<%BANNER%>

Soil Moisture Spatial Variability under Florida Ridge Citrus Tree Canopies and Identifying Alternative Methods of Schedu...

Permanent Link: http://ufdc.ufl.edu/UFE0024532/00001

Material Information

Title: Soil Moisture Spatial Variability under Florida Ridge Citrus Tree Canopies and Identifying Alternative Methods of Scheduling Irrigation Based on Tree Canopy Stress
Physical Description: 1 online resource (95 p.)
Language: english
Creator: Waldo, Laura
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: camera, candler, canopy, citrus, cwsi, entisols, greenseeker, infrared, irrigation, kriging, moisture, multispectral, ndvi, radiometers, reflectance, soil, spatial, stress, tdr, variability, water
Soil and Water Science -- Dissertations, Academic -- UF
Genre: Soil and Water Science thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Due to the increasing demand for water in Florida, the supply of water available for irrigation is decreasing and irrigation costs are rising. Conventionally, irrigation is scheduled based on soil moisture status; however the irrigation system, soil type, and canopy interception of both rainfall and irrigation all increase the variability of moisture under citrus tree canopies. Soil moisture measurements were taken 10 cm apart in a 1.5 by 1.5 m grid under citrus trees using a Time Domain Reflectometry probe equipped with 20 cm probe rods. Soil moisture measurements were analyzed using geostatistics in order to create kriging-interpolated spatial maps depicting variations in soil moisture under mature and young citrus trees. The minimum number of TDR sensor measurements required to estimate soil water content in the topsoil with an accuracy of 90% of plant available water (PAW) 95% of the time was 20 to 289. Therefore, due to site specific variation, single point soil moisture measurements were not accurate enough for triggering irrigation to avoid drought stress or leaching. Accurately matching irrigation scheduling with tree-specific needs may require plant based measurements. Several methods for measuring tree canopy stress were tested both in a controlled greenhouse environment and in the field. These methods include infrared radiometers for thermal infrared canopy temperature measurements, multispectral cameras for water stress index using reflectance, and the GreenSeeker? sensor for measurements of canopy normalized difference vegetation index (NDVI) as an indicator of plant water status. All of the measurements taken were plotted against stem water potential (SWP) as a reference of plant water stress. Results indicate that the thermal infrared radiometer is capable of characterizing water stress using the crop water stress index (CWSI) in a greenhouse setting, however field infrared measurements were less accurate due to instrument sensitivities to wind and low solar radiation levels during cloudy periods. The multispectral camera was able to accurately predict plant water status (R^2 = 0.90***) using a ratio of reflectance at the 840 nm and 670 nm wavelengths. Similar results were found using the commercially available GreenSeeker for NDVI, however the regression analysis showed that while significant, it was less accurate than the multispectral camera (R^2 = 0.31***). Therefore, soil moisture spatial variability is too high to accurately measure soil water status using a single point measurement for triggering irrigation, whereas tree based measurements in the greenhouse using the infrared radiometer could predict water stress using the CWSI, furthermore the multispectral camera and the GreenSeeker were effective at determining water stress in field trees.
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 Laura Waldo.
Thesis: Thesis (M.S.)--University of Florida, 2009.
Local: Adviser: Schumann, Arnold W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024532:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024532/00001

Material Information

Title: Soil Moisture Spatial Variability under Florida Ridge Citrus Tree Canopies and Identifying Alternative Methods of Scheduling Irrigation Based on Tree Canopy Stress
Physical Description: 1 online resource (95 p.)
Language: english
Creator: Waldo, Laura
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: camera, candler, canopy, citrus, cwsi, entisols, greenseeker, infrared, irrigation, kriging, moisture, multispectral, ndvi, radiometers, reflectance, soil, spatial, stress, tdr, variability, water
Soil and Water Science -- Dissertations, Academic -- UF
Genre: Soil and Water Science thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Due to the increasing demand for water in Florida, the supply of water available for irrigation is decreasing and irrigation costs are rising. Conventionally, irrigation is scheduled based on soil moisture status; however the irrigation system, soil type, and canopy interception of both rainfall and irrigation all increase the variability of moisture under citrus tree canopies. Soil moisture measurements were taken 10 cm apart in a 1.5 by 1.5 m grid under citrus trees using a Time Domain Reflectometry probe equipped with 20 cm probe rods. Soil moisture measurements were analyzed using geostatistics in order to create kriging-interpolated spatial maps depicting variations in soil moisture under mature and young citrus trees. The minimum number of TDR sensor measurements required to estimate soil water content in the topsoil with an accuracy of 90% of plant available water (PAW) 95% of the time was 20 to 289. Therefore, due to site specific variation, single point soil moisture measurements were not accurate enough for triggering irrigation to avoid drought stress or leaching. Accurately matching irrigation scheduling with tree-specific needs may require plant based measurements. Several methods for measuring tree canopy stress were tested both in a controlled greenhouse environment and in the field. These methods include infrared radiometers for thermal infrared canopy temperature measurements, multispectral cameras for water stress index using reflectance, and the GreenSeeker? sensor for measurements of canopy normalized difference vegetation index (NDVI) as an indicator of plant water status. All of the measurements taken were plotted against stem water potential (SWP) as a reference of plant water stress. Results indicate that the thermal infrared radiometer is capable of characterizing water stress using the crop water stress index (CWSI) in a greenhouse setting, however field infrared measurements were less accurate due to instrument sensitivities to wind and low solar radiation levels during cloudy periods. The multispectral camera was able to accurately predict plant water status (R^2 = 0.90***) using a ratio of reflectance at the 840 nm and 670 nm wavelengths. Similar results were found using the commercially available GreenSeeker for NDVI, however the regression analysis showed that while significant, it was less accurate than the multispectral camera (R^2 = 0.31***). Therefore, soil moisture spatial variability is too high to accurately measure soil water status using a single point measurement for triggering irrigation, whereas tree based measurements in the greenhouse using the infrared radiometer could predict water stress using the CWSI, furthermore the multispectral camera and the GreenSeeker were effective at determining water stress in field trees.
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 Laura Waldo.
Thesis: Thesis (M.S.)--University of Florida, 2009.
Local: Adviser: Schumann, Arnold W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024532:00001


This item has the following downloads:


Full Text

PAGE 1

SOIL MOISTURE SPATIAL VARIABILIT Y UNDER FLORIDA RIDGE CITRUS TREE CANOPIES AND IDENTIFYING ALTERNA TIVE METHODS OF SCHEDULING IRRIGATION BASED ON TREE CANOPY STRESS By LAURA JEANNE WALDO A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Laura Jeanne Waldo 2

PAGE 3

Your dedication is type d here. It should begin with the word To. (To my Mom is a typical dedication) 3

PAGE 4

ACKNOWLEDGMENTS I would like to extend a special thanks to my advisor Arnold Schumann and my committee Jim Syvertsen and Kelly Morgan. I would also like to thank Apogee Instruments for donating the thermal infrared sensors, along with the Hunt Brothers Coop. and the Southwest Florida Water Management District for funding my research. I would al so like to thank Marjie Cody, Kevin Hostler, Kirandeep Mann, Sherrie Buch anon, Roy Sweeb, Eric Whaley, and Robert Spitaleri for their help with data collection. A spec ial thanks to my friends and colleagues at the Citrus Research and Education Center in Lake Alfred, Florida for their support from start to finish. Finally, I would like to extend a very spec ial thanks to my parents and grandparents for all of their emotional and fina ncial support over the past 28 years of my life. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4 LIST OF TABLES ...........................................................................................................................7 LIST OF FIGURES .........................................................................................................................8 ABSTRACT ...................................................................................................................... .............10 CHAPTER 1 LITERATURE REVIEW .......................................................................................................12 Introduction .................................................................................................................. ...........12 Soil-Based Measurements ......................................................................................................1 2 Direct Soil Moisture Measurement .................................................................................13 Indirect Soil Moisture Measurement ...............................................................................14 Dielectric Methods ..........................................................................................................15 Tensiometric Soil Moisture Measurement ......................................................................17 Tree Based Measurements ......................................................................................................1 8 Canopy Temperature Measurements ...............................................................................19 Multispectral Imagery .....................................................................................................20 Hypothesis and Research Objectives ......................................................................................22 Hypothesis .......................................................................................................................22 Research Objectives ........................................................................................................22 2 SPATIAL VARIABILITY OF SOIL WATE R UNDER CITRUS TREE CANOPIES IN CENTRAL FLORIDA ............................................................................................................24 Introduction .................................................................................................................. ...........24 Hypothesis ..............................................................................................................................25 Objectives .................................................................................................................... ...........26 Materials and Methods ...........................................................................................................26 Isotropic Semivariograms ................................................................................................28 Kriging Interpolation .......................................................................................................31 Results and Discussion ........................................................................................................ ...33 3 IDENTIFYING AND TESTING ALTERNA TIVE METHODS OF DETERMINING WATER STATUS FOR IRRIGATION USIN G TREE CANOPY MEASUREMENTS ......53 Introduction .................................................................................................................. ...........53 Canopy Temperature Measurements ...............................................................................54 Multispectral Imagery .....................................................................................................56 Hypothesis ..............................................................................................................................58 Objectives .................................................................................................................... ...........58 5

PAGE 6

Materials and Methods ...........................................................................................................58 Thermal Infrared Canopy Measurements ........................................................................58 Multispectral Camera Imaging ........................................................................................63 GreenSeeker NDVI ......................................................................................................66 Results and Discussion ........................................................................................................ ...66 Infrared Canopy Measurements ......................................................................................67 Multispectral Camera Imaging ........................................................................................70 GreenSeeker NDVI ......................................................................................................73 4 SUMMARY OF RESULTS ...................................................................................................88 LIST OF REFERENCES ...............................................................................................................92 BIOGRAPHICAL SKETCH .........................................................................................................95 6

PAGE 7

LIST OF TABLES Table page 2-1 Geostatistics from GS+ isotropic semiva riogram analysis for moisture grid data including several days of drying, rain, and 1 hour of irrigation. .......................................38 2-2 Basic statistical results for moisture grid data including several days of drying, rain, and 1 hour of irrigation. .....................................................................................................39 2-3 Geostatistics from GS+ isotropic semiva riogram analysis for moisture grid data measurements taken 1 h, 4 h, 24 h, and 48 h after irrigation. ............................................40 2-4 Basic statistical results for moisture gr id data including measurements taken 1 h, 4 h, 24 h, and 48 h after a 3 hour irrigation a pplication ............................................................41 2-5 95% confidence levels calculated for th e number of sensors needed for precise soil water estimation .................................................................................................................42 2-6 Number of sensors needed, for accurate estimations of soil water, after 95% confidence levels were calculated for each tree grid. ........................................................43 2-7 Number of sensors needed, for accurate estimations of soil water, after 95% confidence levels were calculated for each tree grid. ........................................................44 7

PAGE 8

LIST OF FIGURES Figure page 2-1 Imposed irrigation wetting pattern from nozzles used in experimental block ...................45 2-2 Example of a tree grid placed under the canopy of a young tree .......................................45 2-3 Field Scout TDR100 Soil Moisture Mete r used for soil moisture measurements .............46 2-4 Irrigation spray from the micro-jet no zzle used in the experimental block .......................46 2-5 Calibration showing linear regressi on used for conversion from period ( s) to volumetric water content (m3/m3) ......................................................................................47 2-6 Example of the spherical model fo r isotropic semivariograms (h = cm) ...........................47 2-7 Example of the exponential model fo r isotropic semivariograms (h = cm) ......................48 2-8 Example of the Gaussian model fo r isotropic semivariograms (h = cm) ..........................48 2-9 Kriging interpolation for a mature tree after rainfall .........................................................49 2-10 Kriging interpolation for a young tree after rainfall ..........................................................49 2-11 Kriging interpolation of a mature tree after 1 hour application of irrigation .....................50 2-12 Kriging interpolation from a young tree grid following a 1 hour application of irrigation including the overl ying pattern of water spray from irrigation nozzle ..............50 2-13 Series of Kriging interpolated spatia l maps showing irrigation nozzle pattern .................51 2-14 Kriging interpolation results from moistu re measurements taken from the blank grids 1 hour, 4 hours, 24 hours, and 48 hours afte r a 3 hour irrigati on application. ...................51 2-15 Kriging interpolation results from moisture measurements taken from the mature tree grids 1 hour, 4 hours, 24 hours, and 48 hours after a 3 hour irrigation application. ..........52 2-16 Kriging interpolation results from mois ture measurements taken from the young tree grids 1 hour, 4 hours, 24 hours, and 48 hours after a 3 hour irrigation application. ..........52 3-1 Apogee Instruments Inc., IRR-PN Infrared Radiometer ...................................................75 3-2 20 Rough Lemon trees, planted in potted gr ove soil used for greenhouse experiments ...75 3-3 Aqua-pro capacitance soil moisture probe used for soil moisture monitoring ..................76 3-4 IRR-PN on ring stand pointed at tr ee canopy during long term measurements. ...............76 8

PAGE 9

3-5 Regression analysis from non-stressed tree data for baseline or lower limit ( TLL) calculation ................................................................................................................... .......77 3-6 Soil covered with Tyvek HomeWrap (DuPont, Wilmington, Delaware) ....................77 3-7 Graphical presentation of stem water pot entials versus the stress-degree-day (Tc Ta) results from measurements taken in th e second infrared radiometer experiment .......78 3-8 Graphical presentation of soil volumetric water content versus the stress-degree-day (Tc Ta) results from measurements ta ken in the second infrared radiometer experiment..................................................................................................................... .....78 3-9 Graph of measurements take n during IRR-PN experiment 2 ............................................79 3-10 Crop Water Stress Index and ir rigation application by date ..............................................79 3-11 Comparison of measurements taken with the infrared radiometer in cloudy and sunny conditions in the greenhouse. .............................................................................................80 3-12 Regression analysis showing Tc Ta ve rsus stem water potential results from the third infrared radiometer experiment. ................................................................................80 3-13 Regression analysis comparing measuremen ts of Tc Ta, taken with the infrared radiometer in both windy and wind free conditions versus stem water potential. .............81 3-14 Comparison of the regression analys is of the CWSI normalized IRR-PN measurements taken in windy and windless c onditions versus stem water potential ........81 3-15 Correlation analysis of Crop Stress I ndex ratios using the 840 nm and 670 nm wavelengths........................................................................................................................82 3-16 Analysis comparing the stressed and non-stressed leaves .................................................82 3-17 Regression analysis of water stress index ratios ................................................................83 3-18 Screen capture of data processing fr om a small greenhouse tree showing the large amount of background interference. ..................................................................................83 3-19 Regression from Clear Sky conditions s howing outliers B. E. as a result of improper imaging ...............................................................................................................84 3-20 Images captured by th e multispectral camera ....................................................................85 3-21 Regression analysis from measuremen ts taken on a sunny day with average wind speed of 1.65 m/s and max gusts of 4.02 m/s. ...................................................................86 3-22 Regression analysis from m easurements taken on a cloudy day .......................................86 3-23 Regression analysis from measurem ents taken with the GreenSeeker ...........................87 9

PAGE 10

Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science SOIL MOISTURE SPATIAL VARIABILIT Y UNDER CITRUS TREE CANOPIES AND IDENTIFYING ALTERNATIVE METHODS OF SCHEDULING IRRIGATION BASED ON TREE STRESS By Laura Jeanne Waldo May 2009 Chair: Arnold W. Schumann Major: Soil and Water Science Due to the increasing demand for water in Fl orida, the supply of water available for irrigation is decreasing and irri gation costs are rising. Conventi onally, irrigation is scheduled based on soil moisture status; however the ir rigation system, soil type, and canopy interception of both rainfall and irriga tion all increase the variability of moisture under citrus tree canopies. Soil moisture measurements were taken 10 cm apart in a 1.5 by 1.5 m grid under citrus trees using a Time Domain Reflectometry probe equi pped with 20 cm probe rods. Soil moisture measurements were analyzed usi ng geostatistics in order to create kriging-interpolated spatial maps depicting variations in soil moisture under mature and young citr us trees. The minimum number of TDR sensor measuremen ts required to estimate soil wate r content in the topsoil with an accuracy of 90% of plant available water (PAW) 95% of the time was 20 to 289. Therefore, due to site specific variation, si ngle point soil moisture measur ements were not accurate enough for triggering irrigation to avoid drought st ress or leaching. Accurately matching irrigation scheduling with tree-specific need s may require plant based meas urements. Several methods for measuring tree canopy stress were tested both in a controlled greenhouse environment and in the field. These methods include infrared radiomet ers for thermal infrared canopy temperature 10

PAGE 11

11 measurements, multispectral cameras for water stress index using reflectance, and the GreenSeeker sensor for measurements of canopy normalized difference vegetation index (NDVI) as an indicator of plant water status. All of the measurements taken were plotted against stem water potential (SWP) as a reference of plan t water stress. Results indicate that the thermal infrared radiometer is capabl e of characterizing water stress using the crop water stress index (CWSI) in a greenhouse setting, however field infrar ed measurements were less accurate due to instrument sensitivities to wind and low solar radiation levels during cloudy periods. The multispectral camera was able to accurately pred ict plant water status (R^2 = 0.90***) using a ratio of reflectance at the 840 nm and 670 nm wa velengths. Similar results were found using the commercially available GreenSeeker for NDVI, how ever the regression analysis showed that while significant, it was less accurate than th e multispectral camera (R^2 = 0.31***). Therefore, soil moisture spatial variability is too high to a ccurately measure soil water status using a single point measurement for triggering irrigation, whereas, tree based measurements in the greenhouse using the infrared radiometer could predict water stress us ing the CWSI, furthermore the multispectral camera and the GreenSeeker were e ffective at determining water stress in field trees.

