1 C OMPARING GRID AND EC DIRECTED ZONE SAMPLING SCHEMES FOR SOIL FERTILITY MANAGEMENT IN FLORIDA By ASHLEY MASON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMEN TS FOR TH E DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2013
2 2013 Ashley Mason
3 ACKNOWLEDGMENTS I would like to extend grateful thanks to the faculty and staff within the University of Florida Agronomy department, partic ularly on campus and at the West Florida Research and Education Center. I would specifically like to thank Dr. Schnell his advisement and for making this opportunity available and Dr. Ferrell for his support and guidance.
4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 5 LIST OF FIGURES ................................ ................................ ................................ .......... 6 LIST OF ABBREVIATIONS ................................ ................................ ............................. 7 ABSTRACT ................................ ................................ ................................ ..................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 10 Literature Review ................................ ................................ ................................ .... 13 Materials and Methods ................................ ................................ ............................ 26 2 RESULTS ................................ ................................ ................................ ............... 35 Field Level Soil Chemical and Physical Properties ................................ ................. 35 Measurement of Soil EC a ................................ ................................ ........................ 36 Validation of Sampling and Interpolation Techniques ................................ ............. 37 Relating EC a to S oil Properties ................................ ................................ ............... 37 Soil Properties and Yield by Zone ................................ ................................ ........... 38 Evaluating Grid and Zone Sampling Techniques ................................ .................... 39 3 DISCUSSION AND CONCLUSION ................................ ................................ ........ 61 Discussion ................................ ................................ ................................ .............. 61 Conclusion ................................ ................................ ................................ .............. 65 LIST OF REFERENCES ................................ ................................ ............................... 66 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 70
5 LIST OF TABLES Table page 1 1 Percent soil type by field characterized by USGS soil survey map ..................... 31 1 2 Average fertilizer and lime application per hectare for all study fields ................ 34 2 1 Mean and standard error from all available field soil samples, representing the average and variability in our nutrients and texture properties. P, K, Mg and Ca represent Mehlich 1 extrac nutrients. ................................ ..................... 41 2 2 Summary of linear regression of soil properties and yield to soil EC a and yield (p < 0.05). ................................ ................................ ................................ ........... 49 2 3 Summary of averag e P, K, Ca, Mg by EC a Zone and whole field with SE by ................................ .......... 51 2 4 Average yield (kg/ha) for all study fields, separated by zone and summar ized by whole field mean for each. ................................ ................................ ............. 54 2 5 The percent (%) of field area predicted within 1 standard deviation of the control by field and nutrient. Mehlich 1 extrac phosphorus (P), potassiu m (K), calcium (Ca), and magnesium (Mg). Grid no interpolation (GNI), grid with interpolation (GWI), zone polygon (ZP) and zone interpolated all. ..................... 56 2 6 Results from the subsample least squ ares mean comparisons. Treatment occurrence in table indicates it was similar to the control dataset (p<0.05). ....... 58 2 7 Average value from 10 x 10 meter subsample procedure. Average is p rovided by treatment for P, K, Mg, Ca. ................................ ............................. 58
6 LIST OF FIGURES Figure page 1 1 Example of EC a data points collected by the Veris 3150 sensor cart, field 21 prior to interpolation. ................................ ................................ ........................... 32 1 2 After interpolation of EC a points, the maps were reclassified into three classes. These classes indicated three management zones consisting of low, medium and high EC a zones. Within these zones points were placed where soil samples were taken to capture variability within each zone. ........................ 33 1 3 Grid samples sites were identified by overlaying a one hecta re grid on each field. Soil samples were taken from the middle of each grid. .............................. 34 2 1 EC a means by EC a zones field 1, variation show using a 95% confidence interval. ................................ ................................ ................................ ............... 42 2 2 EC a means by EC a zones for field 10, variation shown using a 95% confidence interval. ................................ ................................ ............................. 43 2 3 EC a means by EC a zones for field 14, variability shown usi ng a 95% confidence interval. ................................ ................................ ............................. 44 2 4 EC a means by EC a zones for field 18, variability shown using a 95% confidence interval. ................................ ................................ ............................. 45 2 5 EC a means by EC a zones for field 19, variability shown using a 95% confidence interval. ................................ ................................ ............................. 46 2 6 EC a means by EC a zones for field 20, variability shown using a 95% confidence interval. ................................ ................................ ............................. 47 2 7 EC a means by EC a zones for field 21 variability shown using a 95% confidence interval. ................................ ................................ ............................. 48 2 8 Field 18 linear regression example o f soil test phosphorous and soil EC a P < 0.05. ................................ ................................ ................................ ................ 50 2 9 Field 20 linear regression example of percent sand and soil EC a P < 0.05. ..... 50 2 10 Example of a final difference raster from the standardized control minus treatment method for comparing our four sample schemes ............................... 55 2 11 Average field area predicted within one standar d deviation of the control (%) by our four treatment methods, GNI, GWI, ZIA and ZP. ................................ ..... 57
7 LIST OF ABBREVIATIONS Ca Calcium CEC Cation Exchange Capacity EC a Apparent Electrical Conductivity EC Electrical Conductivity GNI Grid No Interpolation GWI Grid With Interpolation ha Hectare K Potassium Mg Magnesium pH Acidity or A lkalinity P Phosphorous RMSE Root Mean Square Error ZIA Zone Interpolated All ZP Zone Polygon
8 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for th e Degree of Master of Science COMPARING GRID AND EC DIRECTED ZONE SAMPLING SCHEMES FOR SOIL FERTILITY MANAGEMENT IN FLORIDA By Ashley Mason December 201 3 Chair: Ronald Schnell Cochair: Jason Ferrell Major: Agronomy Site specific crop management involves identifying spatial variability within a field, and using that information to implement efficient management practices. On the go measurement of soil el ectrical conductivity (EC) is one method of delineating management zones that seeks to improve upon the grid sampling method for nutrient ma nagement ( Corwin and Lesch, 2005). However, the relationship between spatial variability of soil EC and crop nutrien ts is unclear (Johnson et al ., 2001; Rysan and Sarec, 2008). In this project comparisons were made between grid and soil EC a delineated zones sampling strategies for fertility management in peanut production. Our first objective was to relate soil chemical or physical properties to on the go measurements of soil EC. Our second was to compare grid and zone soil testing methods to determine which is most accurately characterizing nutrient variability. These included comparison of grid sampling by polygon, gri d sampling with interpolation, zone sampling by polygon and zone sampling with interpolation. From our results we conclude that while EC a variation existed within our fields, the variation in EC a had no relation to soil factors of agronomic importance. We conclude that while EC a zone soil
9 sampling for productivity management has proved appropriate in other locations with different landscapes and soil type, (Johnson et al. 2001; Kaffka et al. 2005; Kitchen et al. 2005; Sudduth et al., 2005 ) zone soil samp ling using EC a measurements is not an improvement upon the traditional one hectare grid soil sampling in Northwest Florida.
