to 7 4 % for this treatment and from 6 6 to 79% for I+NoF. For the fertilized treatment s, the total stolon root length decreased from 2009 to 2011, mainly due to a big reduction in the first layer, which is the one with highest amount of stolon roots (Table 3 9 ). As the amount of stolon roots decreased for the first 15 cm, its distribution in the soil profile was also affected. For the I+F treatment in 2009, the first 0 15 cm layer contributed to 49% of the total while in 2011 it was 43%. However, in 2009 the 15 30 cm layer contributed with only 9% of the total amount of roots, while it increased to 3 1 % in 2011. For the NoI+F treatment, there was a big reduction in stolon roots length in the first layer and no difference in the other layers. This treatment caused changes in stolon root distribution (Table 3 10 ). For the I+NoF treatment, the amount of stolon roots slightly increased in the 0 15 cm layer from 2009 to 2011 a nd decreased in the other layers. In the 15 30 cm layer it was reduced by 50% (Table 3 9 ). It resulted in higher relative amount of stolon roots in the first layer and lower in the other ones (Table 3 10 ). For NoI+NoF treatment the only big variation was i n the 15 30 cm. Due to an increase of stolon and roots in this layer, the relative amount decreased in the other layers. It is important to highlight that in 2009 there was a stronger effect of the management adopted in the previous 5 years when the field was harvested fewer times a year and no irrigation or fertilization were applied. This management is similar to the one used for the NoI+NoF treatment, the only difference is the number of harvests

PAGE 51

51 during the year. That is why the NoI+NoF treatment had les s variation between the two samplings and I+F had the biggest variation. The temporal variability of stolon roots mass was also observed by Blue (1973). These results cannot be extrapolated to other months, since stolon root mass is affected by the time of the year (Rymph, 2004) Soil Water Content In general during the growing season, the amount of water had a great variation for each treatment and for each year ( Fig. 3 7 and 3 8 ). As expected, the variation on the irrigated treatments was lower than the n on irrigated ones. The lowest values of soil moisture occurred on the rainfed treatments, but the highest values occurred for both irrigated and non irrigated treatments Overall, 2009 was the year with the highest soil moisture for all treatments and 2010 the year with the lowest. This can be attributed to rainfall, which was the highest in 2009 and the lowest in 2010. The I+NoF treatment had the smallest variation in the first 37.5 cm of the soil profile through the whole experimental period (Fig. 3 8) T his was mainly due to the fact that irrigation provided high amount of water to the soil and lower evapotranspiration occurred in this treatment due to low plant growth. The NoI+F treatment caused higher variability of soil water through the years (Fig. 3 7 ) This was caused by the higher water demand of this treatment in comparison with the NoI+NoF one and due to water shortage in comparison with the irrigated treatments. Fo r this experiment, average soil moisture for the each growing cycle did not have a linear relationship with the differences in yield for each growing cycle between irrigated and non irrigated treatments. For some growing cycles, there was a large difference in yield between the two fertilized treatments; however, the difference in avera ge soil moisture between them was small, and vice versa ( Fig. 3 9 ) For example,

PAGE 52

52 for the late June 2011 harvest the difference in yield between I+F and NoI+F was 1942 kg of DM ha 1 and the difference in daily average soil water content during the correspondent growing cycle was 0.00 3 cm 3 cm 3 For the September 2010 harvest the difference in yield between these two treatments was 389 kg of DM ha 1 while the difference in daily aver age soil moisture content was 0.0 32 cm 3 cm 3 This indicates that the average soil moisture during the growing cycle did not completely explain the whole yield variability. It is important to consider its interaction with stolon roots mass, which varies t h roughout the year (Rymph, 2004), variability of soil moisture and t he interaction of weather conditions (temperature, solar radiation and photoperiod) Conclusion s Under the conditions of this experiment, fertilization, irrigation and time of the year as w ell their interaction affect yield. For non fertilized treatments, the use of irrigation caused small or no effect in yield, since N and K were limiting the growth. For only two out of 23 harvests the I+NoF and NoI+NoF were significantly different. For fer tilized treatments, low or no N and K stress occurred, therefore the use of irrigation caused increase in yield. Fertilization had stronger impact in yield than irrigation. The use of fertilization increased 3.5 times the yield, while irrigation only incre ased 1.2 times. The average yield for fertilized treatments was 3059 kg of DM ha 1 while for non fertilized treatments it was 855 kg of DM ha 1 For irrigated treatments, the average yield was 2124 kg of DM ha 1 whilst for non irrigated it was 1790 kg of D M ha 1 When nutrients or water were not limiting growth the June/early July harvest had the highest yield of each year. It was mainly due the ideal weather conditions

PAGE 53

53 (temperature near to the optimum abundant solar radiation and long daylength ) during t he growing cycle. For the I+F treatment, the peak production occurred from early June to late August. For fertilized treatments, growth starts to decline in late August, when the daylength is around 12.5 h due to changes in partitioning caused by daylength Crude protein concentration was affected by irrigation, fertilization and time of the year. Fertilization had stronger effect than irrigation, as observed for DM accumulation. For fertilized treatments, the average CP concentration was 15. 3 % while for non fertilized ones it was 7. 3 %. For irrigated treatments the average CP concentration was 11. 1 % and for no n irrigated treatment was 11. 5 %. The highest values of CP concentration were observed for the last harvest of the year (late November or early Decemb er) for the I+F treatment. It shows that the combination of irrigation, fertilization and time of the year impacted CP concentration. In general, most part of the stolon roots was found in the first 15 cm of the soil profile. 4 1 to 63% of the total stolon roots of the soil profile were found in the 0 15 cm layer, 58 to 79% in the 0 30 cm layer, 10 to 20% in the 30 60 cm layer and 6 to 22% in the 60 90 cm layer. Higher amount of stolon roots were found in the non fertilized treatments. The stolon roots lengt h in the I+F treatment was11 2 cm cm 3 while in the I+NoF it was 13 7 cm cm 3 For NoI+F, stolon roots length was 132 cm cm 3 while for NoI+NoF it was 154 cm cm 3 As the first stolon roots sampling happened in October 2009 and the second one happened two ye ars later, the first sampling was stronger affected by the management adopted before the beginning of the field trial, what was similar to the management

PAGE 54

54 used in the NoI+NoF treatment. Thus, the NoI+NoF had the lowest variation between the two sampling and I+F had the greatest. In general, soil water content had larger variation in the non irrigated treatments. It occurred due to increase of soil moisture during a rainfall event and decrease for a long period of water shortage. Overall, 2009 was the year wi th the highest soil water content and 2010 with the lowest. It was mainly due to the highest amount of rainfall observed in 2009 and the lowest in 2010. The treatment with lowest variation of soil moisture content was the I+NoF. It was caused by the high i nput of water due to rainfall and irrigation and low water demand due to low plant growth. Soil moisture did not have straight relationship with yield variability. For some growing cycles, the difference in yield between the fertilized treatments was large and the difference in soil moisture was small and vice versa. It is important to consider the interaction of soil water and stolon roots distribution, which varies throughout the year. Also, the interaction of weather conditions with soil water is importa nt. If soil water content is low and atmospheric conditions are ideal or vice versa, growth will be reduced.

PAGE 55

55 Table 3 1. Monthly averages of minimum, maximum and average temperature, solar radiation and total rainfall and total rainfall + irrigation in C itra, FL Year Month Avg. Temp. ( C ) Solar Radiation Rainfall Rainfall+Irrigation Min. Max. Avg. (MJ m 2 day 1 ) (mm) (mm) 2009 Jan 6.7 20.7 13.7 10.9 54.4 54.4 Feb 6.0 21.6 13.8 15.3 34.0 34.0 Mar 10.9 25.5 18.2 17.9 42.4 42.4 Apr 12.9 27.1 20.0 20.7 61.2 61.2 May 18.3 30.2 24.3 18.2 240.5 245.6 Jun 22.0 32.7 27.4 20.8 149.4 168.4 Jul 22.1 31.9 27.0 18.4 107.2 148.6 Aug 22.0 32.4 27.2 16.9 117.9 145.5 Sep 21.2 31.2 26.2 15.7 119.9 139.4 Oct 17.4 29.1 23.3 13.4 35.1 37.8 Nov 11.7 23.7 17.7 10.7 63.0 90.7 Dec 9.5 19.9 14.7 7.3 66.8 75.2 Avg. 15.1 27.2 21.1 15.5 1091.7 1243.3 2010 Jan 2.9 16.9 9.9 10.8 55.1 65.5 Feb 4.0 16.6 10.3 12.9 110.5 118.9 Mar 7.8 21.0 14.4 16.0 128.0 169.2 Apr 13.8 28.0 20.9 20.9 25.7 73.7 May 20.1 32.2 26.2 21.6 94.7 127.0 Jun 22.5 34.7 28.6 21.7 93.5 173.5 Jul 23.5 34.1 28.8 20.2 169.4 199.1 Aug 24.5 33.3 28.9 17.1 31.8 36.6 Sep 21.2 32.9 27.1 16.5 40.9 124.5 Oct 13.8 29.8 21.8 16.8 0.0 66.5 Nov 9.9 25.4 17.6 12.0 35.8 48.3 Dec 1.1 17.1 9.1 10.9 117.1 117.1 Avg. 13.8 26.9 20.3 16.4 902.5 1319.8 2011 Jan 5.0 18.6 11.8 10.8 53.1 53.1 Feb 10.1 22.7 16.4 12.6 0.0 0.0 Mar 11.2 25.9 18.5 17.1 103.4 163.6 Apr 14.1 30.0 22.1 21.9 54.9 119.6 May 16.7 32.0 24.3 23.4 35.6 86.9 Jun 20.7 34.3 27.5 19.3 136.7 159.5 Jul 22.6 33.5 28.0 18.2 80.0 205.5 Aug 23.6 34.3 29.0 17.9 184.9 310.6 Sep 21.0 31.9 26.5 15.2 114.3 138.7 Oct 13.6 26.7 20.2 13.3 176.8 181.1 Nov 11.2 24.5 17.9 10.9 46.7 46.7 Dec 9.6 23.1 16.4 9.0 6.1 6.1 Avg. 15.0 28.1 21.5 15.8 992.4 1471.4

PAGE 56

56 Table 3 2. Analysis of variance summary for dry matter (kg of DM ha 1 ) as affected by fertilization (F), irrigation (I), week of the year (W), and repetition (R) Source of variation Degree s of freedom Sum of square Mean square F value p value Fertilization (F) 1 558551108 558551108 4047.85 <0.01 Repetition (R) 4 790413 197603 1.43 0.3 7 F x R Error 1 4 551946 137987 Irrigation (I) 1 12769534 12769534 143.97 <0.01 I x R Error 2 4 354774 88694 F x I 1 2929262 2929262 21.56 <0.01 F x I x R Error 3 4 543434 135859 Week (W) 8 201831049 25228881 78.49 <0.01 I x W 8 4396710 549589 1.71 0.09 F x W 8 130897339 16362167 50.91 <0.01 F x I x W Error 4 8 3443160 430395 1.34 0.22 Residuals 408 131141467 321425 4047.85 Table 3 3. Average dry matter (kg of DM ha 1 ) as affected by fertilization and irrigation Irrigated Non Irrigated Average Fertilized 3305 aA 2812 aB 3059 a Non Fertilized 942 bA 768 bA 855 b Av g. 2124 A 1790 A 1957 Different capital letters within lines and small letters within columns are significantly different by Bonferonni test (P<0.05). Table 3 4. Average dry matter (kg of DM ha 1 ) as affected by fertilization and week of the year Week Fertilized Non Fertilized Average 18 2653 cA 425 cB 1539 b 22 3723 bA 465 cB 2094 ab 26 4422 aA 902 bB 2662 a 30 3812 abA 1029 abB 2420 ab 34 4020 abA 1163 abB 2592 a 38 2949 cA 1282 aB 2115 ab 42 1871 dA 1063 abB 1467 b 46 1152 deA 496 cB 824 bc 48 350 eA 111 cB 230 c Avg. 3059 A 855 B 1957 Different capital letters within lines and small letters within columns are significantly different by Bonferonni test (P<0.05).