PAGE 12

CHAPTER 1 LITERATURE REVIEW Introduction Typically, rainfall in Florida exceeds the amount of water lo st through evapotranspiration (Morgan, et al., 2006). Annual rainfall, however, is not uniformly distributed throughout the year causing a surplus in the summer (w et) season, and a deficit in the winter and spring (dry) seasons (Morgan, et al., 2006). Due to this fact, supplemental irrigation is necessary to maintain fruit-set and increase yield. Studies have shown that improved management of irrigation is critical in preventing nitrogen and other chemicals from leach ing into the surficial aquifer (Morgan, et al., 2006). As a result of these studies, irrigation ma nagement is an essential component of the Florida Ridge regions citrus best management practices (BMPs). In orde r to achieve the desired crop response and reduce irrigation applied, a site-specifi c irrigation schedule is needed. Sitespecific irrigation is matched to the actual cr op needs at the smallest manageable unit level (Cohen, et. al., 2005). Identifying the appropriate time to irrigate requires monitoring these needs intensively. Conventionally this has been achieved through so il water monitoring and soil water balance calculations (Jones, 2004). Recently, newer methods of scheduling irrigation have been proposed which use canopy measurements to iden tify the level of water stress based on tree canopy measurements (Jones, 2004). Soil-Based Measurements Conventionally crop water status has been m easured using soil-based measurements. These measurements are done by sampling a specific amount of soil and removing the water (direct methods), or by measuring some other soil property such as electrical conductivity (indirect methods) (Muoz-Carpena, 2004). 12

PAGE 13

Direct Soil Moisture Measurement There are two types of direct methods for meas uring the water content of the soil, thermogravimetric and thermo-volumetric methods. The direct thermo-gravimetric method or gravimetric method of measuring the water content of the soil is determined using the difference in the weight before and after drying a soil sa mple (Charlesworth, 2000). This method measures the water content of th e soil by weighing a fresh sample, dryi ng it at 105 C to obtain the mass of dry soil, then expressing the weight of the water in the soil over the weight of the dried soil (g/g), resulting in the gravimetric water content (Muoz-Carpena, 2004). One problem with the gravimetric method is the varying density of diffe rent soils, which mean s a unit weight of soil may occupy a different volume (Charlesworth, 20 00). In order to compare the different water contents of different soils as well as to calculate the amount of wa ter to add to th e soil to satisfy the plants requirements a measure of the soil vol umetric water content is needed (Charlesworth, 2000). Likewise, the direct thermo-volumetric method, which is similar to the gravimetric method, uses moist and dry soil weight. However, factoring in the soils bulk density (mass per unit volume of an undisturbed soil core) and the de nsity of water results in the volume of water over the volume of oven-dried und isturbed sample soil core (m3/m3) (Muoz-Carpena, 2004). While these methods (gravimetric and volumet ric) are accurate and inexpensive, both are slow, taking two days minimum. These direct methods are also destructiv e since the soil sample must be removed from the field a nd brought to a lab to be dried in an oven. As a result, repeated sampling from the same location is no t possible (Muoz-Carpena, 2004). These two disadvantages make the direct method of measur ing soil moisture impracticable for large scale measurements or for circumstances where soil moisture results are needed quickly, for example, grove irrigation scheduling. 13

PAGE 14

Indirect Soil Moisture Measurement Similar to the direct methods, there are also two types of indirect methods. The first are volumetric methods, since they measure the volume tric soil moisture, and secondly tensiometric methods, which are those that yield the soil suct ion or water potential (Muoz-Carpena, 2004). Indirect methods used to estimate soil moisture involve a calibrated relatio nship with some other measurable variable. The suitabil ity of these indirect methods is dependent on issues such as cost, accuracy, response time, installation, management and durability (Muoz-Carpena, 2004). Indirect methods, both volumetric and tens iometric, are relate d through a soil water characteristic curve which is specific to a gi ven soil; therefore each soil type must have a different calibration since the methods will not respond the same way for all soil types (Haman and Izuno, 1993). The effectivenes s of the indirect measuremen t device is dependent on the physical properties of the soil, as well as th e goal of the soil moisture measurement (MuozCarpena, 2004). While volumetric devices result in a more intuitive quantity, they may not be as useful as some tensiometric devi ces which relate to the energy that plants have to invest to extract water from the soil, especially in fine textured soils which can hold water tightly to the soil particles making it unavailable for plan t absorption (Muoz-Carpena, 2004). Another important factor for selecting a suitable device for soil moisture measurements is the response time, or time it takes to obtain th e result. Some sensors take more time than others because they require the soil moisture to equilibrate with the sensor matrix (Muoz-Carpena, 2004). Also soil physical properties such as textur e may influence the suitability of the sensor due to instrument contact with the soil, and maintenance that may need to be done due to environmental circumstances (Muoz-Carpena, 2004). There are several types of volumetric indirect soil moisture measurement methods. These include dielectric, neutron m oderation, and ground penetrating radar methods. Also there are 14

PAGE 15

several methods for the tensiometric indirect soil moisture measurements. The tensiometric methods include tensiometers, resistance blocks heat dissipation, and so il psychrometers. Dielectric Methods Dielectric methods estimate the soil water content by measuring the soil bulk permittivity or dielectric constant, Kab, which for the time-domain methods determines the velocity of an electromagnetic wave or pulse through the soil (Muoz-Carpena, 2004). Since soil is made up of different components, which include minerals, air, and water, the value of the permittivity is determined by the relative proportion of each of th e components. In addition, since the dielectric constant of water (Kaw = 81) is much larger than that of the other soil constituents (Kas = 2-5 for soil minerals and 1 for air), the total permittivity of the soil, or bulk permittivity, is mainly governed by the presence of water (Topp et al ., 1980). According to Topp et al. (1980), a common approach to establis h the relationship between Kab and volumetric soil moisture (VWC) is through the relationship seen in Equation 1-1. The relationship works fo r most mineral soils and for moisture below 50% (0.5 m3/m3). Also the relationship depends on the frequency of the electromagnetic signal sent by the sp ecific device (Topp et al., 1980). VWC = -5.3 x 10-2 + 2.29 x 10-2Kab 5.5 x 10-4Kab 2 + 4.3 x 10-6Kab 3 (1-1) Some of these dielectric methods are Time Domain Reflectometry (TDR), Frequency Domain Reflectometry (FDR), Amplitude Doma in Reflectometry (ADR), Phase Transmission, and Time Domain Transmission (TDT). All of these methods, except the FDR method, involve an electrical pulse or wave traveling along a transmission line. In the TDR method the pulse travels down and reflects back along the tran smission line, whereas in the TDT method the electromagnetic pulse only travel s in one direction and has an electrical co nnection at both ends rather than at a single end (Muoz-Carpena 2004). In the ADR me thod the electromagnetic pulse that is traveling along the transmission lin e will encounter an area with different impedance 15

PAGE 16

in which part of the wave will be reflected back to the transmitter; as the reflected wave interacts with the incident wave it will change the amp litude of the wave along the transmission line (Wijaya et al., 2002). Similarly in the phase transmission method a sinusoidal wave will show a phase shift relative to the phase at the origin after having traveled a fixed distance along a transmission line (Muoz-Carpen a, 2004). Unlike the methods mentioned previously, the FDR method uses a soil capacitor (the soil between two electrodes) together with an oscillator to form an electrical circuit; changes in the soil mois ture can be detected by changes in the circuit operating frequency (Robinson et al., 1999). Many of these methods are affected negatively by high salinity conditions in the soil. The conditions could be as a result of fertilization or from the use of saline irrigation water. In either case this could lead to an underestimation of the water needed for th e trees, and ultimately result in a reduction of the yield. Anot her problem with many of these methods is the small volume of soil measured for water content. For example, some TDR sensors have a sensing volume along their length defined by only a three centimeter radius, while the ADR method can have a sensing volume as little as 4.4 cubic centimeters (Muoz-Carpena, 2004). Since the phase transmission and the TDT methods have larger volumes of soil that are measured they also re quire permanent installation, which disrupts the soil profile and leads to inaccurate measurements due to the channeling of water around the sensor or differe nt soil densities around th e sensor that are not representative of the area in question. Two additional volumetric field methods in clude ground penetrating radar (GPR) and electromagnetic induction (EMI) (Muoz-Carpena, 2004). GPR is based on the same principle as TDR; however, it does not requ ire any direct contact between the soil and the sensor. GPR is mounted on a vehicle or a sled close to the so il surface and has the poten tial of providing rapid, 16

PAGE 17

non-disturbing, soil moisture measurements over a large area (Davis and Annan, 2002). EMI is specifically suited for measurements over large areas. EMI uses two antennae to transmit and receive electromagnetic signals that are reflect ed by the soil (Dane and Topp, 2002). A problem with EMI is that it does not measure water content directly, but through soil electrical conductivity. As a result, a known calibration rela tionship between the sensor and the soil is needed, which is site specific and can not be assumed (Dane and Topp, 2002). Tensiometric Soil Moisture Measurement Unlike dielectric measurements where the water content is measured, in the tensiometric methods the soil water matric potential is meas ured. The matric potentia l of a soil includes both the capillary and adsorption effects of the soil. All of the tensiometric methods include a porous material placed in contact with the soil, which allows water entry. The basic principle of this method is that in a dry soil the water will be removed from the porous medium, and in a wet soil water will travel into the porous material (Charlesworth, 2000). Due to the fact that this method measures the matric potential of the soil, a site specific soil calibration is not needed, however the major drawbacks to this method are permanent installation and the periodic maintenance that is required (Muoz-Carpena, 2004). As with tensiometers; resistance blocks, h eat dissipation, and soil psychrometers all employ porous material which allows the move ment of water. Tensiometers and soil psychrometers are the most common instruments for the measure of the energy status of a soil (Jones, 2007). While tensiometers directly use the soil matric potential resistance blocks use electrodes within the porous material for m easurement based on electrical resistance (Charlesworth, 2000). Along with the use of a porous medium, heat dissipation also makes use of the thermal conductivity of a dry soil versus that of a wet soil (Muoz-Carpena, 2004). Heat capacity is the amount of heat energy needed to increase the temperature of a quantity of water 17

PAGE 18

by one degree Celsius (Charlesworth, 2000). The so il psychrometer utilizes a porous medium for the flow and entry of moisture; however it m easures the relative humidity in a chamber as a result of vapor movement (Campbell and Gardner, 1971). One advantage to these methods is that they are less susceptible to high salt levels in the soil. Unlike tensiometers and resistances blocks, which can require more maintenance, he at dissipation requires very little maintenance (Muoz-Carpena, 2004). A problem w ith tensiometric methods, lik e the volumetric methods, is the generally small sensing volumes in some methods and the high soil disruption of the other methods; both of which could lead to either under or over estimation of soil water and result in over or under irrigation. While a drawback to some of these sensors is the small sensing volume or sphere of soil, which only allows for an accurate reading from very near the probe location, the sensors that have a larger sensing volume have the drawback of needing to be permanently installed and thus disturbing the soil profile during installation. As a result of some of these drawbacks and disadvantages, some new methods have been iden tified that rely on measurements of tree canopy water stress rather than so il moisture measurements. Tree Based Measurements Several methods of measuring crop water st ress have been tested for accuracy in measuring plant water status. One of these met hods is crop temperature measurement by thermal infrared thermometers, which have been found to be reliable in some crops as well as noninvasive. Another method tested canopy reflectance using multispectral cameras, which utilize a ratio of the canopy reflectance at two specific wavelengths. 18

PAGE 19

Canopy Temperature Measurements Plant temperature has long been recognized as an indicator of water av ailability (Jackson et al., 1981). Until thermal infrared thermometers became available, most plant temperature measurements were made with contact sensors on or embedded in leaves (Jackson et al., 1981). The theory behind the use of canopy temperature is that for any given environmental conditions, the leaf or canopy temperature is directly relate d to the rate of evapotranspiration from the canopy surface (Leinonen and Jones, 2004). Therefore, as plants transpire the temperature of the canopy is lowered by evaporative cooling, effectively making the canopy temperature cooler than the air temperature. However, when a plan t is water stressed, transpiration becomes limited which results in an increase in temperature th at can match or exceed that of the air. The usefulness of canopy temperature as a measure of crop water stress was recognized in the 1960s (Mller et al., 2007). Jackson (1981) derive d the use of canopy temperature minus air temperature (Tc Ta), from the energy balance for a crop canopy (Equation 1-2). Rn = G + H + E (1-2) Where Rn is the net radiation (W/m2), G is the heat flux below the canopy (W/m2), H is the sensible heat flux (W/m2) from the canopy to the air, E is the latent heat flux to the air (W/m2), and is the heat of vaporization Hope and Jackson (1989) used Tc Ta, of a wheat crop, as an index of crop water status. The difference (Tc Ta or T) is also called the stress-degree-day when the T is summed over a period of time. Later Tc Ta and vapor pressure deficit (VPD) data for several crops showed a linear relationship fo r well watered crops under clear sky conditions (Jackson et al., 1981). Jackson et al. (1988), iden tified the use of upper and lower limits for calculating the crop water stress index (CWSI) (Equation 1-3), a nd described the purpose of the upper ( TUL) and lower ( TLL) limits was to form bounds by which the measured temperature can be normalized. 19

PAGE 20

CWSI = (( T TLL)/( TUL TLL)) (1-3) According to Monteith and Unsworth (2008), th ere is a dependence of transpiration rate on radiation and vapor saturation deficit. When leav es are in their natural environment, stomatal aperture depends strongly on sola r radiation; in the absence of light, stomata are usually closed, making transpiration effectively zero (Monteith and Unsworth, 2008) There is also substantial evidence both from the field and from work in controlled environments which reveals that many plants close their stomata as saturation deficit or VPD increases, which is presumably a mechanism for conserving water (Monteith and Unsworth, 2008). Multispectral Imagery Electromagnetic radiation that is reflected or emitted from the Earths surface can be recorded by a sensor from the gr ound, aircraft, or sate llite (Curran, 1983). Today an increase in knowledge of the way in which electromagnetic ra diation interacts with our environment, has enabled scientists to use such remotely sensed da ta to determine the amount of soil moisture in a field or the amount of suspended sediment in estuarine waters (Curran, 1983). Some of the solar irradiance that is impinging upon a vegetation canopy is reflected, while the rest is either transmitted and/or is ab sorbed (Curran, 1983). The intensity with which radiation is reflected at any particular wavelength is dependent on both the spectral properties and also the area of the leaves substrate, and shadow (Curra n, 1983). Leaves usually reflect weakly in the blue and red wavelengths due to the absorption by photosynthetic pigments, and likewise they reflect strongly in the near-infrared (N IR) wavelengths due to cellular refraction. The most widely used green vegetation indices ar e formed with data from discrete red and NIR bands (Elvidge and Chen, 1995). A ratio calle d the normalized difference vegetation index (NDVI) or Vegetation index (Iv) (Equation 1-4) is one of the more popular (Curran, 1983). Iv = (Rir Rr)/(Rir + Rr) (1-4) 20

