|Table of Contents|
Table of Contents
Chapter 1. Introduction
Chapter 2. Literature review
Chapter 3. Soil spatial variability of selected chemical properties in the Everglades agricultural area: I. Semi-variograms
Chapter 4. Soil spatial variability of selected chemical properties in the Everglades agricultural area: II. Block Kriging
Chapter 5. Regional variation of selected soil chemical properties in the Everglades agricultural area
Chapter 6. Nitrogen and phosphorus mineralization in organic soils of the Everglades agricultural area
Chapter 7. Summary and conclusions
Appendix A. Summary statistics of experimental soils
Appendix B. Isotropic and anisotropic semi-variograms
Appendix C. Regional contour maps
INFLUENCE OF SOIL SPATIAL VARIABILITY AND SOIL-WATER
CONDITIONS ON SELECTED HISTOSOLS IN THE
EVERGLADES AGRICULTURAL AREA
ORLANDO ANTONIO DIAZ
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1990
I wish to express my sincere appreciation to Dr. Edward A. Hanlon and Dr. David L. Anderson, chairman and cochairman of my graduate committee, respectively, for their full support, constant encouragement, guidance, and friendship during the realization of this study. I owe to them a great debt of gratitude for their support in my field and laboratory studies, as well as their editorial review of this dissertation. I am fortunate to have studied under their guidance and I am appreciative of their personal generosity and understanding.
I am grateful to Dr. T.L. Yuan for his continuous moral support and encouragement to finish my dissertation. As a member of my graduate committee, he gave me constructive guidance and criticism to my laboratory experiments. I would like to thank the other members of my graduate committee, Dr. M.E. Collins and Dr. K.M. Porter for their suggestions, support, and editorial comments. A special note of gratitude is extended to Dr. W.G. Blue for substituting Dr. T.L. Yuan in the final examination.
During the course of this project I have been fortunate to receive a great deal of assistance. Therefore, I would like to extend my sincere appreciation to Mr. Modesto Ulloa, ii
Mr. Robert Underbrink, and Mr. Apelgren for all their support and interest in this project. A special note of gratitude is extended to the Sugar Cane League for their support and advice in the early stages of the project. I would also like to thank Mr. Stephen Cox and Mr. Sam McCollum from the Soil Conservation Service, Greenacres, FL, for their assistance in soil classification of experimental sites.
I also wish to thank Dr. K.R. Reddy and Dr. D.A. Graetz from the Wetland Soils Research Laboratory for their assistance in the incubation study. A particular note of appreciation is extended to Mr. John R. Richardson from the Cooperative Fish and Wildlife Research Unit, for his valuable assistance in the mapping procedures of the Everglades Agricultural Area.
I also want to thank the staff of the Analytical
Research Laboratory of the University of Florida and Soil Testing Laboratory in Belle Glade for their help in some soil chemical analyses.
Special gratitude is given to Mr. Jorge Gonzalez for his friendship and cooperation throughout the program and laboratory work. Special thanks are also given to the Tergas family for their friendship and encouragement. Gratitude is extended to my friends Mr. S.K. Patel, Mrs. Yu Wang, and Miss J. Eastridge for their friendship and help in certain topics of the research.
Finally, I wish to express the deepest gratitude to my family in El Salvador and California for their encouragement, support, and patience.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ........ ................... ii
ABSTRACT ......... ...................... vii
INTRODUCTION ........... ..................... 1
LITERATURE REVIEW 4.. .........4
The Everglades Agricultural Area 4.........4
Geology ........... ................... 5
Formation Processes ....... ............. 7
Soils in the EAA ........ .............. 8
Organic Soil Subsidence ... ........... 12
Soil Variability ...... ................. 14
Geostatistics ....... .................. 19
Semi-variograms ..... ............... 21
Kriging Statistics .... ............. 28
SOIL SPATIAL VARIABILITY OF SELECTED CHEMICAL PROPERTIES IN THE EVERGLADES AGRICULTURAL AREA. I. SEMI-VARIOGRAMS 35
Introduction ........ ................... 35
Materials and Methods .............. 38
Collection and Preparation of Soil Samples. 38
Laboratory Analysis .... ............. 42
Statistical Analysis .... ............ 43
Results and Discussion .......................... 45
Soil Series Variation .... ........... 45
Semi-variogram Analysis ... .......... 48
Conclusions ....... .................. 68
SOIL SPATIAL VARIABILITY OF SELECTED CHEMICAL PROPERTIES IN THE EVERGLADES AGRICULTURAL AREA. II. BLOCK KRIGING 72
Introduction ........ ................... 72
Materials and Methods ..... .............. 74
Statistical Analyses .... ............ 74
Geostatistical Analyses ... ........... 75
Results and Discussion ..... .............. 77
Spatial Analysis ........................ 77
Block-kriging and Mapping .... ..... 78
Soil Sampling . . . . . . . . 96
Conclusions . . . . . . . . . 103
REGIONAL VARIATION OF SELECTED SOIL CHEMICAL PROPERTIES IN THE EVERGLADES AGRICULTURAL AREA . 106
Introduction . . . . . . . . . 106
Materials and Methods . . . . . . 108
Data Collection and Laboratory Analysis . 108 Statistical and Geostatistical Analyses . 109 Mapping . : . . . . . . . 110
Results and Discussion . . . . . . 112
Spatial Analysis . . . . . . 112
Regional Variation . . . . . . 120
Conclusions . . . . . . . . . 135
NITROGEN AND PHOSPHORUS MINERALIZATION IN ORGANIC SOILS OF THE EVERGLADES AGRICULTURAL AREA . . . 138
Introduction . . . . . . . . . 138
Materials and Methods . . . . * 141
Collection of Soils and Preparation of
Columns . . . . . . . 141
Incubation and Leaching Procedures . . 144 Laboratory Analysis . . . . . . 146
Statistical Analyses . . . . . 148
Results and Discussion . . . . . . 148
Nitrogen . . . . . . . . 148
Phosphorus . . . . . . . . 158
conclusions . . . . . . . . . 166
SUMMARY AND CONCLUSIONS . . . . . . . 170
Introduction . . . . . . . . . 170
semi-variograms . . . . . . . . 171
Block Kriging . . . . . . . . 172
Regional variability . . . . . . . 173
Nitrogen and Phosphorus Release . . . . 174
SUMMARY STATISTICS OF EXPERIMENTAL SOILS . . . 178
ISOTROPIC AND ANISOTROPIC SEMI-VARIOGRAMS . . . 183
REGIONAL CONTOUR MAPS . . . . . . . . 190
REFERENCES . . . . . . . . . . 193
BIOGRAPHICAL SKETCH . . . . . . . . 206
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INFLUENCE OF SOIL SPATIAL VARIABILITY AND SOIL-WATER
CONDITIONS ON SELECTED HISTOSOLS IN THE
EVERGLADES AGRICULTURAL AREA By
Orlando Antonio Diaz
Chairman: Dr. Edward A. Hanlon Cochairman: Dr. David L. Anderson Major Department: Soil Science
Chemical properties of Histosols in the Everglades Agricultural Area (EAA) are affected by the inherent variability, management practices, and soil-water conditions. Soil spatial variability is a limiting factor affecting the reliability of predictions concerning soil properties, soil behavior, and land-use performance. Recognition of the importance of spatial variability on land use has led to the study of soil heterogeneity in greater detail. This study was conducted to evaluate the influence of field and regional soil spatial variability and soil water conditions on Histosols in the EAA. The structure of spatial dependence of selected soil chemical properties was studied by the use of semi-variograms. Block kriging was used to map field variability of four major soil series and
the regional variability of important soil chemical properties across the EAA. The effect of intermittent flooding and drained conditions on N and P release from surface soils were determined by column leaching studies.
Results showed that selected soil chemical properties in the EAA are spatially dependent. Both road and ditch spoils greatly alter soil chemical properties, although anisotropic semi-variograms showed that road spoils have a greater influence on soil variability. The range of spatial dependence of the majority of the soil properties was > 100 m at all locations.
Kriged field contour maps indicated that an area of
approximately 40 to 50 m from the road and 25 to 30 m from each side of the ditch should be avoided during soil sampling. Soil variability among fields was large as judged by the number of samples required to produce repeatable test values.
Kriged regional contour maps indicated that soil
variability in the EAA is largely a result of soil type and management. Estimation variance maps indicated areas where further sampling will provide additional information to improve prediction confidence levels.
In a related laboratory experiment, the amount of total N released under drained conditions was approximately 60% higher than that released under intermittent flooding conditions. However, the amount of total P released under
intermittent flooding was two to six times higher than total P released under drained conditions.
The Everglades Agricultural Area (EAA) contains one of the richest agricultural regions in the United States. Revenues from the EAA and adjacent areas stimulate over one billion dollars per year into the economy of South Florida. The EAA produces a large share of the sugarcane and winter vegetable consumed in this country. In addition, this region supports a sizable area of sod, rice, citrus, and cattle production.
The EAA is a relatively flat area with most soils
containing 85% or more of organic matter by weight. The organic soils from this area have been extensively studied; however, there is little information regarding their nutrient distribution and spatial variability.
Soil variability is the product of soil-forming factors functioning and interacting over a continuum of spatial and temporal scales. Natural processes, such as climate, mineral and organic parent materials, flooding, and weathering, influence soil variability over long time periods. soil variability in the EAA has been further increased by drainage, agricultural land preparation, and the extensive construction of roads and canals.
Soil variability is a major problem affecting the
reliability of soil testing for fertilizer recommendations. Reliability of a soil test result largely depends on whether or not the sample used for soil testing is actually representative of the field being sampled. Whenever soil heterogeneity increases, the precision of statements concerning their properties, behavior, and land use decreases (Trangmar et al., 1985).
One tool used to measure and describe the spatial
variability of soils is geostatistics. This technique has the ability to consider directly the spatial dependence of soil properties during sample interpolation. It also provides the ability to evaluate and map soil variability of selected properties. Two of the main analyses in geostatistics are semi-variogram calculations and kriging statistics. Semi-variograms are the graphic representation of the spatial variability between any two samples as distance between samples changes. Once an appropriate semivariogram has been calculated, values at unsampled locations can be estimated through kriging.
The broad objective of this research was to obtain
information of field and regional soil spatial variability of the Histosols in the EAA. The specific objectives were to (i) use semi-variograms to determine the structure of spatial dependence of selected soil chemical properties in the Histosols of the EAA; (ii) use block kriging techniques
to map and evaluate within field variability in four major soils series; (iii) summarize and map regional variation of agronomically important soil chemical properties in the EAA using geostatistical methods; and, (iv) to measure the effects of intermittent flooding and drained conditions on N and P release into drainage water of five typical Histosols from the EAA.
The Everglades Agricultural Area
The Everglades is the largest contiguous body of
organic soils in the continental United States (Hammar, 1929; Stephens, 1956). The Everglades accounts for approximately 75% of the organic soils in Florida (Davis, 1946) and about 14% of the total national deposits (Stephens, 1969). Peat and muck soils occupy approximately 778,000 ha of the Everglades (Jones, 1942). A portion of the Everglades was drained at the beginning of the century for agricultural purposes, becoming what is known today as the Everglades Agricultural Area (EAA). The EAA contains approximately 245,930 ha of organic soils (Histosols), forming one of the richest agricultural regions in the United States (McCollum et al., 1978). This area is intensively cultivated with annual cash receipts averaging more than 500 million dollars and stimulating over 1.2 billion dollars into the economy. The main crops grown in the BAA are sugarcane (Saccharum spp.) and winter vegetables. Sod, especially St. Augustine grass [Stenotaphrum secundatum (Walt.) Kuntze], rice (Oriza sativa L.), and cattle production accounts for the remaining
agricultural production in the EAA. Sugarcane production in South Florida occupies an area of approximately 173,344 ha yr-1 (Anderson, 1990a). From this total area, 89.8% (155,660 ha) is produced on the Histosols of the EAA (Coale, 1990). Winter vegetables are the second major crop grown in the EAA, utilizing an area of approximately 25,000 ha of organic soils. Rice is a relatively new crop introduced in the area during the past 10 years. In 1988, rice was grown on approximately 5850 ha in the EAA (Alvarez et al., 1989).
The Everglades is primarily a great sawgrass (Cladiun
iamaicense Crantz) marsh approximately 65 km wide and 160 km in length that extends from Lake Okeechobee to nearly the end of the Florida peninsula (Loveless, 1959; Zelazny and Carlisle, 1974). Both sides are bounded by low sandy ridges. The depth of these organic soils varies from north to south. Near the east side of Lake Okeechobee the depth of the organic materials is 2.5 to 3.5 m, but in the southern part of the Everglades they are quite shallow (Davis, 1946; Snyder et al., 1978). The depth of organic materials over a large part of the area is now less than 1 in (Soil Conservation Service, 1988; Anderson, 1990a).