PAGE 57

57 Table 3 5. Analysis of variance summary for crude protein concentration (%) as affected by fertilization (F), irrigation (I), week of the year (W), and repetition (R) Table 3 6. Average crude protein concentration (%) as affected by fertilization and irrigation Irrigated Non Irrigated Average Fertilized 15. 0 aA 15.5 aA 15.3 a Non Fertilized 7.2 bA 7.4 bA 7.3 b Average 11.1 A 11.5 A 11.3 Different capital letters within lines and small letters within columns are significantly different by Bonferonni test (P<0.05). Table 3 7 Average crude protein concentration (%) as affected by fertilization and week of the year Week Fertilized Non Fertilized Average 18 17.2 aB 8.8 bA 13 .0 ab 22 15.3 aBC 6.8 bB 11.1 b 26 12.9 aC 6.8 bB 9.9 b 30 13.6 aC 7.4 bB 10.5 b 34 13.7 aC 7.1 bB 10.4 b 38 15.5 aBC 7.1 bB 11.3 b 42 17.3 aB 7.5 bB 12.4 b 46 17.3 aB 7.3 bB 12.3 b Avg. 15. 3 A 7. 3 B 11. 3 Different capital letters within lines and small letters within columns are significantly different by Bonferonni test (P<0.05). Source of variation Degree s of freedom Sum of square Mean square F value p value Fertilization (F) 1 3992.9 3992.9 26619.3 <0.01 Repetition (R) 2 0.3 0.1 1.0 0. 51 F x R Error 1 2 0.3 0.2 Irrigation (I) 1 11.6 11.6 38.7 <0.05 I x R Error 2 2 0.6 0.3 F x I 1 1.7 1.7 3.1 0.22 F x I x R Error 3 2 1.1 0.6 Week (W) 8 254.9 31.9 14.0 <0.01 I x W 8 45.1 5.6 2.5 <0.05 F x W 8 121.3 15.2 6.7 <0.01 F x I x W Error 4 8 26.5 3.3 1.5 0. 18 Residuals 220 501.6 2.3

PAGE 58

58 Table 3 8 Average crude protein concentration (%) as affected by irrigation and week of the year Week Irrigated Non Irrigated Average 18 13.2 aAB 12.9 aA 13 .0 ab 22 10.6 aAB 11.2 aA 11.1 b 26 9.0 aB 10.5 aA 9.9 b 30 10.3 aAB 10.6 aA 10.5 b 34 10.1 aB 10.6 aA 10.4 b 38 11.0 aAB 11.5 aA 11.3 b 42 12.4 aAB 12.3 aA 12.4 b 46 13.0 aAB 11.6 aA 12.3 b Avg. 11. 1 A 11. 3 A 11. 3 Different capital letters within lines and small letters within columns are significantly different by Bonferonni test (P<0.05). Table 3 9 Bahiagrass stolon root length (cm cm 3 ) in October 2009 and 2011 in different soil depths for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Depth I+F I+NoF NoI+F NoI+NoF 2009 2011 2009 2011 2009 2011 2009 2011 0 15 4.4 2.5 4.2 4.9 6.0 4.1 5.0 4.7 15 30 0.8 1.8 2.6 1.3 1.2 1.3 1.6 2.4 30 60 0.9 0.6 0.9 0.4 0.6 0.7 1.0 0.9 60 90 1.0 0.2 0.9 0.4 0.4 0.8 0.8 0.7 Table 3 10 Bahiagrass stolon root distribution (%) in October 2009 and 2011 in different soil depths for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Depth I+F I+NoF NoI+F NoI+NoF 2009 2011 2009 2011 2009 2011 2009 2011 0 15 49% 43% 41% 62% 65% 49% 49% 45% 15 30 9% 31% 25% 17% 13% 15% 16% 23% 30 60 20% 20% 17% 11% 13% 17% 19% 17% 60 90 22% 6% 17% 10% 9% 19% 16% 15%

PAGE 59

59 Figure 3 1. Split plot design of field trial Figure 3 2. Linear move system used

PAGE 60

60 Figure 3 3. Impact sprinkler gun used to irrigate the field after problems with the linear move system Figure 3 4 Time domain reflectometry (TDR) probes a ) at five depths for fertilized treatments and b ) at three depths for non fertilized treatments Figure 3 5. Dry matter (kg of DM ha 1 ) for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF ) a) b)

PAGE 61

61 Figure 3 6. Crude protein concentration (%) for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Figure 3 7. Average soil moisture content (cm 3 cm 3 ) in the first 37.5 cm of the soil profile for fertilized treatments *Bars indicate one standard deviation from the mean.

PAGE 62

62 Figure 3 8 Average soil moisture content (cm 3 cm 3 ) in the first 37.5 cm of the soil profile for non fertilized treatments *Bars indicate one standard deviation from the mean. Figure 3 9 Observed yield (kg of DM ha 1 ) and average soil water content (cm 3 cm 3 ) in the first 67.5 cm of the soil profile for the correspondent growing cycle of each harvest for fertilized treatments

PAGE 63

63 CHAPTER 4 MODELING OF BAHIAGRA SS ( PASPALUM NOTATUM ) GROWTH USING CROPGRO FORAGE MODEL AND THE AGRICULTURAL R EFERENCE INDEX FOR DROUGHT The Decision Support System for Agrotechnology Transfer (DSSAT) (Jones et al., 2003; Hoogenboom et al., 2004) is a worldwide used software that integrates knowle dge about soil, climate, crops and management. DSSAT incorporates models of 16 crops and offers a user friendly interface that allows users to simulate options for crop management over a number of years to assess the risks associated with each option (Jone s et al. 2003). In DSSAT, bahiagrass and other forage species growth can be simulated using the CROPGRO which is a mechanistic model that integrates plant, soil, management, and weather inputs to predict crop growth and composition (Boote et al., 1998a; b) This model also simulates crop growth, development, and yield as well as changes in soil water, carbon and nitrogen that occur during crop development. Simulations using CROPGRO allow users to predict crop development for several years under different ma nagement strategies. The CROPGRO Forage version was created to simulate perennial tropical forage growth. In this version, the model was modified to simulate the C4 photosynthetic pathway as well as to add a perennial storage organ (stolon) a dormancy p rocess to alter partitioning of growth and mobilization of N during short daylength periods. Also, it modifies the freeze damage routine to allow partial loss of green structures due to freezing (Rymph, 2004). The forage version was calibrated and used to simulate growth of some forage species such a s bahiagrass ( Paspalum notatum Fluegg) (Rymph, 2004), Tifton 85 bermudagrass ( Cynodondactylon (L.) Pers.) (Alderman, 2008 ),

PAGE 64

64 palisadegrass ( Brachiaria brizantha ) (Pedreira et al., 2011), and guineagrass ( Panicu m maximum ) (Lara et al., 2012). As CROPGRO simulates a high number of processes and the soil plant atmosphere interaction, it is then very complex and, thus, requires a high number of inputs to make more accurate predictions Therefore, its use is restrict ed to locations where data are available. Simpler models that take into account less processes and need fewer inputs can be useful tools for decision makers. The Agricultural Reference Index for Drought (ARID) is an example of a simple agricultural drought index based on a reference crop (grass) that is calculated by subtracting the ratio of actual to potential transpiration from 1 (Woli et al., 2012). As ARID was based on physiological principles frequently used in crop models to reduce growth when root 2012), it shares similarities with soil water balance used in the DSSAT. The main differences are that ARID does not consider the division of soil into layers nor root growth and distribution. In CRO PGRO the soil is divided into many layers in which the water availability varies between the permanent wilting point and saturation (Boote et al., 1998a; Jones et al., 2003). If the soil water content is higher than field capacity, the water is drained to the next layer, while in ARID the drained water is lost from the system because only one large layer is taken into account (Woli et al., 2012). Also, ARID treats the entire root system as a single unit and assumes a removal of water from the root zone as a whole, where the water uptake depends only on the atmospheric demand and soil moisture (Woli et al., 2012). Conversely, in CROPGRO, the roots have daily variability and the daily water supplying capacity of the soil root

PAGE 65

65 system depends on the root length and soil water capacity in each soil layer (Boote et al., 1998a). ARID was successfully used to simulate yield losses caused by drought (Woli, 2010) since there is a straight relationship between crop transpiration and crop dry matter accumulation (Jensen, 1968; Hanks, 1974; Doorenbos and Kassam, 1979; Jones et al., 2003; Woli, 2010) It relates the amount of total dry matter produced per unit of transpired water, which is the water use efficiency (Doorenbos and Kassam, 1979). As the water use efficiency is assumed to be constant for a given crop in a given year, actual yield can be ex pressed as potential yield times the ratio of actual to potential evapotranspiration (Hanks, 1974). This approach using the ratio of actual to potential evapotranspiration is commonly used in crop models. If this ratio is lower than 1, it indicates that st omatal conductance was decreased at some time of the day to avoid plant desiccation thus water stress occurred (Jones et al., 2003). The hypothesis of this study was that t he Agricultural Reference Index for Drought (ARID) can be used as a water stress factor indicating crop losses due to water stress. The main objectives were to validate the CROPGRO Forage version to simulate bahiagrass growth under different managements and to use the combination of CROPGRO Forage model and ARID to predict bahiagrass y ield. Material and Methods Validation of the CROPGRO Forage Model Model evaluation was based on a field trial conducted between 2009 and 2011 at the University of Florida at the Plant Science Research and Education Unit (PSREU), Citra, F L A complete descr iption of this experiment is given in Chapter 3. Briefly, in this experiment four treatments using two irrigation levels (irrigated and non irrigated) and

PAGE 66

66 two fertilization levels (0 or 90 kg ha 1 of N and 45 kg ha 1 of K) were compared. It resulted in fou r treatments: 1) irrigated and fertilized (I+F); 2) irrigated and non fertilized (I+NoF); 3) non irrigated and fertilized (NoI+F); and 4) non irrigated and non fertilized (NoI+NoF). 23 fresh biomass harvests happened during the experiment period. Dry matte r (DM) and crude protein (CP) data were measured for each treatment in each harvest, except for the July 29, 2009 harvest, when CP was not available Soil water content was measured using time domain reflectometry (TDR) probes at five depths (5, 15, 30, 45 and 60 cm) in each of fertilized plot and at three depths (5, 15, and 30 cm) in each of the non fertilized plots. Rainfall and irrigation plus rainfall were measured using two rain gauges (one for rainfed treatments and one for irrigated treatments). Add itional weather data necessary (minimum and maximum air temperature, solar radiation, wind speed and relative humidity) were retrieved from the closest Florida Automated Weather Network (FAWN) located at the PSREU farm, approximately 700 m from the experim ental field. XBuild which is the DSSAT crop management data editing program (Uryasev et al., 2003b) was used to create a simulation experiment file based on the field experiment described on Chapter 3. This file contains simulation and management details. The soil file was created through Sbuild (Uryasev et al., 2003 a) to input soil data for each layer. This file contains all information related to each soil layer. It includes soil texture, lower limit or wilting point (SLLL), drained upper limit or field capacity (DSUL), saturation (SSAT), bulk density (SBDM), sat urated hydraulic conductivity (SSKS), and root regrowth factor (SRGF).

PAGE 67

67 SDUL was determined under field conditions at the end of the experiment by using the TDR located on the fertilized treatments. These treatments were chosen because they have TDR in five different depths. The determination was done after adding water to the soil to reach soil water content higher than the DSUL, then covering the soil to avoid evapotranspiration (Fig. 4 1) and monitoring changes in soil water content. SLLL was determined b ased on rainfed treatments after a long period of water restriction (from April to June 2011). SSAT and SSKS were calculated based on soil texture using Saxton and Rawls method (Saxton et al., 1986) SBDM was determined through field measurements. As soil organic matter at the beginning of the experiment was not measured, values reported by Beaty and Tan ( 1972) were used to determine the organic matter on the soil file The SRGF was calculate as de scribed by Uryasev et al. (2003a) Table 4 1 shows the soil profile used in the simulation. Two weather files were created using Weatherman software (Wilkens, 2003) w hich is the DSSAT software for weather data. Both files had the same solar radiation, maximum and minimum temperature, wind speed, and relative humidity data. The only difference between these files was rainfall input data. The first one (for irrigated tre atments) used data collected from the rain gauge located at irrigated treatments and included data of rainfall plus irrigation. The second one (rainfed treatments) used data collected from the rain gauge located at the rainfed treatments and did not includ e irrigations. Therefore, the input of irrigation was done through the weather file and not using fileX. The CROPGRO Forage model is constantly transformed and adjusted; therefore, the set of species and cultivar parameters for each crop is not well defin ed yet. For this

PAGE 68

68 reason, the bermudagrass version presented by Alderman (2008) has been constantly improved by the author and the latest version was selected to be used in this study. A bermudagrass model is acceptable for simulating bahiagrass growth beca use both species are C4 plants with similar characteristics and both have a storage organ. To facilitate the simulation of each harvest, the MOW file (Pedreira et al., 2011) was used. This file contains information related to harvest conditions, including harvests date, amount of stubble remaining after harvest (MOW), the leaf fraction in the stubble (RSPLF) and number of leaves left on a primary tiller axis after harvest (MVS). As stubble mass was not measured after each harvest, it was estimated based on the literature (Beaty et al., 1968; Rymph 2004) and few field samples. For fertilized treatments, MOW was set at 1900 kg of DM ha 1 and for non fertilized treatments, it was set at 600 kg ha 1 RSPLF was not estimated, therefore 99 was used and MVS was set as 1 for all harvests and all treatments. The term herbage mass will be used to indicate the total aboveground biomass, which is the harvested biomass plus estimated stubble. Also, during the field trial, there was no differentiation between alive and dead material. Therefore, simulated values of aboveground biomass and dead canopy were added to make a comparison with the observed biomass. The standard output of simulated soil water is done using fixed layers (0 5, 5 15, 15 30, 30 45, 45 60, and 60 90 cm). However, the TDR measurements were punctual at five depths (5, 15, 30, 45, and 60 cm) for fertilized treatments and at three depths (5, 15, and 30 cm) for non fertilized treatments. Therefore, depending on the soil layer and the treat ment, the simulated values were directly compared with the TDR

PAGE 69

69 measurements or with the average of measurements of two depths. Table 4 2 shows how these comparisons were done. Daily average soil moisture was calculated by multiplying the average soil moist ure by the depth of the layer. Estimation of Yield under Non Nitrogen Stress Conditions It was assumed that no N stress occurred on the fertilized treatments, since it was applied at least 630 kg of N ha 1 year 1 This has been reported as enough amount of N fertilize r for maximum bahiagrass growth (Overman et al., 1990; Overman and Stanley, 1998; Rymph, 2004) Also, this amount of N fertilize r is more than what is currently recommended by IFAS /UF (Institute of Food and Agriculture/University of Florida) specialists (Newman, 2007; Hanlon et al., 2009; Mylavarapu et al., 2009; Newman e t al., 2010) Furthermore, the observed concentration of N on the plant tissues was within or higher than the recommended sufficiency range (Mackowiak et al., 2008) Due to this, two different approaches to simulate water stress were tested and compared to predict yield of fertilized treatments. The first one was using ARID to calculate losses caused by lack of water in the soil To do this, it was first necessary to estimate bahiagrass monthly potential yield (Y P ) in Citra, Florida and then use ARID to estimate relative losses due to water stress The second approach was using CROPGRO Forage model without simulating N stress; therefore lack of water was the only stress factor influencing plant growth. The results obtained in these simulations were compared with the observed biomass from the fertilized treatments. Potential y ield (Y P ) To estimate Y P for each month in Ci tra, Florida, the bermudagrass CROPGRO Forage model was used to simulate bahiagrass maximum growth for 11 consecutive years. The options for water and N stress were turned off in the fileX. Thus, only