PAGE 21

Where Rir is NIR reflectance and Rr is red reflectance Spectrometers with measurement ranges beyond 1000 nm have been used to determine water stress in plants by analyz ing reflectance measurements at several key wavelengths called water bands (Dallon and Bugbee, 2003). The mo st prominent water bands are at 1400 and 1900 nm and reflectance at these wavelengths has been shown to correspond to water content in plant tissue (Dallon and Bugbee, 2003). Unfortunately natural sunlight reaching the surface of the earth has low intensities at these wavelengths due to absorptive filtering by water in the atmosphere. Spectrometers capable of m easuring radiation beyond 1000 nm are also considerably more expensive than those measur ing in the visible and short wavelength NIR ranges (i.e. 400 1000 nm). At the 970 nm wavele ngth there is another water band, however it has historically been considered too small to accurately measure water stress. Dallon and Bugbee (2003) found that if using an accurate spectromete r that can measure wavelengths up to 1000 nm, accurate estimates of wate r stress can be measured at the 970 nm water band. In order to test the use of the 970 waveband Dallon and Bugbee (2003) used three indices to analyze the various water bands. The first of the indices used, th e reflectance water index, is a ratio between the reflectance at a water band to a nearby reference wavelength that is unaffected by water content variability. The second of the indi ces used is the band depth analys is, which uses a process called continuum removal where a linear continuum line is approximated across an area of absorption, connecting two unaffected points of the spectrum; a nd the third of the indices is the first-order derivative green vegetation index (1DGVI), wh ich is based on a complex calculus formula involving integrated derivatives that reduces down to a simple difference between a wavelength within the water band that is subtracted from a reference point wavelength. 21

PAGE 22

Applying a similar index to images taken with a multispectral camera, Schumann et al. (2007) found that by photographing a canopy at specific wavelengths with a multispectral camera can result in a yield index and a canopy st ress index. Using a multispectral camera fitted with a filter wheel, Schumann et al. (2007) used grayscale values for each pixel from citrus canopy images taken at 840 nm and 670 nm and applie d these pixel values to a ratio of 840 nm / 670 nm. Hypothesis and Research Objectives Given that there are several fact ors which lead to variability in soil moisture, it is possible that using single point measurements will not be sufficient enough to accurately trigger irrigation. As a result, irrigati on managers will need to rely on some other method of scheduling irrigation. Thermal canopy temperature measurements and multispectral camera imaging, when applied to tree canopy measurements, have a pos sibility of detecti ng plant water stress. Presumably using the trees canopy for water stress measurements would be more accurate because of the trees ability to integrate water from the entire root system. Using the canopy for measuring stress should remove the possibility for variation due to single soil based measurements. Hypothesis The spatial variability of water in the rootzone under citrus tr ee canopies at a given time is excessive and limits the use of single-sensor soil moisture data for triggering irrigation events. Canopy measurements taken with an infrar ed radiometer, multispectral camera, or GreenSeeker can be used to estimate the curre nt water status of a citrus tree with the possibility of being used for irrigation triggering. Research Objectives To investigate the soil moisture variability under citrus tree canopies using a TDR moisture sensor and spatially mapping the moisture result s under several citrus tr ees before and after rain or irrigation. 22

PAGE 23

23 To test several alternative methods of measuring tree water stress based on canopy measurements in order to determine if thes e methods could lead to optimal irrigation scheduling which would effectively supply wate r to the trees and mi nimize the leaching of nutrients due to over irrigation.

PAGE 24

CHAPTER 2 SPATIAL VARIABILITY OF SOIL WATE R UNDER CITRUS TREE CANOPIES IN CENTRAL FLORIDA Introduction Irrigation is normally carried out according to recomme ndations based on potential evapotranspiration, and crop coefficients, with adju stment according to soil moisture and rainfall (Cohen et. al., 2005). Due to the possible spatial variability of soil moisture under a citrus tree canopy it may not be practical to use soil meas urements for accurate irrigation scheduling. As mentioned in Chapter 1, soil moisture is ca lculated using direct measurements of soil water by weighing and drying a soil sample and indi rect measurements which include dielectric and tensiometric methods. The disadvantages of these methods include the amount of time required for the one time sampling from a single point, as in the case of direct measurements, and the small sensing volume, maintenance, and soil disruption of the indirect methods (MuozCarpena, 2004). Sampling a small volume or sphere of soil with indirect methods would not be a problem if the soil moisture was spatially unifo rm. However, due to several factors the soil moisture under a citrus tree canopy is spatially vari able. This spatial variab ility of soil moisture results from the variability of soil properti es both horizontally and with depth, non-uniform interception of rainfall by the tree canopy, micr o-topography, and irriga tion methods (Cohen et al., 2005). Irrigation methods used in citrus produ ction in Florida include micro-jet, drip, or overhead sprinkler. Soil types in the Ridge citrus production area consist primarily of highl y drained fine sands in the Entisol soil order. The tr ee canopy affects where rainfall will make contact with the soil (Alva et al., 1999). Rain collect s on leaves and branches, running along stems and the canopy edge increasing the affects of rainfall in some areas more than others. Irregularities from irrigation application will cause variability; for example, irrigation nozzles impose a pattern in 24

PAGE 25

the application of the water (Figure 2-1). Irrigation nozzles that are designed to have an evenly wetted area are disrupted by low hanging branches the tree trunk, and wind, which can affect the wetted pattern on the ground surface. The some of the indirect moisture sensing devices that are available today are adversely affected by the salinity of the soil, either as a result of fertilization or saline irrigation water (Boman and Stover, 2002). As ionic strength in the soil solution increases so does apparent soil water content measured with these sensors. Increased salinity also affects the osmotic potential in the rhizosph ere relative to unaffected soil, thus, affecting the uptake of water (Muoz-Carpena, 2004). These sens ors may show that ther e is adequate water in the soil for the tree to absorb wh ile in reality that moisture may not be sufficient. This is due to the fact that the indirect methods rely on th e dielectric constant of the measured media, measuring in essence, the electrical conductivity of the soil, which is strongly affected by ions in solution. The sensor circuits output in volts or millivolts is converted to soil moisture based on a specific calibration in order to make the measurem ent useful for moisture sensing. An increase in salts in the soil, due to fertilization or the use of saline irrigation water will increase the electrical conductivity of the soil solution cau sing the sensor to report more soil moisture than was is actually present in the soil (Boman and Stover, 2002). In order to overcome this problem, moisture sensors would need to be calibrated over all possible combinations or soil water and salinities. The combination of these issues makes it necessary to map the soil moisture under citrus trees to find out whethe r single point soil moisture meas urements are accurate enough to predict the water requirements of a tree without either over or under irrigation. Hypothesis The spatial variability of water in the rootzone under citrus tr ee canopies at a given time is excessive and limits the use of single-sensor soil moisture data fo r triggering irrigation events. 25

PAGE 26

Objectives To investigate the soil moisture variability under citrus tree canopies using a TDR moisture sensor and spatially mapping the moisture results under several citrus trees before and after rain or irrigation. Materials and Methods A one hectare citrus grove (Block 9B) loca ted at the Citrus Research and Education Center (CREC), in Lake Alfred, Florida was used for this study. According to the Southwest Florida Water Management Districts GIS websit e (2002), the soil found in this block is Candler Fine Sand, 0 to 5 Percent slopes. The Candler seri es is typically found on the Florida Lake Wales Ridge and consists of excessively drained soils that formed in sandy marine or aeolian deposits (USDA, NRCS, 1990). Six trees were used in th is study; three mature trees (average of 21 m3) and three young reset tr ees (average of 9.25 m3). Square grids of plastic mesh measuring 1.5 x 1.5 m were placed under each trees canopy on the sa me side as the micro-jet irrigation nozzle (Figure 2-2). Three additional grids were placed ov er bare soil, to illustrate the effects of irrigation without tree canopy or root interference. Measurements of soil moisture were taken at uniform 10 cm spacing within the grid, totali ng 225 sample points per tree. Soil moisture measurements were taken after rainfall even ts of 6.25 mm and 11.25 mm and after irrigation events of one hour, equivalent to approxima tely 2.15 mm, and three hours, equivalent to approximately 6.45 mm. A micro-jet irrigation system was used for the irrigation events, with a violet Maxi-jet fill in 360 nozzle delivering 51 L/h and with a wetted diameter of 5.5 m. (Figure 2-4). The soil moisture was measured using a Time Domain Reflectometry (TDR), Field Scout Soil Moisture Meter (Spectrum Tec hnologies, Plainfield, Illinois) w ith 20 cm probe rods (Figure 2-3). 26

PAGE 27

Since the TDR instrument is an indirect measurement of soil moisture, a site specific calibration was performed to prope rly calibrate the TRD probe for accurate measurements of volumetric soil water content (VWC). A large (a pproximately 15 L), representative, sample of soil was removed at the 0-20 cm depth from the field site and dried in the oven at 105C for 48 hours. Two liters of mixed, dry soil was used for calibration; de-ionized water was added in known increments by volume and mixed thoroughly w ith the soil. For each water increment, the moistened soil was packed in a 10 cm diameter PV C tube with end cap, compacted to the 2 L fill mark, and measurements were taken with the TDR meter. The TDR periods, in microseconds, were calibrated with the calculated volumetric water contents by linear regression (Figure 2-5). This linear regression was used to convert th e microsecond period to volumetric water content (m3/m3) for all of the moisture data collected. Sin ce the irrigation water used is high quality and has an EC of 0.38 dS/m no other calibrations fo r salt sensitivity were conducted. The soil found in the block where measurements will be take n is a Candler Sand, which is a hyperthermic, uncoated Typic Quartzipsamments, which is lo w in organic matter and excessively drained, making it fairly resistant to salt buildup. Given th e soil type and high quality irrigation water an extra calibration for salts will not be necessary. Si nce the variability of the soil is what is in question, any added variability due to salts added by fertilizer a pplications during the study will only add to the increased amount of variability. The moisture measurements were geospatia lly analyzed using GS+ Geostatistics for the Environmental Sciences (Gamma Design Software, Plainwell, MI). Semivariograms and Kriging analysis was used to measure the extent and natu re of the VWC variability within each measured grid area. The data from root zones under the tr ee canopies were compared with irrigated bare soil in the same grove. 27

PAGE 28

Three types of VWC measurements were taken: 1) soil moisture after several days of drying, 2) soil moisture one hour after one hour of irrigation (2.15 mm) as well as zero hours, four hours, 24 hours, and 48 hours after three hours of irrigation (6.45 mm); and 3) soil moisture immediately after and 20 hrs af ter 6.25 mm and 11.25 mm of rain fall, respectively. Soil water data from the blank soil grids, which were away from any trees and should have received no interference from tree canopy or root water uptake were used to compare to the soil water data from the root zone of the mature and young trees. After each rainfall or irrigation event, the soil moisture was analyzed geospatially for variab ility using Isotropic Semivariograms and Kriging interpolation. Isotropic Semivariograms GS+ automatically chooses the model that best fits the data being an alyzed. The isotropic models used for the analysis of the moisture data that were collected include the Spherical, Exponential, and Gaussian models All of the models use the same coefficients which include nugget (Co), effective range (A), range parameter (Ao), and sill (Co + C). The nugget (Co), is the y-intercept of the model, and is a measure of th e amount of variance due to errors in sampling, measurement, and other unexplai ned sources of variance (Mulla & McBratney, 2000). The range parameter (Ao) is used in the isotropic variogram to calculate the effective range (A), which is the distance at which samples become spatially independent and uncorrelated with one another (Mulla & McBratney, 2000). At separation distances greater than the range, sampled points are no longer spatially correlated. Th is means that as a region is being sampled, in order to understand the spatial pattern of a given property, which in this case is soil moisture, it is advisable that the sampling design use separation distances that ar e at most, no greater than the value for the range parameter of the semivariogr am (Mulla & McBratney, 2000). It is preferable that the sample spacing be from one quarter to one half of the range (Mulla & McBratney, 2000). 28

PAGE 29

Finally the sill (Co + C) is the model asymptot e. The sill represents spatially-independent variance. Data locations sepa rated by a distance beyond which semivariance does not change (after the model asymptote or sill) are spatia lly independent of one another (Gamma Design Software, LLC, 2004). Theoretically the sill is equivalent to the variance of the sampled population at a large separation distance if the data have no trend (Mul la & McBratney, 2000). The Spherical isotropic model (Figure 2-6) is a modified quadratic function for which at some distance (Ao), pairs of points will no long er be autocorrelated and the semivariogram reaches an asymptote or sill (Co + C) (G amma Design Software, LLC, 2004). GS+ uses Equation 2-1 for Spherical models. y(h) = Co + C [1.5(h/Ao) 0.5(h/Ao)3] for h Ao (2-1) y(h) = Co + C for h > Ao Where: y(h) = semivariance for interval distance class h, h = the lag distance interval, Co = nugget variance 0, C = Structural variance Co, and Ao = range parameter, the effective range A = Ao The second model which GS+ uses is the Expon ential isotropic model (Figure 2-7), which is similar to the Spherical model in that it appro aches the sill (Co + C) gr adually, but is different from the Spherical model in the rate at which the si ll is approached and in the fact that the model and the sill never actually converge (Gamma Design Software, LLC, 2004). GS+ uses Equation 2-2 for Exponential models. y(h) = Co + C [1 exp(-h/Ao)] (2-2) Where: y(h) = semivariance for interval distance class h, h = the lag distance interval, Co = nugget variance 0, C = Structural variance Co, and 29

PAGE 30

Ao = range parameter, in the case of the exponential model the effective range A = 3Ao, which is the distance at which the sill (C o + C) is within 5% of the asymptote. The third model GS+ uses is the Gaussian is otropic model or hyperbolic isotropic model (Figure 2-8), which is similar to the exponentia l model in that the model and the sill never actually converge but different in that the m odel assumes a gradual rise for the y-intercept (Gamma Design Software, LLC, 2004). GS+ us es Equation 2-3 for Gaussian models. y(h) = Co + C [1 exp(h2/Ao2)] (2-3) Where: y(h) = semivariance for interval distance class h, h = the lag distance interval, Co = nugget variance 0, C = Structural variance Co, and Ao = range parameter, in the case of the Gaussian model the effective range A = 30.5Ao, which is the distance at which the sill (C o + C) is within 5% of the asymptote. GS+ uses the statistical minimum Residual Sum of Squares (RSS) to select the model that best fits the data. The RSS provides an exact m easure of how well the m odel fits the variogram data. The lower the RSS the better the mode l fits. GS+ also gives the coefficient of determination (R2), which provides an indication of how well the model fits the variogram data. The R2 value is not as precise as the RSS; however it is still useful in evaluating the model fit. Also along with the RSS and R2, a proportion of C/(CO + C) is calculated. This statistic provides a measure of the proportion of sample va riance that is explained by spatially structured variance (Gamma Design Software, LLC, 2004). This value will be equal to1.0 for a variogram with no nugget variance (the curve passes through the origin); conve rsely it will be equal to 0 where there is no spatially depe ndent variation at the range sp ecified (Gamma Design Software, LLC, 2004). 30

PAGE 31

Kriging Interpolation After the semivariogram has been calculated the Kriging inter polation can then be applied to the grid data. Kriging provides a means of interpolating values for points not physically sampled using knowledge about the underlying spatial relationships in the data set (Gamma Design Software, LLC, 2004). The variograms provided this underlying knowledge. Kriging is based on regionalized variable theory and is supe rior to other means of interpolation because it provides an optimal interpolation estimate fo r a given coordinate location (Gamma Design Software, LLC, 2004). There are several types whic h include point and block kriging. For this analysis block kriging was used because the samp le points only measure a small area. According to Burrough and MacDonnell (2004), given the larg e amplitude, short-range variation of many natural phenomena like soil or water quality. Ordi nary point kriging would result in many sharp spikes or pits at the data points, which can be overcome by modifying th e kriging equations to estimate average values of z for experimental pl ots of a given area. Block kriging, which was used for the analysis of the moisture data, prov ides an estimate for a discrete area around an interpolation point (Gamma Design Software, LLC, 2004). Equation 2-4 shows the estimation of value z for block B, which is used in bloc k kriging, while equation 2-5 shows the minimum variance (Burrough & McDonnell, 2004). n i iixz Bz1)( )( (2-4) Where: z(B) is the variable (z, in this case soil VWC) over block B, i = weights needed for local interpolation, and z(xi) = variable z at known sampling point x at location i. ),(),()(1 2BB Bx Bn i ii (2-5) 31