The EAA is located in the upper Everglades, extending from south of Lake Okeechobee to the Broward County line (Fig. 2-1). The majority of the soils within the EAA are
Fig. 2-1. Location of the study area and surroundings.
underlain by the Pleistocene-age Fort Thompson formation consisting of alternating beds of limestone, shell, sand, and marl, which are often perforated by solution holes. Near the southern border of the EAA, this rock formation grades into another formation of softer and more porous rock called Miami oolite. Along the western edge the organic soils are underlain by sandy material (Cooke, 1945; Snyder et al., 1978).
In past geologic times, Lake Okeechobee, a circular
fresh water lake approximately 1800 kMn2, overflowed its south and eastern rims during the rainy season each year. This overflow, together with an annual rainfall of about 60 in (152 cm), inundated the Everglades basin and provided a suitable environment for accumulation of the present organic soils of the Everglades. The majority of these soils were derived from growth of emergent reed and sedge-like plants, principally sawgrass. These soils developed into light, brown colored, fibrous peats, which when drained, develop into excellent field soils with a black, finely-fibrous, well-decomposed organic surface.
Histosols in the EAA began to form approximately 4400 B.C. in the late Hypsithermal period (McDowell et al., 1969). Initially, approximately 500 to 4000 years were required to develop about 7.6 cm of the basal muck peat,
composed of a mixture of marl and organic matter. However, between 3500 to 1200 B.C. peat development proceeded at a rate of about 7.3 cm per century. It was during this period that the existing peat, measuring about 1.37 m in thickness, was developed. The Everglades basin before the initiation of the drainage programs in the early 1900's was inundated for a large part of the year. Water was high enough to maintain anaerobic conditions permitting continuation of peat development until shortly after 1906 when the Everglades Drainage District began construction of the first drainage canals. By 1914, the peat had developed to an average depth of 3.65 m (Stephens, 1956). This represents an average of peat development of about 8.4 cm per century. Although, the early attempts at draining the area were unsatisfactory for commercial farming, the water balance was disturbed to a point that aerobic conditions prevailed. After this time, the process of peat accumulation was reversed and destruction of the soil by microbial oxidation began (Stephens, 1956).
Soils in the EAA
The soils in the EAA are organic (Histosols), generally containing more than 85% organic matter by weight derived from hydrophytic vegetative residues. Except for Histosols found adjacent to Lake Okeechobee, their mineral content is less than 35%. In the current soil classification system,
organic soils are defined as soils containing at least 12 to 18% organic carbon (20 to 30% organic matter) by weight, depending on the clay content of the mineral fraction (Soil Survey Staff, 1975).
The major Histosol series and some of their most important characteristics are as follows.
Torry mucks (euic, hyperthermic Typic Medisaprist) are very poorly drained, deep, organic soils with a high content of fine textured mineral material. These soils have an organic layer > 130 cm over limestone. Permeability of these soils is moderate (1.5 to 5.1 cm hr-I) to a depth of 91 cm and rapid (15.2 to 50.8 cm hr-) at subsequent depths. Torry soils have a high available water capacity (0.2 to 0.3 cm cm"1) in all layers with a high natural fertility. A representative pedon of these soils has a black muck (sapric material) surface layer about 30 cm thick with a mineral content of about 70%. The next layer is a sticky black muck that extends to about 91 cm with about 60% of mineral material. Below 91 cm, there is a black muck that has a mineral content of about 35% and extends to a depth of about 165 cm. Torry mucks account for about 7.1% of the organic soils in the EAA.
Terra Ceia mucks (euic, hyperthermic Typic Medisaprist) are very poorly drained and deep organic soils. These soils have an organic layer > 130 cm over limestone. A representative pedon of a Terra Ceia muck has a black muck
layer of sapric material in the upper 20 cm of soil with a dark reddish brown muck underneath that to a depth of 163 cm or more. Terra Ceia mucks account for about 9.5% of the organic soils in the EAA.
Pahokee mucks (euic, hyperthermic Lithic Medisaprist)
are poorly drained organic soils with organic layers from 91 to 130 cm thick over limestone. A representative pedon of a Pahokee series has a black muck layer in the upper 71 cm of soil with a dark reddish brown muck underneath that extends to the limestone bedrock at a depth of between 91 to 130 cm. Pahokee mucks account for about 27.4% of the organic soils in the EAA.
Lauderhill mucks (euic, hyperthermic Lithic
Medisaprist) are poorly drained organic soils with dark organic layers from 51 to 91 cm over limestone. In a representative pedon, the surface layer is a black granular muck about 20 cm thick. Below, there is a layer of black muck about 25 cm thick that is slightly more fibrous. The next layer is dark reddish brown fibrous muck that extends to a maximum depth of 91 cm. Lauderhill mucks account for about 39.6% of the organic soils in the EAA.
Terra Ceia, Pahokee, and Lauderhill mucks are similar, differing only in the thickness of the organic material over the limestone bedrock. These soils contain less than 35% mineral matter by weight. Under natural conditions, these soils are flooded, or they have a water table within 25 cm
from the surface for 6 to 12 months. These soils have rapid permeability and very high available water capacity with a moderate natural fertility.
Dania mucks (euic, hyperthermic, shallow Lithic
Medisaprist) are very poorly drained soils with an organic layer < 51 cm over sand and limestone. A representative pedon of this soil series has a black well-decomposed muck layer in the surface 10 cm of soil. The next layer is dark reddish brown muck about 30 cm thick. Below the organic material, there is a very thin layer of light gray sand resting over the limestone bedrock. Like the soils discussed above, this soil has rapid permeability when drained and a high available water capacity with moderate natural fertility. Most areas of these soils are cleared and used for improved pasture or sod production. Dania mucks account for about 10.2% of the organic soils in the EAA.
Okeechobee mucks (euic, hyperthermic Hemic Medisaprist) are poorly drained, deep organic soils formed in thick deposits of hydrophytic plant remains. A representative pedon of this soil has a black granular muck (sapric material) surface layer about 20 cm thick. Below there is a layer of black muck about 50 cm thick. The next layer is a dark reddish brown fibrous mucky peat (hemic material) about 55 cm thick. Finally, there is a dark reddish brown muck layer that extends to a depth of 165 cm or more. When
drained, permeability and available water capacity of these soils is high with moderate natural fertility. Okeechobee mucks account only for about 2.6% of the area in the EAA.
Okeelanta mucks (sandy, siliceous, euic, hyperthermic Terric Medisaprist) are very poorly drained soils, with an organic layer about 102 cm thick over sand. Permeability of these soils is very high when drained. The available water capacity is high in the organic layer and low in the underlying sandy parent material. Their natural fertility is low to moderate. In a representative pedon of this soil, the surface layer is black (sapric material) and about 20 cm thick. Below, there is a layer of dark reddish brown muck that extends to a depth of about 79 cm. Underneath the organic material there is a thick layer of very dark gray sand that changes to light gray at a depth of about 140 cm. Okeelanta mucks account for only 3.6% of the total area in the EAA (McCollum et al., 1976; McCollum et al., 1978; Soil Conservation Service, 1988; Anderson, 1990a). Organic Soil Subsidence
Drainage of organic soils deposits for agricultural
purposes results in the loss of soil through rapid breakdown of organic matter. Subsidence of the organic soils in the EAA is defined as a loss of soil depth and volume due to shrinkage, compaction, and biological oxidation (Clayton, 1943; Thomas, 1965b; Smith, 1990). Initially, the most
rapid subsidence occurs from shrinkage as the peat dries out and compaction due to the use of farm equipment. These two processes do not involve the direct loss of soil material. Biological oxidation refers to the microbial oxidation of the organic material. Products of organic matter oxidation include CO 2 and water, various form of N such as NH 4 + I N'03and soluble organic N, Ortho-P, and soluble organic P. Other factors affecting subsidence are burning, wind erosion, character of soil material, and cropping system (Thomas, 1965b; Lucas, 1982; Smith, 1990). While all the listed factors are important, the metabolic activity of microorganisms has been singled out as the largest factor affecting subsidence. The reason is because microbial activity is dynamically indefinite and largely uncontrollable. This process continues as long as the soil is drained, and eventually most of the organic material is destroyed, leaving little soil material over the underlying rock. Microbial oxidation accounts for 70% of soil subsidence in the EAA (Volk, 1973). The other factors are considered to be determinate and to some extent manageable (Volk, 1973; Snyder et al., 1978).
Subsidence has been reported to be greatest during the first few years after drainage (Clayton et al., 1952; Thomas, 1965b). During this time, shrinkage, compaction, and oxidation occur at a higher rate. Although rate of subsidence may vary with soil type and imposed crops, the
fact remains that organic soil losses due to oxidation are appreciable on any drained and cultivated organic soil. Measurements of subsidence in the EAA have shown that soil is lost at an average rate of 3 cm yr-I (Stephens, 1956; Thomas, 1965b). Assuming present rates of subsidence and the fact that oxidation will occur as long as soils remain drained for cropping, Stephens (1956) predicted that by the year 2000 only about 13% of the soils in the EAA will be deeper than 91 cm, and 45% will be less than 30 cm deep. These predictions appear to be correct (Snyder et al., 1978; Soil Conservation Service, 1988). However, these soils can be conserved through better water management practices, such as using higher water tables and flooding during fallow (Snyder et al., 1978; Forbes, 1981).
Some soils are not homogeneous, but rather a
heterogeneous bodies of materials. Because of this heterogeneity, methods have been developed to delineate soil classification units. One type of soil variation is the variation among the several units which have been classified as homogeneous. For example, poorly drained soils formed from recent alluvium are usually different in most of their properties from well-drained soils formed from residual parent material (Petersen and Calvin, 1986). Because of the nature of soil-forming processes, distinct boundaries
between soil classification units are rare. Although two adjacent soil series may be distinctly different, there is usually a gradual transition in the field between one series and another. However, local variations within a particular soil series exist.
The nature of soil variability is scale-dependent based on soil-forming factors and processes interacting over many different spatial and temporal scales (Burrough, 1983a). Therefore, the nature of soil variability identified by spatial studies of soil properties depends largely on the scale of observation, soil properties, and methodology used to conduct the investigation (Wilding and Drees, 1983). variation in soil properties from one point in the landscape may result from natural causes such as vegetation, topographic changes, or from man-made variations such as fertilizer application (Beckett and Webster, 1971).
Wilding and Dress (1978) pointed out that spatial soil variability can be grouped in to two broad categories, systematic and random. systematic variability is expressed as gradual or distinct changes in soil properties that can be explained by the soil-forming factors or processes at a given scale of observation. sources of systematic variation may range from differences in topography, lithology, climate, biological activity, and age of soils in regional studies (Van Wambeke and Dudal, 1978) to differences in microfabric and physicochemical composition when soils are
studied on a micro level (Miller et al., 1971). Associated with systematic variation are differences in observed soil properties that cannot be related to a known cause. There are also spatial, temporal, and measurement sources of variation which cannot be explained by the scale of investigation (Ball and Williams, 1968). This unexplained heterogeneity is termed random, noise, or chance variation (Burrough, 1983b; Wilding and Drees, 1983).
Soil variability is the product of soil-forming factors operating and interacting over a continuum of spatial and temporal scales. Processes that operate over large distances such as climate produce gradual soil changes, although abrupt changes may occur between two climatically determined plant communities (Beckett and Webster, 1971). other processes, such as parent material and soil weathering influence soil variability over long time periods. According to Robinson and Lloyd (1915), soils formed on transported materials tend to be more variable than those weathered from bedrock in situ. Even within an outcrop of an apparently uniform sedimentary rock, there can be regional soil differences, as a result of geochemical gradients at the time of deposition (Ulrich, 1949). Regional contrasts in topography can produce regional differences in soil properties. Within a region, dissection and the associated deposition of eroded materials can produce recurrent patterns of land forms, and as a result,
recurrent patterns of dissimilar soils (Beckett and Webster, 1971).
Within the soil profile some physical and chemical
properties tend to increase lateral variability, such as the development of Igilgail or of frost-wedges (Crompton, 1956). Many biological activities increases local variability also. Ebersohn and Lucas (1965) reported that the localized uptake of nutrients and water or their concentration beneath the tree canopy increase localized soil variability beneath the canopy. Similarly, tree-throw, burrowing or wallowing animals, termite and ant mounds, produce soil heterogeneity (Lyford and MacLean, 1966).