PAGE 70

70 maximum and minimum temperature, solar radiation, photo period, CO 2 concentration in the atmosphere, and crop characteristics influenced plant growth. In this simulation, weather data (maximum and minimum temperature and solar radiation) from 2001 to 2011 obtained from the FAWN weather station located in Citra were used and the weather file was created using the Weatherman (Wilkens, 2003) Harvests were simulated using MOW file. Stubble mass was set at 1900 kg ha 1 RSPLF was not estimated ( 99 was used) and MVS was set at 1 for each harvest. For each growing cycle, 11 simulations were done and the highest value for each growing cycle was set as the potential for the corresponding period. For instance, the highest simulated value for May occurred in 2002 and for June it occurred in 2011, therefore these values were used as maximum yield for the respective months. Agricultural Reference Index for Drought (ARID) The Agricultural Reference Index for Drought (ARID) (Woli et al., 2012) was developed as a simple method to quantify water deficit in crops. This drought index is non crop specific since it is based on a reference crop (grass) that completely covers the soil surface and is actively growing in a well drained soil. ARID uses a simple soil water balance for the reference grass having a 400 mm soil layer with evenly distributed roots. Daily ARID was calculated using the following equation (Woli et al., 2012) : where T i is the transpiration (mm day 1 ) and ET O,i is the reference evapotranspiration (mm day 1 ) on day i th. Since T i is always lower than or equal to ET O,i ARID i values range from 0 to 1, with 1 indicating full water deficit and 0 no deficit at all. Between these two extremes, water stress decreases linearly with the increase in actual T i when

PAGE 71

71 ET O,i is constant. When calculated for a period of time that includes s everal days, ARID values can decrease from one day to the next due to rainfall events that increase the available water content in the soil. In the absence of rainfall ARID values were not allowed to decrease from one day to the next due to lower ETo cause d by cooler conditions or cloudy sky. In that case the value of the index is kept the same as in the previous day. ET O,i was calculated as described by (Allen et al., 1998) : i is the slope of the saturation vapor pressure versus air temperature curve (kPa o C 1 ); R n,i is the daily net radiation (MJ m 2 day 1 ); G is the soil heat flux density (MJ m 2 day 1 ), considered as null for da ily estimates; is the psychrometric constant (0.0677 kPa o C 1 ); T i is the daily mean air temperature ( o C) at 2 m, based on the average of maximum and minimum temperatures; u 2,i is the wind speed (m s 1 ) at 2 m height; and e s,i and e a,i represented the sa turation and actual vapor pressures (kPa), respectively. T i was calculated using the following equations (Woli et al., 2012) : which is equal to 400 mm in ARID; RWU is the soil limited root water uptake (mm day 1 representing the maximum fraction of available water extracted in a day, which was 0.097 in this study; and is the p lant available water content (mm mm 1 ) after deep drainage at the end of the previous day.

PAGE 72

72 ARID uses a budget based simple soil water balance, in which the input of water is done through rainfall and irrigation and output due to transpiration, deep drainag e and runoff (Fig. 4 2). Therefore, the available water for certain day is the result of available water in the previous day plus the input minus the output of water. The available water content on the i th day (mm mm 1 ) was calculated as follow: where W i is the amount of available water on the i th day and W i 1 on the previous day (mm); P i is precipitation, I i is irrigation, D i is deep drainage and R i is runoff. If soil water content does not exceed soil water holding capacity (difference between field capacity and permanent wilting point) drainage is zero, otherwise it occurs. In this case, it is calculate using the follow equation: inage coefficient (the fraction of drainable water that can be drained during one day); is the available soil water content on the i th day before the drainage (mm mm 1 m is soil water holding capacity (mm mm 1 was 0.58, an m was 0.06 mm mm 1 since field capacity was 0.105 and permanent wilting point was 0.045 mm mm 1 Runoff only happens when rainfall rate is greater than the infiltration rate. That is, the runoff only occurs after the initial demands of interception, in filtration and surface storage have been satisfied (Woli et al., 2012) Therefore, runoff is calculated using the Soil Conservation Service ( 1972) equation:

PAGE 73

73 where I a is initial abstraction including interception, retention, and infiltration (mm day 1 ); S is the potential maximum retention (mm day 1 61 in this study. Thus, runoff only occ urs when rainfall exceed s 32.5 mm. Actual y ield u sing ARID The ratio of actual evapotranspiration (ET A ) to potential evapotranspiration (ET O ) is commonly used as a depletion factor for growth (Jones et al., 2003) As ARID is calculated as a function of TR and ET O and TR is assumed to be similar to ET A the depletion factor can be calculated by subtracting ARID from 1 (Woli, 2010) However, the effect of water stress on growth and yield depends on the type of the crop, timi ng of the water stress on crop growth and intensity of this stress (Doorenbos and Kassa m, 1979) Due to this, it was necessary to adjust the effect of the water deficit to the crop. ried depending on its intensity, t hus a weighted ARID was calculated. If ARID was lower than 0.3, the negative impact of water stress on crop growth was null. If ARID was higher than 0.3, its negative impact on crop growth was also higher, being then multiplied by a weight factor k dependi intensity. Thus, actual yield ( Y A ) for each growing cycle was calculated by multiplying Y P by the average weighted ARID for the growing cycle as follows:

PAGE 74

74 where i is days after the harvest; N is number of the days of the growing cycle; and k is the coefficient of adjustment. If ARID is lower than 0.3, k = 0; if ARID is equal or higher than 0.3 and lower than 0.7, k = 0.25; and if ARID is equal or higher than 0.7, k = 0.60. CROPGRO Forage without simulate ni trogen s tress To simulate a non nitrogen stress situation using CROPGRO Forage model, simulations were done similar to what was previously described in the CROPGRO Forage m odel section. The only difference was the N stress option, which was turned off in the fileX. Therefore, it was assumed that only water stress affected plant growth. In CROPGRO, the water stress is calculated by the ratio of ET A to ET O (Jones et al., 2003) similar to ARID. In CROPGRO, the impact of water stress is different for each process. If root water uptake is not able to meet evapotranspiration demand, total crop photosynthesis and transpiration are reduced in equal proport ion to the decrease in water uptake. However, other processes that are more sensitive to water stress are reduced when the ratio of ET A to ET O is lower than 1.5 (Boote et al., 1998a) Statistical Analysis For all simulations, predicted values were compared with observed data using visual aspects (graphical results), ratio simulated to observed, coefficient o f determination (R 2 ), root mean square error (RMSE) and the Willmott agreement index (D Stat) (Willmott, 1981) using the followin g equations as suggested by (Wallach, 2006b)

PAGE 75

75 where D i is the model error; Y i i of the Y i values. A better model prediction has small RMSE and ratio, R 2 and D Stat close to 1 (Pedreira et al., 2011) Results and Discussion Validat ion of the CROPGRO Forage M odel Yield The simulations conducted using the bermudagrass CROPGRO Forage model underestimated DM accumulation for fertilized treatments (ratio = 0. 75 ) especially in 2010 (Fig. 4 3 a and c). For these treatments, the R 2 was 0.65, RMSE was 1358 kg DM ha 1 cut 1 and D Stat was 0.76. For I+F treatment, the observed annual average yield was 25340 kg DM ha 1 year 1 while the simulated was 19403 kg DM ha 1 year 1 For this treatment, the RMSE was 1267 kg DM ha 1 cut 1 what is higher than previously reported in the literature (Rymph, 2004; Alderman, 2008) However, the D Stat was 0.79 and R 2 was 0.71 what is relatively high, thus indicating a good relationship between predicted and observed values. For NoI+F, the observed annual average yield was 21561kg DM ha 1 year 1 and the simulated was 15549 kg DM ha 1 year 1 The ratio was 0.73, RMSE was 1443 kg DM ha 1 cut 1 and D Stat was 0.70. For these two treatments, the predictions for 2010 were poor. However, for fall 2009 and summer and fall 2011 the results were more acc urate than for the rest of the period, especially for the I +F treatment (Fig 4 3 a and c).

PAGE 76

76 For the non fertilized treatments, the simulation overestimated observed values (ratio = 1.57), especially in 2009 (Fig 4 3 b and d). For these treatments, the R 2 wa s low (0.21), the RMSE was 890 kg DM ha 1 cut 1 and D Stat was also low (0.47). For I+NoF treatment, the observed annual average yield was 7219 kg of DM ha 1 year 1 while the simulated was 11279 kg of DM ha 1 year 1 For this treatment the ratio was 1.57, RMSE was 934 kg of DM ha 1 year 1 R 2 was 0.23 and D Stat was 0.49. For NoI+NoF, the ratio was 1.57, RMSE was 846 kg of DM ha 1 R 2 was 0.10 and D Stat was 0.40. For these treatments, the RMSE was lower than for fertilized treatment, but it is relatively higher since the average yield for non fertilized treatment was more than two times lower than for fertilized treatments. The R 2 and EF indicate that the predictions for these treatments were low correlated with the observed values. Table 4 3 shows the results for all treatments. E ven if the model underestimated biomass for the I+F treatment the results were reasonable. For this treatm ent, late summer and fall predictions were more accurate than for the rest of the period. For the treatments where irrigation, fertilizer, or both were not applied, the predictions were not as good, indicating that the water and N stress were not well simu lated. For the non fertilized treatments, the graphs show that the predictions were similar to many growing cycles. However, the model occasionally overestimated the yield resulting in low D Stat and R 2 Crude p rotein For the prediction of CP concentration the model showed low performance for all treatments during the whole experiment period (Fig. 4 4). For the fertilized treatments, the model underestimated CP concentration (ratio = 0. 77 ). The relationship between observed and predicted values was low (R 2 = 0. 25 ), RMSE was high ( 4.44 %), and D

PAGE 77

77 Statistic was also low (0. 4 6 ). The average observed CP concentration for I+F treatments was 1 5.0 % while the simulated concentration was 11.0 %. For NoI+F, the observed CP concentration was 15. 5 % and the simulated 1 2. 4 %. For non fertilized treatments, the model overestimated CP content (ratio = 1. 2 6 ) and there was a small relationship between observed and simulated values (R 2 = 0. 2 9 ). The RMSE was 2. 3 8 % and D Stat was 0. 39 indicating poor predictions. The average observed CP concentration for I+NoF was 7. 2 % and the estimated was 9.0 %. For NoI+NoF, the observed CP was 7. 4 % and the estimated was 9.3 %. The model did not simulate variations in CP due to fertilization well (Fig. 4 4). The average CP concentration for fertilized treatments was 15. 3 %, which corresponds to more than double the CP concentration of non fertilized treatments (7. 3 %). The average simulated CP content for the fertilized treatment was 11.7 % while fo r non fertilized it was 9.2 %. The results obtained in this study were not as good as the ones previously observed for simulated bahiagrass and using the bermudagrass version of the model. Rymph (2004) and Alderman (2008) found relatively high rela tionship between observed and simulated CP concentration and plant nitrogen. Table 4 4 shows the results for all treatments. Soil m oisture The simulations underestimated soil water content (Fig. 4 5). For all treatments, the ratio of simulated to observed was lower than 1. However, the predictions were highly correlated and with high D Stat (Table 4 5). All treatments had relatively similar results, but rainfed treatments had slightly higher R 2 (0. 64 ) than irrigated ones (0. 60 ). The differences between the observed and the simulated were due to higher variability of the predicted soil moisture. In the simulations, the soil water content

PAGE 78

78 increased and decreased more rapidly than what was observed in the field. As the input of soil parameters (soil saturation soil organic matter and saturated hydraulic conductivity) were estimated using other models, they might inaccurate. It increased the errors associated with the simulations. O bserved and simulated soil moisture and rainfall + irrigation data from a 30 day period for the I+F is demonstrated in Figure 4 6. This figure shows that simulated soil moisture had greater variability than the observed sample. This was mainly due to faste r response to a rainfall or irrigation event. Also, it is likely due to quicker drainage in the simulated soil water than the one observed in the field. Thus, it indicates that in the simulations the water infiltration rate was higher than the observed one A small rainfall event that did not affect observed soil moisture, resulted in increased simulated soil water content, which resulted in overestimation of soil moisture by the model. On the other hand, as in the simulations, the water was drained faster than in the field; the simulated soil water underestimated the values after a few days of water absence. It occurred mainly due to the overestimation of the drainage. E rrors observed in the biomass prediction could be attributed to errors in the soil moist ure simulations, since weaker predictions occurred for the rainfed treatments. Estimation of Y ield u nder N on N itrogen S tress C ondition s Overall, when the N fertilizer was assumed to meet plant demand, and consequently, N stress was not simulated, the predicted biomass values were more accurate than when N stress was simulated. When ARID was used to calculate biomass, the predicted values were similar to the observed ones. For both treatments, the ratio simulated to observe d was 1.02, RMSE 774 kg of DM ha 1 cut 1 the R 2 was 0.83 and D Stat was 0.91. Overall, the predictions were more accurate when no or low