PAGE 32

Where: 2(B) is the estimation of variance for the Block (B), i is the weight needed for local interpolation, (xi,B) is the estimated average semivariogram between sampled point i and the block, is the Langrangian multiplier, and (B,B) is the estimated average semivariogram within the block When these equations (2-4 and 2-5) are used, th e resulting smoothed interpolated surface is free from pits and spikes resulting from point kriging (Burrough & McDonnell, 2004). According to Schmitz and Sourell (2000), it is possible to show how the measurement uncertainty or variability is related to the required number of sensors. Through the use of a 95% confidence level the number of sensors can be de termined. Schmitz and Sourell (2000) used an acceptable 10% measurement error of the plant available water (PAW) to create the threshold value which determines the number of sensors or measurements that would be needed. The PAW for the Candler soil present in this experimental grove block is between field capacity, which is 0.085 m3/m3 volumetric water content (VWC), and pe rmanent wilting point (PWP), which is 0.02 m3/m3 VWC. Therefore, the range for the PAW is 0.065 m3/m3 VWC. At 10% of the PAW the threshold will be 0.0065. For added accuracy using a threshold of 5% PAW would set the threshold at 0.00325. A simple confidence interv al was calculated with the Excel (Microsoft Corporation, Redmond, Washi ngton) function CONFIDENCE (E quation 2-6). Using the terms alpha, which is equal to 0.05; the standard deviation of a population ( ) of moisture readings, which is the 225 sample points from one individual grid; size, wh ich is the number of sensors that would be required (N). Microsoft Excel uses the standard Confidence Level statistical equation (Equation2-7). CONFIDENCE(alpha,standard_devi ation,size) (2-6) 95% Confidence level = .96 ( /SQRT(N)) (2-7) 32

PAGE 33

Results and Discussion After analysis of the data using GS+, the so il moisture under citrus trees was found to be highly variable. The results from the isotropic semivariograms show a wide range of variability within the tree sizes as well as variation te mporally or across the moisture events. Semivariogram results from the dry, rainfall, and irrigation events (T able 2-1), show the Effective Range (A), in centimeters, which vari es from tree to tree. The average A for the young trees for the dry grids (after several days of drying) was 72 cm, while the average A for the mature trees was 50 cm. This difference in eff ective range could be due to the greater root density under the mature tree and the area of the grid covered by the trees canopy. The coefficient of variation (CV) for the dry events also shows a range of variation, with the low being 0.26 and a high of 0.56, showing a dispersion of the moisture within the grid and from grid to grid. After a moisture event such as rainfall of 6.25mm and 11.25mm, the coefficient A varied even more than it did with the dry events. The minimum range for both rainfall events was 16 cm from a mature tree grid, while the maximum A was 104 cm from a young tree grid (Table 2-1). The difference in A can be explained by canopy interception of rainfall causing water to temporarily pool in areas directly below low hangi ng branches and at the trunk of the tree. The spatial map from the kriging interpolation for this mature tree (Figure 2-9), shows the dark blue areas which represent the areas where water pooled, creating regions of greater moisture (up to 0.3 m3/m3 volumetric water content), while leaving other areas nearly dry (as low as 0.02 m3/m3 volumetric water content). Root de nsity could also play a role in increasing the variability due to the high active root area under a mature tree. In the case of the large A value from the young tree grid, this could be due to less coverage of the grid by the small tree canopy allowing a more even distribution of rain water to contact the so il. The spatial map from the kriging interpolation 33

PAGE 34

for this young tree (Figure 2-10), sh ows the larger areas of similar moisture concentrations, with only some minor pooling of rainwater due most lik ely to soil micro-topography. The CV range for the rainfall events was wide with a low of 0. 19 and a high of 0.55 (Table 2-2), and Standard deviation ( ) tended to be higher with the larger rainfall event (Table 2-2) averaging 0.04 m3/m3 for the 11.25 mm rainfall versus the 0.02 m3/m3 for the 6.25 mm rainfall event, which indicates that after the increased rainfall the scatter of the data was greater. Soil moisture after irrigation events also show ed a high range of variability from grid to grid. After a one hour irrigation application, A va ried from a low of 26 cm to a high of 100 cm, again with the low A being from a mature tree (Figure 2-11) and the high A from the young tree (Figure 2-12). The moisture variability in this situation can be explaine d by the type of nozzle used, soil micro-topography, root density and canopy size. Figure 2-11, illustrates the ponding locations where the irrigation water pattern was intercepted by low hanging branches, while Figure 2-12 illustrates the nozzle pattern due to incr eased water infiltration within the pattern of the irrigation spray. When taking into consideration the irrigation spray pattern on the ground (Figure 2-1), the spray pattern of the irrigation nozzl e (Figure 2-4), and the sandy nature of the soil, it stands to reason that a similar pattern would also be obt ained through soil moisture measurements. Several of the moisture grids yielded a similar pattern as a result (Figure 2-13). Effective ranges measured one hour after the three hour irrigation event va ried a great deal ranging from as low as 34 cm to as high as 116 cm in distance (Table 23). Even after 48 hours, the effective range of measured soil moisture is still varied, ranging from a low of 20 cm to a high of 182 cm in distance (Table 2-3). 34

PAGE 35

The variation in effective ranges can also be explained by the inter ception of irrigation water by low hanging branches, especially within the mature tree grids. In some cases the low hanging branches collected irrigation water and caused pooling of water below these branches, increasing the variability of the wetted patt ern. While the blank grids showed moisture concentrated in the nozzle wetting pattern (Figure 2-14), the mature trees showed the moisture in the nozzle wetted pattern as well as in spots wher e drip apparently was excessive from branches (Figure 2-15). The younger trees also show some br anch interception of ir rigation (Figure 2-16), but not to the extent that this interfer ence is seen from th e mature trees. The CV data from these grids also confirms th at there is excessive spatial variability in measured soil moisture. The minimum CV for the 3 hour irrigation data was 0.23 or 23% and the maximum was 0.48 or 48% (Table 2-4), whic h shows that not only is the soil moisture variable from grid to grid and event to ev ent, but also within each individual grid. After calculating the 95% confidence level fo r the moisture results from each grid, the number of sensors or measurements needed to produce a reliable estimate of the soil moisture status was found to be high. This calculation estimates the minimum number of sensors that would obtain an accurate (acceptable error set at 10% or 5% of the PAW) average soil moisture measurement to reliably trigger irrigation for ma intaining the correct soil moisture levels, 95% of the time. Using the confidence function in Excel, the 95% confidence level was calculated for each tree grid using in the calculation the standard deviation of a grid population and the number of sensors (N) and a probability (alpha) of 0. 05. The thresholds calculated using both 10 % and 5% of the PAW were used to identify the number of sensors that would be appropriate for the data set. Table 2-5 shows the calculation for one of the dates when the young and mature trees were sampled. The yellow highlight shows the point at which the threshold was reached for the 35

PAGE 36

10% PAW, while the orange highlight shows the point at which the threshold for the 5% PAW was reached. These thresholds establish the numbe r of sensors that would be required for an accurate estimate of the soil moisture. After all the calculations were completed, at the 10% PAW level, the number of sensors needed after rainfall ranged from a low of 20 measurements, young tree with an evenly distributed rainfall pattern, to a high of 289 measurements, matu re tree with a high amount of canopy interference (Table 2-6). After several da ys of drying the range for 10% PAW was from 34 to 167 measurements (Table 2-6). One hour fo llowing an irrigation event of 3 hours (Table 27) had a range of measurements that was rath er high, with a low of 120 to 168 measurements needed. However 48 hours after the same irrigation event the range of measurements needed was quite a bit lower ranging from 14 to 53 measurements. This sh ows that a lower number of sensors would be appropriate for triggering the start an irrigation cycl e but not for triggering when the appropriate amount of water has been applied, which would require a far greater number of measurements or sensor s. Thus, when using soil moisture sensors to trigger irrigation systems, the risk of over-irrigating due to spatially variable soil moisture distribution is higher than the risk of under-irrigating. Using 5% of the PAW would allow for a more accurate application of water, since the soil would not be outside of 5% of the PAW, but would require a substantially large number of measurements or sensors. After rainfall the number of measurements needed at 5% PAW would range from 80 to 1,168 measurements (Table 2-6), which is an unreasonably high number of measuremen ts. Likewise after irri gation the number of measurements needed was also high, with a low of 347 measurements to a high of 675 measurements (Table 2-7). However 48 hours afte r the same irrigation event, the number of 36

PAGE 37

sensors needed for the 5% PAW was much lowe r ranging from 54 to 287 measurements, still in many cases unfeasibly high for triggering irrigation. Soil moisture results indicate hi gh variability of soil moisture under citrus trees after all moisture events as well as after several days drying. These results show that the spatial variability is excessive and ma kes single point measurement, as is often done for irrigation scheduling, inaccurate. Knowing the best place fo r a single moisture sensor under a tree canopy for accurate irrigation scheduling is impossible, and would lead to situations of either over or under irrigating trees, which can lead to the prob lems of nutrient and pesticide leaching as well as possible yield reduction due to water stress, as in the case w ith under irrigation. Since the aim of the Ridge Citrus BMPs is to reduce the amount of nitrogen reaching the groundwater and eventually the aquifer, more efficient irrigation triggers are necessary, and based on these results, this accuracy will not come from single point soil moisture measurements. 37

PAGE 38

Table 2-1. Geostatistics from GS+ isotropic semi variogram analysis for moisture grid data including several days of drying, rain, and 1 hour of irrigation. Several Days of Drying Grid # Parameter Range (Ao) Effective Range (A) Nugget (Co) Sill (Co+C) R2 RSS C/(Co+C) Model Y1 40 119 2.6 5.9 0.850 0.73 0.563 Exponential Y2 57 57 11.9 23.8 0.883 7.61 0.500 Spherical Y3 39 39 1.5 4.4 0.598 1.37 0.662 Spherical M1 12 36 1.7 12.6 0.812 2.53 0.865 Exponential M2 12 21 0.8 10.3 0.905 0.90 0.919 Gaussian M3 93 93 1.4 4.5 0.998 0.01 0.695 Spherical Rainfall (6.25mm) Y1 10 18 0.1 2.1 0.633 0.09 0.947 Gaussian Y2 93 93 2.3 4.9 0.990 0.05 0.524 Spherical Y3 100 100 1.0 5.2 0.996 0.05 0.813 Spherical M1 14 24 1.1 7.4 0.834 1.21 0.856 Gaussian M2 9 28 0.3 5.0 0.725 0.42 0.944 Exponential M3 23 68 0.2 2.6 0.998 0.00 0.931 Exponential Rainfall (11.25mm) Y1 13 40 0.7 9.3 0.938 0.59 0.927 Exponential Y2 72 72 9.3 18.7 0.992 0.40 0.505 Spherical Y3 60 103 10.6 28.8 0.997 0.89 0.631 Gaussian M1 12 36 7.6 32.9 0.834 11.30 0.767 Exponential M2 9 16 1.0 18.1 0.579 4.51 0.945 Gaussian M3 62 62 4.2 11.4 0.979 0.54 0.634 Spherical After 1 Hour Irrigation Y1 46 46 1.5 11.5 0.983 0.57 0.873 Spherical Y2 67 67 3.1 15.1 0.990 0.77 0.780 Spherical Y3 33 100 4.2 12.5 0.990 0.26 0.666 Exponential M1 17 29 6.1 22.1 0.905 7.23 0.723 Gaussian M2 9 26 1.8 17.7 0.694 5.25 0.897 Exponential M3 24 73 1.6 14.1 0.997 0.12 0.890 Exponential B = Blank, Y = Young tree, M = Mature tree, R2 = Coefficient of determination, RSS = Residual Sum of Squares 38

PAGE 39

Table 2-2. Basic statistical results for moisture gr id data including several days of drying, rain, and 1 hour of irrigation. Several Days of Drying Grid # Mean Median Mode Min Max SEM CV Y1 0.09 0.09 0.09 0.03 0.28 0.02 0.0016 0.26 Y2 0.08 0.07 0.05 0.02 0.29 0.04 0.0029 0.56 Y3 0.06 0.05 0.05 0.02 0.20 0.02 0.0013 0.34 M1 0.07 0.06 0.06 0.03 0.24 0.03 0.0023 0.49 M2 0.06 0.05 0.06 0.02 0.24 0.03 0.0021 0.52 M3 0.06 0.06 0.07 0.02 0.13 0.02 0.0013 0.33 Rainfall (6.25mm) Y1 0.08 0.08 0.08 0.05 0.13 0.01 0.0010 0.19 Y2 0.07 0.07 0.08 0.02 0.13 0.02 0.0014 0.30 Y3 0.07 0.07 0.06 0.02 0.13 0.02 0.0014 0.31 M1 0.07 0.07 0.04 0.02 0.13 0.03 0.0017 0.39 M2 0.06 0.06 0.07 0.02 0.15 0.02 0.0015 0.37 M3 0.06 0.06 0.05 0.02 0.10 0.02 0.0010 0.27 Rainfall (11.25mm) Y1 0.11 0.11 0.11 0.05 0.23 0.03 0.0020 0.27 Y2 0.11 0.11 0.07 0.03 0.22 0.04 0.0028 0.37 Y3 0.10 0.09 0.06 0.03 0.27 0.05 0.0032 0.49 M1 0.10 0.08 0.06 0.03 0.32 0.06 0.0038 0.55 M2 0.09 0.08 0.06 0.03 0.31 0.04 0.0029 0.50 M3 0.10 0.11 0.11 0.03 0.20 0.03 0.0022 0.31 After 1 Hour Irrigation Y1 0.13 0.14 0.15 0.05 0.24 0.03 0.0023 0.27 Y2 0.10 0.09 0.06 0.04 0.20 0.04 0.0025 0.39 Y3 0.10 0.10 0.11 0.05 0.19 0.03 0.0022 0.32 M1 0.10 0.09 0.05 0.03 0.24 0.05 0.0030 0.46 M2 0.09 0.08 0.05 0.03 0.27 0.04 0.0028 0.47 M3 0.10 0.09 0.05 0.04 0.21 0.04 0.0024 0.37 B = Blank, Y = Young tr ee, M = Mature tree, = Standard deviation, SEM = Standard Error of the Mean, and CV = Coefficient of Variation 39

PAGE 40

Table 2-3. Geostatistics from GS+ isotropic semi variogram analysis for moisture grid data measurements taken 1 h, 4 h, 24 h, and 48 h after irrigation. 1 Hour after 3 Hour of Irrigation Grid # Parameter Range (Ao) Effective Range (A) Nugget (Co) Sill (Co+C) R2 RSS C/(Co+C) Model B1 34 34 0.7 10.5 0.935 1.16 0.938 Spherical B2 34 116 3.8 13.8 0.970 1.15 0.724 Exponential B3 35 35 0.5 10.2 0.844 3.30 0.948 Spherical Y1 34 34 0.0 14.2 0.980 0.88 0.999 Spherical Y2 71 71 2.8 19.3 0.982 2.78 0.856 Spherical Y3 73 73 6.2 20.6 0.992 0.91 0.699 Spherical M1 17 50 1.1 20.3 0.995 3.42 0.947 Exponential M2 12 20 1.6 15.6 0.848 2.68 0.901 Gaussian M3 95 95 3.5 15.7 0.987 1.31 0.777 Spherical 4 Hours after 3 hour Irrigation B1 16 48 0.6 5.9 0.983 0.09 0.892 Exponential B2 95 95 2.4 6.7 0.993 0.08 0.641 Spherical B3 22 66 0.6 4.5 0.909 0.41 0.865 Exponential Y1 30 30 0.2 7.4 0.915 0.63 0.974 Spherical Y2 28 83 0.0 8.6 0.987 0.34 0.999 Exponential Y3 86 86 5.1 16.8 0.997 0.31 0.694 Spherical M1 20 86 1.9 13.6 0.970 1.07 0.862 Exponential M2 14 41 0.8 8.1 0.976 0.18 0.900 Exponential M3 60 103 3.2 10.2 0.989 0.47 0.689 Gaussian 24 Hours after 3 hour Irrigation B1 39 39 0.2 2.6 0.985 0.02 0.931 Spherical B2 51 152 0.7 4.3 0.994 0.03 0.832 Exponential B3 22 67 0.3 2.5 0.934 0.09 0.882 Exponential Y1 13 23 0.4 4.2 0.932 0.13 0.921 Gaussian Y2 80 80 1.1 5.0 0.989 0.10 0.780 Spherical Y3 32 95 0.8 10.5 0.997 0.12 0.925 Exponential M1 22 38 3.7 8.6 0.917 0.95 0.569 Gaussian M2 11 32 0.5 5.9 0.874 0.30 0.910 Exponential M3 115 115 1.4 6.6 0.992 0.14 0.789 Spherical 48 Hours after 3 hour Irrigation B1 44 44 0.3 1.7 0.975 0.01 0.832 Spherical B2 61 182 0.7 3.7 0.987 0.04 0.806 Exponential B3 49 49 0.6 1.7 0.970 0.01 0.630 Spherical Y1 12 20 0.3 3.4 0.939 0.05 0.912 Gaussian Y2 73 73 1.0 3.6 0.967 0.14 0.731 Spherical Y3 41 122 2.0 9.2 0.992 0.16 0.782 Exponential M1 15 44 1.0 6.6 0.806 1.15 0.846 Exponential M2 12 36 0.6 3.1 0.963 0.03 0.811 Exponential M3 95 95 1.6 4.7 0.997 0.02 0.663 Spherical B = Blank, Y = Young tree, M = Mature tree, R2 = Coefficient of determination, RSS = Residual Sum of Squares 40