The effects of all these factors are superimposed,
affecting the overall soil variability. Processes that give rise to soil differences over short distances introduce variability within all sampling areas regardless of their size. Long-range soil changes caused by soil processes will on the average make noticeable contributions only to the variability of larger sampling areas. Therefore, the overall variability of the soil within an area depends strongly on the environment, but soil variability is likely to increase with the size of the area sampled (Beckett and Webster, 1971).
Variability of soil chemical properties can be
increased on cultivated soils. Intense grazing produces dung or urine patches rich in P and K, respectively (Friesen
and Blair, 1984). Uneven fertilizer or manure application, row cultivation, and the growth of row or tree crops, all tend to superimpose additional heterogeneity on soil chemical properties. Melted and Peck (1973) reported that where fertilizers have been applied, large differences in nutrient levels can be found in different parts of the same field. These differences are not necessarily sampling errors, but a reflection of the true soil variation within the field.
Soil properties vary on both the horizontal and
vertical planes. Horizon boundaries may be more distinct than surface boundaries within a soil classification unit. This should be taken into consideration when sampling soils. As the soil heterogeneity increases, the precision of statements concerning their properties, behavior, and land use performance decreases (Trangmar et al., 1985). Therefore, researchers have suggested the subdivision of soil populations, both horizontally and vertically, into sampling strata which are as homogeneous as possible. In addition, all possible sources of variation within the population should be sampled if valid inferences are to be made about the population from the sample (Petersen and Calvin, 1986).
Most soil properties are spatially correlated, but
cannot always be spatially measured or recorded. Therefore, in order to interpolate the values of individual variables or class types at unsampled locations, variability information recorded at sampled sites is used (Burgess and Webster, 1980a; Oliver, 1987). In soil science, spatial variation has been defined largely by the spatial classification of either individual properties or of class types. The classical approach in the field is to group soils together in similar units or lay out small plots with the assumption that variability within the plots is purely random. These assumptions have held more or less at the broad scale. However, as attention has increasingly focused at the local scale, and the need has arisen for quantitative estimates of individual variables spatial classification has not worked efficiently (Webster, 1985; Oliver, 1987).
When spatial dependence of soil properties exits in most sampling units, the classical statistical model is inadequate for interpolation of spatially dependent variables. Classical statistical model assumes random variation and takes no account of spatial correlation and relative location of the sample (Trangmar et al., 1985; Oliver, 1987). Recent development in statistical theory has enabled spatial dependence of soil properties to be directly considered for sample interpolation. These developments are
based on the theory of regionalized variables (lMatheron, 1963). A variable is considered to be a regionalized variable if it varies from one place to another with continuity. Often the regionalized variable cannot be represented by traditional statistical functions (Davis, 1973). Regionalized variable theory was developed from empirical ideas of D.G. Krige in the gold mines of South Africa. He suggested that the spatial estimation of gold content could be improved by taking the degree of similarity, or autocorrelation, between samples into consideration. The regionalized variable theory takes into account both the random and structured characteristics of spatially distributed variables to provide quantitative tools for their description and optimal unbiased estimation. This theory now forms the basis of procedures for analysis and estimation of spatially dependent variables. These procedures are known collectively as geostatistics (Journel and Huijbregts, 1978).
Geostatistics was primarily developed for the mining industry (Matheron, 1963). Geostatistics was very useful for engineers and geologists for studying the spatial distribution of important properties such as grade, thickness, or accumulation of mineral deposits. However, it has been emphasized that the estimation technique can be used wherever a continuous measure is made on a sample at a particular location in space or time. Especially where
sample estimates are expected to be affected by their position and their relationships with their neighbors (Clark, 1979a). Geostatistics is also helpful in quantifying spatial and inter-variable correlations, designs of optimum interpolation schemes, and consideration of scale of sampling. In addition, new sampling locations can be defined to improve estimates for a total population or location (Warrick et al., 1986). Semi-variograms
Normally, samples taken in close proximity will be more related than samples taken farther apart. In geostatistical terms, if the samples are highly related, then the variance of the distribution of their differences will be low, and vice versa (Davis, 1973, Clark, 1979b). Consequently, this variance is a measure of the influence of samples over neighboring areas. The variance between samples as it relates to distance from each other is represented by a variogram, a graph of sample variance vs. the distance h. Huijbregts (1975) mathematically described a variogram, 2r(h), as the average quadratic deviation between values,
(Y), at two points, x and x+h, or space
21(h) = E([Y(x+h) Y(x)]2) (1)
However, the half-variogram or semi-variogram, r(h), is more commonly used (Clark, 1979b).
The application of the regionalized variable theory assumes that the semi-variance between any two samples in the study region depends only on the distance and direction of separation between each other and not on their geographic location. Based on this assumption, the average semivariogram for each lag distance h, can be estimated for a given value of three-dimensional space (Trangmar et al., 1985).
The equation used in calculating a semi-variogram is
r(h) = i/2N(h) Z [Y(x+h) Y(x)]2 (2)
where F(h) is the semi-variogram,
N(h) is the number of pairs of points used in the
Y represents the measured values in space
separated by a distance along the distance h.
The semi-variogram is defined as one-half of the
variance of the differences between points separated by a distance h. The semi-variogram represents the average rate of change of a property with distance. The shape of the semi-variogram describes the pattern of spatial variation in terms of its magnitude, scale, and general form. The shape of the experimental semi-variogram may take many forms, depending on the data and sampling interval used. Ideally, the experimental semi-variogram should pass through the origin when the distance of sample separation is zero.
However, many soil properties have nonzero semi-variances as the distance (h) approaches zero.
A nonzero semi-variance is called the "nugget variance" or "nugget effect" (Journel and Huijbregts, 1978). The nugget variance (Co) represents unexplained or random variance caused by measurement error. It may also be that the sampling scheme was too far apart to eliminate all positional variability. Ideally, when the distance becomes very large, samples are independent of one another. The semi-variogram value will then become more or less constant since it represents the difference between sets of independent samples. The distance at which samples become independent of another is called the "range" of spatial dependence (A). The semi-variance value at which the graph becomes constant is called the sill (CCo) of the semivariogram (Fig. 2-2).
When a well-defined model cannot be fit to the
calculated semi-variance values, a pure nugget effect exists in the data. A pure nugget effect means that all the calculated r(h) values are equal to the sill at all values of h. The pure nugget effect is indicative of a completely random distribution of the variable at the sampling interval used. Decreasing the sampling distance will often reveal structure in the apparently random effects of the pure nugget variances (Burrough, 1983a).
cz (D co
0 CD + 0)
The semi-variogram is central to geostatistics and the single most important tool in geostatistical applications to soils. Mathematical functions for semi-variograms must be positive-definite functions to insure that estimation variances will be nonnegative (McBratney and Webster, 1986; Marx and Thompson, 1987; Oliver, 1987). Only models with functions that are positive-definite in up to three dimensions are considered in this chapter. The most commonly used models for the semi-variograms are given below.
1. Linear model.
F(h) = Co + wh for h > 0 (3)
r(0) = 0 (4)
where r semi-variance
Co intercept or nugget variance w = slope
h = lag distance
B. Transitive models that reach a sill.
1. Spherical model reaches a sill at a definite range.
rF(h) = Co + w [1.5 (h/a) 0.5 (h/a)3] (5)
for 0 < h a
r (h) = Co + w for h > a (6)
r(h) = 0 (7)
where a = range
Co + w = sill
2. Exponential model reaches a sill asymptotically.
r (h) = CO + w [1 exp (-h/a)] for h > 0 (8) r(h) = 0 (9)
In the unbounded models, the variance appears to
increase without limit. The linear model is the most common in this group. The other major group is the transitive models which use a finite variance derived from movingaverage processes. The spherical and exponential models are the most commonly used transitive models (McBratney and Webster, 1986).
Examples of these semi-variograms are shown in Fig. 2-3. The semi-variogram can be estimated in different directions to detect anisotropy in the variation. It can also be estimated from irregularly scattered two-dimensional data by grouping lag intervals by distance and direction (Webster, 1985).
Semi-variograms can be used to optimize sampling
(Campbell, 1978; Burgess et al., 1981; McBratney et al., 1981). Campbell (1978) used semi-variograms to study spatial variation of sand and pH measurements within each of two sampling areas displaying contrasting pattern of variation. He was one of the first to use geostatistical methods in soil science. McBratney et al. (1981) explained that an optimal sampling scheme is based on the theory of
C)0 rC) x
regionalized variables that assumes that spatial dependence can be expressed quantitatively in the form of the semivariogram. Assmus et al. (1985) used semi-variograms to describe the field spatial variability of P. These authors concluded that if a strong relationship exists, the minimum sampling distance should be equal to the range of the semivariogram model.
Semi-variograms are also used to study spatial
variability of soil properties over small distances (Vieira et al., 1981), and over large distances (Yost et al., 1982a; Uehara et al., 1985; Xu and Webster, 1984). Vieira et el. (1981) used a sampling distance from 1 to 19 m to measure the spatial variability of infiltration rates in Entisols. Semi-variograms indicated that samples separated by 50 m or more were unrelated. Yost et al. (1982a) collected soil samples along transects on the Island of Hawaii at 1 and 2 km intervals to determine spatial dependence of soil chemical properties over long distances. They concluded that soil chemical properties were spatially dependent. Kriginq Statistics
The estimation procedure incorporated into the
regionalized variable theory is known in soil science as kriging, after D.G. Krige. Krige was the scientist that empirically devised the kriging technique for use in the South African goldfields (Webster, 1985). Kriging is a
method used to spatially predict soil properties. It is an optimal geostatistical procedure in the sense that it provides estimates of values at unrecorded places without bias and with minimum and known variance. Kriging is essentially a means of local estimation in which each estimate is a weighted average of n observed values (Burgess and Webster, 1980a; Webster and Burgess, 1983). The kriging technique outside the mining industry is relatively new; however its popularity in other disciplines has increased in the last 15 years. Kriging has been used successfully in mining (David, 1977), hydrology (Delhomme, 1978), and soil science (Vieira et al., 1981; Yost et al., 1982b). The simplest forms of kriging are point and block kriging. Both techniques assume that the sample data are normally distributed and stationary (Henley, 1981).
Point estimation is considered to be the most common
kriging technique used in soil science (Burgess and Webster, 1980a; Vieira et al., 1981). The interpolation of the regionalized variable Z at a location x0 is
-(X) Z fi Z(xi) (10)
where Z(x0) = unknown parameter at location x0,
f. = weights associated with data points,
Z(xi) = value of a property at a point xi. The weights are dependent on the semi-variogram, and the configuration of sampling points, with more weight being
given to nearby points (Oliver, 1987). The weights are chosen so that the estimate Z(X0) of the true value Z(X0) is unbiased and the estimation variance is minimized (Trangmar et al., 1985).
Most of the early applications of kriging in soil
science involved simple point estimation for isoproperty mapping (Burgess and Webster, 1980a; Vieira et al., 1981; Webster and McBratney, 1987). Webster and McBratney (1987) used point kriging to map extractable P, exchangeable K, and pH in the surface soils from the Broom's Barn Experimental Station. An area of 77 ha was used in this study, with samples taken at 40-m intervals. Semi-variograms of the properties were determined and kriged values estimated at 10-m intervals using a square sampling grid. These values were contoured to produce maps of the properties. Point kriging has been found to be one of the most practical methods for mapping soil pH with high precision (Laslett, et al., 1986; Webster and McBratney, 1987). The most attractive feature of kriging for mapping is that besides the unbiased estimation of values, it provides an indication of the reliability of the estimates.
Point kriging is also used in the estimation of
variances for designing sampling schemes for future kriging operations (Burgess et al., 1981; Marx and Thompson, 1987). In addition, point kriging has the ability to define the direction and magnitude of soil variability using a minimum
of subsamples. This technique may have its greatest advantage in the determination of plot size and shape in selection of field research areas (Sabbe and Marx, 1987).
When an estimation is made from averages, the procedure is called block kriging (Burgess and Webster, 1980b). Block kriging has the capability to interpolate an average value for an area or block larger than the soil area actually sampled. This procedure can avoid some of the weakness of point kriging, resulting in smaller estimation variances and smoother maps (Burgess and Webster, 1980b; Trangmar et al., 1985). The estimation variance of block kriging is always less than that of the point kriging because the within-block variance is removed from the error term. Burgess and Webster (1980b) reported up to 20-fold improvements in average estimation precision using block kriging compared to point kriging.