PAGE 79

79 water deficit was observed. For I+F treatment, the predictions had higher relationship and lower error than for NoI+F (Table 4 6). For both treatments, better predictions were obtained in 2011. In 2009, the biomass was relatively well estimated except for the first harvest of the year (Fig 4 7). Due to this over prediction, the results for 2009 were not as good as the re sults in 2011. The results obtained using CROPGRO Forage and ARID were as good as the ones obtained in the calibration of the first CROPGRO Forage used to simulate bahiagrass growth (Rymph, 2004) and the following version used to simulated bermudagrass growth (Alderman, 2008) Woli (2010) obtained similar values of D Stat usi ng ARID to estimate cotton, maize, soybean and peanut yield in different locations in Florida and Georgia. During summer, observed biomass values had higher variability between growing cycles than simulated ones (Fig. 4 7 ). This lack of variability in the simulated values can be attributed mainly to the lack of interaction between drought and plant processes since Y A was calculated by multiplying Y P by average weighted ARID. This approach did not take into account the cumulative effect of drought from past growing cycles on the plant or the interaction between drought and other environmental factors. Also, in a smaller degree of impact, the use of fix values of Y P for each growing cycle reduced yield variability caused by variations of temperature and solar radiation. When the CROPGRO Forage model with no N stress was used, the model overestimated the biomass accumulation and the results were not as good as the ones observed using ARID. For both treatments, the ratio simulated to observed was 0.89, RMSE 1116 kg of DM ha 1 cut 1 the R 2 was 0.72 and D Stat was 0.83. Better results

PAGE 80

80 were obtained in 2009, when lower variability on the observed values during summer occurred. As observed using ARID, the treatment with low or no water stress had better results (Table 4 6 ). The problem of low variation due to lack of interaction between drought and environmental factors, and due to lack of variability of the Y P was not observed. However, the simulated variability did not represent reality well (Fig. 4 8) and the correlation was lower than the one observed when ARID approach was used. Conclusion s Due to the high number of uncertainties associated to the field experiment, to the model itself (the model was calibrated for bermudagrass and it is still being improved) and to the model inputs (stubble mass, soil organic matter, wilting point, ratio leaf:stem), the biomass predictions in case of I+F treatment were considered satisfactory, due to low or water and N stress. For this treatment, the D Stat was 0.79 and R 2 w as 0.71. Better predictions were obtained for late summer and fall 2009 and 2011 than during the remaining period s For the treatments in which water, fertilizer or both were not applied, the predictions were less accurate, especially in the non fertilized treatments. This indicates that the water and N stress were not well simulated. The CP concentration was not well predicted in these simulations. The model did not simulate variations in CP due to fertilization well. High RMSE and low R 2 were obtained for all treatments. For fertilized treatments, the predictions overestimated the CP values, but for non fertilized treatments, the simulations overestimated it. The simulated values of soil water content were underestimated in all treatments. However, the cor relation between observed and simulated values was high, indicating good model performance. The difference in simulated and observed is mainly due to

PAGE 81

81 overestimation of water infiltration that caused faster increase and decrease of the simulated soil moistu re. To improve all predictions, it would be necessary to improve model initialization through measurements of soil characteristics, such as soil organic matter that affects N availability, and saturated hydraulic conductivity that influences water movement in the soil. The predictions considering that no N stress occurred on the field and lack of water was the only stress affecting forage growth showed similar values to the observed ones. Better predictions were obtained when low or no water stress occurre d. Even though the approach using ARID and a fixed value of Y P for each growing cycle was simpler and was not very sensitive to yield variations between growing cycles, the results were more accurate than the ones using water stress calculated by CROPOGRO Forage model with or without N stress. When the stress factor was based on ARID, the RMSE was 774 kg of DM ha 1 cut 1 the R 2 was 0.83 and D Stat was 0.91, showing good performance. When CROPGRO Forage model was used without simulating N stress, the RMSE w as 1116 kg of DM ha 1 cut 1 the R 2 was 0.72 and D Stat was 0.83. The approach using bermudagrass CROPGRO Forage model and ARID as the water stress factor had the best results. However, this approach was calibrated to field trial conducted in Citra. Becaus e of this, it would be necessary to validate for other regions and for more years. be used in case of low soil data availability. However, as ARID is simpler than the

PAGE 82

82 app roach used in CROPGRO to calculate water stress and its tests were restricted, data from other locations and years should be used to validate the approach.

PAGE 83

83 Table 4 1. Soil sand, clay, silt and organic carbon content, lower limit (SLLL), drained upper limit (DSUL), saturation (SSAT), saturated hydraulic conductivity (SSKS), bulk density (SBDM), and root regrowth factor (SRGF) used to create the soil file Depth Sand Clay Silt Org. C SLLL DSUL SSAT SSKS SBDM SRGF cm % % % % cm 3 water cm 3 soil cm hr 1 g cm 3 0 15 93.4 3.4 3.2 1.02 0.048 0.115 0.33 11.70 1.49 1.00 15 30 92.4 4.4 3.2 0.27 0.043 0.104 0.35 9.36 1.50 1.00 30 45 92.0 5.4 2.6 0.11 0.047 0.097 0.36 7.64 1.51 1.00 45 60 92.9 4.5 2.6 0.62 0.044 0.091 0.35 9.51 1.50 0.80 60 90 90.8 6.4 2.8 0.28 0.039 0.076 0.37 6.16 1.52 0.80 Table 4 2. Layer of simulated soil water and the corresponding TDR probe used for comparison Layer of simulated soil water (cm) Fertilized treatments TDR probes depth Non fertilized treatments TDR probes depth 0 5 5 cm 5 cm 5 15 Average of 5 and 15 cm Average of 5 and 15 cm 15 30 Average of 15 and 30 cm Average of 15 and 30 cm 30 45 Average of 30 and 45 cm 30 cm 45 60 Average of 45 and 60 cm No comparison 60 90 60 cm No comparison Table 4 3. Ratio simulated to observed, root mean square error (RMSE), coefficient of determination (R 2 ) and D Statistic (D Stat) for biomass (kg of DM ha 1 cut 1 ) predictions based on the bermudagrass CROPGRO Forage model for four different treatments: i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Treatment Ratio RMSE (kg ha 1 ) R 2 D Stat I+F 0.77 1267 0.71 0.79 I+NoF 1.57 934 0.23 0.49 NoI+F 0.73 1443 0.55 0.70 NoI+NoF 1.57 846 0.10 0.40

PAGE 84

84 Table 4 4. Ratio simulated to observed, root mean square error (RMSE), coefficient of determination (R 2 ) and D Statistic (D Stat) for crude protein content (%) predictions based on the bermudagrass CROPGRO Forage model for four different treatments: i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilize d (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Treatment Ratio RMSE (%) R 2 D Stat I+F 0. 7 4 4.68 0. 39 0.4 7 I+NoF 1.2 6 2.2 8 0.3 3 0.4 0 NoI+F 0. 80 4.18 0.09 0.4 3 NoI+NoF 1.2 5 2.4 6 0.2 5 0.3 6 Table 4 5. Ratio simulated to observed, root mean square error (RMSE), coefficient of determination (R 2 ) and D Statistic (D Stat) for soil moisture predictions (cm 3 cm 3 ) based on the bermudagrass CROPGRO Forage model for four different treatments: i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) Treatment Ratio RMSE ( cm 3 cm 3 ) R 2 D Stat I+F 0.88 0.044 0.54 0.68 I+NoF 0.81 0.048 0.70 0.63 NoI+F 0.81 0.043 0.60 0.74 NoI+NoF 0.92 0.042 0.64 0.73 Table 4 6. Ratio simulated to observed, root mean square error (RMSE), coefficient of determination (R 2 ) and D Statistic (D Stat) for biomass (kg of DM ha 1 cut 1 ) predictions based bermudagrass CROPGRO Forage and ARID and bermudagrass CROPGRO Forage without simulating nitrogen N stress for two treatments: i) irrigated and fertilized (I+F);and ii) non irrigated and fertilized (NoI+F) CROGPRO and ARID CROPGRO N off I+F NoI+F I+F NoI+F Ratio 0.99 1.05 0.93 0.85 RMSE (kg ha 1 ) 741 806 1007 1215 R 2 0.85 0.80 0.76 0.65 D Stat 0.92 0.89 0.86 0.78

PAGE 85

85 Figure 4 1. Ground covered to avoid evapotranspiration after adding water Figure 4 2. Diagram for the soil water balance of a reference crop (grass)

PAGE 86

86 Figure 4 3. Observed vs simulated herbage mass (kg of DM ha 1 ) using bermudagrass CROPGRO Forage for a) irrigated and fertilized (I+F); b) irrigated and non fertilized (I+NoF); c) non irrigated and fertilized (NoI+F); and d) non irrigated and non fertilized (NoI+NoF) treatments

PAGE 87

87 Figure 4 4. Simulated vs observed crude protein content (%) using bermudagrass CROPGRO Forage model for a) irrigated and fertilized (I+F); b) irrigated and non fertilized (I+NoF); c) non irrigated and fertilized (NoI+F); and d) non irrigated an d non fertilized (NoI+NoF) treatments

PAGE 88

88 Figure 4 5. Simulated vs observed soil moisture content (mm 3 mm 3 ) using bermudagrass CROPGRO Forage model for a) irrigated and fertilized (I+F); b) irrigated and non fertilized (I+NoF); c) non irrigated and f ertilized (NoI+F); and d) non irrigated and non fertilized (NoI+NoF) treatments *Top 90 cm of soil for fertilized treatment and top 45 cm for non fertilized treatment

PAGE 89

89 Figure 4 6. Rainfall + irrigation from a 30 day period sample and the respectively a) observed and b) simulated soil moisture content (mm 3 mm 3 ) for the I+F treatment

PAGE 90

90 Figure 4 7. Observed vs simulate yield (kg of DM ha 1 ) using ARID as the water stress factor for a) irrigated and fertilized (I+F); b) non irrigated and fertili zed (NoI+F) treatments

PAGE 91

91 Figure 4 8. Observed vs simulate herbage mass (kg of DM ha 1 ) using bermudagrass CROPGRO Forage model with no nitrogen N stress for a) irrigated and fertilized (I+F); b) non irrigated and fertilized (NoI+F) treatments

PAGE 92

92 C HAPTER 5 EFFECTS OF EL NIO S OUTHERN OSCILLATION ON THE AGRICULTURAL REFERENCE INDEX FOR DROUGHT (ARID) IN FL ORIDA Drought indices are used to quantify drought intensity, compare current conditions to previous droughts, and provide a regional overview of potential impacts of droughts (Woli et al., 2012). A drought index is a scientifically based numerical index associated to some cumulative effects of an extended and anomalous moisture deficiency and provides quantification of a drough t situation (Lourens and Jager, 1997; World Meteorological Organization, 1984). The Agricultural Reference Index for D rought (ARID) was presented by Woli et al. ( 2012). This is a simple and non crop specific drought index that takes into account the soil p lant atmosphere relationship in a daily time step. ARID is based on a reference crop (grass) actively growing in a well drained soil as well as completely covering the soil surface. The only inputs to calculate ARID are reference evapotranspiration, rainfa ll and few soil characteristics, thus it can be widely calculated for different regions and time periods. Also, as ARID influenced by weather variable, it is impacted by climate variability. The El Nio Southern Oscillation (ENSO) is a coupled ocean atmosphere phenomenon that is considered the main source of interannual cl imate variability in the world (Ropelewski and Halpert, 1996; Fraisse et al., 2006) In the s outheast USA, ENSO has a strong influence during cold months while this influence is much lower during warm months (Winsberg, 2003). The main hypothesis of this study is that El Nio Southern Oscillation (ENSO) influences the variability of soil moisture in Florida causing lower water stress during El Nio and higher stress during La Nia. The main goal of this study was to understand

PAGE 93

93 the spatial and temp oral variability of ARID in Florida. Specific objectives included to identify the influence of the ENSO phenomenon on the temporal and spatial variability of ARID in Florida Material and Methods Study Area and Data The study area was the state of Florida which is located between 803'W, 2457'N and 8738'W, 310'N with a total area of 170,312 km 2 and an elevation varying between 0 and 105 m above sea level. According to Kppen classification Florida is classified as humid subtropical (central, north and pa nhandle regions) and tropical savannah (south region) (Henry et al., 1994). Daily minimum and maximum temperatures and rainfall data were obtained from 56 National Weather Service COOP (Cooperative Observer Program) weather stations located in Florida, 6 i n Georgia and 5 in Alabama. Table 5 1 shows the geographic characterization of the stations used with their spatial distributed portrayed on Fig 5 1. Calculation of ARID Values The complete description of ARID was done on Chapter 4. Daily ARID was calculated using the following equation (Woli et al., 2012): where T i is the transpiration (mm da y 1 ), calculated as described by Woli et al. (2012); and ET O,i is the reference evapotranspiration (mm day 1 ) on day ith. Since Ti is always lower than or equal to ET O,i ARID i values range from 0 to 1 based on E quation ( 5 1) with 1 indicating full water deficit and 0 no deficit at all. When calculated for a period of time that includes several days ARID values can decrease from one day to the next due

PAGE 94

94 to rainfall events that increase the available water content in the soil. In the absence of rainfall ARID values were not allowed to decrease from one day to the next due to lower ETo caused by cooler conditions or cloudy sky. In that case the value of the index is kept the same as in the previous day. Myakka fine sand is the m ost common soil type found in Florida and according to Natural Resources Conservation Service (NRCS 5 1 ) this is the official state soil of Florida and it is the most extensive soil in Florida The values of wilting point and field capacity vary respectivel y around 0.04 and 0.11 m 3 m 3 (Univ ersity of Florida Institute of Food Agricultural Sciences, 1985) ; therefore, these theoretical values were used to calculate ARID in this study. To estimate daily ARID values ET O is normally calculated using the FAO 56 PM equation described by Allen e t al. (1998) which requires air temperature, wind speed, solar radiation and relative humidity. However, due to restricted availability of long term weather data, in this study ET O was calculated using an empiric method that requires only the available wea ther data, i.e., maximum and minimum daily temperature and rainfall. In a previous study by (Gelcer et al., 2010a) to evaluate the best empirical equation to estimate ET O in Florida, when only minimum and maximum temperature and rainfall data are available, it was found that the Prie stley and Taylor (PT) equation (Priestley and Taylor, 1972) using estimated solar radiation and dew point temperature was the mo st accurate. The PT equation is: 5 1 http://www.mo15.nrcs.usda.gov