PAGE 41

Table 2-4. Basic statistical results for moisture gr id data including measurements taken 1 h, 4 h, 24 h, and 48 h after a 3 h our irrigation application 1 Hour After 3 Hour Irrigation Grid # Mean Median Mode Min Max SEM CV B1 0.07 0.06 0.04 0.01 0.18 0.03 0.0022 0.47 B2 0.07 0.06 0.03 0.02 0.20 0.03 0.0023 0.48 B3 0.06 0.06 0.05 0.01 0.18 0.03 0.0021 0.48 Y1 0.13 0.14 0.17 0.05 0.21 0.04 0.0025 0.28 Y2 0.10 0.08 0.07 0.03 0.19 0.04 0.0027 0.43 Y3 0.13 0.14 0.15 0.05 0.27 0.04 0.0029 0.32 M1 0.11 0.11 0.07 0.04 0.22 0.04 0.0029 0.38 M2 0.10 0.09 0.07 0.03 0.20 0.04 0.0026 0.38 M3 0.11 0.11 0.06 0.04 0.19 0.04 0.0024 0.35 4 Hours After 3 Hour Irrigation B1 0.05 0.05 0.03 0.01 0.18 0.02 0.0015 0.45 B2 0.05 0.05 0.03 0.01 0.14 0.02 0.0016 0.48 B3 0.05 0.05 0.05 0.01 0.11 0.02 0.0013 0.40 Y1 0.11 0.11 0.12 0.04 0.19 0.03 0.0018 0.25 Y2 0.07 0.07 0.05 0.02 0.14 0.03 0.0018 0.36 Y3 0.11 0.11 0.13 0.03 0.21 0.04 0.0026 0.35 M1 0.09 0.09 0.11 0.03 0.19 0.03 0.0023 0.39 M2 0.08 0.07 0.05 0.02 0.14 0.03 0.0019 0.36 M3 0.08 0.08 0.11 0.03 0.15 0.03 0.0019 0.34 24 Hours After 3 Hour Irrigation B1 0.04 0.04 0.03 0.01 0.09 0.02 0.0010 0.37 B2 0.04 0.04 0.03 0.01 0.09 0.02 0.0012 0.42 B3 0.04 0.04 0.04 0.01 0.09 0.01 0.0010 0.34 Y1 0.09 0.09 0.09 0.03 0.15 0.02 0.0014 0.36 Y2 0.07 0.06 0.06 0.02 0.11 0.02 0.0014 0.34 Y3 0.09 0.09 0.10 0.03 0.18 0.03 0.0020 0.32 M1 0.08 0.08 0.04 0.01 0.16 0.03 0.0019 0.24 M2 0.07 0.07 0.07 0.03 0.14 0.02 0.0016 0.31 M3 0.07 0.08 0.09 0.02 0.14 0.02 0.0016 0.33 48 Hours After 3 Hour Irrigation B1 0.04 0.04 0.03 0.01 0.08 0.01 0.0008 0.30 B2 0.04 0.04 0.04 0.01 0.09 0.02 0.0011 0.38 B3 0.04 0.04 0.04 0.01 0.08 0.01 0.0008 0.29 Y1 0.08 0.08 0.08 0.04 0.14 0.02 0.0012 0.23 Y2 0.06 0.06 0.05 0.02 0.12 0.02 0.0012 0.29 Y3 0.08 0.08 0.08 0.03 0.17 0.03 0.0019 0.33 M1 0.07 0.08 0.08 0.03 0.16 0.02 0.0016 0.33 M2 0.07 0.07 0.08 0.03 0.13 0.02 0.0012 0.26 M3 0.07 0.07 0.07 0.02 0.14 0.02 0.0014 0.30 B = Blank, Y = Young tr ee, M = Mature tree, = Standard deviation, SEM = Standard Error of the Mean, and CV = Coefficient of Variation 41

PAGE 42

Table 2-5. 95% confidence levels calculated fo r the number of sensors needed for precise soil water estimation N Y1 Y2 Y3 M1 M2 M3 1 0.04648 0.08448 0.03767 0.06783 0.06291 0.03920 5 0.02079 0.03778 0.01685 0.03033 0.02814 0.01753 10 0.01470 0.02671 0.01191 0.02145 0.01990 0.01240 20 0.01039 0.01889 0.00842 0.01517 0.01407 0.00877 34 0.00797 0.01449 0.00646 0.01163 0.01079 0.00672 36 0.00775 0.01408 0.00628 0.01130 0.01049 0.00653 50 0.00657 0.01195 0.00533 0.00959 0.00890 0.00554 51 0.00651 0.01183 0.00528 0.00950 0.00881 0.00549 60 0.00600 0.01091 0.00486 0.00876 0.00812 0.00506 80 0.00520 0.00945 0.00421 0.00758 0.00703 0.00438 90 0.00490 0.00890 0.00397 0.00715 0.00663 0.00413 93 0.00482 0.00876 0.00391 0.00703 0.00652 0.00406 100 0.00465 0.00845 0.00377 0.00678 0.00629 0.00392 108 0.00447 0.00813 0.00362 0.00653 0.00605 0.00377 120 0.00424 0.00771 0.00344 0.00619 0.00574 0.00358 130 0.00408 0.00741 0.00330 0.00595 0.00552 0.00344 134 0.00402 0.00730 0.00325 0.00586 0.00544 0.00339 140 0.00393 0.00714 0.00318 0.00573 0.00532 0.00331 146 0.00385 0.00699 0.00312 0.00561 0.00521 0.00324 160 0.00367 0.00668 0.00298 0.00536 0.00497 0.00310 167 0.00360 0.00654 0.00292 0.00525 0.00487 0.00303 180 0.00346 0.00630 0.00281 0.00506 0.00469 0.00292 200 0.00329 0.00597 0.00266 0.00480 0.00445 0.00277 204 0.00325 0.00591 0.00264 0.00475 0.00440 0.00274 360 0.00245 0.00445 0.00199 0.00357 0.00332 0.00207 374 0.00240 0.00437 0.00195 0.00351 0.00325 0.00203 380 0.00238 0.00433 0.00193 0.00348 0.00323 0.00201 400 0.00232 0.00422 0.00188 0.00339 0.00315 0.00196 420 0.00227 0.00412 0.00184 0.00331 0.00307 0.00191 435 0.00223 0.00405 0.00181 0.00325 0.00302 0.00188 500 0.00208 0.00378 0.00168 0.00303 0.00281 0.00175 600 0.00190 0.00345 0.00154 0.00277 0.00257 0.00160 670 0.00180 0.00326 0.00146 0.00262 0.00243 0.00151 674 0.00179 0.00325 0.00145 0.00261 0.00242 0.00151 N = Number of sensors, Y = Young tree, M = Mature tree 42

PAGE 43

Table 2-6. Number of sensors needed, for accu rate estimations of soil water, after 95% confidence levels were calculated for each tree grid. Several Days of Drying Grid 10% PAW 5% PAW Y1 51 204 Y2 167 674 Y3 34 134 M1 108 435 M2 93 374 M3 36 146 Rainfall 6.25 mm Y1 20 80 Y2 41 165 Y3 40 158 M1 62 249 M2 47 187 M3 22 88 Rainfall 11.25 mm Y1 83 335 Y2 157 635 Y3 210 847 M1 289 1168 M2 170 688 M3 95 383 After 1 Hour Irrigation Y1 109 441 Y2 124 499 Y3 100 403 M1 184 744 M2 153 619 M3 115 466 43

PAGE 44

Table 2-7. Number of sensors needed, for accu rate estimations of soil water, after 95% confidence levels were calculated for each tree grid. 1 Hour After 3 Hours of Irrigation Grid 10% PAW 5% PAW B1 93 376 B2 106 426 B3 86 347 Y1 130 524 Y2 154 597 Y3 167 675 M1 168 667 M2 139 561 M3 120 486 4 Hours After 3 Hours of Irrigation B1 50 202 B2 55 221 B3 36 142 Y1 65 261 Y2 64 257 Y3 135 546 M1 109 441 M2 71 285 M3 73 293 24 Hours After 3 Hours of Irrigation B1 22 87 B2 30 119 B3 20 79 Y1 38 153 Y2 38 151 Y3 79 318 M1 72 289 M2 52 209 M3 50 197 48 Hours After 3 Hours of Irrigation B1 14 54 B2 24 97 B3 14 57 Y1 31 124 Y2 29 116 Y3 71 284 M1 53 215 M2 28 110 M3 39 157 44

PAGE 45

Figure 2-1. Imposed irrigation we tting pattern from nozzles used in experimental block Figure 2-2. Example of a tree grid placed under the canopy of a young tree 45

PAGE 46

Figure 2-3. Field Scout TDR100 Soil Moisture Meter used for soil moisture measurements Figure 2-4. Irrigation spray from the micro-je t nozzle used in the experimental block 46

PAGE 47

Figure 2-5. Calibration showing linear regr ession used for conversion from period ( s) to volumetric water content (m3/m3) Figure 2-6. Example of the spherical model for isotropi c semivariograms (h = cm) 47

PAGE 48

Figure 2-7. Example of the exponential model for isotropic semivariograms (h = cm) Figure 2-8. Example of the Gaussian model for isotropic semivariograms (h = cm) 48

PAGE 49

Figure 2-9. Kriging interpolation for a mature tree after rainfall Figure 2-10. Kriging interpolati on for a young tree after rainfall 49

PAGE 50

Figure 2-11. Kriging interpolation of a mature tree after 1 hour application of irrigation (circles denote approximate locations of low hanging branches) Figure 2-12. Kriging interpola tion from a young tree grid follow ing a 1 hour application of irrigation including the overlyi ng pattern of water spray fr om irrigation nozzle (lines indicate the irrigati on wetting pattern) 50

PAGE 51

A. B. C. Figure 2-13. Series of Kriging in terpolated spatial maps showi ng irrigation nozzle pattern. A) Moisture results from a blank grid, B) mo isture results from a young tree grid, and C) moisture results from a mature tree grid. ( lines indicate the irrigation wetting pattern) Figure 2-14. Kriging interpolation results from moisture measurements taken from the blank grids 1 hour, 4 hours, 24 hours, and 48 hours after a 3 hour irrigation application. 51

PAGE 52

Figure 2-15. Kriging interpolation results from moisture measurements taken from the mature tree grids 1 hour, 4 hours, 24 hours, and 48 hours after a 3 hour irriga tion application. Figure 2-16. Kriging interpolation results from moisture measurements taken from the young tree grids 1 hour, 4 hours, 24 hours, and 48 hours after a 3 hour irri gation application. 52

PAGE 53

CHAPTER 3 IDENTIFYING AND TESTING ALTERNATIVE METHODS OF DETERMINING WATER STATUS FOR IRRIGATION USIN G TREE CANOPY MEASUREMENTS Introduction Irrigation scheduling typically strives to achieve an optimum water supply for maintaining crop productivity with the ultimate aim of soil water content being maintained close to field capacity (Jones, 2004). Even today, many farmers still irrigate by the cal endar, applying water whether the crop needs it or not (Jackson, 1982). The farmers that do not irrigated based on the calendar lean towards the use of soil based meas urements for irrigation scheduling. However, due to the increasing de mand of water for general purposes, the supply of water available for irrigation is decreasing and irrigation costs are rising (Gonzlez-Dugo et al., 2006). The shortcomings of soil sensors in assessing the wate r requirements of whole trees were discussed in Chapter 2 of this thesis. As a result, irriga tion managers must come up with a method of irrigating that is based on plant specific needs in order to reduce the excess use of water while ensuring the tree has enough to remain productive. The classical methods of monitoring crop wate r stress include measurements of soil water content, plant properties, or mete orological variables to estimate the amount of water lost from the plant-soil system during a given period of time (Gonzlez-Dugo et al., 2006). These methods have been found to be time consuming and in the case of soil moisture methods, too variable for precise irrigation applications with the goal of minimizing water usage. Several methods for irrigation scheduling ar e being explored in many different crops, however many of these new methods have not been tested on citrus trees. These plant-based methods include canopy temperature measurements and multispectral imagery. 53

PAGE 54

Canopy Temperature Measurements Plant temperature has long been recognized as an indicator of water av ailability (Jackson et al., 1981). Until infrared thermometers became ava ilable, most plant temperature measurements were made with contact sensors on or imbedded in leaves (Jackson et al., 1981). As plants transpire the temperature of the canopy is lowere d by evaporative cooling, effectively making the canopy temperature cooler than the air temperatur e. However, when a plant is water stressed, transpiration becomes limited resulting in an increase in temperature that can reach and go beyond that of the air. The usefulness of canopy temperature as a measure of crop water stress was recognized in the 1960s (Ml ler et al., 2007). Jackson et al (1981) derived the use of canopy temperature minus air temperature (Tc Ta), from the energy balance for a crop canopy (Equation 3-1), where Rn is the net radiation (W/m2), G is the heat flux below the canopy (W/m2), H is the sensible heat flux (W/m2) from the canopy to the air, E is the latent heat flux to the air (W/m2), and is the heat of vaporization. Rn = G + H + E (3-1) Hope and Jackson (1989) used Tc Ta, of a wheat crop, as an index of crop water status. The difference (Tc Ta or T) when summed over a period of time is also called the stress-degreeday. For several crops the relationship between Tc Ta and VPD, for well watered crops under clear sky conditions is linear (Jackson et al., 1981). Jackson et al. (1988), identified the use of upper and lower limits for calculating the crop wa ter stress index (CWSI) (Equation 3-2), and described the purpose of the upper ( TUL) and lower ( TLL) limits was to form bounds by which the measured temperature can be normalized. CWSI = (( T TLL)/( TUL TLL)) (3-2) According to the U.S. Water Conservation Laboratory (2001), the TLL can be calculated using the slope and intercept from a linear regression of a baseline created using measurements of Tc 54

PAGE 55

Ta from non-stressed plants at varying VPD (E quation3-3) (US Water Conservation Laboratory, 2001). TLL = intercept + slope(VPD) (3-3) The upper limit represents a fictitious temperatur e difference that would occur if the canopy were instantly desiccated (Jac kson et al., 1988). The upper limit is calculated using the slope and intercept from the regression of the baseline, a nd includes the saturation va por pressure (VPs) for a temperature limit that is chosen (Equation 3-4) (US Water Conser vation Laboratory, 2001). TUL = intercept + slope(VPs{Ta} VPs{Ta + intercept}) (3-4) According to Monteith and Unsworth (2008), th ere is a dependence of transpiration rate on radiation and saturation deficit. When leaves are in their natural environment, stomatal aperture depends strongly on solar radiation; in the abse nce of light, stomata are usually closed, making transpiration effectively zero (M onteith and Unsworth, 2008). There is also substantial evidence both from the field and from work in controlled environments which reveals that many plants close their stomata as saturation deficit or VPD increases, which is presumably a mechanism for conserving water (Monteith and Unsworth, 2008). Calculating the VPD of the air is done using the air temperature ( C) and relative humidity (RH). The Saturated vapor pressure (es) is calculated using the air te mperature (Ta) (Equation 35) (NOAA, Southern Region Headquarters, 2008). Since the RH can be calculated using es and the actual vapor pressure (e) (Equation 3-6a), this eq uation can be modified to calculate the actual (e) using RH (Equation 3-6b) (NOAA, Southern Region Headquarters, 2008). Finally since the VPD is a difference or deficit between the es and e, it is calculated by subtraction (Equation 3-7). es = 6.11 x 10^((7.5 x Ta)/(237.7 + Ta)) (3-5) RH = (e / es) x 100 (3-6a) 55

PAGE 56

e = .01 x RH x es (3-6b) VPD = es e (3-7) Multispectral Imagery Electromagnetic radiation that is reflected or emitted from the Earths surface can be recorded by a sensor from the ground, aircraft, or sa tellite (Curran, 1983). Today an increase in knowledge of the way in which electromagnetic ra diation interacts with our environment, has enabled scientists to use such remotely sensed da ta to determine the amount of soil moisture in a field or the amount of suspended sediment in estuarine waters (Curran, 1983). Some of the solar irradiance that is impi nging upon a vegetation canopy is reflected, while the rest is either transmitted and/or is abso rbed (Curran, 1983). The intensity with which radiation is reflected at any pa rticular wavelength is dependent on both the spectral properties and also the area of the leaves substrate, and shadow (Cu rran, 1983). According to Curran (1983), leaves usually reflect weakly in the bl ue and red wavelengths due to the absorption by photosynthetic pigments, and likewise they refl ect strongly in the near-infrared (NIR) wavelengths due to cellular refraction. The most widely used green vegetation indices are formed with data from discrete red and NIR bands (Elvidge and Chen, 1995). A ratio called the normalized difference vegetation in dex (NDVI) or Vegetation index (Iv) (Equation 3-8) is one of the more popular (Curran, 1983). Iv = (Rir Rr)/(Rir + Rr) (3-8) Where Rir is NIR reflectance and Rr is red reflectance Spectrometers with measurement ranges beyond 1000 nm have been used to determine water stress in plants by analyz ing reflectance measurements at several key wavelengths called water bands (Dallon and Bugbee, 2003). The mo st prominent water bands are at 1400 and 1900 nm and reflectance at these wavelengths has been shown to correspond to water content in plant 56

PAGE 57

tissue (Dallon and Bugbee, 2003). Unfortunately natural sunlight reaching the surface of the earth has low intensities at these wavelengths due to absorptive filtering by water in the atmosphere. Spectrometers capable of m easuring radiation beyond 1000 nm are also considerably more expensive, such as the SVC HR-1024 (Spectra Vista Corp., Poughkeepsie, New York) which costs about $70,000, than those measuring in the visible and short wavelength NIR ranges (i.e. 400 1000 nm). At the 970 nm wavelength there is another water band, however it has historically been considered too small to accurately measure water stress. Dallon and Bugbee (2003) found that if using an accurate spectrometer that can measure wavelengths up to 1000 nm, accurate estimates of water stress can be measured at the 970 nm water band. In order to test the use of the 970 waveband Da llon and Bugbee (2003) used three indices to analyze the various water bands. The first of the indices used, the reflectance water index, is a ratio between the reflectance at a water band to a nearby reference wavele ngth that is unaffected by water content variability. The second of the indices used is the band depth analysis, which uses a process called continuum removal where a linear continuum line is approximated across an area of absorption, connecting two unaffected points of the spectrum; and the third of the indices is the first-order derivative green vegetation index (1DGVI), which is based on a complex calculus formula involving integrated de rivatives that reduces down to a simple difference between a wavelength within the water ba nd that is subtracted from a reference point wavelength. Applying a similar index to images taken with a multispectral camera, Schumann et al. (2007) found that by photographing a canopy at specific wavelengths with a multispectral camera can result in a yield index and a canopy st ress index. Using a multispectral camera fitted with a filter wheel, Schumann et al. (2007) used grayscale values for each pixel from citrus 57

PAGE 58

canopy images taken at 840 nm and 670 nm and applie d these pixel values to a ratio of 840 nm / 970 nm. When applied to tree canopy measurements, both thermal and multispectral methods have a possibility of detecting plant water stress. Wh ile both the thermal and mu ltispectral procedures have been tested out on several different crops, su ch as wheat, soy and turf, they have not been tested thoroughly for citrus with the purpose of being used for irrigation scheduling. Hypothesis Canopy measurements taken with an infrared radiometer, multispectral camera, or GreenSeeker can be used to estimate the current wa ter status of a citrus tree with the possibility of being used for irrigation triggering. Objectives To test several alternative methods of measuring tree water stress based on canopy measurements in order to determine if these me thods could lead to optimal irrigation scheduling which would effectively supply water to the trees and minimize the leaching of nutrients due to over irrigation. Materials and Methods Several experiments were carried out to thoroug hly test the alternativ e methods discussed. These methods include thermal infrared canopy m easurements and multispectral images of both greenhouse and grove trees. Additi onal test of the commerci ally available GreenSeeker instrument which measures the ND VI of a tree canopy was conducted. Thermal Infrared Canopy Measurements Three different experiments were carried out to test the infrared canopy measurements. All three experiments used the precision narrow infrared radiometer (IRR-PN) from Apogee Instruments, Inc. (Roseville, California) (Figure 3-1). The IRR-PN has an 18 degree half angle 58

PAGE 59

and can respond to temperature changes in less th an one second. For the temperature range of 10 to 65 C the sensor has a .2 C absolute accuracy, .1 C uniformity, and .05 C repeatability and uniformity. The field of view fo r this sensor depends on the angle and distance from the sensor to the canopy. Th e target temperature reading that is displayed by the sensor is an average of temperatures sensed in the field of view, meaning that all of the leaves in the field of view will be measured, which includes sunlit and shaded leaves. For all of the experiments testing this sensor, only full-sun portions of th e tree canopies were used for measurements. The first of three experiments was done w ithin a greenhouse for controlled environment settings, in order to test the effects of wind on the sensor output. This experiment used 20 nongrafted, Rough Lemon citrus trees (Citrus jambhiri) (Figure 3-2). The trees were transplanted from the potting soil and into Candler Fine Sand, 0 to 5 Percent slopes soil taken from a 0-15 cm depth in a citrus grove at the Citrus Research and Education Center, Lake Alfred, Florida. This soil series was used to represent the soil that would typically be found in a Florida Ridge citrus grove. Of the 20 trees, eight were planted with AquaPro direct burial cap acitance soil moisture probes (AquaPro Sensors, Ducor, California) (Fi gure 3-2). The voltage outputs from the sensors were logged using an XR5 data logger (Pace Sc ientific Inc., Mooresville, North Carolina). Infrared canopy temperature measurements were taken using the IRR-PN which was logged using a CR10X data logger (Campbell Scientific Inc., Logan, Utah). The IRR-PN sensor was placed 30 cm away from, and perpendicular to, a tr ee canopy, which makes the field of view for the sensor approximately 300 cm2. The sensor simultaneously measured the air temperature with a built-in transducer which was also logged to the CR10X data logger. Along with infrared measurements, greenhouse climate conditions were also measured. These measurements were logged on an XR5 data logger, and included gr eenhouse air temperature and relative humidity, 59

PAGE 60

using a TRH-160 temperature/RH probe (Pace Scient ific, Inc., Mooresville, North Carolina) and short wave solar radiation using a PYR-PA2. 5 high output pyranometer (Apogee Instruments, Inc., Roseville, California). Measurements with the IRR-PN of the 20 tree canopies were taken along with short wave solar radiation, air temperature and relative humidity while under varying soil moisture conditions, ranging from very dry (0.03 to 0.04 m3/m3) to excessively moist (0.20 m3/m3) and higher. These measurements were taken both with and without a desktop fan running to test the effect of wind. Wind speed measurements were taken using a Kestrel 3000 Pocket Weather Station (Forestry Suppliers In c, Jackson, Mississippi). Using the T method, the air temperatures were s ubtracted from the canopy temperatures and graphed with soil moisture, stem water potentia l, and VPD in order to test if the IRR-PN could yield measurements that are related to plant water status. Stem water potentials were measured using a Scholander-type pressure chamber (PMS instrument, Corvallis, Oregon) (Scholander et al., 1965). According to McCutchan and Shackel (1992), covering a leaf with a refl ective bag stops the transpiration of the leaf and eliminates any gradients of water potential within the leaf. When bagged leaves remain attached to the tree for at least 1 hour, the water potential of the leaf would be expected to equilibrate with water potential of the stem and, therefore, be a meas urement of the stem water potential (McCutchan and Shackel, 1992). Bags were placed on two leav es per tree in the morning, covered to the petiole attachment point on the stem, and measur ed for stem water pote ntial in the afternoon, allowing enough time for the water potential in the l eaf to equilibrate with the water potential of the stem, and therefore yielding the stem water potential (SWP). 60

PAGE 61

The second experiment that was carried out with the thermal infrared radiometer was used to create the crop water stress i ndex, as well as to see if cont inuous measurements of Tc Ta could result in better irrigation triggering. Two of the grove soil potted trees were used. Each tree had an IRR-PN sensor, which was connected to a CR10X (Logan, Utah) data logger, aimed at the edge of the canopy from a 30 cm distan ce and at a 30 degree angle (Figure 3-4). This angle provides an average of the leaf temperatur e over the entire field of view for the sensor which is approximately 500 cm2 (at 30 degrees and 30 cm distance). Measurements were taken every minute for several days to monitor Tc Ta values over time and as canopy stress increased. Simultaneously measurements of gree nhouse air temperature, relative humidity, short wave radiation and volumetric so il water content were also m easured every minute. The canopy temperature measurements were compared with th e soil moisture status and VPD to test whether the sensor could be used in a stationary setting fo r tree water status monito ring, with the ultimate result being whether or not it coul d adequately trigger irrigation. Research by Jackson (1982) s hows that calculations for the crop water stress index can be calculated using the stress-degree-day calculations (Tc Ta). According to the U.S. Water Conservation Laboratory (2001), the upper ( TUL) and lower ( TLL) limits for T can be calculated using the vapor pressure deficits and Tc Ta. In Or der to calculate the CWSI the upper limits and lower limits must be known. Calculating the TUL requires knowing the value of TLL. The baseline or TLL can be found by taking measurements from non stressed trees and graphing the Tc Ta against the VPD. Severa l days of non stressed measurements were used from the two trees that were placed in front of th e IRR-PN sensor in order to find the baseline for the TLL. These measurements from each day were then averaged into hourly periods and plotted to create the baseline linear regression (Figure 3-5). The slope and intercept from this linear 61

PAGE 62

regression were then used to calculate the upper limit (Equation 3-4) and the lower limit (Equation 3-3), which were subsequently used to calculate the CSWI (Equation 3-2). Once the CWSI values were calculated they were averag ed and graphed by day along with the irrigation applications. The average for the CWSI uses only the measurements taken between the daylight hours of 10:00 to 15:00 with short wave radi ation intensity of greater than 300 W/m2. This time frame was chosen based on the observations by Idso et al. (1981), which show that for a nonstressed plant, during the daylight hours, the VPD will lie on or near the baseline. Using the results from these data, a vast majority of the measurements for the non-stressed trees fell on or near the baseline during this five hour time frame. The last experiment with the thermal infrar ed radiometer used field conditions and 20 trees, 12 of which were mature trees and the re maining eight were young re set trees. A 0.75 ha block of Hamlin orange trees grafted onto Swingle rootstock (block 8A), on the Citrus Research and Education Center (UF, IFAS) campu s in Lake Alfred, Florida was used for this experiment. According to the Southwest Florid a Water Management Di stricts GIS website (SWFWMD, 2002), the soil found in this block is Ca ndler Fine Sand, 0 to 5 Percent slopes. The Candler series, which belongs to the Entisol so il order, is typically found on the Florida Lake Wales Ridge and consists of excessively drained soils that formed in sandy marine or aeolian deposits (USDA, NRCS, 1990). Tyvek HomeWrap (DuPont, Wilmington, De laware) plastic fabric was placed under the canopy of six of the mature trees and four of the young trees in order to keep rainfall from penetrating the soil where the majority of the tr ee roots are located; these trees also had the irrigation nozzle blocked to avoid wetting the soil from irrigation. Of the remaining un-covered trees, three of the large trees a nd two of the small trees had thei r irrigation nozzles blocked to 62

PAGE 63

avoid irrigation, and the remaining trees were controls, receiving both irrigation and rainfall. Once visible wilting was observed on the covered soil trees, measurements were taken with the IRR-PN approximately 30 cm away from the tree canopy and held parallel to the surface of the ground at a height of approximate ly 140cm. This distance and angle yields an average canopy temperature over the field of vi ew which is approximately 300 cm2. Directly after the measurements for canopy and air temperatures were recorded, the SWPs for the trees were measured. Measurements were taken under both clear and cloudy sky conditions and while the wind was at a minimum. For SWP measurements in this experiment, four leaves on each tree were covered to the petiole attachment point on the stem with plastic bags that were covered with aluminum foil. Bags were placed on the leaves in the morning and measured for stem water potential in the afternoon, allowing enough time for th e water potential in the leaf to equilibrate with the water potential of the stem, and theref ore yielding the stem water potential (SWP). Multispectral Camera Imaging Two different experiments were carried ou t to test the use of multispectral canopy measurements in identifying tree water stress. Both experiments used the Eye Complementary metal-oxide-semiconductor (CMOS) Monochrome Camera from Imaging Development Systems GmbH (Oberslum, Germany). Attached to the ca mera is a motorized six filter wheel (Thorlabs, Inc., Newton, New Jersey), which holds six different wavelength filters for 450, 550, 670, 710, 840, and 970 nm. The procedure for the camera operation and analysis in both experiments follow a similar procedure that was used during an experime nt summarized by Schumann et al. (2007). The camera and filter wheel were controlled by so ftware developed by Schumann et al. (2007), running a laptop computer using the USB-2 and RS-232 ports and mounted to a golf cart for portable use in the field. Since th e camera is monochrome, the imag es are in grayscale with pixel 63

PAGE 64

values ranging from 0 to 255, representing the refl ectance (R) of that ligh t wavelength from the tree canopy, and a frame size of 480 by 480 pixels. The software was written to analyze each pixel grayscale value and calcula te a ratio of the pixel from the 840 nm wavelength image with the corresponding pixel from the 670 nm wavelength image. These values were used because according to Schumann et al. (2007), during a wintertime drought experiment using mature Valencia orange trees on Swingle rootstock, measurements with the multispectral camera that were compared with SWP measurements, yielded a regression analysis of the data that found the pixel grayscale ratio of the reflectance (R) at the 840 nm and 670 nm wavelengths yielded the best canopy stress inde x (CSI) (Equation 3-9). CSI = R(840 nm)/R(670 nm) (3-9) Since Dallon and Bugbee (2003), found that the 970 nm wavelength was a good estimator of water stress, additional analys is in the laboratory was done using a NIR-128 L near infrared spectrometer (Control Development, Inc., South Bend, Indiana). The absorption spectrum was measured using two leaves from each of the 20 grove soil potted Rough Lemon trees (Citrus jambhiri), which were subjected to varying levels of water stress. First, the absorbance at the 670 nm, 840 nm, and 970 nm wavelengths were id entified and calculated into CSI ratios of 970/670 nm wavelengths (Equation 3-3) and 840/670 nm wavelengths (Equation 3-4). This was done in an effort to discover if there would be disadvantage of using th e 840 nm wavelength, as used in the CSI ratio from Schumann et al. (2007), as apposed to the 970 wavelength, which was used in Dallon and Bugbee (2003). Using the Wa ter Index procedure outlined in Dallon and Bugbee (2003), the ratio of the wavelengths 970 nm and 670 nm was plotted against the ratio of the wavelengths 840 nm and 670 nm. Second, the absorption spectrum of a stressed leaf was compared to that of a non-stressed leaf, which was analyzed by subtraction in an effort to 64