The most common use of block kriging has been for the production of isarithm maps of soil properties (Burgess and Webster, 1980b). They found that soil properties mapped in two intensive surveys had large nugget variances leading to large estimation variances and erratic isarithms when mapped by point kriging. They concluded that if the interest is estimation of average values for soil properties over areas rather than the estimation of specific points, then block kriging is more appropriate than point kriging. Previous studies have shown that block kriging produces smoother maps
than punctual kriging by interpolating average values for blocks, with the effect of smoothing local discontinuities (Trangmar et al., 1985). Block kriging has also been applied to interpolate spatial effects of crop response to variability imposed by soil management practices. Tabor et al. (1984) used block kriging to determine the spatial variability of nitrates in cotton petioles. They found that calculated variograms and block-kriged maps of petiole nitrates indicated a strong influence due to cultural practices, such as direction of rows and irrigation.
Intensive sampling is required to obtain good kriged estimates and for some properties this can be to costly. However, if the property of interest is spatially correlated with another property that can be measured easier, then it can be estimated more precisely from fewer observations of the other property. The procedure used in this kind of estimation is called co-kriging (McBratney and Webster, 1983b; Vieira et al., 1983). To apply co-kriging, it is necessary to model variograms for each variable separately as well as cross-variograms for all pairs. The theoretical basis as well as examples showing how to use this technique have been extensively documented (Myers, 1982; McBratney and Webster, 1983a, 1983b; Vauclin et al., 1983).
Some of the assumptions required for kriging is that
the data are stationary or specifically follow the intrinsic hypothesis (Webster and Burgess, 1980; Trangmar et al.,
1985). The intrinsic hypothesis states that the variability between any two samples depend on the distance between them, but not on their spatial position. Therefore, two samples separated by the same distance, should have the same variability (Trangmar et al., 1985). Universal kriging was designed to overcome the effect of non-stationarity and permit the use of kriging in the presence of strong trends (Webster and Burgess, 1980). Universal kriging is a form of interpolation that uses local data trends to minimize the estimate error. The presence of such trends are identified qualitatively, and their form is quantitatively found by structural analysis, which simultaneously estimates semivariances of the differences between the drift and the actual data. The resulting semi-variograms are used for the interpolation.
Webster and Burgess (1980) were some of the first researchers to apply universal kriging in soil science. They applied this method to measurements of electrical resistivity in the soil from samples collected at 1-in intervals. They concluded that universal kriging was not always applicable for soil survey mainly because of the large variances usually encountered. Yost et al. (1982b), working with large areas in the Island of Hawaii pointed out that universal kriging resulted in little improvement over ordinary kriging when comparing the observed and estimated points. Their results showed that ordinary kriging is
useful in summarizing and interpreting soil analyses and that it seems quite robust to certain degrees of nonstationarity. However, other scientists have found the technique useful in evaluating soil variability of selected soil properties. Ovalles and Collins (1988) used universal kriging to study the spatial variation of selected soil properties in northwest Florida. They found that their kriged maps of selected soil properties were helpful in locating areas of large standard errors in NW Florida. These results indicated the need for more intensive sampling at specific locations. The authors suggested that kriged standard error maps can be used to plan future sampling strategies.
SOIL SPATIAL VARIABILITY OF SELECTED CHEMICAL PROPERTIES IN THE EVERGLADES AGRICULTURAL AREA. I. SEMI-VARIOGRAMS
The inherent variability of soil properties is widely recognized (Beckett and Webster, 1971). Recognition of the importance of spatial variability on land use has led to the study of soil heterogeneity, at different levels of generalization. Some studies have looked at spatial variability ranging from zonal or regional differences, such as between soil orders, suborders, or great groups (Van Wambeke and Dudal, 1978; Trangmar et al., 1984), to small changes in physical and chemical properties of surface soils occurring within areas of 4 m'2 (moorinann and Kang, 1978) .
Soil spatial variability is a major problem affecting the reliability of soil testing for fertilizer advisory purposes. A continuous concern within the soil-testing system is whether or not a soil sample represents the field being sampled. Cameron et al. (1971) pointed out that the three major sources of soil variation in soil testing are: laboratory, temporal or seasonal, and spatial or field. They found that spatial or field variation was responsible for the largest and most significant variation in soil
testing. As the heterogeneity of soil increases, the precision of statements about their properties, behavior, and land-use performance decreases. Predictions at unsampled locations can be estimated from information obtained at the known locations. The precision of such extrapolation is strongly influenced by the variability of soils both within sampling units and between locations (Trangmar et al., 1985).
One tool being used to measure and describe soil spatial variability is the semi-variogram. The semivariogram represents the variability between samples across distance and its shape describes the spatial variability in terms of its magnitude, scale, and general form. Semivariograms have been used extensively in the mining field. The interest in the application of this technique to the field of soil science has increased in the last 15 years.
Analysis of spatial dependence using semi-variograms
has contributed to our understanding of many aspects of soil variability, genesis, management, and interpretation (Vieira et al., 1981; Yost et al., 1982a; Ruso, 1984; McBratney and Webster, 1986; Bos et al., 1984). Campbell (1978) computed semi-variograms to describe the spatial variability of soil texture and pH of less and glacial-till-derived soils. Assmus et al. (1985) used semi-variograms to describe field spatial variability of P levels. They concluded that if there is a strong spatial relationship, the minimum sampling
distance should be equal to the range determined by the semi-variogram model. Burgess and Webster (1980a), created semi-variograms to show variability of sodium content, stoniness, and loam thickness from detailed soil surveys of Central Wales and Norfolk, England. Bos et al. (1984) used semi-variograms to study the spatial variation of selected chemical properties of surface soil in reclaimed phosphatemine lands. They found that a spatial correlation generally existed at sampling distances ranging from less than 2.6 to
Semi-variograms have been applied to measure spatial dependence of soil properties at small (Gajem et al., 1981) as well as large distances (Yost et al., 1982a; Xu and Webster, 1984). Gajem et al. (1981) used semi-variograms to determine the spatial dependence of selected physical properties of a Typic Torrifluvent sampled at 0.02-, 0.2-, and 2-m intervals. They reported a range of 0.6 m of spatial dependence of soil properties sampled at 0.2-m intervals. Yost et al. (1982a) determined semi-variograms of several chemical properties over the Island of Hawaii. They found a spatial dependence range of 58 km for P sorbed at 0.2 mg P L"1 from samples taken at 1- to 2-km intervals.
The primary objective of this study was to determine the structure of spatial dependence of selected soil chemical properties in the organic soils of the EAA. The secondary objective was to examine and interpret the use of
semi-variograms as a tool to detect withing-field soil variability.
Materials and Methods
Collection and Preparation of Soil Samples
Surface samples (0 to 15-cm depth) of four different
soil series found in the EAA were used in this study (Table 3-1). Samples from two adjacent fields (A and B) were collected from each soil series during the summer of 1987. Surface samples of Lauderhill muck (euic, hyperthermic Lithic Medisaprist) were collected from both a sod, St. Augustine grass [Stenotaphrum secundatum (Walt.) Kuntze], and a fallow field. The fallow field (A) had been in sugarcane production (Saccharum spp.) for the past 5 years. Pahokee muck (euic, hyperthermic Lithic Medisaprist) samples were collected from two fields that had been in continuous sugarcane production for the past 5 years. Okeelanta muck (euic, hyperthermic Terric Medisaprist) samples were collected from two adjacent fields that had a history of sugarcane production for the previous 5 years. However, at the time of sampling, field A was previously in sweetcorn (Zea mays L.) production (harvested one week before sampling), and field B was in fallow. Torry muck (euic, hyperthermic Typic Medisaprist) samples were collected from sugarcane and fallow fields near the city of Pahokee. Field
Table 3-1. Description of four organic soils from the EAA used in the spatial variability study.
Soil series Classification history Rng Twn Sec
Lauderhill euic, hyperthermic A-Fallow 37 45 19
Lithic Medisaprist B-Sod
Pahokee euic, hyperthermic A-Sugarcane 38 45 6
Lithic Medisaprist B-Sugarcane
Okeelanta euic, hyperthermic A-Sweetcorn 39 43 2
Terric Medisaprist B-Fallow
Torry euic, hyperthermic A-Sugarcane 37 42 29
Typic Medisaprist B-Fallow
Table 3-2. Sampling distances used in the study.
Sampling Size of Sampling
Soil series scheme study area distance
--- ha --- --- m --Lauderhill Large 14.2 44.8
Small 0.8 10.0
Pahokee Large 14.0 44.2
Small 0.8 9.8
Okeelanta Large 13.3 43.3
Small 0.8 9.6
Torry Large 13.2 43.3
Small 0.8 9.8
A has been in continuous sugarcane production for the last 5 yr, while field B was fallow with a history of sweetcorn production.
Two different sampling schemes (Fig 3-1) were used for each location. Samples were taken on triangular grid intercepts with the purpose of increasing the number of equidistant neighboring points. The first sampling scheme included an area ranging from 13.2 to 14.2 ha covering two adjacent fields. This sampling was designed to detect differences due to ditches between fields. Lag distances used in this sampling varied from 43 to 45 m according to the size of each individual field (Table 3-2). The side of each field parallel to the access road was used as a base line for a triangular grid of 10 rows and 10 columns with a total of 100 samples taken from each location.
The second sampling scheme was designed to account for soil spatial variability that could not be measured at the larger distances used in the first sampling scheme. For the second sampling scheme, four small areas of approximately
0.2 ha were randomly located within the larger sampling grid at each location (Fig. 3-1). The lag distance used on this sampling varied from 9 to 10 m depending on the location (Table. 3-2). Extra samples (*) were taken along the ditches to guarantee detection of soil variability due to ditch spoils.
ro :j Z-4- -P
1 1 4 Cl)
Soil samples were composites of 10 cores taken around each sampling point (0.5 M2). Soils were air-dried at 38 0C for 72 h and passed through a 2-mm sieve before chemical analysis.
Soil pH was determined in a 1:2 soil:water suspension (PHw), and 1:2 soil:0.01 M CaCl2 suspension (pH.) by glass electrode. Extractable nutrients were measured using Mehlich-I extracting solution (CaMI, MgMI, I' PmI' ZnMI, CUMI, Mnmi, and FeMI) (Mehlich, 1953) by the IFAS Analytical Research Laboratory, University of Florida, Gainesville (Hanlon and DeVore, 1989). Soil samples were analyzed also by the Everglades Soil Testing Laboratory, Belle Glade, for water and 0.5 M acetic acid-extractable P (Pw and Pat respectively; 4:50, soil:solution by volume) and 0.5 M acetic acid-extractable K, Ca, and Mg (Kai Caa, and Mga, respectively; 10:25, soil:solution by volume) (Thomas, 1965a; Sanchez, 1990).
Ash content was determined by igniting 5 g of ovendried soil at 500 0C for 5 h. Weight loss was considered as the amount of organic matter in the sample and the residue was taken as the mineral content. The residue was dissolved in 2 M HCl and total elements with the exception of P from the Torry muck, were analyzed using the inductively coupled argon plasma (ICAP) technique. Total P content in the Torry
muck was measured by igniting 1 g of oven-dry soil at 550 0C for 2 h (Olsen and Sommers, 1982). Ash was dissolved in 6 M HCl and analyzed colorimetrically (Murphy and Riley, 1962). All laboratory results are reported on a weight basis.
Mean, range, variance, and coefficient of variation were computed using Statistical Analysis System software (SAS, 1982a, 1982b). The Univariate procedure was used to test for normality. Approximation of each variable to normal and log (base e) normal probability distributions were determined using the Kolmogorov-Smirnoff D statistic. Variables that failed to meet the normality test were log transformed before analysis.
Geostatistical approaches applied in this study were
derived from concepts discussed by David (1977) and Journel and Huijbregts (1978). Data from each location were entered in a X, Y, and Z (soil property) format. The program used in the calculation and fitting of semi-variograms was VAR5, Version 2.0. This program was part of a set of geostatistical analysis programs developed by the Department of Agronomy and Soil Science, University of Hawaii (Yost et al., 1989). Semi-variograms were calculated from the formula
r(h) = i/2N(h) E[Y(x) Y(x+h)]2
where N is the number of sample pairs at each distance interval (h), and Y(x) is the value of the soil property at field location (x). The semi-variogram is the plot of r(h) as the ordinate, against the average lag distance (h), as the abscissa (Fig. 2-2). Semi-variances were averaged over 180 degrees (considering all samples) and averaged over 90 degrees in four directions (N-S, NE-SW, E-W, and NW-SE) to determine if roads and ditches had influence in the direction of the semi-variance (anisotropy).