10 CHAPTER 1 INTRODUCTION Cost for fertilizer and lime have increased dramatically in recent years, creating incentive for new strat egies in fertility management. Soil sampling has historically been used to assess fertility needs while managing each field as a homogenous unit (Flowers et al., 2005). Yet considerable variability is known to exist within fields, which creates potential f or over or under application of nutrients. Greater application efficiency may be achieved through site specific nutrient management. Site specific management utilizes rapidly evolving and available electronic information technologies to modify land managem ent as conditions change spatially and temporally (van Schilfgaarde, 1999). In lieu of traditional soil sampling, which has historically been to treat fields as homogenous areas, grid soil sampling identifies within field nutrient variability and initiates nutrient application on a site specific basis (Fleming et al., 2001). A grid size of one hectare (ha) is most prevalent in practice (Fleming and Westfall, 2000). However, research indicates that yield influencing soil variability is not captured in the ty pical grid size of one hectare. While smaller grids clearly better represent within field variability, they are labor and cost prohibitive. There exists a need for nutrient assessment strategies that require fewer samples, yet generate effective field pres criptions for variable rate fertilizer applications (Johnson et al., 2001; Khosla, 1999; Sadler et al., 1998). Zone soil sampling, an alternative to grid sampling, seeks to identify and manage within field variability in a site specific manner. Zone samplin g ideally reduces the number of samples per field, thereby reducing sampling costs while still obtaining within field nutrient variation information. For zone sampling to be useful, identification of agronomically relevant variability within fields is nece ssary. Efforts to identify and
11 manage soil salinity levels in arid, irrigated soils prone to salinity issues have acted as a catalyst for the development of spatial technologies that easily obtain information about within field variability of soil properti es. Corwin & Lesch (2003) note that soil salinity has typically been defined in terms of the electrical conductivity (EC) of a saturated soil paste extract (EC e generally do not have enough moi sture to perform the extraction, thus soil solution extracts are facilitated by adding water to the sample (Corwin & Lesch 2003). The time, labor and capital for analysis of EC e have encouraged exploration of alternative methods of obtaining field salinit y information. Measurement of apparent soil EC (EC a ) is now a popular alternative to EC e for use in identifying within field EC variation. ECa measures the conductivity of the soil profile within the measurement range, providing a sum value of conductance of the soil solution, solid soil particles and exchangeable cations (Corwin & Lesch 2003). Therefore while this measurement is used in lieu of the laboratory analysis of EC e it is actually analyzing the soil solution, solid and cations rather than the soi l solution exclusively. Williams and Baker (1982) found in saline or salt affected soils, 65 70% of variation in EC measurements could be explained by the concentration of soluble salts. However in non saline soils, conductivity variations are mostly attri buted to soil moisture, texture, and CEC variability (Williams and Baker 1982). EC a measurements may be obtained by either a four electrode system (Rhoades et al ., 1990; Halverson and Rhoades 1976), referred to as the wenner array, (Rhoades and van Schil fgaarde 1976) or an electromagnetic induction (EMI) system (Lesch et al ., 2000). The four electrode system requires at least four electrodes be in contact with the soil, during which an electrical current is injected and the reduced voltage is measured
12 (L und et al., 1999). This value is the soil ap parent electrical resistivity. Corwin and Lesch ( 2003) note that in homogeneous mediums, resistivity is the reciprocal of conductivity. A resistivity device converts given measurements of apparent resistivity int o measurements of apparent EC a (Corwin and Lesch 2003). The EMI system obtains EC a data using a transmitter coil to create an electrical field between the transmitter and receiver coil, in which the response is measured (Lund et al., 1999). No contact is required between the receiving instrument and the soil; therefore EM measurements have an advantage of over the fixed four electrode array systems, which produces less reliable measurements in dry or stony soil (Corwin and Lesch 2003). However the advanta ge of the mobilized fixed array system includes equipment simplicity and speed of operation (Carter et al. 1993), which has led to its proliferation in practical field use. Veris Technologies in Salina, Kansas offers commercialized EC a measurement equipme nt, among which is a series of mobilized fixed array systems. These efficient tools for measuring EC a are paired with global positioning system instruments to create manageable EC a datasets. EC a measurements are product of a complex interaction of soil pro perties (Corwin and Lesch, 2003). Soil properties that influence EC a measurements include soil salinity, clay content and cation exchange capacity, clay mineralogy, soil pore size and distribution, soil moisture content, organic matter, bulk density, and s oil temperature (McNeil, 1992; Rhoades et al., 1999a, 1999b; Corwin and Lesch, 2003). Previous studies relating crop yield directly to EC a present inconsistent results (Corwin and Lesch, 2003), and the relationship between spatial crop nutrient variabilit y and soil EC a remains unclear. This is the premise of our comparison of grid soil sampling to EC a directed soil sampling for fertility management in peanut
13 production. Our objectives include relating soil chemical and physical properties to spatial measur ements of soil EC a Our second objective is to compare grid and EC a directed zone soil sampling to determine which most accurately characterizes spatial variability in soil chemical properties. Literature Review Grid soil sampling has been popularly adopt ed over the last decade to identify within field soil variability in effort to improve the efficiency of soil nutrient management compared to the traditional method of managing a field as a homogenous unit (Fleming et al., 2001). While a one hectare grid i s common, research indicates that yield influencing soil variability is not captured in the typical grid size of one hectare. Franzen et al ., (1995) examined several grid sizes in effort to identify the density of sampling that most effectively identifies soil variability. Results from Franzen et al (1995) suggest the use of 220 square foot grids for initially determining soil variability. Field scale yield measurements by Sadler et al (1998) found yield to be quantitatively different in distances as sh ort as 10 m. While smaller grids clearly better represent within field variability, they are labor and cost prohibitive. There is a need for management methods that require fewer samples, yet generate effective field prescriptions for variable rate fertili z er applications (Johnson et al., 2001; Khosla, 1999; Sadler et al., 1998). Strong incentive exists for growers to pursue these management methods that optimize their fertility programs in that variable rate fertility programs have the potential to improve input efficiency, field profitability, and environmental stewardship (Sawyer 2013). For example, in a comprehensive examination of yield limiting factors in a 48 ha peanut production field in the Texas High Plains, Fre et al ., (2009) found Calcium and N itrogen to be most limiting to yield. By recognizing and managing these limiting factors
14 with variable rate fertilizer application, Fre et al ., (2009) reported a savings of $20.41/ha in N and $10.73/ha in Ca as compared to the projected expenditures of a conventional fertility program. The clear limitation of the ability of grid sampling to capture within field variability in a cost effective manner has stimulated interest in managing within field variability by separating fields into areas of homogenous characteristics. This alternative to grid sampling seeks to create within field management zones by identifying homogenous sub regions that possess characteristics that may limit, or relate to yield limiting factors (Vrindts et al., 2005). Assuming gener ation of useful zones is achievable; this would allow field inputs to be managed more efficiently, ideally saving time and money, affirmed by Fleming et al (2000). Multiple methods are available for assessing in field variability. Yield data, when collecte d over multiple years, can be used to identify homogenous zones of crop productivity (Black more, 2000; Fraisse et al., 2001 ). Research conducted by Fleming et al. (2001) concluded that zone management by a combination of aerial photographs, soil color, to pography, and grower field knowledge, creates prescription maps that are equally effective as variable rate maps generated from grid sampling. Topography and soil characteristic data utilized by Reyneirs et al., (2006) successfully created critical zones f or yield variation in response to erosion risks. Additionally, Zone Mapping Application for Precision Farming (ZoneMAP) is a tool accessible online which provide users access to archives of satellite imagery data. This tool uses the fuzzy c means (FCM) alg orithm for generating management zones for given fields; ZoneMAP also has the capacity to determine the optimal number of zones based on the archived data for a given field.
15 Li et al., (2008) found the combination of soil electrical conductivity measureme nts, yield, and normalized difference vegetation index (NDVI) data to accurately characterize spatial variation of soil chemical properties and productivity in cotton. Research completed by Nanna and Franzen (2003) concluded that there are various means of assessing field variability, and that consistency should be valued over high correlations. One year of their study delineating nitrogen management zones found that Order 1 soil survey maps highly correlated with field nitrogen management needs. However in years to follow the soil survey maps did not perform consistently in comparison to zone delineation by topography, yield and NDVI from satellite images. The need for performance consistency and collection efficiency has stimulated interest in zone deline ation by apparent soil electrical conductivity (EC a ) measurements. Soil EC a is a relatively inexpensive measurement when collected using sensor based technology. Electrode based EC a sensors have been used to explore soil variability for several decades and are commercially available today. One of the earliest forms of the (Halvorson and Rhoades, 1976). It was later adapted for mobile use by mounting the electrodes on a tract or and connecting to a GPS receiver and data logger (Carter et al., 1993). The device was then commercialized in 1999 by Veris Technologies (Lund et al., 1999). These electrode based sensors require soil contact, and the current equipment typically uses a set rolling coulters. The Veris 3100 (Veris Technologies, Salina KS) has six rolling coulters that provide two continuous spatially referenced measurements of EC a Depth of measurement is determined by the spacing of the coulters, the larger the spacing th e deeper the measurement (Corwin and Lesch 2003). Of the six coulter
16 electrodes, coulters number 2 and 5 are the current producing electrodes, (Corwin and Lesch 2003) while the others receive the electrical charge. These two sets of electrode arrays allo w for simultaneous measurement of EC a at two soil depths: 0 to 0.3 m and 0 to 0.9 m (Corwin and Lesch 2003). Previous reports have described the theory and principles behind measurement of EC a (Corwin and Lesch, 2003; Friedman, 2005). The injected curren t for measuring EC a using the Wennar array fixed coulter method travels in three pathways within the soil. These include a liquid phase pathway via dissolved solids contained in the soil water occupying the large pores, a solid liquid phase pathway primari ly via exchangeable cations associated with clay minerals, and a solid pathway via soil particles that are in direct and continuous contact with one another (Rhoades et al., 1999). The resulting value of the received electrical charge after traveling throu gh the soil profile provides the soil resistivity. Burger (1992) notes that when measured with the Wennar array the formula for resistivity (p) is: (1 1 ) Where V is the voltage, a is the interelectrode spacing, i is the electrical current, and R is the measured resistance. Being that EC a is the inverse of R, for useful purposes this equation becomes: EC a a R (1 2 ) Soil solution salinity, quantity of exchangeable ions associated with clay particles, and soil particles in direct contact within the zone of measurement all influence EC a readings. Williams and Baker (1982) found in saline or salt affected soils, 65 70% of variation i n EC measurements could be explained by the concentration of soluble salts. However in non saline soils, conductivity variations are mostly attributed to soil
17 moisture, texture, and CEC variability (Williams and Baker 1982).These factors are then influenc ed by a combination of soil physical and chemical properties, including but not limited to; soil water content, bulk density, clay mineralogy and content, organic matter, and soil temperature (Corwin and Lesch, 2005; Rysan and Sarec, 2008). In general, san ds have a low conductivity and will thus influence ECa measurements to be lower, while silts have a medium conductivity and clays have a higher conductivity (Lund et al., 1999). Because of within field variability of soil properties, correlation between EC a and specific soil properties may be quite different from field to field (Corwin and Lesch, 2003). Additionally, soil and environmental conditions and methods employed during collection of EC a data can also influence measurements (Sudduth et al., 2005). Soil electrical conductivity has not been found to have a consistent or direct relation to crop yield, (Rysan and Sarec 2008; Johnson et al., 2003) However, EC a measurements have been found to relate to soil physical and chemical properties that impact yi eld. This is why, as outlined in the compilation of survey protocols by Corwin and Lesch (2005), ground truth samples within EC a delineated zones are a crucial component for the creation of EC a directed prescription maps (Johnson et al ., 2001). In research conducted by Johnson et al (2001) soil samples from EC a directed management zones were analyzed for correlation with EC a Percent clay, bulk density, pH and EC 1:1 was all found to have a positive correlation with the EC a readings. Soil moisture, total an d particulate organic matter, total C, total N, microbial biomass carbon and microbial biomass nitrogen were negatively correlated with EC a Collected surface crop residue was significantly related to EC a zones, and negatively correlated with EC a
18 Johnson et al (2001) thus concluded that soil classification by EC a is an effective basis for delineating field management zones. Sudduth et al (2003) analyzed the relationship of certain soil properties to EC a in four fields, and found soil moisture to be correl ated with EC a in only one field. Clay and CEC were correlated with EC a in all studied fields (Sudduth et al., 2003). Regarding yield influencing nutrients, Heiniger et al (2003) found few significant linear relationships between EC a and soil test levels o f P, K, Ca, Mg, Mn, Zn or Cu. In the instances which a statistically significant relationship existed, if was often weak with an R 2 less than .50. Heiniger et al (2003) identified a few instances of stronger significant relationships between ECa and soil nutrients, in which the R 2 was between .51 and .75, and found this occurred when the nutrient was also closely associated with one of the soil properties that directly influence EC a From this analysis Heiniger et al that EC a can be used to directly determine soil nutrient content across a field. However, EC a can be used along with other measured soil properties in a multivariate analysis to describe key factors influencing changes in nutrient concentrations and to e stablish nutrient management (2003) note that it is important when using EC a measurements to establish management zones, to know and account for site specific changes in volumetric soil moisture content, soil texture, CEC and s alinity. In an analysis of the usefulness of zones established by EC a values obtained by EM, Taylor et al ., (2003) found the zones to be closely correlated with soil series maps and related to yield variations in instances where there existed large differe nce in soil textural classes.