PAGE 65

identify the points along the spectrum of the tw o leaves where wavelengths varied. Finally, the water stress index ratios of wavelengths 840 nm / 670 nm and 970 nm / 670 nm were plotted against stem water potential measurements, in an effo rt to determine if there is a disadvantage to using the 840 nm wavelength over the 970 nm wavelength. The first of the two experiments using the multispectral camera used 20 grove soil potted Rough Lemon (Citrus jambhiri) trees. The computer controlled multispectral camera captured images at each of the selected wavelengths The images were then processed using the wavelengths from the CSI (Equation 3-9). Afte r each tree was imaged, SWP measurements were taken. For this experiment two leaves per tree we re covered to the petiole connection point at the stem with plastic lined foil c overed bags at least two hours be fore SWP measurements were taken. The CSI results from this experiment we re plotted against the measurements of SWP for each tree in order to test whether CSI calcu lated from multispectral camera images can adequately determine the water status of a tree for potential use for triggering irrigation. The second experiment used a 0.75 ha block of Hamlin orange trees grafted onto Swingle rootstock (block 8A), on the same Candler soil described previously. Severa l trees were selected in this block, including 12 mature trees and 8 young reset trees. In orde r to induce water stress, Tyvek HomeWrap (DuPont, Wilmington, Delawa re) was used to cove r the soil under the canopies of several trees as desc ribed previously (Figure 3-6). Measurements were taken with the computer controlled multispectral camera only after visible water stress occurred in the trees with Tyvek covers. During measurements the camera was aimed at a sunlit area of the canopy while the cart was centered in the row middle. The entire field of view for the camera was enti rely covered by tree canopy in both the young and mature trees. Immediately after the camera was finished taking the images, SWP was measured. 65

PAGE 66

For this experiment, four leaves on each tree were covered to the petiole attachment point on the stem with plastic lined, foil covered bags. The le aves were covered in th e morning and measured for stem water potential in the af ternoon, allowing enough time for the water potential in the leaf to equilibrate with the water potential of the st em, and therefore yielding the stem water potential (SWP). In order to test whether the CSI ratio could be used to tr igger irrigation the CSI measurements were plotted against the measured SWP from each tree. GreenSeeker NDVI In addition to measurements taken with the multispectral camera, measurements were also taken using the GreenSeeker RT100 (NTech Industri es, Inc., Ukiah, California). Since at any given moment, the amount of solar radiation rece ived at a location on the Earth's surface depends on the state of the atmosphere a nd the location's latitude using an instrument that provides it own light source could be an advantage. The GreenSeeker, unlike the multispectral camera, provides its own light source and is therefore not affected by vary ing levels of sky irradiance. The output measurement is also different from the multispectral camera in that instead of a water stress index, like the CSI, it uses an NDVI (Equati on 3-8), where Rir (NIR) uses the 774 nm wavelength and Rr (Red) uses the 656 nm wavelength. Fo llowing similar procedures as in the field experiments using the IRR-PN and multispectral camera the GreenSeeker was used to test the use of the NDVI as an indicator of plant water stress. The NDVI va lues calculated were plotted against the average SWP of four leaves per tree. Results and Discussion After collecting and analyzing the data the resu lts show that both the infrared radiometer (IRR-PN) and the multispectral camera estimated water stress in citrus trees, but only under specific circumstances. Both sensors worked better under field conditions than in greenhouse conditions. This is probably due to some issues that occurred with the potted citrus trees and the 66

PAGE 67

artificial greenhouse environment. Since the same soil that is found in the field portion of this research was used in the potted trees, moisture control problems became evident. While some trees had adequate drainage, ot hers suffered from poor drainage of the pots which resulted in stress from lack of oxygen to the roots. Field of view for the sensors was another issue that was encountered during the greenhouse experiments. Due to the small size of the canopy for the potted trees there was quite a b it of background noise signal which was introduced into both the infrared readings as well as the photographs taken by the multispectral camera for image processing. This noise resulted in measuremen t error for the Tc Ta calculation and the CSI because of the temperature and reflecta nce of background objects, respectively. Infrared Canopy Measurements The results from the first IRR-PN experiment, using 20 greenhouse trees, do not conclude that it could be possible to make an accurate measure of the trees water stress using the Tc Ta. When the Tc Ta values are pl otted against the stem water potential (SWP) readings the results do not show a correlation (Figure 3-7), wh ich could be as a result of experimental error. Reduced transpiration, which would result in increased canopy temperature, could be caused by other stresses on the tree canopies. During the experiment there were spider mite infestations, which damaged the leaves, as well as root damage as a result of water logging in the pots. Another possible so urce of error in the experiment could be from the small canopy size which causes background temperature interf erence. Since the IRR-PN sensor takes a temperature reading for its entire field of view it is possible th e objects in the background not obscured by the tree canopy could cause incorrect temperature readings. Further experimental error could have been introduced with problems encountered while taking stem water potential (SWP) readings. Due to the age of the pressure chamber apparatus, there were several leaks 67

PAGE 68

along the gas lines making it difficult to obtain an accurate measure for the pressures status once the leaf pressure was reached. When Tc Ta was plotted against the volumet ric water content (VWC) of the soil in the pots, there was no noticeable co rrelation (Figure 3-8). Experime ntal error could also have affected these results, with the water saturation occurring near the bottom of the pots. Due to the length of the probe rods (20 cm) and the depth of the soil in the pots (22 cm) it is possible that the soil moisture readings were inaccurate due to the high water status at the bottom of the pot, which at times was near saturation. Results from the second greenhouse IRR-PN e xperiment conclude that Tc Ta can be tracked over a period of time and used to crea te a threshold for the crop water stress index (CWSI). Long term measurements were ta ken on two of the greenhouse trees. These measurements included greenhouse air temperatur e, IRR-PN sensor body temperature, IRR-PN target temperature (canopy temperature), greenhous e relative humidity, short wave radiation, and VWC of the soil in the pot of the tree being m easured. When these results are graphed Tc Ta can be seen for the different times of the da y. During daylight hours and while the tree is transpiring the canopy temperature is usually less than the air temperature, however during the night hours, the canopy temperatur e is higher then the green hous e air temperature and the IRRPN sensor body temperature, because there is no photosynthesis and therefore no transpiration, as well as the retention of heat by the plant leaves (Figure 3-9). The VPD and Tc Ta data from the long term measurements were used to create the baseline (Figure 3-5) for the CWSI calculations as described in the Materials and Methods section. The slope and the intercept from this ba seline were used to calculate the upper and lower limits which are then used to calculate the CWSI (Equations 3-3, 3-4, and 3-2, respectively). 68

PAGE 69

Once the CWSI was applied to the measurem ents from the long term study and graphed alongside the irrigation app lications the trend is apparent (Fig ure 3-10, A and B). While there are some decreases in the CWSI between the irrigati on applications, which are due to environmental conditions such as lower average daily short wave radiation intensity, the general trend is still increasing. Based on these results it is possible that a threshold could be created and used to trigger irrigation. For this experiment water, as irrigation, was applied based on soil moisture status and visual symptoms of water stress. With further research it could be possible to establish a threshold using the CWSI to trigger irrigation, and to test whether this would be more accurate than soil based sensors. The data from this second greenhouse experime nt was also used to compare the Tc Ta readings taken from the sunny (SWR > 500 W/m2) and the cloudy conditions (SWR < 300 W/m2). The data points used in this analysis came from a non-stressed da y during the long term measurements, and were from the same peri od of time (10:00 15:00) used in the CWSI analysis. The data from the sunny conditions clearly shows a trend similar to that seen in the non-stressed trees, which was used to create the CWSI. The data from the cloudy conditions illustrates what would be expected, since clouds reflect, absorb and transmit the incoming solar radiation, modifying in this way the amount and spectral quality of the solar radiation reaching the Earths surface (Alados et al., 1999). The cl oud particles are responsible for scattering processes that affect more markedly the shorter wavelengths in the solar spectrum, which include the photosynthetically active radiation spectral range (Alados et al., 1999). Since the clouds reduce the amount of photosynthetically active radiation th at reaches the plants, the photosynthesis would be reduced, whic h in turn, would decrease the amount of transpiration, and therefore increase the canopy temperature, which is reflected by the overall increase in Tc Ta 69

PAGE 70

(Figure 3-11). This clearly shows that conditions must be sunny during measurements taken with the IRR-PN, since transpiration is reduced under cloudy settings. Taking measurements during periods of low light intensity could give an artificial reading of water stress due to the decrease in transpiration. Results from the third experiment using the IRR-PN were varied. The large tree data fit to a regression of Tc Ta versus SWP with a higher R2 than the small trees (R2 = 0.40* and R2 = 0.28NS, respectively), however when placed in the same regression together there was a much poorer fit (R2 = 0.19NS) (Figure 3-12). Measurements were taken on clear sunny days with wind (average wind speed during measurements 1.65 m/s, max wind speed 4.02 m/s) and without wind. After a regression of Tc Ta to SWP, the results from the measurements taken without wind have a better fit (R2 = 0.40*) than the results from measurements taken with wind (R2 = 0.18NS) (Figure 3-13). These results agree with the pr evious research for th e Tc Ta or stressdegree-day measurements done on other crops (J ackson et al., 1988). Even after normalizing the data with the CWSI calculation, which was done using the upper and lower limits from the previous experiment, the results still conclude that the IRR-PN doe s not yield reliable measurements when used under windy conditions (Figure 3-14). All of the results from the IRR-PN experiments show that it would be difficult to obtain accurate results from the stress-degree-day or Tc Ta measurements in the field. The results from the CWSI from the greenhouse however, mi ght yield a reliable system for irrigation triggering of greenhouse trees with more research needed to find the appropriate threshold for the CWSI. Multispectral Camera Imaging In order to choose the best ratio of wavelengths for the CSI resu lts, the analysis of the data from the spectrometer experiments was completed. First, after plotting the two ratios against 70

PAGE 71

each other (R2 = 0.84***) the results lead to the conclusion th at there is no benefit to using the 970/670 nm ratio over the 840/670 nm wavelength ra tio (Figure 3-15). Secondly, based on the results form the subtraction of absorbance spectrums (Figure 3-16, A and B), there is little to no recognizable difference between the wavelengths at 840 nm and 970 nm, which agrees with the water index conclusion that th ere is no disadvantage to usi ng 840 nm wavelength over 970 nm wavelength. In addition, results from the subtraction show that at the 670 nm wavelength there is an increased difference, indicating that the waveba nd that is actually indicating stress in the CSI ratio is the 670 nm wavelength. Finally, the re sults from the regressions of 840 nm / 670 nm and 970 nm / 670 nm (Figure 3-17, A and B) also concl udes that there is no di sadvantage to using the 840 nm wavelength over the 970 nm wavelength, in fact, the regression of the ratio of 840 nm / 670 nm had a better and more significant fit (R2 = 0.23*) than the ratio of 970 nm / 670 nm (R2 = 0.12NS). As a result in these experiments the ratio of 840/670 was used in the CSI equation. It is also possible that due to the nonlinear response characteristics of th e multispectral cameras image sensor, there is less noise at the 840 wavelength than at the 970nm wavelength, which results in the 840/670 nm ratio yielding better re sults than the 970/670 nm ratio. The CMOS type sensor used in this camera has a zero response cu toff near 1000 nm which is close to that of the 970 nm wavelength, increasing the possibili ty of error at this wavelength. The data from the first experiment to test the multispectral camera was inconclusive due to the same sources of experimental error encountere d in the infrared measurements. The sources of error included the overall health of some of the trees. The reduced health is most likely a result of the root rot which was caused by lack of drainage from the pots, or the mite infestations, which damaged the young foliage. Beyond that, expe rimental error could have been caused by 71

PAGE 72

the small canopy size of the trees us ed in the experiment. Since th e canopies were so small there was quite a bit of background noise that was fact ored into the crop stre ss index (CSI) (Figure 318). Since during the image analysis, every pixel is used, even the pixels for the greenhouse wall are calculated, resulting in an increased CS I for a tree that is otherwise non-stressed. Results from the field experiment show a strong connection between the CSI and stem water potential (SWP). CSI from a day of measurements with clear sky conditions and little to no wind shows a strong trend with SWP (R2 = 0.90***) (Figure 3-19). During image processing some of the images were found to cause error due to improper ar eas of the canopy being captured. During image capture with the multispect ral camera the outer uniformly sunlit portions of the canopy are selected (Letter A of Figur e 3-19), however during image processing there were two types of error that were encountered leading to non-uniformly illuminated images; first, interference with young flushes of growth which might artif icially increase the CSI due to the light color of the young leaves or because the young leaves tend to wilt before the mature leaves (letters B and D in Figure 3-19); and second, large portions of shaded gaps in the image of the canopy (letters C and E in Figure 3-19). These problems can be seen in the 710 nm wavelength images (Figure 3-20, A-E). The regre ssion created using these data only includes the CSI that were considered not to be such outliers. Images were also captured for analysis of CSI on a clear day with an average wind speed of 1.65 m/s with gusts up to 4.02 m/s as well as on a cloudy day with little or no wind present. Results show a reduced accuracy of the meas urements on the windy day with a regression R2 of 0.66*** (Figure 3-21). This reduced accuracy could be due the movement of the branches as a result of wind from one wavelength image to th e next. Since the wavelengths used in the analysis are 670 nm and 840nm, the time between th e capture of these images is a few seconds. 72

PAGE 73

During measurements with little to no wind, ther e would be little to no movement of the canopy, however, on a windy day, the branches would be moving during image capture, resulting in some incorrect pixel ratios. Data co llected on a cloudy day also reveal a reduced accuracy of the CSI measurements, with a regression R2 of 0.44** (Figure 3-22). This could be caused by the reduced reflectance from the canopy as a result of the redu ced intensity and altered spectral quality of solar radiation. Even with the error observed from the wi ndy and cloudy CSI measurements the results from the clear conditions conclude that the mu ltispectral camera could be used for irrigation triggering, however more research would need to be done to identify the threshold for the CSI that would be an indicator of the beginning of water stress. Once this is established it is possible that multispectral camera measurements could be more accurate than soil based measurements because the trees should be the best integrators of available water in the entire root system. GreenSeeker NDVI Finally, NDVI measurements from the GreenSeek er instrument were plotted against the SWP measurements taken. The results from this re gression analysis show that as the water stress decreases the NDVI increases. The calculated N DVI fell between 0.646 for a stressed tree (SWP = -2.84 MPa) and 0.967 for a non-stressed tree (SWP = -1.06 MPa). The regression of NDVI to SWP yielded a linear fit with an R2 of 0.31*** (Figure 3.23). While the results show that a prediction of plant water status could be made, more research w ould need to be done to find a threshold NDVI for irrigation. Furt her studies would need to be ca rried out to determine if this sensor would be accurate e nough to trigger irrigation. Based on the results from all of the experiment s carried out to test canopy measurements, it should be possible with further research to develop thresholds that could result in a more accurate trigger for irrigation. In a controlled greenhouse setting, the infrared radiometer could 73

PAGE 74

be used to calculate CWSI which could be us ed to trigger irrigation after a threshold is established. Based on results from the field st udies, the multispectral camera and GreenSeeker outperformed the infrared radiometer; however thre sholds would still need to be identified to determine the proper CSI or NDVI values that would be used to trig ger irrigation. 74

PAGE 75

Figure 3-1. Apogee Instruments Inc ., IRR-PN Infrared Radiometer Figure 3-2. 20 Rough Lemon trees, planted in po tted grove soil used fo r greenhouse experiments 75

PAGE 76

Figure 3-3. Aqua-pro capacitance soil moisture probe used for soil moisture monitoring Figure 3-4. IRR-PN on ring stand pointed at tree canopy during long term measurements. 76

PAGE 77

Figure 3-5. Regression analysis from non-stressed tree data for baseline or lower limit ( TLL) calculation Figure 3-6. Soil covered with Tyvek Ho meWrap (DuPont, Wilmington, Delaware) 77

PAGE 78

Figure 3-7. Graphical presentati on of stem water potentials vers us the stress-degree-day (Tc Ta) results from measurements taken in the second infrared radiometer experiment Figure 3-8. Graphical presentati on of soil volumetric water conten t versus the stress-degree-day (Tc Ta) results from measurements ta ken in the second infrared radiometer experiment 78