PAGE 95

95 i is the slope of the relationship between saturation vapor pressure and air temperature (kPa o C 1 ); R n,i is the daily net radiation (MJ m 2 day 1 ); G is the soil heat flux densi ty (MJ m 2 day 1 constant (0.0677 kPa o C 1 1 ); and k PT is the Priestley and Taylor coefficient, an empirical coefficient that may vary for d ifferent regions, because it is influenced by soil moisture and vegetation types (Priestley and Taylor, 1972; Suleiman and Hoogenboom, 2007; Sentelhas et al., 2010) In this study the value of k PT was fixed to 1.25 and 1.14, respectively for cold (October March) and warm (April September) months (Gelcer et al. 2010 a ). Daily net radiation was calculated using the following equations (Allen et al., 1998): where R ns,i is the net shortwave radiation (MJ m 2 day 1 ); R nl,i is the net outgoing longwave radiation (MJ m 2 day 1 ); SR i is the solar radiation (MJ m 2 day 1 ); is the Stefan Boltzmann constant (4.903 10 9 MJ K 4 m 2 day 1 ); T maxK,i and T minK,i are respectively maximum and minimum absolute temperature (K) during a 24h period; e a,i is the actual vapor pressure (kPa); SR o,i is the solar radiation on a clear sky (MJ m 2

PAGE 96

96 day 1 ); Q o,i is the daily total extraterrestrial insolation (MJ m 2 day 1 ), calc ulated as described by Allen et al. (1998); and T dew,i is dew point temperature ( o C). When solar radiation and dew point temperature data are not available, estimated solar radiation (eSR i ) (MJ m 2 day 1 ) and estimated dew point temperature (eT dew,i ) ( o C) can be used to calculate ET O,i (Allen et al., 1998; Gelcer et al., 2010a; Sentelhas et al., 2010) Therefore the term SR i in the E quations ( 5 2 ) and ( 5 3 ) was replaced by eSRi and T dew,i in E quation ( 5 4 ) by eT dew,i In this study, eSR i was estimated using Bristow and Campbell equation (Bristow and Campbell, 1984) and eT dew,i using the equation described by Hubbard et al. ( 2003) as demonstrated below: where T avg,i T min,i and T max,i are respectively average, minimum and maximum temperature ( o C); T t,i i is the daily range of air temperature ( o o C); and A (0.7), B and C (2.4) ar e empirical coefficients. Daily ARID values were compiled into monthly average values which were clustered according to ENSO phases using the Multivariate ENSO Index (MEI) introduced by Wolt er and Timlin ( 1993) to classify months as El Nio, La Nia or Neutral. Typical ARID monthly maps were also created using all years on record independently of ENSO phase (e.g. January All Years is the average of all Januaries from 1950 to 2010) while EN SO monthly maps were create using only months classified

PAGE 97

97 as certain ENSO phase (e.g. El Nio January is the average of all Januaries classified as El Nio). Table 5 2 shows the monthly classification in accordance with the ENSO phases. ENSO Deviation Monthly deviation maps were created by subtracting All Years maps (average maps) from the El Nio/La Nia maps. Therefore, a negative or positive value shows that a particular month having that ENSO phase has lower or higher water stress than average. Spat ial Interpolation ARID values were interpolated using universal kriging to enhance the visualization of its spatial variability by predicting values at locations where data were not available to calculate ARID. Kriging was selected because it is considered as one of the most advanced interpolation method (Burrough and McDonnell, 1998) that accounts for the spatial correlation structure of the variable of interest (G oovaerts, 1997) Kriging required the assumption of stationarity that is ARID was described as a random process with a constant mean over the whole study area. To meet this stationarity requirement, the mean ARID for a combination of ENSO phase and mont h was modeled as polynomials of third degree using longitude and latitude as predictors. The procedure of kriging was implemented in three steps that are described next for a set of spatially distributed ARID values pertaining to a specific ENSO phase and month. First, the spatial variation of ARID was represented by the sample variogram which depicted the relationship between the estimated semivariance and separation distance (Webster and Olivier, 2007)

PAGE 98

98 where z (x i ) and z (x i + h) represent ARID values measured at a pair of locations separated by the distance h; and n(h) is the number of pairs of data points corresponding to the lag h. The second step involved the generalization of the sample variogram into one of the authorized variogram models to provide estimates of sem ivariances at all lags where predictions of ARID were needed. Weighted least squares were used to fit a spherical model with nugget effect to the sample variogram (Diggle and Ribeiro Jr., 2007) In the last step, the interpolated ARID values were obtained as weighted linear combinations of calculated ARID values in the study area. All analyses were i mplemented using the gstat package (Pebesma, 2004) in the R language (Venables et al., 2011) Results and Discussion Historical M aps The maps presented in Fig. 5 2 show the result of spatial and temporal variability of ARID throughout a typical year. They represent the interaction between evapotranspiration and rainfall during an average year. These maps show that during cold months there is a strong gradient of A RID in the northwest southeast direction. During this period, the combination of higher temperature and lower rainfall in the southern portion of the state caused by the low latitude and a warm and dry air that flows out of the Bermuda Azores high pressure system (Henry et al., 1994; Winsberg, 2003) led to soil water deficit because of higher evapotranspiration and lower input of water into the soil. ARID during this period varied from 0.4 and 0.8 in this region. The

PAGE 99

99 interaction of cooler temperatures and h igher rainfall in North Florida, as a result of cold drafts from the interior part of the country and midlatitude cyclones that advects warm and cold air masses to North Florida (Henry et al., 1994; Winsberg, 2003), resulted in lower water stress and conse quently lower ARID (from zero to 0.3) when compared with the northern region. This indicates that northwestern Florida was subjected to lower water stress than the southern part during cold months. Even having lower rainfall in cold months compared to warm months, January and December showed the lowest values of ARID (Fig. 5 2). Since January is the coldest month in Florida (Henry et al., 1994), it is reasonable to assume that the low temperature in these cold months results into low evapotranspiration, whi ch was also found in the previous study by Gelcer et al. (2010 a ). As during warm months the geographical variation of daily maximum temperatures is small throughout the state (Henry et al., 1994; Winsberg, 2003), ARID values are more consistent in the whol e state and temperature loses importance due to the high values and uniformity. Thus rainfall gains importance, because of its high frequency and volume, particularly in the peninsula. The interaction between temperature and rainfall results in more unifor m soil moisture and consequently similar ARID values throughout the state. During this period ARID values were slightly lower for the southern regions and slightly higher for the panhandle. During summer, convective rainfall is dominant in Florida (Henry e t al., 1994) and as it is normally localized, the use of weather stations could represent a limitation since some of them may be located in places where the highest rainfall occurs on different days. However, as ARID is also strongly influenced by soil wat er holding capacity and

PAGE 100

100 evapotranspiration (Woli et al., 2012), long term monthly average of ARID showed similar values for each region. Overall in Florida, November is the month with the least rainfall, when the average total is less than 56 mm. In the pa nhandle, the average total rainfall in this month is around 100 mm and less than 30 mm in the southwest part of the state (Henry et al., 1994). However, the evapotranspiration during this month is also low (Gelcer et al., 2010 a ), resulting in low values of ARID in the north and panhandle regions and slightly higher in the southern region. In average, precipitation in May is 45% lower than in June (Henry et al., 1994). The high values of ARID estimated for May are explained by the small number of fronts affe cting Florida, resulting in higher incoming solar radiation contributing to increased air temperature resulting in higher evapotranspiration (Winsberg, 2003; Gelcer et al., 2010a) Even if in May, temperatures are lower than during summer, the combination of low water input through rain and high water loss resulted in high ARID values (Fig. 5 2). ENSO Effects on ARID The variability of rainfall between neutral, El Nio and La Nia phases influences the incoming solar radiation and consequently the temperature. The temperatures in January and February present greater departures from normal for both phases. The variability of temperature and rainfall caused by ENSO does not influence the spatial distribution of ARID in Florida. The influence of EN SO is only observed in the range of ARID values in cold months. The historical maps for El Nio (Fig. 5 4) and La Nia (Fig. 5 5) phases showed the same gradient in the northwest southeast direction during cold months and high uniformity during warm months For Januaries El Nio, ARID values range from 0 to 0.1 in the panhandle and from 0 to 0.4 in the peninsula and for

PAGE 101

101 Januaries La Nia it ranges from 0 to 0.2 in the panhandle and from 0.3 to 0.7 in the peninsula. The findings 99) demonstrating that the critical period of soil water deficit in Florida is from March to May. During El Nio, May had the highest values of ARID and January the lowest. During La Nia, the highest values of ARID occurred in April and May in the norther n region of the state and in March in the southern portion. Lowest values were found in January for the northern part of the state and in September in the southern part. The combination of decrease in temperature and increase in rainfall during El Nio eve nts caused lower ARID values in cold months, which resulted in negative deviation values (Fig. 5 6). During La Nia, the deviation was positive from November to March showing higher values in the first three months of the year (Fig. 5 7). For both ENSO pha ses, the deviation is not greater than 0.05 from May to September showing that ENSO has stronger effect during cold months and weaker influence in warm months as values in the central southern part of the state were more influenced by ENSO due to rainfall variability caused by this phenomenon, showing a variation of 0.2 in this region. The panhandle was less influenced by ENSO, where the deviation for the whole year was not greater than 0.1. Conclusions The results demonstrate that the ENSO phenomenon affects soil moisture conditions across Florida, particularly during winter months leading to different ARID levels.

PAGE 102

102 During winter there is a gradient in the north south direction in which the northern region had lower values of ARID. For the average year, ARID values ranged from zero to 0.3 in the panhandle and 0.4 to 0.8 in the peninsula during this period, indicating lower water availability in the southern region. Low er temperature and solar radiation and consequently lower evapotranspiration are the main factors in determining ARID. During summer, temperature and solar radiation are high and more uniform throughout the state and rainfall is localized. Therefore, the i nput of water in the soil is the main factor driving ARID values. April and May are the driest months with ARID values ranging from 0.4 to 0.8 in the average year while December and January had lower values of ARID (from zero to 0.7) ENSO has limited effe cts on the spatial distribution of ARID values across Florida. Different phases impact mainly the range of values during winter months. During El Nio ARID is smaller and the opposite happens during La Nia, when ARID is higher due to higher temperature an d lower rainfall. For example, in January, ARID values range from 0 to 0.4 for El Nio and from 0 to 0.7 for La Nia, showing that El Nio events result in lower soil water stress. This effect of ENSO phase on ARID is stronger in the central part of the st ate when compared to the panhandle. Since rainfall is localized during summer due to its convective nature, interpolation techniques have some limitations of failing to capture its full spatial variability. However, it can be expected that longer term mon thly averages of ARID values are found to be similar for each region. In the case of daily values or weekly averages of ARID, a denser network of weather stations or grid based products should be used to better characterize the spatial variability of rainf all.

PAGE 103

103 The results of this study will allow comparison of current condition with long term average values of ARID, assisting with the characterization of situations of lower or higher crop water stress. In addition, the forecast of ENSO phases combined with related ARID maps can be used to anticipate crop water stress conditions helping agricultural managers plan for the season.

PAGE 104

104 Table 5 1. County, latitude, longitude and number of years of data of each of the NWS Cooperative Observer Program (COOP) stations used to calculated historical ARID County State Latitude Longitude Starting Year Number of Years Franklin FL 29.73 85.02 1950 61 Desoto FL 27.22 81.87 1950 61 Highlands FL 27.60 81.53 1950 61 Polk FL 27.90 81.85 1950 61 Hernando FL 28.62 82.37 1950 61 Sumter FL 28.67 82.08 1950 61 Palm Beach FL 26.87 80.63 1950 61 Washington FL 30.78 85.48 1950 61 Lake FL 28.45 81.75 1950 61 Putnam FL 29.42 81.52 1950 61 Okaloosa FL 30.78 86.52 1950 61 Dixie FL 29.65 83.17 1950 61 Walton FL 30.75 86.08 1950 61 Volusia FL 29.02 81.32 1950 61 Hendry FL 26.60 81.13 1956 55 Collier FL 25.85 81.38 1950 61 Putnam FL 29.75 81.53 1950 61 Nassau FL 30.67 81.47 1950 61 Broward FL 26.10 80.20 1950 61 Lee FL 26.58 81.87 1950 61 St. Lucie FL 27.47 80.35 1950 61 Alachua FL 29.70 82.28 1960 51 Baker FL 30.27 82.18 1950 61 Dade FL 25.82 80.28 1950 61 Alachua FL 29.83 82.60 1950 61 Citrus FL 28.80 82.32 1950 61 Duval FL 30.50 81.70 1950 61 Hamilton FL 30.52 82.95 1950 61 Osceola FL 28.28 81.42 1950 61 Columbia FL 30.18 82.60 1950 61 Suwannee FL 30.28 82.97 1950 61 Madison FL 30.45 83.42 1950 61 Lafayette FL 30.05 83.18 1950 61 Brevard FL 28.10 80.65 1950 61 Jefferson FL 30.57 83.87 1950 61 Glades FL 26.83 81.08 1950 61 Sarasota FL 27.25 82.32 1950 61 Collier FL 26.17 81.72 1950 61 Marion FL 29.08 82.08 1950 61 Okeechobee FL 27.20 80.83 1950 61 Orange FL 28.43 81.33 1952 59 Manatee FL 27.62 82.35 1950 61