19 Following collection of EC a data, various procedures can be used to delineate management zones. The first step is to interpolate vector (point) EC a data to produce a raster image, or matrix of grid cells representing the sur face of the field. Various options for interpolation exist. McCoy and Johnston (2001) identify two groups of interpolation techniques in ArcGIS (ESRI, city, state) spatial analyst. These include deterministic and geostatistical. Deterministic is defined a s using mathematical functions for interpolation, while geostatistical is defined by use of statistical and mathematical methods. Inverse Distance Weighting is a deterministic interpolator, while krigging is a geostatistical interpolator. Inverse Distance weighting operates with the assumption that points close to each other are more alike than points further apart. Values determining unknown points are weighted based on their proximity to the unknown point. There are various kinds of geostatistical interp olators; however they are all related to kriging. Kriging also uses the same assumption as IDW, however includes the use of statistics. Kriging quantifies the study area by fitting a spatial dependence model to the data. To make predictions, kriging uses t he fitted model, spatial configurations, and values at measured sample points near the prediction location (McCoy and Johnston 2001). There are three steps required for successful Kriging. n and Unwin (2010) and include: 1. ) Producing a description of the spatial variation in the sample control point data. 2.) Summarizing this spatial variation by a regular mathematical function. 3.) Using this model to determine Step one is accomplished by creatin g a semivariogram in attempt to understand
20 application of exactly the same idea of control point data, with the additional provision that we wish to estimate its value at (2010) as zj)2 (1 3 ) The right side of the equation refers to the sum of squares of all the pairs of control point development of the semivariogram idea by George Matheron, French geostatistician in 1963 Important language describing complexities of semivariograms include the notion term to point at which semivariance is the highest. The distance at which semivariance is the is the Unwin (2010) explain that this step is what distinguishes Kriging from IDW, in that it makes use of what is understood, or can be inferred, about the spatial stru cture of the existing control data. This theory was developed by Georges Matheron in 1963, and Unwin humorously describe the step of estimating an appropriate mathematical function familiarity with the variable in question, and specify that this is not a step that can be
21 specify that in preparing a semivariogram, there exist the need for some arbitrary decision making, like distance intervals and nugget values. While an educated decision is optimal, Chiles and Delfiner (1999) stated that a crudely determined set of weight s can provide excellent results when interpolating a dataset. The next step is computationally intensive, in which the selected model is used to interpolate the control data points, with and Unwin 2013). In research completed by Robinson and Metternich (2006 ) inverse distance weighting, ordinary kriging, lognormal kriging, and splines were compared for interpolation accuracy of seasonally stable soil properties. Robinson and Metternich ( 2006 ) defined seasonally stable soil properties to include pH, electrical conductivity, and organic matter. To compare the different interpolation method s, Robinson and Metternich (2006 ) used cross validation to examine the differences between known and p redicted data points. Cross validation assesses the accuracy of an interpolation method by eliminating known information, allowing the interpolation to predict the missing value, and computing the difference between the known and predicted value. The root mean square error values, average kriging standard error, root mean square standardized prediction error, and mean standardized prediction error values were considered. Their results indicated that ordinary kriging performed best for topsoil pH, and logno rmal kriging provided the best results in EC a interpolation. In their study IDW best predicted subsoil pH. The spline method best predicted soil organic matt er (Robinson and Metternich 2006 ). In summary of this point, cross validation is a common
22 and relia ble way to assess the success of an interpolation method exercised in a specific dataset (Voltz and Webster 1990). Following interpolation of EC a data, values are typically classified into groups (zones) that minimize variability within the group and maxim ize variability between the groups. There are various ways this can be accomplished. Frigden et al (2004) developed a software package that uses fuzzy c means unsupervised clustering to delineate potential management zones. Fraisse et al (2001) also used an unsupervised clustering algorithm, ISODATA (Iterative Self Organizing Data Analysis Technique); referred to as the ISOCLUSTER function in the Arc/Info GIS software to create management zones from EC a and yield data. ISOCLUSTER uses a modified iterative optimization procedure, or migrating means technique. Arbitrary means are assigned by the software per user defined cluster; each cell is then assigned to the closest of those means. Means are recalculated for each cluster based on the distances of cells belonging to the cluster after iteration, and the process is repeated until the movement of cells from one cluster to another. The repetition of mean recalculation is called an lasses, minimum cells and sampling interval (Fraisse et al., 2001). In creation of management zones using EC a data, Johnson et al (2003) interpolated the EC a data using IDW with the nearest neighbor technique. The resulting maps were reclassified into fou r management zones. Additionally, Lesch et al (2000) developed a software package (ESAP) which identifies optimal location for zone sampling sites from EC a data, using a response surface sampling design. This software is available through the USDA Agricul tural Research Service. Johnson et al (2001), in evaluation of zone soil sampling,
23 mos t common method used in practical EC a zone sampling situations. It is clear that various mathematical techniques can be applied to determine the number of management zones assigned to a particular field. However, in many commercial situations the number o f zones used is often arbitrary and may lack agronomic significance. Although various methods exist for collecting and processing spatial data to establish soil sampling procedures, the goal of our study is to assess the performance of grid and zone soil nutrient management on Florida soils using methods currently in use for fertility management in Florida. Researchers have used various strategies for quantifying differences in grid and zone soil test data. Flowers et al (2005) compared yield based manag ement zones to three grid sampling methods. Zones were generated by culminating the previous four years yield data into one raster. Management zones were created using four strategies. Three of the strategies yielded maps with 20 by 20 meter spatial manage ment units. The fourth method allowed larger zones, more appropriate for practical use. Their grid sampling methods included grid cell, grid center, and grid center with kriging. The grid cell method uses the mean value of all data points contained within the respective grid. The grid center method used the value of the data point nearest to the center of the grid. Grid center with kriging took the value of the data point nearest to the center of the grid and interpolated it using kriging. Control regions w ere constructed by randomly dividing each field into areas of equal size and number related to the grid sampling and yield based zone methods. For
24 each assigned control region, fertilizer recommendations and mean soil test P, K and pH values were calculate d. Flowers et al (2005) note that a method of determining if the residual variance of two treatments is statistically difference is not available. Therefore, to compare the sampling strategies residual variance to the variance of the corresponding control s, Flowers et al (2005) concluded a difference of 15% signified that the treatment data was a result of the sampling scheme, not density of sample data. Additionally, a whole field soil test value was generated from the mean of all data points available in each field. This yielded separate whole field values for soil test P, K and pH, and recommendations for P, K and lime applications. These values served as the baseline to which the sampling schemes were compared to, in that the whole field average met hod was assumed to generally have the highest residual variance. The mean value for each field was then Weighted variances for each sampling scheme were calculated and compared to the whole field average method described earlier. The values found for the whole field average method were considered to be 100% variance. The residual variances of the sampling schemes were found by comparing weighted variance to the 100% variance value. Their results indicated that the 68 meter grid cell sampling method best ca ptured within field nutrient variability. Mallarino and Wittry (2004) exhibit another method of comparing data from grid and zone soil sampling. Management zones were determined in such a way that integrated soil survey map information, elevation data, yi eld, information from aerial photographs and grower experience. Grids were created by systematic arrangement of
25 cells ranging 1.2 to 1.6 hectares in size. The baseline for comparison was created by soil test values derived from a 0.2 hectare grid sampling scheme. An index of efficacy was created for each sampling method. Within group variability was compared to between group variability. Higher between group variability suggests a successful sampling scheme, whereas higher within group variability can be as sumed product of a flawed sampling design. Additionally, P and K fertilizer recommendations were generated for each sampling approach. By comparing the fertilizer recommendations for the intense 0.2 hectare sampling data to the other sampling methods, Mall arino and Wittry (2004) were able to determine the field proportion that would be correctly, over, and under fertilized for each method. Their results indicated that grid cell and zone sampling strategies best identified within field nutrient variability w hen compared to methods using exclusively soil survey maps or elevation zones. Additionally, Fleming et al (2004) used analysis of variance (ANOVA ) to compare nutrient and yield sample data between zones delineated by EC a and soil color with farmer input. They found EC a to be more useful in identifying areas of spatial variability within fields.