PAGE 79

Figure 3-9. Graph of measurements taken during IRR-PN experiment 2 A. B. Figure 3-10. Crop Water Stress Inde x and irrigation application by da te. A) Rep 1, and B) Rep 2 79

PAGE 80

Figure 3-11. Comparison of measurements taken with the infrared radiometer in cloudy and sunny conditions in the greenhouse. Figure 3-12. Regression analysis showing Tc Ta versus stem water potential results from the third infrared radiometer experiment. 80

PAGE 81

Figure 3-13. Regression analysis comparing measurements of Tc Ta, taken with the infrared radiometer in both windy and wind free conditions versus stem water potential. Figure 3-14. Comparison of the regression an alysis of the CWSI normalized IRR-PN measurements taken in windy and windless cond itions versus stem water potential 81

PAGE 82

Figure 3-15. Correlation analysis of Crop Stress Index ratios using the 840 nm and 670 nm wavelengths A. B. Figure 3-16. Analysis comparing the stressed a nd non-stressed leaves, A) measured absorption of stressed vs. non-stressed leaves and B) subtraction of abso rption spectra (non-stressed stressed) 82

PAGE 83

A. B. Figure 3-17. Regression analysis of water stress index ratios A) 840 nm / 670 nm and B) 970 nm / 670 nm. Figure 3-18. Screen capture of data processi ng from a small greenhouse tree showing the large amount of background interference. 83

PAGE 84

Figure 3-19. Regression from Clear Sky conditions showing outliers B. E. as a result of improper imaging 84

PAGE 85

A. B. C. D. E. Figure 3-20. Images captured by the multispectral camera showing normal image A and outliers B E. A) Normal capture of a uniformly sunlit canopy, no shaded holes or interfering branches, B) young light-green flush interfering with image, C) large shaded hole captured in image, D) young water stressed fl ush interfering with remaining image, and E) Large shaded hole in canopy. 85

PAGE 86

Figure 3-21. Regression analysis from measurements taken on a sunny day with average wind speed of 1.65 m/s and max gusts of 4.02 m/s. Figure 3-22. Regression analysis from measurements taken on a cloudy day 86

PAGE 87

Figure 3-23. Regression analys is from measurements taken with the GreenSeeker 87

PAGE 88

CHAPTER 4 SUMMARY OF RESULTS As water resources become limited, it is importa nt to irrigate in the most efficient manner possible, while still maintaining a suitable crop yield. Throughout much of Florida, citrus is irrigated based on soil moisture status and evapotranspiration. However, there are several factors that contribute to a high level of variability in moisture within the rootzone. These include soil type, micro-topography, the type of irrigation system, root density, and canopy interference with rain and irrigation water, all of which make it diff icult to irrigate at a high level of efficiency. Additionally, although soil moisture sensors are precise, they lack accuracy in the field due to several factors which include a small sensing vol ume and disruption of the soil profile required during installation. All of these factors together make questionable the a ccuracy of a single point soil moisture measurement to be used for irriga tion scheduling. These questions of efficiency and accuracy led to the need for comprehensivel y mapping soil moisture variability to find how many soil based sensors would be required to effici ently irrigate trees and maintain yield. After mapping, there was a high amount of variab ility in soil moisture under citrus trees after varying amounts of rainfall an d irrigation, as well as after se veral days of drying. This spatial variability makes single point measurements inaccurate. Identifying the best position for a single moisture sensor under a citrus tree canopy for accurate irrigation sc heduling is impossible, and would lead to situations of either over irri gation and leaching or possible yield reduction as a result of under irrigation. In general, recently wett ed soils were more variable than drier soils. Calculations to find the number of sensors that would be needed to accurately trigger irrigation within 95% confidence of not having more than a 10% error in the plan t available water (PAW) measurement, revealed that between 14 and 53 measurements were needed for dry soil conditions whereas a maximum of 2 89 measurements were needed for a moist soil after a rainfall 88

PAGE 89

event. The lower variability in dr ier soils means that the moisture sensor trigger which starts the irrigation would be more reliable than the one whic h stops it. This also shows that the variability in the soil moisture would more often lead to situations of over irriga tion rather than under irrigation since the tree is able to incorporate mo isture from its entire root area. The tree would be able to utilize the water in the wetter areas of the soil even though there may be areas of soil that are dry beyond the permanent wilting point. Based on these results, using soil moisture sensors to measure volumetric water content fo r accurate irrigation is cost prohibitive and unfeasible. Since the aim of the Ri dge Citrus BMPs is to reduce the amount of nitrogen reaching the groundwater and eventually the aquifer, more efficient irrigation triggers are necessary. Based on the mapping results, this level of irrigati on control can not come from single point soil moisture measurements. As a result, canopy-base d water stress methods were investigated. Infrared canopy temperature measurements and multispectral images were tested in a controlled greenhouse environment and in a small citr us field block. In addition to testing these methods, a commercially available GreenSeeker instrument, which is equipped with its own light source and unaffected by the amount of solar radi ation was used to test the efficiency of the normalized difference vegetation index (NDVI), which is the a reflectance ratio of near infrared reflectance minus red reflectance over the near infrared reflectance plus the red reflectance (see equation, Chapter 3), in predicting plant water stress of the field trees. Results from the thermal infrared radiometer experiment in the greenhouse, using a single air temperature (Ta) measurement subtra cted from average canopy temperature (Tc) measurements (Tc Ta) or T, plotted against stem water potential (SWP), showed these measurements to be inaccurate for measuring wa ter stress. However, due to experimental error and plant health issues, it is not possible to conclude that Tc Ta measurements would be 89

PAGE 90

inappropriate in controlled environments. After an experiment in the greenhouse to test the long term use of the infrared radiometer, it is possible to identify the water stress of the tree using the normalized Tc Ta calculated using the crop wa ter stress index (CWSI) (see equations, Chapter 3). Once a threshold of CWSI to plant water stress has been established, it is possible that this method could be used to schedule irrigation for potted citrus trees in a greenhouse. In the field measurements with the infrared radiometer a nd calculations of CWSI, variations in wind and cloud cover made reliable measurements of CWSI with thermal infrared radiometry on citrus trees impossible. Testing the multispectral imaging method requ ired subsequent leaf absorbance wavelength spectrum analysis. This was done to identify th e appropriate ratio of reflectance wavelengths necessary for data analysis usi ng the crop stress index (CSI), whic h is a ratio of reflectance at near infrared over reflectance at red. After id entifying the appropriate wavelength ratio of 840 nm / 670 nm, analysis was applie d to the captured imag es of both the greenhouse and field citrus trees using the multispectral camera. Due to the same sources of error as experienced in the greenhouse with the infrared radiometer, the multis pectral camera and CSI was not effective for characterizing stress in greenhous e trees. Furthermore, the field of view for the multispectral camera at a 30 cm distance also cau sed error in the CSI ratio mos tly due to the large amount of the background in the image that was not c overed by the tree canopy in the foreground. However, results from the field experiment unde r clear, wind free conditions, showed that the CSI ratio was reliable at predicting the water stress of the trees (R2 = 0.90***). When tested under cloudy and windy conditions, the CSI ratio still provided a r easonable estimate of the tree water stress, however not to the same accuracy as seen during the clear wind free conditions (R2 90

PAGE 91

= 0.44** and 0.66***, respectively). It can be concluded that the multispectral camera and CSI ratio could be used to trigger irrigation once an appropriate threshold of th e CSI was established. The commercially available GreenSeeker with NDVI output, which is unaffected by the varying levels of solar ra diation, could also predict the water stress of citrus trees in the field with a statistically significant linear fit (R2 = 0.31***). It could be possibl e to schedule irrigation based on these NDVI measurements after a thres hold of NDVI to SWP has been identified. Since the tree canopy is able to integrate total water use fr om within the rootzone, canopy based water stress measurements would be more accurate for water stre ss measurements for the purpose of scheduling irrigation. It can be conclu ded that the infrared radiometer in the longterm greenhouse experiment and the multispectral camera and GreenSeeker experiments in the field are more accurate for predicting water stre ss than using single sensor based soil moisture measurements to schedule irrigation. 91

PAGE 92

LIST OF REFERENCES Alados, I., F.J. Olmo, I. Foyo-Moreno, and L. Alados-Arboledas. 2000. Estimation of photosynthetically active radi ation under cloudy conditions. Agricultural and Forest Meteorology. 102: 39-50. Alva, A.K., O. Prakash, A. Fares, and G. Hornsby. 1999. Distribution of rainfall and soil moisture content in the so il profile under citrus tree canopy and at the dripline. Irrigation Science. 18: 109-115. Boman, B.J. and E.W. Stover. 2002. Managing salinity in Florida citrus. University of Florida, IFAS Cooperative Extension Service. CIR 1411. Burrough, P.A. and R.A. McDonnell. Principles of Geographical Information Systems. 333 pp. Oxford University Press Inc., New York. 2004. Campbell, G.S. and W.H. Gardner. 1971. Psychr ometric measurement of soil water potential: temperature and bulk density effects. Proceedings of the Soil Science Society of America. 35: 8-12. Charlesworth, P. 2000. Soil water monitoring. National Program for Irrigation Research and Development. 101 pp. CSIRO Land and Water, Canberra, Australia. Cohen, Y., V. Alchanatis, M. Meron, Y. Saranga and J. Tsipris. 2005. Estimation of leaf water potential by thermal imagery and spatial analysis. Journal of Experimental Botany 56(417): 1843-1852. Curran, P.J. 1983. Multispectral remote sensing for the estimation of gr een leaf area index. Philosophical Transactions of the Royal Society A 309:257-270. Dallon, D. and B. Bugbee. 2003. Measurement of water stress: comparison of reflectance at 970 and 1450. Utah State University, Crop Physiology Laboratory. http://www.usu.edu/cpl/PD F/Water%20Stress_Dallon.pdf Last accessed March 4, 2009. Dane, J.H. and G.C. Topp. 2002. The soil solution phase. J.H. Dane and G.C. Topp. Methods of soil analysis, Part 4: Physical methods. Soil Science Society of America, Inc. Madison, WI. Davis, J.L. and A.P. Annan. 2002. Ground penetra ting radar to measure soil water content. 446463. J.H. Dane and G.C. Topp. Methods of soil analysis, Part 4: Physical Methods. Soil Science Society of America, Inc. Madison, WI. Elvidge, C.D. and Z. Chen. 1995. Comparison of broad-band and narrow-band red and nearinfrared vegetation indices. Remote Sensing of Environment. 54:1 38-48. Gamma Design Software. 2004. GS+:Geostatistics for the Environmental Sciences. Gamma Design Software, Plainwell, Michigan USA. 92

PAGE 93

Gonzlez-Dugo, M.P., M.S. Moran, L. Mateos, and R. Bryant. 2006. Canopy temperature variability as an indicator of crop water stress severity. Irrigation Science 24(4):233-240. Haman, D.Z. and F.T. Izuno. 1993. Soil plant wate r relationships. Univ. of Florida, IFAS Coop. Ext. Serv.CIR1085. Hope, A.S. and R.D. Jackson. 1989. Early mo rning canopy temperatures for evaluating water stress in a wheat crop. Water Resources Bulletin: Americ an Water Resources Association 25(5):1009-1014. Idso, S.B., R.D. Jackson, P.J. Pinter, Jr., R.J. Reginato and J.L. Hatfie ld. 1981. Normalizing the stress-degree-day parameter for environmental variability. Agricultural Meteorology. 24: 45-55. Jackson, R.D. 1982. Canopy temperature and crop water stress. Advances in Irrigation 1:43-85. Jackson, R.D., S.B. Idso, R.J. Reginato, and P. J. Pinter, Jr. 1981. Canopy temperature as a crop water stress indicator. Water Resources Research 17(4):1133-1138. Jackson, R.D., W.P. Kustas, and B.J. Choudhury. 1988. A reexamination of the crop water stress index. Irrigation Science 9:309-317. Jones, H.G. 2004. Irrigation scheduling: advantages and pitfalls of plant-based methods. Journal of Experimental Botany. 55(407):2427-2436. Jones, H.G. 2007. Monitoring plant and soil water status: established and novel methods revisited and their re levance to studies of drought tolerance. Journal of Experimental Botany.58(2): 119-130. Leinonen, I., and H.G. Jones. 2004. Combini ng thermal and visible imagery for estimating canopy temperature and identifying plant stress. Journal of Experimental Botany. 55(401): 1423-1431. McCutchan, H. and K.A. Shackel. 1992. Stem-water potential as a Sensitive indicator of water stress in prune trees (Prunus domestica L. cv. French). Journal of the American Society for Horticultural Science. 117(4): 607-611. Mller, M., V. Alchanatis, Y. Cohen, M. Meron, J. Tsipris, A. Naor, V. Ostrovsky, M. Sprintsin, and S. Cohen. 2007. Use of thermal and visible imagery for estimating crop water status of irrigated grapevine. Journal of Experimental Botany. 58(4): 827-838. Monteith, J.L. and M.H. Unsworth. Principles of Environmental Physics 3rd Edition. 418 pp. Elsevier, London. 2008. Morgan K., H. Beck, J. Scholberg, and S. Gr unwald. 2006. In-season irrigation and nutrient decision suport system for citrus pro duction. World Congress on Computers in Agriculture, Orlando, FL, July 24-26, 2006. 93

PAGE 94

94 Mulla, D.J. and A.B. McBratney. 2000. Soil spatial variability. A:321. M.E. Summer. Handbook of Soil Science. CRC Press. Boca Raton. Muoz-Carpena, R. 2004. Field Devices for monitoring soil water content. Department of Agricultural and Biological Engineering. University of Florida Cooperative Extension Service. Bulletin 343. NOAA, Southern Region Headquarters. 2008. Vapor Pressure. http://www.srh.noaa.gov/sju/Climate/wx calc2go/formulas/vaporPressure.html Last accessed March 19, 2009. Robinson, D.A., C.M.K. Gardner and J.D. Coope r. 1999. Measurement of relative permittivity in sandy soils using TDR, capacita nce and theta probe: comparis on, including the effect of bulk soil electrical conductivity. Journal of Hydrology. 223: 198-211. Schmitz, M. and H. Sourell. 2000. Variability in soil moisture. Irrigation Science. 19: 147-151. Scholander, P., H. Hammel, E.Y. Bradstreet, and E. Hemmingsen. 1965. Sap pressure in vascular plants. Science. 37: 247-274. Schumann, A.W., K. Hostler, J.C. Melgar, and J. Syvertsen. 2007. Georeferenced ground photography of citrus orchards to estimate yield and plant stress for variable rate technology. Proceedings of the Florida State Horticulture Society. 120: 56-63. Southwest Florida Water Management District. 2002. GIS Data, SOILS. http://www.swfwmd.state.fl.us/data/gis/libraries/physical_dense/soils.htm Last accessed March 19, 2009. Topp, G. C., J. L. Davis, and A. P. Annan. 1980. Electromagnetic Determination of Soil Water Content: Measurements in Coaxial Transmission Lines, Water Resource Research. 16(3): 574. Wijaya, K., T. Nishimura and K. Makoto. 2002. Estimation of bulk density of soil by using amplitude domain reflectometry (ADR) probe. 17th World Congress of Soil Science. Thailand. 385. U.S. Water Conservation Laboratory. 2001. Thermal Crop Water Stress Indices. http://www.plantstress.com/articles /drought_i/drought_i_files/CWSI_phoenix.pdf Last accessed March 10, 2009 USDA-NRCS. 1990. Soil survey of Polk County, Florida. 235. USDA

PAGE 95

BIOGRAPHICAL SKETCH Laura Waldo was born in Madison, Wisconsin and moved to Florida in 1982. Upon graduating from Dr. Phillips High School in Orla ndo, FL, she attended Valencia Community College and earned a general education Asso ciate in Arts degree in 2001. Immediately following, she attended Florida Southern Colle ge-Lakeland, FL and earned a Bachelor of Science degree in horticulture science in 2003. In April of 2004 she began work on the Ridge Citrus nitrogen BMP verification study at the UF IFAS Citrus Research and Education CenterLake Alfred, FL. In the fall of 2006, she began her Master of Science degree at the University of Florida-Gainesville, FL in th e department of Soil and Water Science as a distance education student, while working in Lake Alfred. In 2007, she became a Graduate Assistant while finishing her thesis research. She completed her Master of Science degree in Soil and Water Science in 2009. 95