A reliable semi-variogram is obtained when intervals
are chosen such that the number of pairs is large enough to ensure accurate definition of each point on the semivariogram. Intervals distances (h) that were selected provided at least 25 pairs of points for each interval. This number of paired points has been found to be necessary to provide stable estimates for the semi-variogram. Another important practical consideration is that experimental semivariograms should be considered only for small distances (h < L/2), where h is the lag distance and L is the dimension of the field on which the semi-variogram has been computed (Journel and Huijbregts, 1978). Weighted least squares methods using the SAS nonlinear method were used to model linear and spherical semi-variograms.
Results and Discussion
Soil Series Variation
The mean, range, variance, and % CV of selected soil chemical properties from the surface 15 cm of each soil series are presented in Appendix A. Means of selected soil chemical properties are shown in Table 3-3. Samples close to the road were not included in calculations. Soil pHW varied from 4.8 to 6.3, with Lauderhill muck showing the
highest pHW and the Okeelanta muck the lowest pHW. The higher pHW of Lauderhill muck may be due to the greater effect of the limestone bedrock as the organic layer gets shallower due to subsidence. Flooding, high water tables, and calcareous deposits along roads and canal are additional factors that tend to increase soil pH over time in the organic soils of the EAA (Lucas, 1982). The acidity of organic soils is due to the presence of organic compounds, exchangeable hydrogen and aluminum, iron sulfide, and other sulfur minerals. Okeelanta muck showed the lowest pHW and the FeMI concentration. Under natural conditions most of these organic soils are unfertile and acidic, requiring lime and fertilizer application for commercial crop production.
The presence of soluble salts influences soil pH, and to obtain a true estimate of the acidity of the soil these salts must be removed before a pH determination. Schofield
and Taylor (1955) suggested the use of 0.01 M CaC12 for soil
Table 3-3. Means of soil chemical properties from the surface 15 cm of four organic soils from the EAA.
Variable Lauderhill Pahokee Okeelanta Torry
Analytical Research Laboratory, Gainesville+ pH-H20 6.3 5.3 4.8 5.4
pH-CaCl2 6.0 5.1 4.4 5.0
Ca, g kg-1 9.4 10 8.6 6.8
Mg, g kg"' 1.5 1.3 0.58 0.72
K, g kg"1 0.36 0.22 0.37 0.46
P, mg kg"I 46 49 32 94
Zn, mg kg-I 1.2 1.8 11 10
Cu, mg kg"1 0.33 0.41 0.55 0.28
Mn, mg kg- 0.95 7.3 10 28
Fe, mg kg"1 4.2 2.9 20 7.4
------- Soil Testing Laboratory, EREC+--------Pw, mg kg-1 12 41 37 25
Pa' mg kg-1 77 65 49 78
K, g kg-1 0.29 0.35 0.29 0.29
Ca, g kg"I 6.4 4.0 3.0
Mg, g kg"1 1.1 0.35 0.48
No. samples 184 186 181 185
+ Mehlich I-extractable nutrients.
+ Pw (water-extractable P), Pa, K, Ca, and Mg (0.5 M acetic
Mehlich I-extractable Zn and Cu from Lauderhill muck were
calculated from 84 observations.
pHs measurements. This method estimates the activity of Hions in a soil suspension in the presence of the salt added (0.01 M CaCl2) to approximate a constant ionic strength of the soil regardless of past management, mineralogical composition, and natural fertility level (McLean, 1982). Measurement of soil pHs with 0.01 M CaCl2 decreased soil reaction in all samples 0.3 to 0.4 pH units (Table 3-3).
The Mehlich-I procedure extracted higher amounts of Ca, Mg, and K, than the 0.5 M acetic acid from the EREC Soil Testing Laboratory (Table 3-3). Soil x concentrations varied from 0.22 to 0.46 g kg"', while Ka varied from 0.29 to 0.35 g kg"1. Soil PMI concentrations varied from 32 to 94 mg kg"1 and Pa varied from 49 to 78 mg P kg"1 with Torry muck showing the highest and Okeelanta the lowest concentration. Torry muck is one of the most productive soils in the EAA. Total P content in these soils (2310 mg P kg'I of soil) is three to four times higher than total P content in Lauderhill and Pahokee mucks, and more than six times higher than total P content in the Okeelanta muck (Appendix A). Water extractable-P ranged from 12 to 41 mg kg-I with Pahokee showing the highest and Lauderhill the lowest concentrations. The low Pw values shown by the Lauderhill muck is probably due to the closeness of the limestone bedrock to the soil surface increasing sorption and precipitation of P by exchangeable Ca or CaCO3.
Most of the micronutrients (B, Cu, Fe, Mn, Mo, and Zn) have been found to be deficient in organic soils for plant growth and their availability is greatly affected by soil pH (Lucas, 1982). Soil ZnMI concentrations ranged from 1.2 to 11 mg kg"1 with Lauderhill muck yielding the lowest concentration. Lauderhill muck also had the highest pHw. Zinc deficiency has been identified in several crops grown on organic soils, which is more evident when the pHw is above 6.5 or when acid organic soils are limed (Lucas, 1982). Soil CuMN concentrations ranged from 0.28 to 0.55 mg kg"1. Copper is usually deficient when virgin Histosols are brought into production (Allison et al., 1927). Mehlich Iextractable Mn and Fe varied from 0.95 to 28 and 2.9 to 20 mg kg'I, respectively, with Lauderhill muck showing the lowest concentration, as expected. Manganese is greatly dependent on soil reaction. Serious Mn deficiencies in many crops occur at pH ranges of 6.3 to 7.5. Anderson and Ulloa (1989) reported Mn deficiency in sugarcane fields above pH
Sample probability distributions of the majority of the parameters were lognormal as determined by the KolmogorovSmirnoff D statistics. Therefore, most of the geostatistical analyses were performed on log-transformed data. Isotropic (direction-independent) and anisotropic
(direction-dependent) semi-variograms were calculated for each parameter at each location. Key parameters for direction-dependent and direction-independent semivariograms for the four soil series studied are given in Tables 3-4 through 3-6, and Tables 3-7 through 3-10, respectively. The intercept (nugget variance) is the estimate of r at h = 0 and provides an indication of shortdistance variation. The range is the distance (h) at which r reaches the maximum value (sill). The sill often approximates the sample variance (Journel and Huijbregts, 1978). The range (distance, m) is interpreted as the diameter of the zone of influence and represents the average maximum distance over which a soil property of two samples is related (Yost et al., 1982a). The range provides an estimate of areas of similarities.
Ideally the semi-variogram should pass through the origin when the distance h = 0. However, many soil properties, such as pHw, MnmI and Cm, have nonzero semivariances as h decreases to zero (Tables 3-5, 3-6, and 3-8). This nonzero variance is called the nugget variance or nugget effect (Journel and Huijbregts, 1978). The nugget effect represents unexplained variance, usually caused by measurement error or microvariability that could not be identified at the scale of sampling used. In this study, the nugget effect gives an indication of the variability that occurs in the field a distances shorter than 10 m. The
Table 3-4. Spatial dependence of selected soil chemical properties from the surface 15 cm. of a Lauderhill muck.
Direction+ Intercept+ Sill variance (in)
------------------- pH-H20---------------------Isotropic -0.0042 0.0592 0.0668 96
N S -0.0001 0.0514 0.0668 74
NE SW -0.0046 0.0550 0.0668 89
E W -0.0102 0.0708 0.0668 118
NW SE 0.0032 0.0688 0.0668 135
------------------ pH-Ca12--------------------Isotropic -0.0045 0.0017 0.0710 97
N S -0.0018 0.0515 0.0710 69
NE SW -0.0058 0.0571 0.0710 90
E W -0.0127 0.0757 0.0710 118
NW SE 0.0019 0.0744 0.0710 147
------- Log Mehlich I-extractable Mg-------Isotropic -0.0010 0.0218 0.0218 107
N S -0.0014 0.0136 0.0218 60
NE SW -0.0020 0.0282 0.0218 120
E W -0.0072 0.0326 0.0218 125
NW SE 0.0021 0.0182 0.0218 122
+ Spherical semi-variograns were calculated from 184
+ The negative intercepts may be the result of extrapolations
of the semi-variograms at distances smaller than 10 mn.
Table 3-5. Spatial dependence of selected soil chemical properties from the surface 15 cm of a Pahokee muck.
General Range 5%- of+
Direction+ Intercept Sill variance (in) Sill
------------------- pH-H20---------------------Isotropic -0.0033 0.0542 0.0607 109
N S 0.0094 0.0349 0.0607 141 27
NE SW -0.0028 0.0503 0.0607 117
E W -0.0133 0.0700 0.0607 114
NW SE -0.0019 0.0576 0.0607 102
------------------ pH-Ca12--------------------Isotropic -0.0054 0.0475 0.0524 106
N S 0.0092 0.0524 Unbounded
NE SW -0.0032 0.0423 0.0524 113
E W -0.0152 0.0643 0.0524 108
NW SE -0.0022 0.0521 0.0524 108
---------- Mehlich I-extractable Mn----------Isotropic 2.172 4.522 4.400 102 48
N S 2.271 3.602 4.400 59 63
NE SW 2.294 4.531 4.400 93 51
E W 2.539 5.641 4.400 115 27
NW SE 2.446 4.400 Unbounded
+ spherical semi-variograms were calculated from 186
+ Percent of sill of semi-variograms with negative intercepts
were not calculated. The negative intercepts may be the
result of extrapolations of the semi-variogram at
distances smaller than 10 m.
Table 3-6. Spatial dependence of selected chemical properties from the surface 15 cm of an Okeelanta muck.
Nugget General Range %of+
Direction+ variance Sill variance (m) sill
------------------- pH-H20--------------------Isotropic 0.0057 0.0450 0.0443 81 13
N S -0.0037 0.0378 0.0443 52
NE SW 0.0074 0.0489 0.0443 103 15
E W 0.0054 0.0560 0.0443 104 10
NW SE 0.0013 0.0417 0.0443 64 3
------------------ pH-Ca12-------------------Isotropic -0.0015 0.0433 0.0403 67
N S -0.0043 0.0357 0.0403 52
NE SW 0.0030 0.0488 0.0403 97 6
E W 0.0010 0.0541 0.0403 89 2
NW SE -0.0053 0.0405 0.0403 52
-------- Log Nehlich I-extractable Fe---------Isotropic 0.0130 0.2810 0.2678 105 5
N S -0.0031 0.0169 0.2678 70
NE SW 0.0003 0.3290 0.2678 121 < 1
E W -0.0061 0.3610 0.2678 127
NW SE 0.0435 0.1340 0.2678 81 32
+ Spherical semi-variograms were calculated from 181
+ Percent of sill from semi-variograms with negative
intercepts were not calculated. Negative intercepts may be the result of extrapolation of the semi-variograms at
distances smaller than 10 m.
Table 3-7. Spatial dependence of selected soil chemical properties of a Lauderhill muck based on isotropic semivariograms, 0-15 cm.
Equation Range % of
Variable+ type Intercept Sill Slope (M) sill
-- Analytical Research Laboratory, Gainesville+ -Log Ca Spherical -0.0043 0.0272 130
Log K Linear 0.0268 0.0007 Unbounded
Log P Linear 0.0180 0.0008 Unbounded
Log Mn Spherical 0.0544 0.1027 103 53
Log Fe Spherical 0.0079 0.0636 145 12
--------- Soil Testing Laboratory, EREC--------Log Pw Linear 0.0196 0.0011 Unbounded
Log Pa Linear 0.0075 0.0006 Unbounded
Log K Linear 0.0246 0.0005 Unbounded
+ Semi-variograms were calculated from 184 observations. + Mehlich I-extractable nutrients. Pw (water-extractable P), Pa and K (0.5 M acetic acid
Table 3-8. Spatial dependence of selected soil chemical properties of a Pahokee muck based on isotropic semivariograms, 0-15 cm.
Equation Range % of
Variable+ type Intercept Sill Slope (m) sill
-- Analytical Research Laboratory, Gainesville+ -Log Ca Spherical -0.0026 0.0239 95
Log Mg Spherical -0.0010 0.0279 127
Log K Linear 0.2550 0.0003 Unbounded
Log p Spherical 0.0486 0.1083 115 45
Log Zn Spherical 0.0499 0.0989 72 50
Log Cu No pattern 0.0132 0.0132 < 10 100
Log Fe Spherical 0.0016 0.0180 64 9
--------- Soil Testing Laboratory, EREC--------PW Spherical 60.97 112.9 122 54
Log Pa Spherical 0.0375 0.0834 128 45
Log K Linear 0.2473 0.0003 Unbounded
Ca Spherical 0.0772 0.2604 87 30
Log Mg Spherical 0.0038 0.0215 149 18
t Semi-variograms were calculated from 186 observations. + Mehlich I-extractable nutrients. Pw (water-extractable P), Pa' K, Ca, and Mg (0.5 M acetic
Table 3-9. Spatial dependence of selected soil chemical properties of an Okeelanta muck based on isotropic semivariograms, 0-15 cm.