26 Materials and Methods The study site was located at the West Florida Research and Education Center in Jay, Florida (30.78, 87.14) and used seven fields, four in 2012, and three in 2013. All fields were in a peanut cotton rotation. The most common soil type among the seven fields was Red Bay Sandy l oam and Lucy loamy sand (Table 1 1.). Peanuts were grown in each field during their respective year of study. In M arch of 2012 and 2013, a grid size of 1 ha was assigned to each field using Farmworks software (Hamilton, IN) installed on a Trimble Nomad handheld computer. Soil EC was collected using the Veris 3150 sensor cart (Salinas, KS). The Veris 3150 traversed each field at an 18.3 m swath width, collecting on average one data point every 6.2 meters. Shallow EC a measurements were continuously collected from 0 to 0.3 m depth of soil and were be used to generate three management zones for each field. Management zo nes were delineated in Agjunction.com by interpolation of the EC a measurements to create a raster file that was classified into three zones using a clustering algorithm (proprietary information was unavailable). The use of three zones was pre determined an d considered to be consistent with the number of zones that are typically used by growers given logistical and financial restraints. Soil from the four peanut production fields (3 to 14 ha) was sampled in March of 2012 according to grid and soil EC a direct ed management zones prior to tillage, fertilizer, lime and planting operations. This was repeated for the three fields (4.5 to 8.5 ha) in 2013 with sampling occurring in March. Following delineation of zones and 1 hectare grids for each field, soil sample s were collected and sent to Waters Agricultural Laboratories (Camilla, Ga) for analysis. Analysis included pH, texture and Mehlich 1 extractable P, K, Ca, and Mg. Grid
27 samples were derived from a composite of 8 cores (15 cm depth) collected from within 3 m radius of the center of each grid. The Trimble Nomad handheld computer was used to locate sampling points at the center of each grid. Zone composite samples were derived from a minimum of 8 cores (15 cm depth) collected from random locations from with in each zone. Random soil core locations selected for zone composite samples were spatially referenced, sampled and analyzed independent of the composite sample. Software is available for identification of optimal sampling locations within each zone, howev er to mimic practical sampling scenarios we randomly choose sampling locations with the objective of capturing variability within each zone. Following collection of soil samples, uniform application of P, K and lime were applied to whole fields according to grid or zone recommendations in 2012 and grid recom men dations in 2013. (See Table 1 2 for average P, K and lime applications) In 2012 and 2013, lime and fertilizer was incorporated by shallow tillage and seedbeds were prepared on 91.4 cm centers. Prowl H2O (pendimethalin: N (1 ethylpropyl) 3,4 demethyl 2,6 dinitrobenzenamine) was applied pre plant at a rate of 0.63 kg ha 1 Peanut Georgia 06G was seeded in early May of 2012 and 2013. Additional weed and pest control was applied as needed according to I FAS guidelines. Peanuts were inverted and field dried for 3 5 days before threshing. Yield was determined by harvesting and measuring fresh weight of five meters of two parallel rows centered over each georeferenced zone and grid sample point. One way ANO VA with comparison of means with Tukey Kramer HSD in JMP was used to determine difference in soil variable levels by field and year. ArcMap 10.0 was used for mapping, interpolation and analysis of soil nutrient data. Kriging and
28 Inverse Distance Weighting interpolation methods were implemented using all soil sampling points (grid and zone combined) and compared based on cross validation of known sample values to predicted values of soil parameters, including pH, and Mehlich 1 extractable P, K, Ca, and Mg. T he best interpolation method was determined by selecting the method that provided the smallest root mean square prediction error, a root mean square standardized prediction error value closest to 1 and a mean prediction error near 0. Ordinary kriging was s elected as the method providing the lowest root mean square error. Ordinary kriging model was selected for each nutrient by the model which provided the lowest RMSE (Robinson and Metternicht 2006). The selected map, or control map, was used as a best esti mation of variation of soil nutrients within each field. In addition, interpolation was used to generate the treatment raster maps of soil parameters using grid center points (grid point), 1 ha grid polygons (grid polygon), composite values assigned to the entire zone (zone polygon), and interpolation of all zone points (zone interpolated all). Examples of EC a data found in Figure 1 1, Example of EC a zon es found in Figure 1 2 Example of 1 ha grid polygon found in Figure 1 3 Known versus predicted nutrient values were compared for the purpose of determining if the efficacy of interpolation was impacted by our sample distribution and density. Known sample values for P, K, Mg, and Ca were compared to the predicted value of the same spatial location after inte rpolation. This predicted value was obtained from the interpolated nutrient rasters by using the ArcMap10 spatial join function with the initial soil test point file. The known and predicted nutrient values were compiled for each field and separated by gri d or zone treatment and analyzed using a t test in JMP Pro 10.
29 Linear and quadratic regression analyses in JMP Pro 10 were used to identify relationships between soil chemical and physical properties and their corresponding EC a and yield values. Soil test variables lacking normality were given a log10 transformation. Regression equations were considered significant when p < 0.05 and coefficient of determination ( R 2 ) was reported. EC a data was evaluated in SPSS using a univariate one way analysis of varia nce with LSD for separation of means. This evaluation identified the presence or lack of statistical difference between EC a values between management zones. Amount of variation between zones may explain success or lack of the management zones ability to id entify variation related measured nutrients. Similarly, mean values for P, K, Ca, and Mg were compared across management zones for each field. Accuracy of the four treatment methods was assessed by comparison to the control raster. The control raster for each field was generated by optimally interpolat ing all available data points. Control and treatments rasters were then standardized by (raster values mean)/raster standard deviation.) These maps were converted to integer maps for their use in the spatial analyst map algebra tool in ArcMap 10. The Map standardized control and standardized treatment raster maps derived from grid or zone sample points for P, K, Mg and Ca. The r esulting difference raster maps were compared based on the percent area that was predicted within 1 standard deviation (1SD) of the intensively sampled control (Berry 2011). Treatment method accuracies were further compared using a 10 x 10 meter subsampl ing method. This was accomplished by first creating the 10 x 10 grid shapefile
30 using the grid index features tool in ArcMap 10. Selected field border was specified as the extent. Each previously generated treatment and control raster for P, K, Ca, and Mg w ere converted to an integer raster; means from each integer raster were extracted using the Spatial Analyst Zonal Zonal Statistics tool in ArcMap 10, specifying the grid file for extraction of means from the raster. Resulting file was converted back to a n integer file and converted to a polygon. Data from attribute table was exported to excel to be analyzed in SAS. Analysis was completed using PROC glimmix with adjustment for location considered by specifying location in random residuals. P and t values were considered from Least Squares Means Dunnett Adjustment for Multiple Comparisons. P < 0 .05 was considered significant and indicated the treatment is statistically alike the control.
31 Table 1 1. Percent soil type by field characterized by USGS soi l survey map Soil Type Field 1 Field 10 Field 14 Field 18 Field 19 Field 20 Field 21 Dothan fine sandy loam 8.50% 32.60% Orangeburg sandy loam 19.20% 3.20% 24.90% 0.60% Red Bay sandy loam 72.20% 57.80% 57.90% 1.20% 30% 54.40% 93% Dotha n fine sandy loam 39.90% 7.20% Tifton sandy loam 2.20% Lucy loamy sand 21.60% 49.90% 35.80% 7% Fuquay loamy sand 9.50% 45% 0.30% Troup loamy sand 20.10% 8.90%
32 Figure 1 1. Example of EC a data points collected by the Veris 3150 sensor cart, field 21 prior to interpolation.
33 Figure 1 2. After interpolation of EC a points, the maps were reclassified into three classes. These classes indicated three management zones consisting of low, medium and high EC a zones. Within these zones points were placed where soil samples were taken to capture variability within each zone.