PAGE 105

105 Table 5 1. Continued County State Latitude Longitude Starting Year Number of Years Escambia FL 30.48 87.18 1950 61 Taylor FL 30.10 83.57 1950 61 Hillsborough FL 28.02 82.15 1950 61 Charlotte FL 26.92 82.00 1965 46 Gadsden FL 30.60 84.55 1968 43 Pasco FL 28.33 82.27 1950 61 Seminole FL 28.80 81.27 1950 61 Martin FL 27.20 80.17 1950 61 Leon FL 30.40 84.35 1950 61 Pinellas FL 28.15 82.75 1950 61 Levy FL 29.42 82.82 1956 55 Indian River FL 27.65 80.42 1950 61 Hardee FL 27.55 81.80 1950 61 Gulf FL 30.12 85.20 1950 61 Covington AL 31.30 86.52 1950 61 Escambia AL 31.18 87.43 1950 61 Baldwin AL 30.88 87.78 1950 61 Escambia AL 31.05 87.05 1950 61 Henry AL 31.37 85.33 1950 61 Mitchell GA 31.18 84.20 1950 61 Miller GA 31.17 84.77 1956 55 Charlton GA 30.73 82.13 1950 61 Clinch GA 31.08 82.80 1956 55 Brooks GA 30.78 83.57 1950 61 Thomas GA 30.87 83.93 1950 61

PAGE 106

106 Table 5 2. Monthly ENSO phase classification according to the Multivariate ENSO Index (MEI) (Wolter and Timlin, 1993) wh ere 0 is Neutral, 1 El Nio and 2 La Nia Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1950 2 2 2 2 2 2 2 2 2 0 2 2 1951 2 2 2 2 0 1 1 1 1 1 1 1 1952 1 0 0 0 0 2 0 0 0 0 0 0 1953 0 0 0 1 1 0 1 0 1 0 0 0 1954 0 0 0 2 2 2 2 2 2 2 2 2 1955 2 2 2 2 2 2 2 2 2 2 2 2 1956 2 2 2 2 2 2 2 2 2 2 2 2 1957 2 0 0 0 1 1 1 1 1 1 1 1 1958 1 1 1 1 1 1 1 1 0 0 1 1 1959 1 1 1 0 0 0 0 0 0 0 0 0 1960 0 0 0 0 0 0 0 0 2 0 0 2 1961 0 0 0 0 0 0 0 0 0 2 2 2 1962 2 2 2 2 2 2 2 2 2 2 2 2 1963 2 2 2 2 2 0 1 1 1 1 1 1 1964 1 1 0 2 2 2 2 2 2 2 2 2 1965 2 0 0 0 1 1 1 1 1 1 1 1 1966 1 1 1 1 0 0 0 0 0 0 0 0 1967 2 2 2 2 2 0 2 2 2 2 2 0 1968 2 2 2 2 2 2 2 0 0 1 1 0 1969 1 1 1 1 1 1 1 0 0 1 1 0 1970 0 1 0 0 0 2 2 2 2 2 2 2 1971 2 2 2 2 2 2 2 2 2 2 2 2 1972 2 2 0 0 1 1 1 1 1 1 1 1 1973 1 1 1 1 0 2 2 2 2 2 2 2 1974 2 2 2 2 2 2 2 2 2 2 2 2 1975 2 2 2 2 2 2 2 2 2 2 2 2 1976 2 2 2 2 2 0 1 1 1 1 1 1 1977 1 0 0 1 0 1 1 1 1 1 1 1 1978 1 1 1 0 0 2 0 0 0 0 0 1 1979 1 0 0 0 1 0 0 1 1 1 1 1 1980 1 1 1 1 1 1 1 0 0 0 0 0 1981 0 0 1 1 0 0 0 0 0 0 0 0 1982 0 0 0 0 1 1 1 1 1 1 1 1 1983 1 1 1 1 1 1 1 1 1 0 0 0 1984 0 2 0 0 0 0 0 0 0 0 0 2 1985 2 2 2 2 2 0 0 0 2 0 0 0 1986 0 0 0 0 0 0 0 1 1 1 1 1 1987 1 1 1 1 1 1 1 1 1 1 1 1 1988 1 1 1 0 0 2 2 2 2 2 2 2 1989 2 2 2 2 2 0 2 2 0 0 0 0 1990 0 1 1 0 1 0 0 0 0 0 0 0 1991 0 0 0 1 1 1 1 1 1 1 1 1

PAGE 107

107 Table 5 2. Continued Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 1992 1 1 1 1 1 1 1 1 1 1 1 1 1993 1 1 1 1 1 1 1 1 1 1 1 1 1994 0 0 0 1 1 1 1 1 1 1 1 1 1995 1 1 1 1 1 1 0 0 2 2 2 2 1996 2 2 0 2 0 0 0 2 2 0 0 0 1997 2 2 0 1 1 1 1 1 1 1 1 1 1998 1 1 1 1 1 1 0 2 2 2 2 2 1999 2 2 2 2 2 2 2 2 2 2 2 2 2000 2 2 2 2 0 0 0 0 0 0 2 2 2001 2 2 2 0 0 0 0 0 0 0 0 0 2002 0 0 0 0 1 1 1 1 1 1 1 1 2003 1 1 1 0 0 0 0 0 1 1 1 0 2004 0 0 0 0 1 0 1 1 1 1 1 1 2005 0 1 1 1 1 1 1 0 0 0 0 2 2006 2 2 2 2 0 1 1 1 1 1 1 1 2007 1 1 0 0 0 0 0 2 2 2 2 2 2008 2 2 2 2 0 0 0 0 2 2 2 2 2009 2 2 2 0 0 1 1 1 1 1 1 1 2010 1 1 1 1 1 2 2 2 2 2 2 2 Figure 5 1. NWS Cooperative Observer Program (COOP) stations used to calculate historical ARID

PAGE 108

108 Figure 5 2. Spatial distribution of historical monthly average ARID values for Florida

PAGE 109

109 Figure 5 3. Spatial distribution of monthly average ARID values for Florida du ring Neutral phase

PAGE 110

110 Figure 5 4. Spatial distribution of monthly average ARID values for Florida during El Nio phase

PAGE 111

111 Figure 5 5. Spatial distribution of monthly average ARID values for Florida during La Nia phase

PAGE 112

112 Figure 5 6. Spatial distribution o f monthly deviation in ARID values during El Nio events from All Years

PAGE 113

113 Figure 5 7. Spatial distribution of monthly deviation in ARID values during La Nia events from All Years

PAGE 114

114 CHAPTER 6 SUM M ARY The first hypothesis of this study was that ARID can be used as an indicator of crop losses due to water stress. The objective was to adapt a crop model that uses ARID as a water stress factor. Field work included established Pensacola bahiagra ss that was subjected to two irrigation and two fertilization levels. Yield, CP concentration and soil moisture were monitored from 2009 to 2011, and stolon root length was measured twice Additionally, simulations using CROPGRO Forage model and ARID were done to validate the CROPGRO Forage and verify the relationship between ARID and crop lo s ses. The second hypothesis was that the ENSO influences the variability of soil moisture in Florida, causing lower stress during El Ni o and higher stress during La Ni a. The objectives related to this were to understand spatial and temporal variability of ARID in Florida and to identify the ENSO influence on them The results of the field trial showed that fertilization, irrigation and time of the year as well their in teraction affected crop yield. In this field experiment, fertilization had a stronger impact than irrigation. When N fertilizer and irrigation were applied, the peak of yield occurred from early June to late August. It was caused by the ideal temperature, solar radiation and daylength during this period. Decrease of biomass production was observed when the daylength was decreased. Crude protein concentration was influenced by irrigation, fertilization and time of the year. As observed for DM accumulation, C P concentration was strongly affected by fertilization. Lower values of CP concentration were observed during summer. Most part of stolon roots was found in the first 15 cm of the soil profile. The amount of stolon roots in the 0 15 cm layer varied from 4 0 to 63%. Higher amount of

PAGE 115

115 stolon roots were found in the non fertilized treatments. The soil water moisture had higher variability in the rainfed treatments. It was mainly due to an increase of soil moisture in case of a rainfall event and extreme reducti on in the case of a long period of water shortage. Soil water content did not have linear relationship with yield variability. For some growing cycles, the difference in yield between fertilized treatments was large even if soil moisture difference was sma ll and vice versa. The comparison between the bermudagrass CROPGRO Forage and field observations showed that in case of absence or low water and N stress, the biomass predictions were satisfactory. However, if fertilizer, water, or both were not applied, the results obtained were not as good. It indicates that water and N stress were not simulated well. Variations of CP content between growing cycles and treatments were not simulated well and poor predictions occurred for all treatments. Simulated soil wat er content was underestimated for all treatments. However, the relationship between observed and simulated was high. These differences were mainly due to over prediction of water infiltration that caused faster increase and decrease of the simulated soil w ater. To improve all predictions, it would be necessary to improve model inputs through measurements of soil characteristics, such as soil organic matter and saturated hydraulic conductivity. The predictions considering that no N stress occurred on the fie ld and lack of water was the only stress affecting forage growth showed similar values to the observed ones. Even though the approach using bermudagrass CROPGRO Forage to calculate potential yield and ARID as the water stress factor was very sensitive to v ariations

PAGE 116

116 between growing cycles, the results were more accurate than the ones using water stress calculated by CROPOGRO Forage model with or without N stress. be used in the case of low soil data information However, as ARID is simpler than the approach used in CROPGRO to calculate water stress and these tests were restricted, data from other locations and years should be used to validate the approach. The study to relate spatial and temporal variability of ARID in Florida and to identify the ENSO influence on it showed that independently of the ENSO phase there is a gradient in the north south direction during winter. In this gradient, the northern region had lower values of ARID. During this period, lower temperature and solar radiation are the main factors determining ARID. However, during summer, rainfall is localized due t o its convective nature, and temperature and solar radiation are more uniform throughout the state. Therefore, during this period, rainfall is the main factor driving ARID values. April and May were the driest months and December and January the months wit h lower values of ARID. ENSO phenomenon affects soil moisture conditions across Florida, particularly during the cold months leading to reduction of ARID during El Nio and increase during La Nia. The effect of ENSO phase s on ARID are stronger in the cent ral part of the state when compared to the panhandle. The results of this study show that ARID can be used to estimate yield in case of low soil and crop data availability. Therefore, the spatial and temporal variability of yield as well ENSO influence on it can be estimated using ARID. It will allow comparisons of current condition s with long term average values of ARID, assisting with the characterization of situations of lower or higher crop water stress and yield. In addition,

PAGE 117

117 the forecast of ENSO phase s combined with related ARID maps can be used to anticipate crop water stress conditions helping agricultural managers plan for the season.

PAGE 118

118 A PPENDIX FIELD DATA Table A 1. D ry matter (kg of DM ha 1 ) for four treatments i) irrigated and fertilized (I +F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF ) I + F I + N o F N o I + F N o I + N o F 5/6/2009 1931 509 863 573 6/3/2009 4663 388 3026 180 7/1/2009 4727 536 4052 541 7/29/2009 4538 737 5052 774 8/26/2009 4649 1250 3828 1091 9/23/2009 2674 1223 2957 1013 10/21/2009 1291 619 1311 763 12/2/2009 448 118 252 104 6/2/2010 4293 605 3811 633 6/29/2010 5272 1650 4762 813 7/29/2010 2787 778 2691 583 8/25/2010 4483 1266 4132 1357 9/22/2010 3026 1390 2637 1541 10/20/2010 2105 1456 1719 1184 11/17/2010 997 837 638 372 5/6/2011 3773 279 4045 338 6/2/2011 3691 545 2856 438 6/30/2011 4829 991 2887 879 7/28/2011 4006 1796 3795 1506 8/24/2011 3565 1130 3465 883 9/21/2011 3814 1497 2587 1026 10/19/2011 2851 1637 1948 719 11/16/2011 1606 421 1369 353

PAGE 119

119 Table A 2. Crude protein concentration (%) for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) I + F I + N o F N o I + F N o I + N o F 5/6/200 9 19.76 9.60 16.64 8.95 6/3/2009 14.84 7.04 18.31 8.19 7/1/2009 12.51 7.41 14.07 8.17 8/26/2009 13.11 6.80 14.47 6.91 9/23/2009 16.55 6.97 19.45 7.47 10/21/200 9 19.69 8.59 18.96 8.70 6/2/2010 14.63 6.79 14.65 7.39 6/29/2010 9.55 5.03 10.61 4.90 7/29/2010 14.29 6.65 14.71 7.13 8/25/2010 13.47 6.83 13.30 7.26 9/22/2010 14.62 5.62 13.30 6.65 10/20/2010 17.19 6.20 15.27 6.36 11/17/2010 20.93 7.56 15.97 7.01 5/6/2011 15.06 8.18 17.27 8.62 6/2/ 2011 14.71 5.84 14.85 5.75 6/30/ 2011 12.36 7.08 18.53 8.06 7/28/ 2011 12.45 7.82 13.07 7.88 8/24/ 2011 12.93 7.20 15.06 7.88 9/21/ 2011 13.99 8.44 15.02 7.63 10/19/ 2011 15.22 7.38 17.23 7.81 11/16/ 2011 16.47 7.21 15.70 7.47