Equation Range % off
Variable+ type Intercept Sill Slope (M) sill
-- Analytical Research Laboratory, Gainesville+ -Ca Spherical -0.3960 3.770 103
Log Mg Spherical -0.0416 0.227 84
Log K Spherical 0.0703 0.260 157 27
Log P Spherical -0.0283 0.545 138
Zn Spherical 3.509 18.8 115 19
Log Cu Spherical 0.0356 0.134 75 27
Log Mn Spherical -0.0670 0.479 156
--------- Soil Testing Laboratory, EREC--------Log PW Linear 0.0031 0.0025 Unbounded
Log Pa Spherical 0.0313 0.183 133 17
Log K Linear 0.0940 0.0010 Unbounded
Log Ca Spherical 0.0044 0.859 104 < 1
Log Mg Spherical 0.0044 0.094 il 5
+ Semi-variograms were calculated from 181 observations. + Mehlich I-extractable nutrients. P, (water-extractable P), Pa' K, Ca, and Mg (0.5 M acetic
Percent sill from semi-variograms with negative intercepts were not calculated. Negative intercepts may be the result of extrapolation of the semi-variograms at distances smaller
than 10 m.
Table 3-10. Spatial dependence of selected chemical properties of a Torry muck based on isotropic semi-variograms, 0-15 cm.
Equation Nugget Range % of,
Variable+ type variance Sill Slope (M) sill
-- Analytical Research Laboratory, Gainesville+ -pH-H20 Spherical -0.0250 0.1062 135
pH-CaCl2 Spherical -0.0285 0.1081 129
Ca Linear 0.0710 0.0016 Unbounded
Log Mg Spherical -0.0059 0.0333 138
K Linear 0.0076 0.0001 Unbounded
P Spherical 52.17 291.7 144 18
Zn Linear 0.3033 0.0568 Unbounded
Log Cu Spherical 0.0003 0.0701 83 < 1
Log Mn Spherical -0.0713 0.4122 131
Fe Spherical 0.6979 2.675 136 26
Soil Testing Laboratory, EREC--------Pw Spherical 2.468 39.58 131 6
Log Pa Spherical -0.0033 0.0445 141
Log K Linear 0.0548 0.0004 Unbounded
Log Ca Spherical -0.0154 0.0689 151
Log Mg Spherical -0.0149 0.0699 142
+ Semi-variograms were calculated from 185 observations. + Mehlich I-extractable nutrients. Pw (water-extractable P), Pa, K, Ca, and Mg (0.5 M acetic
Percent of sill from semi-variograms with negative
intercepts were not calculated. Negative intercepts may be the result of extrapolation of the semi-variograms at
distances smaller than 10 m.
nugget variance can also be expressed as a percentage of the sill (Tables 3-8, 3-9, and 3-10) to compare the relative size of the nugget effect among the chemical properties. Data from those tables show that nugget variances of soil chemical properties from these organic soils range from 0 to 100% of the sill.
Water-pH and pHs were the only well structured semivariograms that showed anisotropy (direction-dependency) in the Lauderhill, Pahokee, and Okeelanta mucks (Tables 3-4, 3-5, and 3-6). Mehlich I-extractable Mg, Mn, and Fe also showed anisotropy in the Lauderhill, Pahokee, and Okeelanta mucks, respectively. Semi-variogram structure occurs when there is an increase of the semi-variance to a maximum value. These semi-variograms have distinctive nugget variances (intercept), ranges, and sills (Figs. 3-2, 3-3, 3-4, and 3-5). Other semi-variograms are shown in Appendix B. Anisotropy indicates that the variability of selected soil properties changed with direction. Semi-variograms in the N-S direction (perpendicular to the road) for pHw, pHs, and Mgm, and MnMI from the Lauderhill and Okeelanta mucks showed the shortest range (74, 69, and 60 m, for Lauderhill and 52, 52, and 70 m for Okeelanta, respectively) than the isotropic and the rest of the directional semi-variograms (Tables 3-4 and 3-6). Soil pHw of samples perpendicular to the road separated by less than 74 m in the Lauderhill muck and less than 52 m in the Okeelanta muck are spatially
0 LO 0
El r-I 0
Z 3 E Q4
0 F-I o 0
Fl -I.cn 44
m 0 >
LO ce) CM 0
q C q q C!
C) 0 0 0 0 0
LO LO z
z 1 13 z
F El 0
co 13 cu fo
C 13 11
El 0 13 0
F1 11 El p
13 cl 44
LO 13 LO 3:
t'- to LO I r cr) CM 0 N (0 LO 'gr M N 0
C) C) 0 0 0 0 0 0 C) C 0 0 0 C) 4-4
ci 6 c; ci ci C5 C5 d C; C5 C5 6 6 6 0
0 0 0
0 0 H
C\j N 4
Cl El El El C
LO El U)
0 1: -P
0 (D 11 a) z
-00 0 w
El 13 a 04
0 0 H
13 LO El Ul) 41
I I I I 1 0 1 1 \ .I
LO Ce) C\j co (D It C\j 0
0 0 0 0 q q q
6 ci C5 ci C; 0 CD 0
m 0 1
LQ LO LO LO cr) LO
U*) cys C\i
eOUBIJEA-IWBS c In
0 04 C\j
C:) CD CD P4
C) C.) ca ca
C) C) 44
F- (D LO N co LO It cr) C\j 0
m 0 0
P4 cz cu
LO LO U) Cf) tn CM
co rl- (0 LO -q (Y) C\l
C OOMIJEA-IWOS cn
dependent, indicating that samples should be separated by at least 74 and 52 m when taken for an unbiased estimate of the average soil pHw. Semi-variograms in the E-W direction (perpendicular to the ditches) for the same chemical properties showed ranges of 118, 118, and 125 m, for the Lauderhill muck and 104, 89, and 127 m for the Okeelanta muck, respectively.
These results suggest that road spoils have a greater effect than ditch spoils over these chemical properties in these two soils. A network of roads, canals, and ditches is generally constructed to facilitate commercial crop production. The spoils from the marl/limestone bedrock that are dug and spread on the surface during ditch and road construction is one of the main controlling factors affecting the behavior of several soil chemical properties.
Semi-variograms of MnMI from the Pahokee muck (Table 3-5) showed the largest unexplained variation of all the anisotropic semi-variograms. Semi-variograms in the N-S (perpendicular to the road) and NE-SW directions showed the shortest ranges of spatial influence (59 and 93 m, respectively) and nugget variances that account for 63 and 51% of the sill, respectively. These nugget variances imply that there is more than 50% of unexplained random variance due to measurement errors or spatial variability at distances shorter than 10 m.
Parameters for isotropic (direction-independent) semivariograms for the four soil series are given on Tables 3-7, 3-8, 3-9, and 3-10, respectively. The range of spatial dependence of PMI ranged from 115 to 144 m, with Pahokee muck showing the shortest range (Fig. 3-6). In contrast, the isotropic semi-variogram for the Lauderhill muck increased unbounded with distance and had no sill. An unbounded semi-variogram means an increase in variation with distance. This kind of semi-variogram indicates the presence of strong directional trends in the property under study (nonstationarity). Water-extractable P, showed wellstructured isotropic semi-variograms only for the Pahokee and Torry mucks, with ranges of spatial dependence of 122 and 131 m, respectively (Fig. 3-7). Isotropic semivariograms for Lauderhill and Okeelanta mucks were unbounded. These results indicate that Pw values from samples taken at distances less than 122 and 131 m from the Pahokee and Torry mucks sites are spatially dependent and will become more alike by reducing the distance between them. In contrast, isotropic semi-variograms from the Lauderhill and Okeelanta mucks are unbounded with distance describing a increasing variation of Pw content across the fields.
Soil 1. and Ka showed unbounded semi-variograms in all soils with the exception of the Okeelanta muck that fitted a spherical model to Km (Table 3-9). Mehlich I-extractable
0 0 4J
0 0 C) 0
0 0 1--- U-) 0
0 LO 0
0 CZ 0
CL 0 4
0 0 0 44
0 0 0 E 0 E M
0 0 r-A
0 C) 0 (D rq
0 C) -0 0 0
0 0 0 0 0 c
OC, 0 0
0 0 C9 C) 4
LO 0 0 LO 0
1\ 0 tp
11 10 0 44
CM co (0 0 0 0 0 0 0 0 0
T7 6 q q q 0 0 w 0 0 0 0 0
C) 0 C) 0 0 0 0 0 N N
0 0 U2
C\l C\l 44
0 0 cz
-0 LO 0
0 0 4
0 0 0 0 tp
0 0 0- 4
0 0 0 0 0 E
0 CD 0 C:) CD
0 cz 0
0 C) C.) 0
-0 C) 0 r-I
0 C: 0 C
0 0 ca
0 00 cl 0 0
0 0 0 rl
0 0 0 0 H
0 C) 0 04
8 LO 0 LO 0
0 0 0 4
000 C, 00 00 00 00 1
(0 IT N M (0 co (D It C\j (D
: -: -: C; q q c; ci d 6
0 0 0 0 0 0 0
0 0 C) 0 r:3
=- U-) C, LO
0 CZ 0 0
CL 0 0 4
0 0 0 0 44
0 0 E 0 E ul
0 0 H
0 0 C) 0 0 C) CD H
0 C) 0 C) C)
0 c 0 c 0
o 0 W W
0 0 0 0
0 0 0 0
0 0 0
0 0 0
0 LO 0 0 LO 4
0 0 0
00 000 00\000
0 0 0 4
0 0 0 44
1 I I I to
0 0 0 0 0 0 0 0 0 0 0 0 0 (4-4
(D C\l 0 CC) (D It LO It cf) N 0
E)OUL'IMA-IMS 8OUUIRA-IWE)S 3
0 CD 0 cz
0 0 LO 0
0 T- 0 a)- LO 4
o 0 CD tyl
CIS 0 0
% 0 H
0 0 0 4
00 E 0 0 E m
0 C) (D 0 CD >
0 C) 0 0
0 C: 0 C)
ca 0 ca
0 -U) 0
0 0 0
0 0 9 0
U) 0 -0 P4
0 0 LO 0
0 0 0 0
I I I N L I 0
ce) LO cli LO LO 0 0
0 0 C; ci
aOU131J8A-lWeS 83UUIJBA-IWE)S Cl)
Zn and Cu from the Pahokee muck showed more variation than those from the Okeelanta and Torry mucks (Tables 3-8, 3-9, and 3-10). Soil ZnMI and CUMI from the Pahokee muck showed ranges of 72 and < 10 m and nugget variances of 50 and 100% of the sill, respectively (Table 3-8). These nugget variances imply that there is 50% measurement errors or short-range variability in the field for Zn and a pure nugget effect for Cu (Fig. 3-8). Pure nugget effect is an indication of a completely random distribution of the variable at the sampling interval used. A well structured semi-variogram with a range and sill may still exist, but samples will have to be taken at shorter distances to detect it.
In contrast, semi-variograms of ZnMI and CUMi from the Okeelanta and Torry mucks showed larger ranges of spatial influence and smaller nugget variances. Soil CUMI from the Okeelanta and Torry muck exhibited ranges of 75 and 83 m and nugget variances of 27 and < 1% of the sill. Nugget variance from the Torry muck suggests that there is little measurement error or short range variability in the field for CuMI.
Zones of spatial dependence shown in the isotropic semi-variograms of MnMI ranged from 102 m in the Pahokee muck to 156 m in the Okeelanta muck (Tables 3-5 and 3-9). Lauderhill and Pahokee mucks showed the shortest ranges (103 and 102 m, respectively) and the largest nugget variances
< C) LO rg
< < -0
< cz 0
< < < 1 4<
< < 0
< 0 Clq
< LO >1
< 0 a)
< LO 4
< < 4
-0 < 41
LO C\j LO LO 0 <
0 < < E 0
eoueum-UGS < CD
C) < 0
< =- Lo <
CZ 0 4
< CL C\j o I T o 0
< C; q q q H
< < E C) C) 0 0 4
< eOUelAeA-IWGS fa
< oa) >
< 4< <
< < <
< -0 H
< < U') P4
< < 4-)
< < 0
C14 co (D It N M 0
0 0 0 0 0
(53 and 48%, respectively) (Figs. 3-4 and 3-9). These results agree with the chemical behavior of this element. Lindsay (1972) reported that Mn availability decreases 100 fold for each unit of pH increase. Page (1962) showed that an increase of pH also enhances the production of Mn soil organic matter complexes which also render Mn less available. Therefore, as expected, the isotropic semivariogram from the Lauderhill and Pahokee mucks showed the largest nugget variance. Although, the overall range of pHW from the Pahokee muck was not as large as that from the Lauderhill muck (Table 3-3), there were several areas across
the f ield that showed high pHW values. Mehlich Iextractable Fe yielded well structured semi-variograms with ranges of spatial dependence from 64 to 145 m and nugget variances less than 30% of the sill.
Several of the soil chemical properties studied showed well structured semi-variograms. These results show that selected soil chemical properties from the organic soils of the EAA are spatially dependent. Understanding of changes on soil chemical properties with distance provide valuable information concerning nutrient variability and crop production. This study shows that semi-variograms can be used to detect within-field variability in the organic soils of the EAA. The structure of spatial dependence displayed
CD 0' 4-)
LO LO 4
o o (1) 0
-0 o L) rA
C) -0 0
LO U) U) IT LO (Y) LO C\l LO It ce) C\l 0
L6 li cli C14 c; C=;
CD C) (Cl
70- LO Lo p
L) (D I
-0 a C:
Iq C\l CC) (C) LL M C) C)
7 7 6 q C 8 r-: C)
C; o c; c; 0)
0 0 0 1
8OUL'IRA-IWE)S E)OUBIMA-IM S (Y)
by semi-variograms gives information about the direction and range of spatial dependence of the soil chemical properties.
Soil pH, was the only property that showed anisotropic variability in all soils with the exception of Torry muck. Other elements that displayed anisotropy were MgM1, MnMI, and FeMI* Results from the anisotropic semi-variograms showed that road spoils had a greater influence than ditch spoils on these properties as judged by the shorter ranges of spatial dependence. Soil MnMI from the Pahokee muck was the element exhibiting the highest spatial variability.
Most of the soil properties studied showed isotropic spatial variability. The structure of spatial dependence displayed by the isotropic semi-variograms varied with each soil series. Information from these semi-variograms can contribute to the understanding of soil genesis and behavior of soil properties in the Histosols of the EAA.
The range of spatial dependence of the majority of the soil properties was > 100 m in all locations. These results suggest that samples collected at a distance of 100 m or less are spatially dependent. In contrast, some properties such as Pw, Pa, and Ka from the Lauderhill muck showed a steady increase in soil variability with distance. In general, Torry muck was the most uniform location, while that of the Lauderhill muck was the most variable.
The data obtained from this study provides us with
important information about the within-field variability of
these soils that can be used to improve certain management practices. These results helped to identify the soil series as well as the particular chemical properties that exhibited higher variability in the field. With this information we have a better guideline to improve soil-sampling designs for soil testing.
SOIL SPATIAL VARIABILITY OF SELECTED CHEMICAL PROPERTIES
IN THE EVERGLADES AGRICULTURAL AREA. II. BLOCK KRIGING
The Everglades Agricultural Area (EAA) comprises an
area of approximately 313,638 ha of fertile organic soils in south Florida. The EAA was drained at the beginning of the century, and a series of roads and canals were built before the area could be used for the commercial production of winter vegetables and sugarcane. Topographically, the EAA is a relatively uniform area with the majority of its fields being leveled before being put into production. However, the introduction of canals and roads as well as the constant process of soil subsidence added additional sources of soil variability to the area.
Soil spatial variability is a naturally occurring feature that is important in the identification of soil properties relative to soil productivity (Ball et al., 1968; Cline, 1944). Since, soil properties gradually change across the landscape, the investigation of the variability of soil chemical properties with distance has become more important over the past few years. When an observation gives some information as to the value or magnitude of its
neighbor, such data are spatially dependent. When variables are spatially dependent, classical statistical analyses are no longer valid. One method of handling spatially dependent variables is using the theory of regionalized variables (Matheron, 1963). The successful application of this theory to problems in mining, geology, and hydrology led to the more popular name of geostatistics, of which kriging is a main branch (Krige, 1966; Delhomme, 1978).
Kriging is a geostatistical technique of making
optimal, unbiased estimates of regionalized variables at unsampled locations using the structural properties of the semi-variogram and the initial set of samples. Kriging is essentially a means of weighted local averages in which the weights are chosen so as to give unbiased estimates while at the same time minimizing the estimation variance. In this sense, kriging is an optimum interpolator (Webster and Burgess, 1983; Webster, 1985). The estimates can be made for points, that is, volumes of soil of the same size and shape as those on which the measurements were made, or for larger blocks of a particular shape. Applications and examples of the use of kriging are extensively discussed by Journel and Huijbregts (1978), Gambolati and Volpi (1979), Vieira et al. (1983), and Trangmar et al. (1985).
In recent years, soil scientists have used kriging to estimate soil parameters (Hajrasuliha et al., 1980; Vieira et al., 1981; Yost et al., 1982b; Xu and Webster, 1984; and
Gaston et al., 1990). Mapping of soil properties is improved by using kriging as a method of estimating soil property values when compared to least-squares estimation techniques (Burgess and Webster, 1980a, 1980b).
The objectives of this study were i) to use
geostatistical methods to map spatial variability of soil chemical properties from selected organic soils of the EAA, ii) to use block-kriged contour maps to evaluate the effect of road and ditch spoils on within-field soil variability, and iii) to determine the minimum number of soil samples required for accurately measuring particular soil properties.
Materials and Methods
Soils samples from the upper 15 cm of four different organic soils were used in this study (Table 3-1). Procedures used in the collection, preparation, and laboratory analyses of soil samples are described in Chapter 3.
Mean, range, variance, and coefficient of variation were computed using the Statistical Analysis System (SAS Institute Inc., 1982a, 1982b). The Univariate procedure was used to test for normality. Variables that failed to meet
the normality test were log transformed before geostatistical analysis.
The data from each location were also analyzed to calculate the number of samples needed to obtain a soil value that can satisfy a prescribed margin of error within an acceptable confidence level (a) (Cochran, 1977). Each field (A, B) within a location was separately analyzed to account for variability due to cropping. Data values close to the road and ditch spoils (approximately 40 and 20 m from roads and ditches, respectively) were not used in these analyses. To estimate the number of samples, the overall mean and the sample standard deviation from all soil parameters in each field were calculated. Using these estimates, the number of samples that need to be taken from the field were calculated by the formula
n =(t*S)2/(d)2 (4-1)
t =Probability (1-a/2) (4-2)
where n = required number of samples
t = standard normal deviate corresponding to the
level of significance a.
S = sample standard deviation
d = sample mean *%relative error Geostatistical Analyses
Geostatistical approaches applied in this study are based on the theory of regionalized variables, developed originally for mining and mineral exploration (Matheron,
1963). A comprehensive study of geostatistics has been published by David (1977) and Journel and Huijbregts (1978). Semi-variograms
The principal tool of geostatistics in the analysis of spatially dependent data is the semi-variogram. Semivariograms were computed using the program VAR5, Version
2.0, which is a part of a geostatistical package developed by the University of Hawaii (Yost et al., 1989). Semi-variograms were calculated from the formula
F(h) = 1/2N (h) Z [Y(x) Y(x+h)]2 (4-3)
where N is the number of sample pairs at each distance interval (h), and Y(x) and Y(x+h) are random variables corresponding to sites separated by a distance h. Semivariograms were averaged over 180 degrees (direction independent) and over 90 degrees in four directions (N-S, NE-SW, E-W, and NW-SE) to test for anisotropy (direction dependent). Weighted least square methods as implemented in SAS nonlinear procedures were used to model linear and spherical semi-variograms.
Once appropriate semi-variograms were determined, interpolations were made by block kriging (Yost et al., 1989). Kriging is the second major tool of geostatistics. In this method, the kriged estimate of a property Z(X0), at a location X0 is the weighed average of the values
Z(0) = Z fiZ(X1) (4-4)
where n is the number of neighboring samples Z(Xi) and f are weights applied to each Z(Xi). The weights are chosen so that (i) they sum to 1, thereby assuring lack of bias, and (ii) the estimation variance is minimized
E [Z(X0)-Z(X0)] = 0 (4-5)
a k = VA[(X0)-Z(X0)] = minimum (4-6)
One of the main advantages of kriging over other interpolation mapping methods is that it provides the interpolation error estimates, which can be used to build reliability maps. Contour maps of selected chemical properties from each location were made using the program SURFER Version 4. To make the contour maps, 340 kriged values for each soil parameter were estimated in blocks 20 x 20 m at intervals of 20 m.
Results and Discussion
Despite uniform field topography in the EAA, there was marked soil heterogeneity among the soil series studied. Description of the soil series and related crops, and sampling pattern used in this study are given in Tables 3-1 and 3-2 and Fig. 3-1, respectively. one of the main objectives of this study was to measure the influence of
road and canal spoils on soil variability. To accomplish that objective, calculation of semi-variograms was based only on the large sampling pattern (14 ha area, 100 samples), including all samples close to the road and canals. The small sampling pattern was not used in these analyses due to limitations in the kriging program regarding the maximum number of neighboring points the program could handle.