34 Figure 1 3. Grid samples sites were identified by overlaying a one hectare grid on each field. Soil samples were taken from the mi ddle of each grid. Table 1 2. Average fertilizer and lime application per hectare for all study fields Fertilizer and Lime kg/ha 1 10 14 18 19 20 21 Lime 2501 2224 1406 1098 1937 1341 1489 DAP (18 46 0) 130 118 108 117 Rainbow/Bonanza (5 15 30) 378 396 384
35 CHAPTER 2 RESULTS Field Level Soil Chemical and Physical Properties A nalysis of soil for four fields in 2012 and three fields in 2013 revealed differences in soil nutrient levels, CEC, and texture across fields. Mean soil chemical and p hysical properties for whole fields were derived from averages of all sampling points for each field. Soil test values for P, K, Ca, Mg, and CEC and soil texture were found to differ across fields (Table 2 1 ). Variability in extractable soil nutrient conce ntration, CEC, and texture within and across fields revealed scenarios where site specific management could be beneficial. The IFAS Soil Testing Laboratory has determined critical values for P, K, Ca, and Mg required for peanut production (Adams et al., 19 94). Average concentration of Mehlich 1 extractable P for all sample points in each field ranged from 9 to 66 mg kg 1 Soil test P was 46% greater (p < 0.05) in field 20 compared to fields 14, 18 and 21. Fields 1, 10 and 19 had similar soil test P levels (p > 0.05) and were below critical values for peanut production (30 mg kg 1 Mehlich 1 P) (Adams et al., 1994). Mehlich 1 extractable K concentration ranged from 35 to 105 mg kg 1 for all fields. Soil test values for K were greater (p < 0.05) in fields 1 0 and 14 while fields 19 and 20 had the lowest concentration of extractable K. Fields 1, 18, 19 and 20 were below critical values (60 mg kg 1 Mehlich 1 K) for peanut production (Adams et al., 1994). Concentration of Mehlich 1 extract able Mg was highest in field 10 at 133 mg kg 1 and variability did exist among fields. However, soil test values for Mg ranged from 64 to 133 mg kg 1, all above the critical value (30 mg kg 1 Mehlich 1) for Mg indicating yield response to fertilizer nutrients are unlikely. Sim ilarly, Mehlich 1 extractable Ca values ranged from
36 384 to 670 mg kg 1, all above the critical value (250 mg kg 1 Mehlich 1 Ca) for peanut production. In addition to variability in soil extractable nutrients, soil CEC and texture could influence EC a meas urements and peanut yield. All fields had a mean CEC value below 8 meq 100 g 1, and is common for coarse textured soils. CEC was greatest (p < 0.05) in field 10 (7.19 meq 100 g 1), the lowest (p < 0.05) in field 19 (3.19 meq 100 g 1). Similar to CEC, vari ation in soil texture was observed among fields. Field 10 had greater (p < 0.05) silt and clay content compared to other fields, as was reflected in the measurement of CEC. The coefficient of variation (CV) by field for soil chemical and physical properti es is a good indication of within field variability. Overall, variation in soil nutrient content and CEC within fields was greater than variation of texture within fields. Considerable variability was observed for soil test P concentration, with CV ranging from 26% to 76 47%), Ca (27 76%) and Mg (17 83%). The presence of within field variability of soil nutrient content was expected to provide the necessary environment for evaluating site specific nutrient management practices and comparison of grid and EC a based zone management techniques. Measurement of Soil EC a Spatial measurement of soil EC a was completed using the Veris 3100 for each field and a clustering algorithm was used to divide each field into th ree distinct management zones. Soil EC a values averaged by field ranged from 0.59 to 2.48. In addition, EC a values were greater (p < 0.05) for fields measured in 2012 compared to 2013. Yet, differences in mean EC a values between fields sampled during the s ame year were not detected (p > 0.05).
37 Although EC a values averaged for whole fields did not differ among fields, mean ECa values did differ (p < 0.05) across the three management z ones for each field (Figures 2 1 through 2 7 ). The clustering algorithm us software to delineation zones was effective in separating fields into three statistically distinct EC a based zones. Establishment of three zones based on differences in EC a was necessary to evaluate zone based soil sampling s chemes and to compare the method to alternative approaches. Validation of Sampling and Interpolation Techniques Following delineation of zones, soil was sampled according to grid based or zone based patterns. Whole field averages for soil nutrient concentr ation and texture derived from grid and zone based sampling schemes were compared using a t test. Grid and zone sample nutrient means were significantly different by field (p < 0.05) in field 1 for Mg and Ca, field 20 for K, field 18 for P and field 21 for Ca and Mg. Similar to analysis of mean values from sample points, comparison of known versus interpolated points was performed for grid and zone sampling schemes. No difference in means for any field were detect for soil test values estimated by interpola tion for each sampling scheme (p > 0.05). Similarity between whole field grid and zone sampling schemes for measured and interpolated values was expected to reduce bias of control maps. Relating EC a to Soil Properties Regression analysis identified weak a nd inconsistent linear relationships between EC a and soil nut rie nts, CEC and texture (Table 2 2 .). Non linear relationships were not observed for EC a and any of the measured soil properties. Significant linear relationships (p < 0.05) between soil test P a nd EC a were observed in 4 of the 7 fields. Yet, relationships were weak in 3 of the 4 fields with R 2 < 0.26. Field 10 had the only
38 strong ( R 2 = 0.49) linear relationship between EC a and P. Similar, significant linear relationships (p < 0.05) were observed for EC a and soil test K in 4 of the 7 fields, although the relationships were weak ( R 2 < 0.25). Relationships of EC a with other soil test nutrient values measured were weak ( R 2 < 0.30) and inconsistent across fields. Summary of the regression information i s found in Table 2 2. An example the linear regression of P and EC a is provided in Figure 2 8. A relationship between EC a and soil texture was observed in only 2 of the 7 fields used in the study. Fields 19 and 21 had weak ( R 2 < 0.33) but significant (p < 0.05) linear relationships for sand, silt and clay and EC a Similarly, soil CEC was found to have a significant yet weak ( R 2 = 0.13) relationship to EC a in only 1 of the 7 fields. An example of the linear regression between soil EC a and percent sand is pr ovided in Figure 2 9. Soil Properties and Yield by Zone S oil properties were compared across EC a derived zones for each field. Differences in concentration of soil test nutrients and soil texture by EC a zones (p < 0.1) were found in field 1 for P, field 1 0 for P and Ca, field 14 for P and K, field 21 for P and Mg. No differences were found in P, K, Mg or Ca by zone for fields 18, 19 or 20. Means separation revealed that differences occurred between only two of the three zones for all fields in which differ ences by zone were indicated (Table 2 3 ) In field 1 differences of sample values between two of the three zones existed for texture variables and P, with P values being 28, 20 and 17 mg/kg respectively, all below the critical value for P (30 mg kg 1 ) In f ield 10 variables % sand, % silt, CEC, P and Ca were different between two zones, with P values 27, 12 and 18 respectively, again all below the critical level for peanut production. Ca was 584, 812 and 603 mg kg 1 all zones being above criti cal Ca peanut l evels (250 mg kg 1 ). In field 14 differences between zones were indicated to
39 exist for K, in which zone 1 was different from zones 2 and 3, however zones 2 and 3 are not different. In field 21 differences between zones were indicated for variables: % silt & clay, % sand, % clay, % silt and P in two respective zones, with P values being 56, 43 and 45, all above critical levels (30 mg kg 1). In field 18 no differences were found between soil variables by zone. In field 19 differences between zones were in dicated for: % silt &clay, % sand and CEC in two respective zones. In field 20 no differences were indicated between zones. Yield was measured at each sampling location and yield was compared among and within fields. (Table 2 4 ) Field 19 was the highest yielding, with a field average of 6,757 kg/ha Similar to soil test nutrient values, yield differences were not detected across zones (p >.05), with the exception of fields 19 and 20, in which yield was different (p <.05) in two of three zones. Evaluating Grid and Zone Sampling T echniques An example of the resulting output from the subtraction of standardized control from treatment is provided in Figure 2 10. Results from the subtraction of standardized control rasters from standardized treatment ra ster 5 Maps generated from the contrasting sampling schemes for P predicted 92 to 95% of the field area within 1SD of the control maps. The four methods predicted 91 to 96% of the field area within 1SD of the control for K, 83 to 9 5% for Ca, and 87 93% for Mg. Yet, no difference (p > 0.10) among methods was detected when comparing treatment to control maps for individual nutrients. However, standardization of P, K, Ca, and Mg levels for each field allows observations of all nutrient s to be pooled for comparison of sampling methods. Comparison of methods using pooled standardized nutrient maps revealed dif ferences
40 among methods (Figure 2 11 ). Grid with interpolation and zone with interpolation sampling schemes resulted in greater (p < 0.1 0 ) field area within 1 standard deviation of the control map compared the zone polygon scheme. In the analysis comparing 10 x 10 meter subsamples from each map generated by the contrasting sampling methods and control maps, some differences in samp ling methods were observed. For soil test P, only ZP was similar to the control (p< 0 .0 5) (Table 2 6 ). ZIA and ZP predicted soil test K values similar (p< 0 .05) to the co ntrol. ZIA predicted Ca similar to the Ca control, and GWI predicted Mg similar to the Mg control (p< 0 .05). Means from treatment and control methods are summariz ed in Table 2 7 Besides the aforementioned results, all treatment 10 x 10 meter subsample results differed from the control (p > 0 .05).