PAGE 120

120 Table A 3. Average soil water content (cm 3 cm 3 ) for four treatments i) irrigated and fertilized (I+F); ii) irrigated and non fertilized (I+NoF); iii) non irrigated and fertilized (NoI+F); and iv) non irrigated and non fertilized (NoI+NoF) I+F I+NoF NoI+F NoI+NoF 6/3/2009 0.131 0.173 0.119 0.114 7/1/2009 0.113 0.150 0.100 0.105 7/29/2009 0.115 0.158 0.106 0.108 8/26/2009 0.156 0.169 0.149 0.087 9/23/2009 0.135 0.155 0.146 0.111 10/21/2009 0.115 0.142 0.096 0.079 12/2/2009 0.153 0.149 0.138 0.083 6/2/2010 0.078 0.145 0.084 0.093 6/29/2010 0.099 0.148 0.061 0.074 7/29/2010 0.102 0.144 0.094 0.096 8/25/2010 0.101 0.145 0.105 0.100 9/22/2010 0.100 0.136 0.068 0.072 10/20/2010 0.121 0.158 0.077 0.066 11/17/2010 0.110 0.144 0.054 0.050 5/6/2011 0.106 0.125 0.081 0.084 6/2/2011 0.082 0.141 0.066 0.064 6/30/2011 0.068 0.140 0.071 0.068 7/28/2011 0.115 0.166 0.108 0.091 8/24/2011 0.124 0.179 0.100 0.086 9/21/2011 0.121 0.166 0.091 0.081 10/19/2011 0.101 0.156 0.107 0.099 11/16/2011 0.115 0.164 0.103 0.102

PAGE 121

121 LIST OF REFERENCES Alderman, P.D. 2008. Simulating the regrowth dynamics of Tifton 85 bermudagrass as FL. Allen, R.G., L.S. Pereira, D. Raes, and M. Smith. 1998. Crop evapotranspiration: Guidelines for computing crop water requirements. FAO Irrig. Drain. Pap. 56. FAO, Rome. Ashley, D.A., O.L. Bennett, B.D. Doss, and C.E. Scarsbrook. 1965. Effect of nitroge n rate and irrigation on yield and residual nitrogen recovery by warm season grasses. Agron. J. 57: 370 372. Bastos, E.A., M.V. Folegatti, R.T. de Faria, A.S. de Andrage Jr., and M.J. Cardoso. 2002. Simulation of growth and development of irrigated cowpea. Pesq. Agropec. Bras. 37: 1381 1387. mechanisms, and species genotype diversity. p. 23 28. In H. Takatoh (ed.) Proc. 6th Inter. Turf. Sci. Tokyo, Japan. Beaty, E.R., Y.C. Smith, and J.D. Powell. 1974. Response of Pensacola bahiagrass to irrigation and time of N. fertilization. J. Range Man. 27: 394 396. Beaty, E.R., R.L. Stanley, and J. Powell. 1968. Effect of height of cut on yield of Pensacola bahiagrass. Agron. J. 6 0: 356 358. Beaty, E.R., and K.H. Tan. 1972. Organic matter, N, and base accumulation under Pensacola bahiagrass. J. Range Man. 25: 38 40. Blue, W.G. 1973. Role of Pensacola bahiagrass stolon root systems in fertilizer nitrogen utilization on Leon Fine san d. Agron. J. 65: 88 91. Blue, W.G., and D.A. Graetz. 1977. The effect of split nitrogen applications on nitrogen uptake by Pensacola bahiagrass from an aeric haplaquod. Soil Sci. Soc. Am. J. 41: 927 930. Boote, K.J., J.W. Jones, and G. Hoogenboom. 1998a. S imulation of crop growth CROPGRO Model. p. 651 692. In R.M. Peart and R.B. Curry (ed .) Agricultural systems modeling and simulation. Marcel Dekker, Inc., New York. Boote, K.J., J.W. Jones, G. Hoogenboom, and N.B. Pickering. 1998b. The CROPGRO model for gr ain legumes. p. 99 128. In G.Y Tsuji, G. Hoogenboom and P.K. Thornton (ed .) Understanding options for agricultural production. Kluwer Academic Publishers, Dordrecht, The Netherlands. Boote, K.J., M.I. Mnguez, and F. Sau. 2002. Adapting the CROPGRO legume model to simulate growth of Faba Bean. Agron. J. 94: 743 756.

PAGE 122

122 Breuer, N., V. Cabrera, P.E. Hildebrand, and J.W. Jones. 2005. Climate based management options for north central Florida beef cattle producers. CIR 1476. Univ of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Bristow, K.L., and G.S. Campbell. 1984. On the relationship between incoming solar radiation and daily maximum and minimum temperature. Agric. For. Meteorol. 31: 159 166. Burrough, P. A., and R.A. McDonnell. 1998. Principles of Geographical Information Systems. Oxford University Press, New York. Burson, B.L., and V.H. Watson. 1995. Bahiagrass, dallisgrass, and other paspalum series. In R.F. Barnes, D.A. Mi ller, and C.J. Nelson (ed.) For ages. Volume I: An introduction to grasslands agriculture. Fifth ed. Iowa State University Press, Iowa. Burton, G.W., J.E. Hook, J.L. Butler, and R.E. Hellwig. 1988. Effect of temperature, daylength, and solar radiation on production of Coastal bermudagras s. Agron. J. 560: 557 560. Can, J., J. Gonzlez, and J. Valds. 2007. Precipitation in the Colorado River Basin and its low frequency associations with PDO and ENSO signals. J. Hydrol. 333: 252 264. Dagel, K. 1997. Defining drought in marginal areas: the role of perceptio n. Professional Geographer 49 : 192 202. Diggle, P.J., and P.J. Ribeiro Jr. 2007. Model based geostatistics. Springer, New York. Doorenbos, J., and A.H. Kassam. 1979. Yield response to water. FAO Irrig. Drain. Pap. 33. FAO, Rome. Doss, B.D ., D.A. Ashley, and O.L. Bennett. 1960. Effect of soil moisture regime on root distribution of warm season forage species. Agron. J. 52: 569 572. Farooq, M., A. Wahid, N. Kobayashi, D. Fujita, and S.M.A. Basra. 2009. Plant drought stress: effects, mechanis ms and management. Agron. Sustain. Dev. 29: 185 212. Fraisse, C.W., N.E. Breuer, D. Zierden, J.G. Bellow, J. Paz, V.E. Cabrera, A. Garcia y 2006. AgClimate: A climate forecast inf ormation system for agricultural risk management in the southeastern USA. Comput. Electron. Agric. 53: 13 27. Fraisse, C.W., K.A. Sudduth, and N.R. Kitchen. 2001. Calibration of the CERES Maize model for simulating site specific crop development and yield on claypan soils. Appl. Eng. Agric. 17: 547 556. Gallaher, R.N., C.O. Weldon, and J.G. Futral. 1975. An aluminum block digester for plant and soil analysis. Sci. Soc. Am. Proc. 39: 803 806.

PAGE 123

123 Gelcer, E.M., C.W. Fraisse, and P.C. Sentelhas. 2010a. Evaluation of methodologies to estimate reference evapotranspiration in Florida. p. 189 195. In Proc. Fla. State Hort. Soc., Crystal River, FL. 6 8 June 2010. FSHS, Lake Alfred, FL. Gelcer, E., P.C. Sentelhas, C.W. Fraisse, and E. Yamada. 2010b. Evaluation of referen ce evapotranspiration estimating methods in Florida, USA. p. 213 214. In Reunin Latinoamericana de Agrometeorologa, Baha Blanca, Argentina. 20 22 Oct. 2010. Asociacin Argentina de Agrometeorologa, Baha Blanca, Argentina. Giraldo, L.M., L.J. Lizcano, A.J. Gijsman, B. Rivera, and L.H. Franco. 2001. Adapting the CROPGRO model of DSSAT to simulate the growth of Brachiaria decumbens In W.T. Bowen, P. Malagamba, R. Quiroz, M. Holle, J. White, C. Leon Vel arde, and H.H. Van Laar (ed.) Proc. The Third Inter. Symp. Syst. Approac. Agric. Develop. Lima, Peru 8 10 Nov. 1999. CIP, Lima, Peru. Goovaerts. 1997. Geostatistics for natural resources evaluation. Oxford University Press, Oxford. Hambleton, L.G. 1977. Semiautomated method for simultaneous determination of phosphorus, calcium and crude protein in animal feeds. J.A.O.A.C. 60: 845 852. Hanks, R.J. 1974. Model for predicting plant yield as influenced by water use. Agron. J. 66: 660 665. Hanlon, E.A., R. Mylavarapu, C.L. Mackowiak, and M.L. Silveira. 2009. Devel opment of bahiagrass fertilization. SL 237. Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Hartkamp, A.D., G. Hoogenboom, and J.W. White. 2002. Adaptation of the CROPGRO growth model to velvet bean ( Mucuna pruriens ) I. Model d evelopment. Field Crops Res. 78: 9 25. Hayes, M. 2006. What is drought? Drought Indices. Available at http://www.drought.unl.edu/whatis/indices.htm (accessed 22 March 201 1 ). University of Nebraska, Nat ional Drought Mitigation Center, Lincoln, NE. Heim Jr., R.R. 2002. A review of twentieth century drought indices used in the United States. Bull. Amer. Meteor. Soc. 83: 1149 1165. Henry, J.A., K.M. Portier, and J. Coyne. 1994. The climate and weather of Florida. Pineapple Press, Sarasota, FL. Hildebra nd, A.E. 2006. Impact of the ENSO climate phenomenon on Florida fresh Hirata, M., and W. Pakiding. 2002. Dynamics in tiller weight and its association with herbage mass and tiller density in a ba hia grass ( Paspalum notatum ) pasture. Trop. Grassl. 36: 24 32.

PAGE 124

124 Hirata, M., and W. Pakiding. 2003. Responses of bahiagrass to nitrogen and defoliation. J. Range Man. 56: 608 615. Hodges, T., D. Botner C. Sakamoto, and H.J. Hays. 1987. Using the CERES Maize model to estimate production for the U.S. Cornbelt. Agric. For. Meteorol. 40: 293 303. Hoogenboom, G., J.W. Jones, C.H. Porter, K.J. Boote, W.D. Batchelor, L.A. Hunt, A.J. Gijsman, P.W. Wilkens, U. Singh, and W.T. Bowen. 2003. DSSAT v4 cropping system simulation model. In G. Hoogenboom, J.W. Jones, C.H. Porter, K.J. Boote, W.D. Batchelor, L.A. Hunt, A.J. Gijsman, P.W. Wilkens, U. Singh, and W.T. Bowen (ed.) Decision Support System for Agrotechnology Transfer Version 4.0. Volume 1. Univ. of Hawaii, Honolulu, HI. Hoogenboom, G., J.W. Jones, P.W. Wilkens, C.H. Porter, W.D. Batchelor, L.A. Hunt, K.J. Boote, U. Singh, O. Uryasev, W.T. Bowen, A.J. Gijsman, A.S. Du Toit, J.W. White, and G.Y. Tsuji. 2004. Dec ision Support System for Agrotechnology Transfer. Version 4.0. Univ. of Hawaii, Honolulu, HI. Hubbard, G.K., R. Mahmood, and C. Carlson. 2003. Estimating daily dew point temperature for the northern Great Plains using maximum and minimum temperature. Agron J: 323 328. Impithuksa, V., W.G. Blue, and D.A. Graetz. 1984. Distribution of applied nitrogen in soil Pensacola bahiagrass components as indicated by nitrogen 15. Soil Sci. Soc. Am. J. 48. Inyang, U., J.M.B. Vendramini, L.E. Sollenberger, B. Sellers, A. Adesogan, L. Paiva, and A. Lunpha. 2010. Forage species and stocking rate effects on animal Sci. 50: 1079 108. Jensen, M.E. 1968. Water consumption by agricultural plants. p. 1 22 In T.T. Kozlowski (ed.) Water Deficits and Plant Growth. Academic Press, New York. Johnson, C.R., B.A. Reiling, P. Mislevy, and M.B. Hall. 2001. Effects of nitrogen fertilization and harvest date on yield, digestibility, fiber, and protein fractions of t ropical grasses. J. Anim. Sci. 79: 2439 2448. Jones, C.A., and J.R. Kiniry. 1986. CERES Maize: A simulation model of maize growth and development. Texas A&M Univ. Press, College Station, TX. Jones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Batchel or, L.A. Hunt, P.W. Wilkens, U. Singh, A.J. Gijsman, and J.T. Ritchie. 2003. The DSSAT cropping system model. Eur. J. Agron. 18: 235 265.

PAGE 125

125 Karnauskas, K.B., A. Ruiz Barradas, S. Nigam, and A.J. Busalacchi. 2008. North American droughts in ERA 40 global and NCEP North American regional reanalyses: A Palmer Drought Severity I ndex perspective. J. Clim 21 : 2102 2123. Keetch, J.J., and G.M. Byram. 1968. A drought index for forest fire control. Res. Pap. SE 38. U.S. For. Serv., Southeast. For. Exp. Stn., Asheville NC. Kelly, T.C. 1995. A bioeconomic system approach to sustainability analysis at the farm level. Ph.D. diss. Univ. of Florida, Gainesville, FL. Kim, K.S. 1987. Comparative drought resistance mechanisms of eleven major warm season tur fgrasses. Ph D Diss Texas A&M University, College Station, TX Lara, M.A.S., C.G.S. Pedreira, K.J. Boote, B.C. Pedreira, L.S.E. Moreno, and P.D. Alderman. 2012. Predicting growth of Panicum maximum : an adaptation of the CROPGRO Perennial Forage Model. Agron. J. 104(3): 600 611. Lpez Cedrn, F.X., K.J. Boote, B. Ruz Nogueira, and F. Sau. 2005. Testing CERES Maize versions to estimate maize production in a cool environment. Eur. J. Agron. 23: 89 102. doi:10.1016/j.eja.2005.01.001 Lourens, U.W., and J.M. de Jager. 1997. A com puterized crop specific drought monitoring system: D esign concepts and initial te sting. Agric. Syst. 53: 303 315 doi:10.1016/S0308 521X(96)00004 2 Mackowiak, C.L., A.R. Blount, E.A. Hanlon, M.L. Silveira, M.B. Adjei, and R.O. Myer. 2008. Getting the most out of bahiagrass fertilization. S L 24 9. Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. e ds). 1992. Forages. Prosea, Bogor, Indonesia. McKee, T.B., N.J. Doesken, and J. Kleist. 1993. The relationship of drought frequency and duration to time scales. p. 179 184. In Proc. Conf. on Applied Climatology, 8th, Anaheim, CA. 17 22 Jan. 1993. Am. Meteorol. Soc., Boston. Mislevy, P., and P.H. Everett. 1981. Subtropical grass species respo nse to different irrigation and harvest regimes. Agron. J. 73: 601 604. Mo, K.C., and J.E. Schemm. 2008. Relationships between ENSO and drought over the southeastern United State s. Geophys. Res. Lett. 35 : 1 5. Mylavarapu, R., D. Wright, G. Kidder, and C.G. Chambliss. 2009. UF/IFAS standardized fertilization recommendations for agronomic crops. SL 129. Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Neitsch, S.L., J.G. Arnold, J.R. Kiniry, and J.R. Williams. 2005 Soil and water assessment tool theoretical document, version 2005 Agric. Res. Serv. and Texas Agric. Exp. Stn ., Temple TX.

PAGE 126

126 Newman, Y.C. 2007. Bahiagrass: A Quick Reference. SS AGR 263 Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Newman, Y.C., C. Mackowiak, R. Mylavaparu, and M. Silveira. 2008. Fertilizing and Liming Forage Crops. SS AGR 176 Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Newman, Y.C., T.R. Sinclair, A.S. Blount, M.L. Lugo, and E. Valencia. 20 07. Forage production of tropical grasses under extended daylength at subtropical and tropical latitudes. Environ. Exp. Bot. 61: 18 24. Newman, Y., J. Vendramini, and A. Blount. 2010. Bahiagrass ( Paspalum notatum ): overview and management. SS AGR 332 Univ. of Florida Electronic Data Information Source (EDIS), Gainesville, FL. Oliver, J.E., and J.J. Hidore. 2002. Climatology: an atmospheric science. 2 nd ed. Prentice Hall, New Jersey. Overman, A.R., and W.G. Blue. 1991. Estimation of dry matter producti on and nitrogen uptake by Pensacola bahiagrass in Florida. Available at http://ufdc.ufl.edu/?b=UF00086518 (accessed 23 Feb. 201 2). Agricultural Experiment Station, Institute of Food and Agricultural Science s, University of Florida, Gainesville, FL. Overman A.R., C.R. Neff, S.R. Wilkinson, and F.G. Martin. 1990. Water, harvest interval, and applied nitrogen effects on forage yield of bermudagrass and bahiagrass. Agron. J. 82: 1011 1016. Overman, A.R., and R.L. Stanley. 1998. Bahiagrass response to applied n itrogen and harvest interval. Commun. Soil Sci. Plant Anal. 29: 237 244. Podest, and D. Letson. 1999. El Nio : effects on agriculture and forestry. Available at http://ufdc.ufl.edu/UF00053822/00001/1j (accessed 15 Apr. 2011). The Florida Consortium (Florida State University, University of Florida, and University of Miami). 2002. Use of CERES Maize to study effect of spatial precipitation variability on yield. Agric. Syst.73:205 225. doi:10.1016/S0308 521X(01)00095 6 Palmer, W.C. 1965. Meteorological drought. Res. Pap. 45. U.S. Gov. Print. Office, Washington, DC. Pebesma, E. J. 2004. Multivariable geostatistics in S: the gstat package. Comput. Geosci. 30: 683 691.

PAGE 127

127 Pedreira, B.C., C.G.S. Pedreira, K.J. Boote, M.A.S. Lara, and P.D. Alderman. 2011. Adapting the CROPGRO perennial forage model to predict growth of Brachiaria brizan tha Field Crops Res. 120 : 370 379. Peel, D. 2002. Economics of stocker production. Vet. Clin. Food Anim. 22: 271 296. Peel, D., and S. Meyer. 2002. Cattle price seasonality. Available at http://www.uwagec.org/marketing/MngTCMkt/March2002Update/cattleprice season ality2002.pdf (accessed 18 Jan. 2012). Livestock Marketing Information Center, Denver, CO. Piechota, T.C., and J.A. Dracup. 1996. Drought and regional hydrologic variation in the Southern Oscillation. Wat er Resour. Res. 32: 1359 1373. Priestley, C.H.B., and R.J. Taylor. 1972. On the assessment of surface heat flux and evaporation using large scale parameters. Mon. Weather Rev. 100:81 92. doi:10.1175/1520 0493(1972)1002.3.CO;2 Ropelewski, C.F., and M.S. Hal pert. 1996. Quantifying Southern Oscillation precipitation relationships. J. Clim 9: 1043 1059. Rymph, S.J. 2004. Modeling Growth and Composition of Perennial Tropical Forage Grasses. Ph.D. diss. Univ. of Florida, Gainesville, FL. Sassendran, S.A., D.C. N ielsen, L. Ma, and L.R. Ahuja. 2010. Adapting CROPGRO for simulating spring canola growth with both RZWQM2 and DSSAT 4.0. Agron. J 102: 1606 1621. Saxton, K.E., W.. J. Rawls, J.S. Romberger, and R.I. Papendick. 1986. Estimating generalized soil water char acteristics from texture. Soil Sci. Soc. Am. J. 50: 1031 1036. Sentelhas, P.C. 2010. Agricultural drought indices in current use in Brazil p. 60 71 In M.V.K. Sivakumar, R.P Motha, D.A Wilhite and D.A. Wood ( e d.) Agricultural drought indices: Proc. of an Expert Meeting. Murcia, Spain. World Meteorological Organization, Geneva, Switzerland Sentelhas, P.C., T.J. Gillespie, and E.A. Santos. 2010. Evaluation of FAO Penman Monteith and alternative methods for estimating reference evapotranspiration with missi ng data in Southern Ontario, Canada. Agric. Water Manage. 97: 635 644. Sinclair, T.R., P. Mislevy, and J.D. Ray. 2001. Short photoperiod inhibits winter growth of su btropical grasses. Planta 213 : 488 491. Sinclair, T.R., J.D. Ray, P. Mislevy, and L.M. Prem azzi. 2003. Growth of subtropical forage grasses under extended photoperiod during short daylength months. Crop Sci. 43: 618 623.

PAGE 128

128 Sivakumar, M.V.K. 2010. Agricul tural drought WMO perspectives. In M.V.K. Sivakumar, R.P Motha, D.A Wilhite, and D.A. Wood (ed. ) Agricultural drought indices: Proc. of an Expert Meeting. Murcia, Spain. World Meteorological Organization, Geneva, Switzerland. Soil Conservation Service. 1972. Estimation of direct runoff from storm rainfall. Ch. 10. In National engineering handbook. S ect. 4. Hydrology. U.S. Gov. Print. Office, Washington, DC. Song, X., G. Saito, M. Kodama, and H. Sawada. 2004. Early detection system of drought in East Asia using NDVI from NOAA/AVHRR data. Int. J. Remote Sens. 25: 3105 3111. Suleiman, A., and G. Hoogenb oom. 2007. Comparison of Priestley Taylor and FAO 56 Penman Monteith for daily reference evapotranspiration estimation in Georgia. J. Irrig. Drain. Eng. 133: 175 182. Sumner, S., W. Wade, J. Selph, J. Southwell, V. Hoge P. Hogue, E. Jennings, P. Miller, and T. Seawright. 1992. Fertilization of established bahiagrass pasture in Florida. Available at http://rcrec ona.ifas.ufl.edu/pdf/publications/fertilization established bahia pasture FL.pdf (accessed 8 Feb 201 2 ). Cooperative Extension Service, IFAS, University of Florida, Circular 196. Thomas, B.P., L. Law, and D.L. Stankey. 1979. Soil survey of Marion County, Florida. US Govt. Printing Office, Washington, DC. Trenberth, K.E. 1997. The Definition of El Nio. Bull. Am. Meteorol. Soc. 78: 2771 2777. Twidwell, E.K. 2011. Bahia grass production and management. Available at http://text.lsuagcenter.com/NR/rdonlyres/82A096A6 7F6F 4306 853B F49A1D98AFEF/78032/pub2697BahiagrassHIGHRES.pdf (accessed 10 Feb. 2012) LSU Agricult ural Center, Pub. 2697. University of Florida Institute of Food Agricultural Sciences. 1985. Characterization data for selected soils. Soil Sci. Res. Rep.No 85 1. Uryasev, O., A.J. Gijsman, J.W. Jones, and G. Hoogenboom. 2003a. DSSAT v4 soil data editing program (Sbuild). In G. Hoogenboom, J.W. Jones, C.H. Porter, P.W. Wilkens, K.J. Boote, W.D. Batchelor, L.A. Hunt, and G.Y. Tsuji (ed.) Decision Support System for Agrotechnology Transfer Version 4.0. Volume 2 Univ. of Hawaii, Honolulu, HI. Uryasev, O., J .W. Jones, G. Hoogenboom, and C.H. Porter. 2003b. DSSAT v4 crop management data editing program (Xbuild). In G. Hoogenboom, J.W. Jones, C.H. Porter, P.W. Wilkens, K.J. Boote, W.D. Batchelor, L.A. Hunt, and G.Y. Tsuji (ed.) Decision Support System for Agrot echnology Transfer Version 4.0. Volume 2. Univ. of Hawaii, Honolulu, HI.

PAGE 129

129 USDA NASS. 2007. Census of agriculture Available at http://www.agcensus.usda.gov/Publications/2007/Full_Report/ (accessed 12 Jan. 2012). USDA NASS, Washington, DC. Venables, W.N., D. M. Smith, and R Development Core Team. 2011. An Introduction to R. R Development Core Team. Wallach, D. 2006a. The two forms of crop model. p. 11 53. In D. Wallach, D. M akowski, J.W. Jon es (ed.) Working with d ynamic c rop m odels : Evaluation, a nalysis, p aram eterization, and a pplications. Elsevier, Amsterdan. Wallach, D. 2006b. Evaluating crop models. p. 478. In D. Wallach, D. Makowski, J.W. Jones (ed.) Working with dynamic crop models: Evaluation, analysis, parameterization, and applications. Elsevier, Amster dan. Webster, R., and M.A. Olivier. 2007. Geostatistics for environmental scientists, 2 nd ed. John Wiley & Sons. Wilhite, D.A. 2007. Preparedness and coping strategies for agricultural drought risk management: recent progress and trends. p. 21 38. In M.V.K. Siv akumar, and R.P. Motha (ed.) Managing Weather and Climate Risks in Agriculture. Springer, Berlim. Wilhite, D.A., and M. Buchanan Smith. 2005. Drought as Hazard: Understanding the natural and social context. p. 3 29. In A. Wilhite (ed.) Drought an d Water Crises: Science, technology and management issues. CRC Press, Tayl or and Francis Group, New York. Wilhite, D.A., and M.H. Glantz. 1985. Understanding the drought phenomenon: the role of definitions. Water Int. 10 : 111 120. Wilkens, P.W. 2003. DSSAT v4 weather data editing program (Weatherman). In G. Hoogenboom, J.W. Jones, C.H. Porter, P.W. Wilkens, K.J. Boote, W.D. Batchelor, L.A. Hunt, and G.Y. Tsuji (ed.) Decision Support System for Agrotechnology Transfer Version 4.0. Volume 2. Univ. of Hawaii, Honolulu, HI. Williams, J.R., C.A. Jones, J.R. Kiniry, and D.A. Spanel. 1989. The EPIC Crop Growth Model. Transactions of the ASAE 32 : 497 511. Willmott, C.J. 1981. On the validation of models. Phys. Geogr. 2: 184 194. Winsberg, M.D. 2003. Florida Weather. Second. University Press of Florida, Gainesville, FL. Woli, P. 2010. Quantifying water deficit and its effects on crop yields using a simple, generic drought index. Ph.D. diss. Univ. of Florida, Gainesville, FL.

PAGE 130

130 Woli, P., J.W. Jones, K.T. Ingram, and C.W. Fraisse. 2012. Agricultural Reference Index for Drought (ARID). Agron. J 104 : 287 300. Wolter, K., and M.S. Timlin. 1993. Monitoring ENSO in COADS with a seasonally adjusted principal component index. Proc. of the 17th Clim ate Diagnostics Workshop: 52 57, Norman, OK NOAA/NMC/CAC, NSSL, Oklahoma Clim. Survey, CIMMS and the School of Meteor., Univ. of Oklahoma. World Meteorological Organization. 1984. Glossary of terms used in agrometeorology. CAgM Repor t No. 20, Geneva, Switze rland.

PAGE 131

131 BIOGRAPHICAL SKETCH Eduardo Monteiro Gelcer was born in 1986 in So Paulo, Brazil. He grew up in the biggest city of the country, but spent his vacations on horses and buffalo farms near So Paulo, making his passion for the country side rise. His degree in a gronomic e ngineering has given him a strong understanding of the importance and management of natural resources. After one semester of studying at the Universidade Federal de Lavras (UFLA) and working as student assistant on coffee irrigation, he was acce pted in o Paulo ( ESALQ/USP ) where he completed his b degree. During this period he was awarded a scholarship from Companhia Nacional de Abastecimento /Empresa Brasileira de Pesquisa Agropecuria/Universidade de Campinas ( CONAB/EMBRAPA/UNICAMP ) to develop undergraduate research project with cotton crop yield modeling. These experiences and others such as the participation in an academic program to promote and develop research, teachin g and extension (PET), and in the Junior Business University, resulted in him gaining a wide range of knowledge and experiences. In 2008, he was admitted as trainee in a consultant company where he was able to apply his knowledge, especially in climatology and crop modeling. In 2009, he went to University of Florida where he developed studies related to Florida climate. In the following year he started his m d egree program at the Department of Agricultural and Biological Engineering. This one focuse d on crop models and agrometeorology. In summer 2012 he completed his M aster of Science (M.Sc) degree at UF.