Calculated semi-variograms did not indicate direction dependency as those shown by semi-variograms of some chemical properties in Chapter 3. Isotropic semi-variograms from each soil series and chemical property are shown in Tables 4-1 to 4-4. For the majority of soil chemical properties analyzed, the model that fit the semi-variances best was the linear semi-variogram, with moderate gradient and large intercepts. Spherical semi-variograms were also fitted to some chemical properties, especially in the Torry muck (Table 4-4). The presence of large nugget variances (intercepts) indicated that a large percentage of the soil spatial variability was not detected at the smaller lag distance used (43 m). Thus, more sampling would be needed to calculate more reliable semi-variograms. Block-kriging and Mapping
Using the structural information derived from the fitted semi-variograms (Tables 4-1 to 4-4), values of
Table 4-1. Effect of road and ditch spoils on the spatial dependence of selected soil chemical properties from the surface 15 cm of a Lauderhill muck. soil Equation+ Nugget Range
property type variance Sill Slope (in)
Analytical Research Laboratory, Gainesville+ pH-H20 Linear 0.0309 0.00097
pH-CaC12 Linear 0.0312 0.00095
Log Ca Linear 0.0042 0.00021
Log Mg Linear 0.0184 0.00011
Log K Spherical 0.0351 0.1542 144
Log P Linear 0.0936 0.00083
Log Zn Linear 0.1630 -0.00067
Log Cu Linear 0.0426 0.00208
Mn Spherical 0.0359 0.0843 138
Log Fe Linear 0.0131 0.00070
------Soil Testing Laboratory, EREC5---------Log PW Spherical 0.0551 0.2296 -135
Log Pa Spherical 0.0171 0.1198 -137
Log K Spherical 0.0325 0.1259 129
Ca Linear 0.1844 0.00113
Log Mg Linear 0.0061 -0.00002
Log Pw/Pa Linear 0.0682 -0.00057
+ Isotropic seini-variograms were calculated from 100
+ Mehlich I-extractable nutrients. Ew (water-extractable) Pa, K, Ca, and Mg (0. 5 M acetic
Table 4-2. Effect of road and ditch spoils on the spatial dependence of selected soil chemical properties from the surface 15 cm of a Pahokee muck. Soil Equation+ Nugget Range
property type variance Sill Slope (M)
-- Analytical Research Laboratory, Gainesville+ pH-H20 Linear 0.1291 0.00092
pH-CaCl2 Linear 0.1322 0.00095
Ca Spherical 0.7722 1.450 105
Mg Linear 0.0323 0.00004
Log K Linear 0.2281 0.00030
Log P Spherical 0.0279 0.1618 129
Log Zn Linear 0.1900 0.00037
Log Cu Spherical 0.0026 0.0118 122
Mn Linear 4.899 0.01674
Log Fe Linear 0.0539 0.00004
-------- Soil Testing Laboratory, EREC-------Pw Spherical 46.32 103.55 127
Log Pa Linear 0.0308 0.00101
Log K Linear 0.1979 0.00034
Ca Spherical 0.1997 0.3281 116
Log Mg Linear 0.0145 0.00004
Log P./Pa Linear 0.0375 0.00124
+ Isotropic semi-variograms were calculated from 100
+ Mehlich I-extractable nutrients. Pw (water-extractable P), Pa, K, Ca, and Mg (0.5 M acetic
Table 4-3. Effect of road and ditch spoils on the spatial dependence of selected soil chemical properties from the surface 15 cm of an Okeelanta muck. Soil property+ Nugget variance Slope
-- Analytical Research Laboratory, Gainesville+ -pH-H20 0.5035 0.00270
pH-CaCl2 0.5239 0.00244
Ca 2.673 0.01361
Log Mg 0.1431 0.00179
Log K 0.2660 0.00096
Log P 0.5510 0.00263
Log Zn 0.9667 0.00408
Log Cu 0.0865 0.00030
Log Mn 0.7142 0.00410
Log Fe 0.5123 0.00208
-------- Soil Testing Laboratory, EREC--------Log Pw 0.3640 0.00209
Log Pa 0.0665 0.00099
K 0.2507 0.00071
Ca 0.5134 0.00397
Mg 0.0455 0.00038
P./Pa 0.0403 0.00010
+ Linear isotropic semi-variograms were calculated from 100
+ Mehlich I-extractable nutrients. Pw (water-extractable P), Pat K, Ca, and Mg (0.5 M acetic
Table 4-4. Effect of road and ditch spoils on the spatial dependence of selected soil chemical properties from the surface 15 cm of a Torry muck. Soil Equation+ Nugget Range
property type variance Sill Slope (m)
Analytical Research Laboratory, Gainesville+ -pH-H20 Linear 0.1404 0.00188
pH-CaC12 Linear 0.1633 0.00189
Ca Spherical 0.0006 0.4601 136
Log Mg Spherical 0.0060 0.0475 124
Log K Linear 0.0751 0.00040
Log P Spherical 0.0749 0.2061 151
Zn Linear 4.621 0.06944
Log Cu Spherical 0.0073 0.0401 156
Log Mn Linear 0.2143 0.00488
Log Fe Linear 0.1176 0.00125
Soil Testing Laboratory, EREC--------Log Pw Spherical 0.0294 0.1484 124
Log Pa Linear 0.0162 0.00022
Log K Spherical 0.0860 0.1679 128
Ca Spherical 0.0086 1.126 142
Log Mg Spherical 0.0077 0.1202 140
Log Pw/Pa Spherical 0.0375 0.1965 141
+ Isotropic semi-variograms were calculated from 100
+ Mehlich I-extractable nutrients. Pw (water-extractable P), P,, K, Ca, and Mg (0.5 M acetic
selected chemical properties were interpolated at 340 locations using block-kriging. Interpolated values were used to produce contour maps of selected chemical properties. Contour maps of pHW f rom each soil series are shown in Fig. 4-1. Soil pHW is relative uniform within the fields, with most of the variability occurring close to the roads and ditches. Lauderhill muck was the only soil that showed some variability within the field, probably due to the fact that these are shallow soils with limestone bedrock close to the surface. However, the effect of a
marl/limestone ditch and road spoils on pHW variability is evident in all soils. Visual analysis of pHW contour maps indicates that road spoils significantly affect a strip of soil approximately 40 to 50 m wide (Fig. 4-1). Soil variability due to ditches is not as obvious as the road spoil effect. Ditch spoils varies from location to location due to the different ditch management practices at each farm. A common practice in the EAA that increases soil variability due to ditches is the regular cleaning of ditches and canals, and the subsequent dumping and spread of the spoils to one side of the field. Contour maps of the Okeelanta and Torry mucks show this effect on the lefthand side of the middle ditches (field A) and close to the ditch on the righthand side of the Torry muck (field B). Soil pHW variability displayed by the Lauderhill muck in the middle ditch is due by a combination of road and ditch spoils.
0 0 0 0 0 0 0 0
0 00 CN 0 d- co N 0
rn N (0 o CN co 0
0 0 Q)
CD U) (0 -P
0 rn n V) (d
0 w 0 0
0 0 0 C
uj 04 C14
0 (::) a o V( C)
01) L 2 00
0 0 .9 a LO
to (D H
0 0 0 0 0 0 0 0 0 0 0
0 1- co N co 0 1- co N to
n 04 n N
(W) aouolslc] (W) @:DUDISIC]
0 0 a 0 0 0 0 p
0- 00 N 0 0 -q- 00 04 0 44
0 0 460 0
0 0 C) 4.50 0 44
C*4 cli "A 'C44 E
Lij 9*9 0 0 C)
0 co Ld
6.io C\ 0 c 0
0 0 0 0 r. .
CD CD (0 00 (D 0 tp
0 0 0 0 0 0 0 0 0 0
0 00 N co 0 4- co C14 co
(W) aouolsl(] (W) GOUD;sl(]
Visual analysis of these contour maps suggests that an area at least 40 to 50 mn from the road and 25 to 30 mn on either side of the ditches should be avoided when samples are taken
to test the pHW of these f ields.
Contour maps of Pw (Fig. 4-2) showed also P variability throughout the field due to road and ditch effect, but not as dramatic as the variability shown by Pm, (Fig. 4-3) and P (Fig. 4-4). Water-extractable P concentrations from the Okeelanta muck varied from about 60 and 20 mng kg'1 from the middle of f ields A and B, to about 10 mng kg-1 in the areas close to the road. Similar results were observed in the other three soils (Fig. 4-2) The drastic decrease in P close to the road is due to the presence of high concentrations of free carbonates (Griffin and Jurinak, 1973; Anderson, 1990b). Effects due to ditches were observed only in the Okeelanta and Torry mucks. These
results agree with the pHW contour maps (Fig. 4-1) that showed more variability due to ditches in the same two soils.
These maps are also helpful in showing P variability due to previous cropping or crops at the time of sampling. Fields A and B from the Lauderhill muck had a history of sugarcane production, but at the time of sampling, field A was in fallow and field B was in sod production. The contour map from this location shows that field B exhibited PW values twice as high as those from field A. Generally,
0 0 0 0 0 0 0 0
0 co N 0 0 14- co N 0
n N (0 0 K) N (0 0 -P
0 0 0 0
(D 7-A., (D 0 oe (D
0 PI) 7p, ::s ro 4-)
c C 0
0 0 0 0
LLI C*4 N >- C14 C14
y ry 8 C)
2 S cl 2-4
0 0 0 Of 0 Q) 4J
= 00 2 0 00 U
n 'A 44
0 0 Oz 0 41 0
0 0 oe
0 0 0
0 0 0 0 0 0 0 0
0 14- co N Q0 0 co 04 o (d
C14 n N
(W) OOUDISI(] (LU) QOUD SIC]
0 't co CN 0 co N 0 7
to 0 r) cli
y 0 0
(D to (D (D
K) r) ro
0 0 0
C14 C4 C, E
bi 0 0 0 0 P4
2 (2 m
0 0 0 0 Z -ri
W (D (0 (0 0 tm
0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 14- Go C14 (0 0 Id- 00 (14 (D -,4
to 04 to C14 ()
(w) GOUDISI(] (W) BOUD;SIC] -P 0
0 0 0 0
0 "o- m N 0 0 co 04 0
V) cq Q0 0 C4 to 0 -P
CD :: -.. (D
to PO -P
0 0 0 0
LLJ N (N 04
CIO C) 00 00 0
a_ X 85 0
04 04 o
CD CD 55 (D En
0 0 0 0
co 0 00 cq CD
n C14 ro C14
(W) @DUIDISIC] (W) aOUDISIC] 0
0 0 a 0 0 0
0 co C'4 0 -o- co C14
0 0 0
w CD ED
PO r) n 4
0 rl\ 0 0
V) L77E(- IVUR n rl) 10 tn a
__j cq .9
'o m UJ co
0 I(MOMOOO)IN k u co C)
0 0 M
0 0 0 C
45 (0 0 tp
0 0 0 0
0 0 0 0
co C14 (JO 0 co N co c4
K) 04 I
(W) @DUDISI(] (W) 9OU04sl(] qr 0
0 0 0 0 rd
0 1.0- 00 N 0 0 'd- 00 N 0 0)
ro C14 4-)
vq 65 ro -P
0 0 0 0
99 V) re) 94
gg ? S p
0 96 0
99 C" N E
ocf -- clll
ft o 0 (D
co 00 0 m 00 U 4-J
< Fa- 0 C)
0 0 0
99 N C14
0 0 0 0
o pVT IN I H
0 0 0 0 0 0 0 0 0 0 0
0 'd- co N o Go C14 (D
n C'4 n C%4
(W) GOUDISIC] (W) DOUD;Slc] 4
0 0 0 0 0 0
0 1- co N 0 0 4- co CN 0
PO C'4 co 0 ro 04 CD 0
0 0 0 0 D
0 cli 0 (a
0 0 C), 0 0
C14 C14 C-4
............... .0 0 0
00 00 LLJ oo 10 0
0 0 0 0 co 4
0 04 04 C14
99 0 zn
lj4 4J Z
co 0 tm
915 A05 U -H
k-q I 4
0 0 0 A4
0 0 0 0 0 0
co C14 co 0 lq- m 04 CD
(W) aouo;sl(j (W) GOUDISI(] 0
sod production requires more P (approximately 50 kg P ha&1 yr'1) than sugarcane production (15 to 40 kg P ha&' yr"') in organic soil (Anderson, 1987; Anderson, 1990a).
Fields from the Okeelanta location had a similar situation. Both fields from this location were fallow, however, field A was previously in sweetcorn production (harvested 1 wk before sampling), while field B was previously in sugarcane production (fallow during summer).
Results from this location show PW concentrations from field A to be three-times higher than those shown in field B. Usually, P application for sweetcorn production in organic
soils is more intensive (70 to 80 kg P ha-1 yr-1) than for sugarcane production.
Contour maps from PMI showed high variability due to roads, ditches, and cropping (Fig. 4-3). Soil PMI concentrations from the Lauderhill muck varied from about 45 and 65 mg P kg_1 across the middle of fields A and B,
respectively, to approximately 20 mg P kg'1 to the areas adjacent to the road and approximately 25 mg P kg_1 to the areas adjacent to ditches. Similar variability patterns are
shown by the Okeelanta and Torry mucks. Soil PMI values from the Pahokee muck were uniform with the exception of areas adjacent to the road and ditches. This location shows
PMI concentrations as high as 200 mg kg'1 in the lower righthand of field B. This corner is a filled area that was probably used as a loading platform for machinery.
soil P. contour maps showed similar variability
patterns as those shown by PmI (Fig. 4-4). Soil P. showed high variability due to roads, ditches, and cropping. Soil P. variability due to the road and the equipment loading area in the Pahokee muck is prominent. These areas showed unusual high Pa (> 200 mg P kg-1) compared to the rest of the field. Soil P and Pa variability due to cropping were also
evident. Average soil P and Pavalues from field B (sod production) of Lauderhill muck were approximately twice as high as those measured in field B (sugarcane fallow).
Similarly, soil PMI and Pa from field A (sweetcorn fallow) of the Okeelanta muck were approximately twice as high as the concentrations measured in field B (sugarcane fallow).
The effect of road and ditch spoils on P availability is clearly shown in the contour maps of the ratio .P (Fig. 4-5). The Pw/P8 ratios across field A of the Okeelanta muck decreased from 0.80 to 0.24 and 0.40 in the areas adjacent to the road and ditches, respectively. Similar variability patterns were shown by the other three soils. Low Pw/P8 ratios close to roads and ditches are due to low PW and high P8 concentrations in those areas as a result of the presence of high concentrations of freecarbonates (Griffin and Jurinak, 1973; Anderson, 1990b). With the continuous process of soil subsidence in the EAA, larger areas will be affected by the limestone bedrock,
0 0 0 0 0 0 0
0 lq- co 04 0 0 d- co N
V) C.0 n Q0
0 0 0 0
to r) n Ind
0 0 0
LLJ ZL'O lq- t F 4J
IN N C14
0 0 () 0
Go C 00 0 00 00 U
6 ,- z Vo z:
ac 70 0
0 0 0 Pe-0 ZZ 0.20
C14 C14 N
0 0 0 .10 4-)
CV 0 44
o 0.6 (0 co .5r, C7 to 0
0 0 0
0 0 0 0 0 0
0 CID N to 0 14- CID N 0
n C14 m
(W) @Our)lsl(] (W) QouDjsl(]
0 0 0 0 0 0
0 co C14 0 0 "- 00 IN 0
to (D 0 to cli to
0 0 44
ro rl) pe)
0 0 0 "1,
0 0 0 0 (44
IN IN E
CD z w
0. a) P4
00 L.Lj 00 00
0 0 o L
Z/,o 0 r .
o 0 tp
0 0 0 0 0 0 0 0 0 0 0 LO
0 Go N o 0 00 N o
ro 04 I
(W) GOUDISI(] (W) aDUDISI(]