41 Table 2 1 Mean and standard e rror from al l available field soil samples, representing the average and variability in our nutrients and texture properties. P, K, Mg and Ca represent Mehlich 1 extractable nutrients. mg kg 1 % meq/100g Field P K Mg Ca Sand Silt Clay CEC Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE Mean SE 1 24 1.8 52 1.7 74 4.2 585 21 80 .6 11 .6 10 .3 6 .2 10 20 2.7 105 5.4 66 4.3 670 39 63 1.4 37 1.7 20 1.7 7 .4 14 44 4.2 89 5.1 78 5.5 597 35 82 .7 11 1.2 7 .6 6 .3 1 8 40 3.4 51 2.3 73 6.8 552 26 82 .6 13 .8 6 .3 5 .2 19 23 2.5 35 2.3 83 9.5 423 19 85 2.3 11 .7 7 .4 3 .3 20 27 6.7 47 1.8 127 9.5 482 22 84 .5 10 .8 6 .5 4 .2 21 49 2.2 67 2.5 80 6.1 384 55 81 .6 11 .9 8 .5 6 .2
42 Figure 2 1 EC a m eans by EC a z one s f ield 1 variation show using a 95% confidence i nterval.
43 Figure 2 2 EC a m eans by EC a z ones for f ield 10 variation shown using a 95% c onfidence i nterval.
44 Figure 2 3 EC a m eans by EC a z one s for f ield 14 variability shown using a 95% confidence i nterval.
45 Figure 2 4 EC a m eans by EC a z one s for f ield 18 variability shown using a 95% confidence i nterval.
46 Figure 2 5 EC a m eans by EC a z one s for f ield 19 variability shown using a 95% confidence i nterval.
47 Figure 2 6 EC a m eans by EC a z o ne s for f ield 20 variability shown using a 95% confidence i nterval.
48 Figure 2 7 EC a m eans by EC a z one s for f ield 21 variability shown using a 95% confidence i nterval.
49 Table 2 2 Summary of linear regression of soil properties and yield to soil EC a and yield (p < 0.05). Field 1 10 14 18 19 20 21 P roperty EC a Yield EC a Yield EC a Yield EC a Yield EC a Yield EC a Yield EC a Yield P 0.23 0.49 0.16 0.26 0.16 K 0.15 0.25 0.16 0.12 Mg 0.18 0.28 0.15 0.27 0.16 Ca 0.17 0.23 CEC 0.22 0.25 0.13 Yield 0.1 7 EC a 0.17 pH 0.3 0.12 %silt 0.21 0.33 0.15 %clay 0.15 0.2 0.17 %sand 0.21 0.25 0.11 0.28 0.13 %siltclay 0.24 0.25 0.14 0.28 0.13 %bsMg 0.34 0.18 % bsCa 0.16 0.23 %bsH 0.3 0.18 %bsK 0.43 0.16 0.36 0.18 0.32
50 Figure 2 8 Field 19 l inear regression example of soil test phosphorous and soil EC a P < 0.05. Figure 2 9 Field 20 l inear regression exampl e of percent sand and soil EC a P < 0.05.
51 Table 2 3 Summary of average P, K, Ca, Mg by EC a Zone and whole field with SE by Field mg Kg 1 Zone 1 Zone 2 Zone 3 Whole Field P 1 Average 28.25 a 20.25 b 17.00 b 24 SE 2.36 2.28 2.9 1.79 10 Average 26.88 a 12.36 b 18.00 ab 19.44 SE 4.06 1.82 6.66 2.67 14 Average 61.62 a 33.64 b 31.86 b 44.97 SE 6.78 3.72 4.85 4.23 21 Average 56.06 a 45.43 b 44.83 b 48 SE 5.1 8 1.95 2.65 1.97 18 Average 27.85 a 48.07 a 46.64 a 40.85 SE 4.11 10.62 9.22 5.87 19 Average 21.70 a 26.26 a 17.18 a 22.63 SE 3.48 4.32 5.39 2.53 20 Average 43.62 a 72.05 a 75.62 a 66.58 SE 3.76 10.51 14.13 6.67 K 1 Average 50.5 55.06 51.33 52.2 SE 2.39 2.05 5.27 1.65 10 Average 106.83 b 114.86 a 87.19 a 104.61 SE 6.67 10.61 7.81 5.36 14 Average 68.23 b 105.82 a 98.00 a 88.29 SE 8.12 6 4.01 5.12 21 Average 57.63 a 71.00 a 69.78 a 67.19 SE 5.78 3.81 3.37 2.68 18 Average 38.04 a 37.46 a 27.12 a 40.857 SE 4.31 3.47 2.82 3.195
52 Table 2 3 Continued Field mg Kg 1 Zone 1 Zone 2 Zone 3 Whole Field P 1 Average 28.25 a 20.25 b 17.00 b 24 SE 2.36 2.28 2. 9 1.79 10 Average 26.88 a 12.36 b 18.00 ab 19.44 SE 4.06 1.82 6.66 2.67 14 Average 61.62 a 33.64 b 31.86 b 44.97 SE 6.78 3.72 4.85 4.23 21 Average 56.06 a 45.43 b 44.83 b 48 SE 5.18 1.95 2.65 1.97 18 Average 27.85 a 48.07 a 46.64 a 40.85 SE 4.11 10.62 9.22 5.87 19 Average 21.70 a 26.26 a 17.18 a 22.63 SE 3.48 4.32 5.39 2.53 20 Average 43.62 a 72.05 a 75.62 a 66.58 SE 3.76 10.51 14.13 6.67 19 Average 38.04 37.47 27.12 35. 3 SE 4.32 3.48 2.82 2.27 20 Average 45.25 44.75 53.188 47.03 SE 3.31 2.66 3.96 1.82 Mg 1 Average 587.21 a 588.81 a 569.17 a 585.41 SE 32.47 29.25 32.74 20.26 10 Average 118.25 a 165.50 a 112.81 a 133.61 SE 16.84 17. 93 7.04 10.22 14 Average 116.38 a 89.45 a 89.50 a 100.76 SE 15.52 7.4 7.58 7.59 21 Average 75.18 a 55.54 a 75.83 a 66.5 SE 12.06 12.05 16.43 8.08 18 Average 89.71 a 63.14 a 91.64 a 81.5 SE 20.51 5.92 15.76 8.88
53 Table 2 3 Continued Field mg Kg 1 Zone 1 Zone 2 Zone 3 Whole Field Mg 19 Average 77.41 a 79.26 a 83.00 a 79.48 SE 8.87 7.4 11.04 4.92 20 Average 68.68 a 59.66 a 69.5 a 63.97 SE 7.44 a 5.65 a 9.60 a 3.87 Ca 1 Average 74.58 a 72.69 a 78.50 a 74.43 SE 6.88 5.33 5.76 4.13 10 Average 583.83 b 811.86 a 602.94 b 669.68 SE 59.94 63.14 33.01 38.39 14 Average 660.96 a 534.18 a 557.93 a 592.71 SE 73.54 34.24 34.35 35.65 21 Average 509.00 a 360.84 a 455.35 a 426.51 SE 99.19 86.68 101.46 56.42 18 Average 597.07 a 488.42 a 580.00 a 555.16 SE 91.88 29.53 55.04 36.718 19 Average 412.91 a 438.3 a 410.31 a 423 SE 25.57 33.55 39.66 18.72 20 Average 434.75 a 461.91 a 570.43 a 482.26 SE 39.13 24.53 65.14 21.92 EC a 1 Average 1.26 a 1.82 b 2.47 c 1.45 SE 0.01 0.03 0.08 0.02 10 Average 1.49 a 2.38 b 1.49 c 2.52 SE 0.04 0.06 0.03 0.05 14 Average 0.69 a 1.32 b 2.01 c 1.2 SE 0.01 0.01 0.03 0.02 21 Average 0.58 a 0.81 b 1.12 c 0.83 SE 0.01 0.01 0.02 0.01 18 Average 0.51 a 0.88 b 1.4 c 0.77 SE 0.01 0.01 0.03 0.01
54 Table 2 3 Continued Field mg Kg 1 Zone 1 Zone 2 Zone 3 Whole Field EC a 19 Average 0.44 a 0.63 b 0.98 c 0 .64 SE 0 0 0.02 0.01 20 Average 0.41 a 0.56 b 0.84 c 0.59 SE 0.01 0 0.01 0.01 Table 2 4. Average y ield ( kg/ha ) for all study fields, separated by zone and summarized by whole field mean for each. Yield (kg/ha) Field Zone 1 Zone 2 Zone 3 Whole Field 1 Average Yield 2571 2646 2447 2586 SE 113 75 111 71 10 Average Yield 4271 4040 4490 4229 SE 133 85 191 81 14 Average Yield 4326 4539 4393 4434 SE 168 187 175 107 21 Average Yield 4265 3947 3764 3976 SE 175 225 290 143 18 Average Yield 6129 5681 6413 6064 SE 687 467 437 240 19 Average Yield 7479 6242 6862 6757 SE 155 190 489 177 20 Average Yield 5830 6833 5926 6231 SE 255 368 282 222
55 Figure 2 10 Example of a final diff erence raster from the standardized control minus treatment method for comparing our four sample schemes
56 Table 2 5 The percent (%) of field area predicted within 1 standard deviation of the control by field and nutrient. Mehlich 1 extractable phosphoru s (P), potassium (K), calcium (Ca), and magnesium (Mg). Grid no interpolation (GNI), grid with interpolation (GWI), zone polygon (ZP) and zone interpolated all. Nutrient Treatment Field 1 Field 10 Field 14 Field 21 Field 18 Field 19 Field 20 Mean P G NI 85 100 99 96 90 96 100 95 GWI 95 100 99 95 80 95 100 94 ZIA 90 90 97 98 92 100 100 95 ZP 90 100 96 99 71 88 100 92 K GNI 91 94 84 100 100 100 91 94 GWI 96 98 99 97 100 100 88 96 ZIA 88 78 96 96 100 100 90 92 ZP 88 91 97 100 80 90 93 91 Ca GNI 64 97 94 64 93 86 99 85 GWI 100 99 94 100 87 86 99 95 ZIA 85 96 93 85 88 100 96 91 ZP 88 38 94 88 86 95 91 83 Mg GNI 65 96 97 94 86 85 84 87 GWI 76 99 98 89 98 88 100 93 ZIA 63 94 99 91 100 100 99 92 ZP 99 81 95 94 87 90 93 91
57 Figure 2 11 Average f ield area p redicted within one standard deviation of the c ontrol (%) by our four treatment methods, GNI, GWI, ZIA and ZP.
58 Table 2 6 Results from the subsample least squares mean comparisons. Treatment occurrence in table indicates it was similar to the control dataset (p<0.05). Field 1 10 14 21 18 19 20 Phosphorous ZIA ZP GWI ZP ZP Potassium ZIA ZP Calcium ZIA, GWI ZIA ZP, GNI Magnesium GWI GNI, GWI Table 2 7 Average value from 10 x 10 meter subsample procedure. Average is provided by treatment for P, K, Mg, Ca. Field Treatment P K Mg Ca 1 GNI 25 53 71 592 GWI 25 51 75 576 ZIA 26 59 70 559 ZP 30 59 68 594 Control 26 49 72 582 Mean 26 54 71 586 10 GNI 27 110 150 725 GWI 22 107 158 748 ZIA 8 85 109 547 ZP 16 92 96 482 Control 16 98 139 104 Mean 18 98 130 521 14 GNI 47 572 116 572 GWI 46 658 113 658 ZIA 53 578 97 578
59 Table 2 7 Continued Field Treatment P K Mg Ca 14 ZP 56 576 99 576 Control 46 604 108 604 Me an 50 87 107 598 21 GNI 50 69 172 588 GWI 46 69 85 556 ZIA 47 62 92 631 ZP 49 68 98 597 Control 47 47 56 380 Mean 48 67 106 550 18 GNI 26 54 73 500 GWI 25 54 79 497 ZIA 42 50 74 566 ZP 38 54 76 368 Control 39 49 82 549 Mean 34 52 76 496 19 GNI 21 34 34 410 GWI 20 36 77 420 ZIA 24 36 79 423 ZP 23 36 75 417 Control 23 36 69 413
60 Table 2 7 Continued Field Tre atment P K Mg Ca 19 Mean 22 36 69 417 20 GNI 83 52 67 511 GWI 66 52 66 510 ZIA 56 43 62 466 ZP 54 54 52 254 Control 67 67 63 478 Mean 69 50 63 444
61 CHAPTER 3 DISCUSSION AND CONCLUSION Discussion Analy sis of soil revealed variation of soil nutrient content, CEC, and texture within and among fields. This suggests that site specific assessment of soil and application of nutrients could be beneficial for the soils evaluated in the study. Moreover, fields w ith soil nutrient content below critical values for peanut production provides an opportunity to observed spatial differences in yield response. Measurement of EC a and delineation of zones that are agronomically significant is essential for site specific management of nutrients. The methods employed to delineate zones according to EC a values did separate fields into three statistically distinct zones. Yet, the agronomic significance was uncertain. EC a values measured in the seven fields were lower than val ues reported for other soil types. Examples include ECa ranges of 4.2 to 22.7 mS m 1 and 7.3 to 40.2 mS m 1 in two Missouri fields with claypan ( Kitchen et al ., 2005). Mollisol soils studied in Iowa by Brevik et al (2006) contained EC a ranges of 20 to 70 m S m 1. At study sights in Colorado, within semi arid great plains containing Platner and Rago loam soils, Johnson et al (2001) found EC a ranges between 0 and 78 mS m 1 Factors influencing ECa values can differ among soil types and environments. In our Northwest Florida study area with soil types being predominately Red Bay, Lucy and Dothan sandy loams or Fuquay loamy sand, EC a values ranged between 0.59 and 2.48 mS m 1. Williams and Baker (1982) found that in non saline soils, bulk soil conductivity var iability is mostly attributed to soil moisture, texture, and CEC. Variability of these factors are influenced by a combination of soil physical and chemical properties,
62 including but not limited to; bulk density, clay mineralogy and content, organic matter and soil temperature (Corwin and Lesch, 2005; Rysan and Sarec, 2008). In general, sands have low bulk soil conductivity and will lower measured values of EC a Silts have medium bulk soil conductivity and clays have higher bulk soil conductivity (Lund et al., 1999). Because of within field variability of soil properties, correlation between EC a and specific soil properties may vary from field to field (Corwin and Lesch, 2003). Regarding EC a relationships to soil nutrients, Johnson et al (2003) found P to b e negatively correlated with EC a and clay content to be positively correlated with EC a in Platner, Weld, and Rago loam soils located in northwest Colorado. Our regression analysis determined soil nutrients in soil did not provide consistent relationships to EC a In addition, texture and CEC variation were not influencing our EC a measurements. Therefore, soil moisture or variations in surface cover of the soil may have influenced EC a variation in our fields. In comparison of soil nutrient means by zone, fo ur of seven fields had soil nutrient values that were different by zone (p< 0 .05), however in every case the difference was between only two of the three zones. Other studies have found significant relationship between soil nutrients and EC a delineated zone s, including Johnson et al (2003) in which extractable P was related to EC a zones (p< 0 .10). Their study area was located in northwestern Colorado, with Platner, Weld, and Rago loam soils. Fleming et al (2004) examined the relationship between soil P and K between EC a delineated zones in two field in northeastern Colorado, finding relationships between K in two of three zones in one field, and in all three zones in their other study field (p< 0 .05). No difference was found between P in field one, and differ ences in two of
63 three zones was found in field two (p< 0. 05). Our nutrients by zone results are similar to that of Fleming et al (2004), in that we identified relationships between nutrients by two of our three EC a zones. In the current project, the numb er of zones was predetermined to reflect the sampling design commonly practiced in the region. Other reports have used from three and up to eight zones. However, three zones did not delineate three areas within the field that had significantly different le vels of soil test nutrients or nutrient requirement. Furthermore, no difference in yield was detected across the three zones in any field. This suggests three zones may not have been appropriate for the fields evaluated in our study. The use of a software program, like Management Zone Analyst (MZA), would be helpful in selecting the optimal number of zones for your respective dataset (Fridgen et al ., 2004). ound no difference (p > 0.10) among methods when comparing treatment to control maps for individual nutrients. Standardization of P, K, Ca, and Mg levels for each field allows observations of all nutrients to be pooled for comparison of sampling methods, generating our conclusion that grid with interpolation and zone with interpolation sampling schemes resulted in greater (p < 0.1 0 ) field area within 1 standard deviation of the control map compared the zone polygon scheme. Fleming et al (2004) evaluated an EC a delineated zone sampling scheme in two study fields by first standardizing collected field EC a data to a mean of zero and standard deviation of one, and thereafter creating three management zones in both study fields and sampling on a 76 meter grid. Samples were measuring for soil variables including P and K.
64 Relationships between zones and soil variables were evaluated using ANOVA (p< 0 .05), finding relationships between zones and soil variables including P in one of two fields, in two of three zones K was related to zones in both fields, in two of three zones and in all three zones (p< 0 .05). Fleming et al (2004) compared their EC a zone relationships to soil variables to the relationship between zones created by soil color, and found that EC a deline ated zones to be more effective at identifying homogeneous in field management zones. Fleming et al (2004) compared 76 meter grid sampling to EC a delineated zone sampling. ANOVA was used to determine relationship between grid and zone soil variables and logistic regression was used to examine how well the zone sample characterization of the soil within each EC a zone predicted the grid sample data in each EC a zone. Performance of the zone was evaluated by the percent of grid samples accurately predicted by the zone sample characterization. Fleming et al ., (2004) found varying differences between grid and zone delineated soil samples by field. 82 to 83% of the grid samples not in the defined 10 meter transition border zone were accurately predicted by EC a zo nes. Fleming et al based soil sampling could provide agronomically useful soil information as compared to grid soil sampling, without having the need to acquire a large number of soil samples as is the case of grid soil sampli ng which is time consuming, labor intensive, and cost
65 Conclusion Corwin & Lesch (2005) note that EC a measurements are complex in the sense of what soil properties are influencing their values, and whether the influencing p roperties are associated to field subareas that may be managed differently for production improvement. Implementation of a fertility program using EC a delineated zones that are not related to agronomically relevant soil factors may result in improper ferti lization and misallocation of resources. From our results we conclude that while EC a variation existed within our fields, this variation was not related to useful fertility management zones. In other words, the variation in EC a for the fields in our study had no relation to soil factors of agronomic importance. Additionally analysis of treatment methods found the grid with interpolation method to perform the same as the zone with interpolation method (p< 0 .05). From our results we conclude that while EC a zo ne soil sampling for productivity management has proved appropriate in other locations with different landscapes and s oil type, (Johnson et al., 2001; Kaffka et al., 2005; Kitchen et al. 2005., Sudduth et al. 2005.) zone soil sampling using EC a measureme nts is not an improvement upon the traditional one hectare grid soil sampling in Northwest Florida.
66 LIST OF REFERENCES AgJunction. Copyright 2010. Available at: http://www .agjunction.com Date accessed: 5/25/2012. Adams, J.F., F.E. Fox, J.G. Davis, G.J Gascho, S.C.Hodges, G. Kidder, C.C. Mitchell and F.M. Rhodes. 1994. Research Based Soil Testing Interpretation And Fertilizer Recommendations For Peanuts On Coastal Plain Soils, in: C. C. Mitchell (Ed.), Southern Cooperative Bulletin. Berry, J.K. 2011. Topic 18: Understanding grid based data. Beyond Mapping III. p. 1 40. Blackmore, S. 2000. The interpretation of trends from multiple yield maps. Comput. Electron. Agr ic 26(1):37 51. Brevik, E.C., Fenton, T.E. and Lazari, A. 2006. Soil electrical conducti vity as a function of soil water content and implications for soil mapping. Precision Agric. 7:393 404. Carter, L.M., J.D. Rhoades, and J.H. Chesson. 1993. Mechanization of soil salinity assessment for mapping. ASAE Paper no. 931557. Proc. 1993 ASAE Winte r Meet., Chicago, IL. 12 17 Dec. 1993. ASAE, St. Joseph, MI. Chils, J. P., and P. Delfiner. 1999 Modeling spatial uncertainty. Geostatistics, Wiley series in probability and statistics. Corwin, D. L., S.M. Lesch P.J. Shouse, R. S oppe, and J.E. Ayars. 2 003 Identifying soil properties that influence cotton yield using soil sampling directed by apparent soil electrical conductivity. Agronomy Journal, 95(2), 352 364. Corwin, D.L. and S.M. Lesch. 2005. Apparent soil electrical conductivity measurements in a griculture. Comput. Electron. Agr ic 46: 11 43. Fleming, K.L., and D.G. Westfall. 2000. Evaluating Farmer Defined Management Zone Maps for Variable Rate Fertilizer Application. Precision Agriculture, 2: 210 215. Fleming, K.L., D.G. Westfall, and W.C. B ausch. 2001. Evaluating management zone technology and grid soil sampling for variable rate nitrogen application. In Precision Agriculture [CD ROM]. Proc. Int. Conf., 5th, Minneapolis, MN. 16 19. July 2000. ASA, CSSA, and SSSA, Madison, WI. Fleming, K.L. D.F. Heermann, and D.G. Westfall. 2004. Evaluating Soil Color with Farmer Input and Apparent Soil Electrical Conductivity for Management Zone Delineation. Agron. J. 96:1581 1587.
67 Flowers, M., R. Weisz, and J.G. White. 2005. Yield Based Management Zones and Grid Sampling Strategies: Describing Soil Test and Nutrient Variability. Agron. J. 97: 968 982. Fridgen, J.J., N.R. Kitchen, K.A. Sudduth, S.T. Drummond, W.J. Wiebold and W. Fraisse. 2004. Management Zone Analyst (MZA): Software for Subfield Managem ent Zone Delineation. Agron. J. 96:100 108. Fraisse, C.W., K.A. Sudduth, N.R. Kitchen. 2001. Delineation of site specific management zones by unsupervised classification of topographic attributes and soil electrical conductivity. Trans. ASAE, 44(1):155 166 Franzen, D.W., and T.R. Peck. 1995. Field Soil Sampling Density for Variable Rate Fertilization. Journal of Production Agriculture. Vol. 8. No. 4. Friedman, S.P. 2005. Soil properties influencing apparent ele ctrical conductivity: a review. Comput. Ele ctron. Ag ric 46.1: 45 70. Halvor son, A.D., and J. D. Rhoades. 1976 Field mapping soil conductivity to delineate dryland saline seeps with four electrode technique. Soil science s ociety of America journal 40.4 : 571 575. Heiniger, R.W., R. G. McBride, and Dav id E. Clay. 2003. Using soil electrical conductivity to improve nutrient managemen t. Agronomy Journal 95.3 : 508 519. Johnson, C.K., D.A. Mortensen, B.J. Wienhold, J.F. Shanahan, and J.W. Doran. 2003. Site Specific Management Zones Based on Soil Electrical Conductivity in a Semiarid Cropping System. Agron. J. 95:303 315. Johnson, C.K., J.W. Doran, H.R. Duke, B.J. Weinhold, K.M. Eskridge, and J.F. Shanahan. 2001. Field scale electrical conductivity mapping for delineating soil condition. Soil Sci. Soc. Am. J 65: 1829 1837. Kaffka, S.R., S.M. Lesch, K.M. Bali, D.L. Corwin. 2005. Site specific management in salt affected sugar beet fields using electromagnetic induction. Comp. Electron. Agr ic ., 46 (2005), pp. 329 350 Khosla, R. and M.M. Alley. 1999. Soil Sp ecific Nitrogen Management on Mid Atlantic Coastal Plain Soils. Better Crops, 83 3. Kitchen, N.R., K.A. Sudduth, D.B. Myers, S.T. Drummond, S.Y. Hong. 2005. Delineating productivity zones on claypan soil fields using apparent soil electrical conductivity. Comp. Electron. Agr ic 46, 285 308. Lesch, S.M., J.D. Rhoades, and D.L. Corwin. 2000. The ESAP Version 2.01r user manual and tutorial guide. Research Report #146. George E. Brown Jr., Salinity Laboratory, Riverside, CA, 153pp.
68 Li, Y., Z. Shi, C.F. Wu, H. Y. Li, and F. Li. 2008. Determination of potential management zones from soil electrical conductivity, yield and crop data. Journal of Zhejiang University Science, 9:68 76. Lund, E.D., C.D. Christy, and P.E. Drummond. 1999 Practical applications of soil e lectrical conductivity mapping. Precision agriculture, 99, 771 779. Mallarino, A.P., and D.J. Wittry. 2004. Efficacy of Grid and Zone Soil Sampling Approaches for Site Specific Assessment of Phosphorus, Potassium, pH, and Organic Matter. Prec. Ag. 5: 131 144. McCoy, J. and K. Johnston. 2001. Using ArcGIS spatial analyst. Redlands, CA, USA: Environmental Systems Research Institute, Inc. McNeill, J.D. 1992. Rapid, accurate mapping of soil salinity by electromagnetic ground conductivity meters. Advances in measurement of soil physical properties: Bringing theory into practice. 209 229. Nanna, T., and D.W. Franzen. 2003. A Weighted Classification Method for Nitrogen Zone Delineation. North Central Extension Industry Soil Fertility Conference. Vol. 19. Des M oines, IA. 2010 Geographic information systems Reyniers, M., K. Maertens, E. Vrindts, J. de Baerdemaeker. 2006. Yield variability related to landscape properties of a loamy soil in central Belgium. Soil Tillage Res., 88 (1 2): 262 273. Rhoades, J. D., and J.V an Schilfgaarde. 1976. An electrical conductivity probe for determining soil salinity." Soil Science Society of America Journal 40.5 (1976): 647 651. Rhoades, J. D., P. J Shouse, W.J. Alves, N.A. Mantegh and S.M. Lesch. 1990 Determining soil salinity from soil electrical conductivity using differe nt models and estimates. Soil Sci. Soc. Am. J., 54: 46 54. Robins on, T.P., and Metternicht. 2006 Testing the performance of spatial interpolation techniques for mapping soil p roperties. Comput. Electron. Agric. 50 .2 : 97 108. Rysan, L., and O. Sarec. 2008. Research of Correlation between electric soil conductivity and yield based on the use of GPS technology. Res. AGR. ENG., 54(3): 136 147. Sadler, J.E., W.J. Busscher, P.J. Bauer, and D.L. Karlen. 1998. Spatial Scale Requirements for Precision Farming: A Case Study in the Southeaster USA. Agron. J. 90: 191 197.
6 9 Sawyer, J.E. 1994. Concepts of variable rate technology with considerations for fertilizer application. Journal of P roduction Agriculture 7.2 195 201. Sudduth, K. A., N.R. Kitchen, G.A. Bollero, D.G. Bullock and W.J. Wiebold. 2003 Comparison of electromagnetic induction and direct sensing of soil electrical conductivity. Agron. J. 95: 472 482. Sudduth, K.A., N.R. Kitche n, W.J. Wiebold, W.D. Batchelor, G.A. Bollero, D.G. Bullock, D.E. Clay, H.L. Palm, F.J. Pierce, R.T. Schuler, K.D. Thelen. 2005. Relating apparent electrical conductivity to soil properties across the North Central USA Comp. Electron. Agric., 46 pp. 263 28 3 Taylor, J. C., G.A. Wood, R. Earl and R.J. Godwin. 2003 Soil factors and their influence on within field crop variability, part II: spatial analysis and determination of management zones. Biosyst. Eng. 84: 441 453. V an Schilfgaarde, J. 1999. Is precision agriculture sustainable? Am. J. Altern. Agric. 14(1):43 46. Voltz, M. and R. Webster. 1990. A Comparison of kriging, cubic splines and classification for predicting soil properties from sample information. J. Soil Sci. 41: 473 490. Vrindts, E., A.M. Moua zen, M. Reyniers, K. Maertens, M.R. Maleki H. Ramon, J. de Baerdemaeker. 2005. Management zones based on correlation between soil compaction, yie ld and crop data. Biosyst. Eng. 92(4): 419 428. Williams, B.G. and G.C. Baker. 1982. An electromagnetic induction technique for reconnaissance su rveys of soil salinity hazards. Soil Research 20.2 (1982): 107 118.
70 BIOGRAPHICAL SKETCH Ashley Mason is from Canton, Illinois. Her relevant work experience includes two years employment with a s oil testing and consultation company in Illinois. She received her Western Illinois University. At Western Illinois University Ashley was active in the National FFA. Part of ogram at the University of Florida was spent at the West Florida Research and Education Center. Her degree program at the University of Florida consisted of a Master of Science in agronomy and an interdisciplinary concentration in geographic information sy stems
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EXFPX2INT_TIGS57 INGEST_TIME 2014-05-13T20:08:04Z PACKAGE UFE0046349_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC