HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (16) Citing Articles via Google Scholar Google Scholar Articles by Mueller, T. G. Articles by Barnhisel, R. I. Search for Related Content PubMed Articles by Mueller, T. G. Articles by Barnhisel, R. I. Agricola Articles by Mueller, T. G. Articles by Barnhisel, R. I. Related Collections Spatial Distribution Statistics Geostatistics Spatial Variability Site-Specific Analysis

SYMPOSIUM PAPERS

Soil Electrical Conductivity Map Variability in Limestone Soils Overlain by Loess

^a Dep. of Agron., Univ. of Kentucky, N-122 Agric. Sci. North, Lexington, KY 40546-0091
^b Dep. of Biosyst. and Agric. Eng., Univ. of Kentucky, 218 C.E. Barnhart, Lexington, KY 40546-0276
^c John Deere Agric. Manage. Solutions, John Deere and Co., 4140 NW 114th St., Urbandale, IA 50322

^* Corresponding author (mueller{at}uky.edu)

Received for publication April 23, 2001.

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Sensors exist that allow rapid mapping of bulk soil electrical conductivity (EC); however, the utility of these sensors for Kentucky producers is unknown. The purpose of this study was to assess the nature and the causes of soil EC variability and to make a first assessment of its potential utility in Kentucky, particularly for fields containing soils derived from limestone residuum overlain by loess. Various geostatistical, correlation, and regression analyses were conducted at seven locations to examine EC map variability. Sensor drift and errors associated with changes in coulter depth were minimal. Bulk soil EC related fairly well with clay content across locations and sample dates (r² = 0.40); however, many site- and time-specific correlations were better. Clay (maximum r² = 0.75), moisture content (maximum r² = 0.76), Ca (maximum r² = 0.67), and Mg (maximum r² = 0.64) were positively correlated with EC, and depth to argillic or cambic horizon (maximum r² = 0.62), depth to fragipan (maximum r² = 0.81), and depth to bedrock (maximum r² = 0.32) were negatively correlated with EC. A multiple-regression model (R² = 0.70) was developed to predict EC that included nine factors: clay, sand, soil moisture, buffer pH, base saturation, Ca, soil temperature, depth to cambic and argillic horizon, and slope. Soil EC variability was spatially structured, and spatial patterns were stable over time; however, the degree to which these patterns could be observed depended on the mapping procedures used. Our research suggested that EC mapping may have utility for Kentucky farmers.

Abbreviations: BpH, buffer pH • BS, base saturation • CEC, cation exchange capacity • DEM, digital elevation model • EC, electrical conductivity • GPS, global positioning system • RSV, relative structural variability

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
SOIL EC IS DETERMINED by standardizing measures of soil conductance (resistance^-1) by the volume of soil through which current travels (Sparks, 1995). Traditionally, soil paste EC has been used to assess soil conductivity (Rhoades et al., 1989), but commercial devices are now available that can be used to rapidly map bulk soil EC across agricultural fields either through direct contact or induction techniques.

To utilize EC maps in agriculture, EC map variability must be understood. Bulk EC map variability has three components: measurement error, variation in bulk soil EC, and variation associated with mapping procedures (e.g., contouring and interpolation). Soil EC measurement error is the difference between the measured EC and true EC of soil and may be caused, for example, by global positioning system (GPS) errors (Sudduth et al., 2001), inconsistent soil–sensor contact, and noise or drift associated with the electronics of the sensor. If the volume of soil through which the current travels varies spatially, measurement error would be introduced because the soil volume in bulk soil EC measurements is assumed constant. The extent to which this basic and critical assumption is met is not known and requires further research. Variation in bulk soil EC may occur at spatial and temporal scales including the microscale (variation in EC at distances or times less than the sampling interval) and macroscale (variation in EC at distances or times equal to or greater than the sampling interval). Errors associated with mapping procedures, such as interpolation, are a function of measurement intensity and have been given considerable attention in the literature for grid soil sampling (e.g., Mueller et al., 2001; Mueller and Pierce, 2003) but not EC mapping. It would be ideal if measurement error, microscale variability, temporal variability, and errors associated with mapping were small and if macroscale spatial variability accounted for the greatest proportion of the EC variability observed.

Many of the factors governing soil EC variability are understood. The spatial and temporal variability of bulk soil EC is affected by the complex movement of electrons through soil. Electrons may travel through the water in soil macropores, along the surfaces of soil minerals (i.e., via exchangeable ions), or through alternating layers of particles and solution (Rhoades et al., 1989). Therefore, multiple factors contribute to soil EC variability, including factors that affect the amount and connectivity of soil water (bulk density, structure, water potential, precipitation, and timing of measurement), soil structure (cementing agents such as clay and organic matter), electrolytes in soil water (salinity, exchangeable ions, and soil water content), and the conductivity of the mineral phase (types and quantity of minerals, degree of isomorphic substitution, and exchangeable ions) (McNeill, 1980), with these factors potentially varying with depth and time.

Because the factors that affect bulk EC are complex and interrelated, interpreting measurements is challenging. However, a number of studies have identified situations where bulk soil EC was related to individual factors that affect soil use, management, and productivity, such as soil water content (Kachanoski et al., 1988; Sheets and Hendrickx, 1995), salinity (Rhoades and Ingvalson, 1971; De Jong et al., 1979; Rhoades and Corwin, 1981; Williams and Hoey, 1987; Hendrickx et al., 1992; McKenzie et al., 1997), soil Ca and Mg concentrations (McBride et al., 1990), clay content (Williams and Hoey, 1987), depth of sand deposition (Kitchen et al., 1996), and depth to claypan (Doolittle et al., 1994). Bulk soil EC map variability has not been adequately described for soils derived from limestone residuum overlain with varying depths of loess, which are characteristic of many agriculturally important soils in Kentucky and throughout the world.

The objectives of this study were to assess soil EC map variability for several agricultural landscapes derived from these parent materials in Kentucky and to evaluate EC mapping as a tool for soil use and management in Kentucky. To be a useful tool, EC must relate with factors of agronomic importance, EC maps must be spatially structured, and overall spatial patterns must have temporal stability. Geostatistical, correlation, and regression analyses were conducted to examine soil EC map variability.

	MATERIALS AND METHODS

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
This research was conducted at several test fields in Kentucky—two in Fayette County, one in Hardin County, three in Shelby County, and one in Trigg County. All of the fields contained soils derived from limestone residuum overlain with varying depths of loess (Table 1). The sites were located in some of the major physiographic regions in Kentucky: the Inner Bluegrass (Fayette-1, 38°1'33'' N, 84°29'2'' W; Fayette-2, 38°1'9'' N, 84°29'59'' W), Outer Bluegrass (Shelby-1, 38°1'32'' N, 85°27'40'' W; Shelby-2, 38°20'26'' N, 85°13'45'' W; Shelby-3, 38°16'48'' N, 85°9'19'' W), and Western Pennyroyal (Hardin-1, 37°31'7'' N, 85°56'13'' W; Trigg-2, 36°53'51'' N, 87°42'24'' W).

View larger version (38K): [in this window] [in a new window]   Fig. 1. Veris Technologies 3100 soil electrical conductivity (EC) mapping system. Adapted from Veris Technologies (2002) with permission.

Stationary Measurements
Temporal variability of soil EC measurements was assessed at the Fayette-1 experimental site (Table 2). One hundred EC measurements were collected over a 16-h period while the sensor was stationary. Empirical temporal semivariograms were calculated using Variowin (Pannatier, 1996) for shallow and deep EC.

Linear transects were established in the Hardin, Shelby-2, and Shelby-3 locations. Shallow and deep soil EC were measured along these transects on the dates indicated in Table 2. Detailed characterizations of several points along each transect were obtained, and the various observed quantities were compared with soil EC measurements. At each point in all transects, five 2.5-cm-diam. soil cores were obtained to a depth of 15 cm. The subsamples were combined, dried under forced air at 25°C for 3 d, and then crushed to pass through a 2-mm sieve. Sand, silt, and clay content (micropipette method); cation exchange capacity (CEC; NH₄ saturation); and percentage base saturation (BS; total bases CEC^-1) for pH (1:1 soil/water mixture), buffer pH (BpH; SMP buffer), P, K, Ca, and Mg (Mehlich III extractable) were determined by the soil-testing laboratory of the Division of Regulatory Services, University of Kentucky. The following quantities were observed at each point in the transect:

• Soil temperature (0–20 cm).
  
• Volumetric water content (0–12 cm) using a HydroSense soil moisture sensor (Decagon Devices, Pullman, WA).
  
• Depth to the argillic (Crider, Faywood, Lowell, Nicholson, Shelbyville, and Vertrees series) or cambic (Lindside and Nolin series) horizons as determined by two soil scientists.
  
• Terrain attributes using real-time kinematic (RTK) GPS. Because the EC measurement points did not always coincide with the 4- by 4-m predicted grid points, bilinear interpolation was used to estimate the terrain attributes for each point.
  
• Depth to bedrock using a tile probe (Shelby-2 only).
  
• Depth to fragipan using a soil probe (Shelby-3 only).
  

Correlation analyses were conducted for each location and each measurement date. Stepwise multiple regression ( = 0.15) (SAS Inst., 2001; PROC REG, SELECTION = STEPWISE) was conducted across dates and locations as the response variable. When shallow EC was the response variable, candidate variables for entry into model included various soil measurements (i.e., moisture, temperature, depth to cambic, and argillic horizon), soil analyses results (i.e., sand, silt, clay, pH, BpH, Ca, Mg, BS, and CEC), and terrain attributes [i.e., slope percentage and aspect (degrees from north)]. Natural log and squares of these variables were also included as candidate regressors.

The final model was developed with a series of executions of the PROC REG procedure. After the first execution, we studied the model and regression procedure to determine whether any variable should be removed from the group of candidate variables during the next successive execution of the PROC REG procedure. The procedure was executed again with the updated group of candidate variables for entry into the model. Only one variable was removed between successive executions of PROC REG. The criteria to remove variables were (i) if at least one intercept-adjusted variance inflation factor was >5, (ii) if the parameters of a particular regression variable were not reasonable, and (iii) if the SAS procedure added and then removed variables to and from the model during the stepwise procedure. The variance inflation criterion was used to detect multicollinearity problems (Gunst and Mason, 1980). Because we included soil variables, natural logs, and squares as candidate variables, many of the early models included unlikely combinations of these variables.

Next, we devised a method to evaluate the relative importance of the different factors or groups of factors in the model. For each regressor variable retained in the final model, we calculated the contribution to the fitted values (i.e., predicted value, ) for each observed value of y (regressor coefficient x value of regressor variable). We will refer to this as a partial estimate. Next, we grouped the regressor variables into factor categories. For example, if multiple soil chemistry variables (e.g., pH, Ca, and BS) were retained in the model, then a partial estimate for the soil chemistry factor categories was defined as the sum of the partial estimates for the regressors. In some cases, the factor categories included only one variable. The empirical variance for these partial estimates for each factor category was calculated, and this was used as an index to determine the relative importance of the factor categories (e.g., soil chemistry, soil moisture, and soil texture). The rational for this analysis was that EC variability is a function of the variability of the partial estimates. Those factor categories that contribute more variability to estimates have a greater role in determining EC values.

Computation of the partial estimates for individual regressors is (square of the partial-regression coefficient) x (sample variance of the regressor variable) or, alternatively, (square of standard partial-regression coefficient) x (sample variance of the response variable). For a factor for which the partial estimate is the sum of partial estimates for a subset of regressor variables, the shortcut computational method is (transpose of vector of partial-regression coefficients for the subset) x (sample variance–covariance matrix of subset of regressor variables) x (vector of partial-regression coefficients for the subset) or, equivalently, [(transpose of vector of standard partial-regression coefficients for the subset) x (sample correlation matrix of subset of regressor variables) x (vector of standard partial-regression coefficients for the subset)] x (sample variance of the response variable). Standard partial-regression coefficients have been used as indicators of relative importance of regressor variables (Snedecor and Cochran, 1967; Damon and Harvey, 1987). This interpretation is precise only when variables are uncorrelated and becomes increasingly ambiguous as multicollinearity increases.

Dynamic Measurements
Whole-field EC data were collected from the Hardin, Shelby-1, Shelby-2, and Shelby-3 fields on the dates indicated in Table 2. Each field was traversed at a speed of 10.5 km h^-1 with approximately 7.5 m between passes. Data points were logged every second (approximately 3 m of travel). Data with negative values indicated poor sensor–soil contact and were removed. Omnidirectional semivariograms were calculated and modeled using Variowin (Pannatier, 1996) for each measurement date for both shallow- and deep-EC measurements. In addition to a nugget variance (measurement error and/or unobserved microscale variability; Cressie, 1993), one or two additional structures (either spherical or exponential) were modeled. The spherical or exponential models were described with two parameters, the range of spatial correlation and the sill. The range of the spatial correlation measures the distance beyond which two observations are spatially uncorrelated. For second-order stationary spatial processes, the sill can regarded as the variance of observations that are separated by a distance exceeding the range of spatial correlation. In models with a nugget effect, the partial sill represents the difference between the sill and the nugget effect. See Isaaks and Srivastava (1989) or Goovaerts (1997) for more thorough discussions of semivariogram parameters. The relative structural variability (RSV) is the ratio of the partial sill to the total sill (Robertson et al., 1993), which indicates the proportion of the variability that is spatially structured. Whole-field EC measurements were interpolated with ordinary kriging.

	RESULTS AND DISCUSSION

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
We first investigated temporal variability associated with stationary measures of EC. As the time between stationary measurements increased, the temporal semivariance also increased (Fig. 2) . The slight upward trend in semivariance may have been attributable to drift associated with the electronics of the device; however, it may also be due in part to the light precipitation that occurred through the duration of the experiment. The size of the semivariance values close to the origin (approximately the nugget variance) was an important quantity in this test because it represented measurement error, small-scale temporal variability, or both (Cressie, 1993). We concluded that for this set of stationary measurements, measurement errors were not significant.

View larger version (20K): [in this window] [in a new window]   Fig. 2. Temporal semivariograms for shallow (ECshallow) and deep (ECdeep) electrical conductivity at the Fayette-1 experimental site.

View larger version (18K): [in this window] [in a new window]   Fig. 3. Semivariogram for shallow (ECshallow) and deep (ECdeep) electrical conductivity with semivariance as a function of the coulter depth lag.

Correlations were generally stronger for shallow (0–30 cm) than deep (0–90 cm) EC (Tables 3 and 4) because chemical analyses, soil moisture, and soil temperature values relied on surface soil sampling or measurements (0–15, 0–12, and 0–20 cm, respectively). Nevertheless, deep-EC measurements were still fairly well correlated with the shallow soil property measurements because surface and subsurface properties often are related. Furthermore, deep-EC measurements are affected by EC both at the surface depths (0–30 cm) and subsurface (30–90 cm). Nevertheless, relationships between shallow soil properties and deep EC measurements were interpreted very cautiously throughout this study.

View larger version (32K): [in this window] [in a new window]   Fig. 4. Shallow electrical conductivity (ECshallow) vs. soil moisture, clay, Ca, and depth to argillic or cambic horizons for four locations. The solid circles indicate the first and open circles the second date for each location listed in Table 2 under the Transect EC Measures column. The coefficient of determination (r2) is given for simple linear regressions. The data used for the regression relationship between EC and depth to argillic or cambic horizon only include those observations where the depth to argillic or cambic horizon was <50 cm (solid circles).

For the Shelby-3 field, EC variability also appeared to be related to depth to fragipan (Fig. 5 and Table 3). Within the Nicholson map units in the Shelby-3 field, depth to fragipan was strongly related to shallow EC on 5 May 2000 but not on 23 Feb. 2001 (Fig. 5). Differences in soil moisture between sampling dates likely explained the change in this relationship. The fragipan had little effect on EC variability in February 2001 when the entire soil profile was very wet (volumetric water content = 40%), but in May 2000, when the soil was considerably dryer (volumetric water content = 30%), a volume of soil directly above the fragipan likely remained wet, acting as a conduit for soil electrical current through the soil. Fragipan depth is an important agronomic variable because it affects soil drainage and N chemistry. Ultimately, it can greatly impact yield (Frye et al., 1983). It may be possible to use EC to make accurate depth-to-fragipan maps in landscapes where fragipan is the dominant factor that explains EC variability; however, timing of measurements will be critical as indicated by the poor temporal stability of the relationship between EC and depth to fragipan (Fig. 5).

View larger version (24K): [in this window] [in a new window]   Fig. 5. Shallow (ECshallow) and deep (ECdeep) electrical conductivity vs. depth to fragipan and rock.

Consistent with the findings of Doolittle et al. (1994) and Sudduth et al. (1999), EC was related to depth to cambic or argillic horizons to approximately 50 cm (Fig. 4). Depth to cambic or argillic horizon is an important soil management variable and of great interest. In Kentucky, depth to the cambic or argillic horizon, also referred to as topsoil depth, was used to establish criterion (<15, 15–20, and >20 cm of topsoil) for variable corn seeding where net returns increased by as much as $80 ha^-1 over fixed-rate seeding treatment (Barnhisel et al., 1996). If EC could be used to produce accurate maps of topsoil depth, it could be a useful tool for variable rate seeding for producers in Kentucky; however, the variation in slope and intercept with time and location (Fig. 4) suggests that site-specific calibration of EC data may be required. It may be possible to predict the slopes of these relationships if soil clay and moisture contents were known with depth; however, additional research is needed to validate this.

The lack of stability in the relationship between EC and soil properties across locations and time was generally disappointing (Table 4); however, shallow EC explained substantial variability in clay content:

[1]

This relationship could be used to obtain a rough estimate of clay content. The clay maps would potentially have large absolute errors, but because the within-field and by-sampling-date r² values were generally large (Fig. 4), relative errors would likely be reasonably small. It would be possible to reduce absolute errors in clay maps if the slope and intercept were calibrated with site- and time-specific soil sampling and particle size analysis. The costs associated with calibration would be reduced if surface texture were assessed with the feel method, which can be done by properly trained personnel.

The adequacy of the correlation and regression relationships for management will be application dependent (e.g., yield map interpretation and co-kriging). Unfortunately, thresholds for adequacy are known for few applications. For example, for co-kriging to be more effective than ordinary kriging, the absolute value of the simple linear correlation coefficients between EC and the secondary data value of interest should be 0.70 (i.e., r² 0.49) (Ahmed and De Marsily, 1987). Based on the within-field and by-sampling-dates analyses (Table 3 and Fig. 4), EC correlations with volumetric water content, clay, depth to cambic or argillic horizon, and depth to fragipan were sufficiently large. Therefore, it is reasonable to expect that co-kriging could be used to enhance spatial estimates of these soil properties. However, less complicated procedures such as multiple regression may perform as well as co-kriging (Mueller and Pierce, 2003). Very accurate maps may be highly desirable for management of high-value crops or for urban land use planning; however, they may be too costly for site-specific management of grain crops.

For yield map interpretation, precise soil property maps may not be necessary. If EC relates to factors that limit soil use and management, and if these factors affect yield, then yield maps may be used to calibrate EC maps. If practitioners were to have a basic understanding of EC variability for a given area, EC mapping would be an even more useful tool for understanding factors that limit yield. We found that areas with thin topsoil tend to be those that have been more highly eroded and, therefore, have higher clay contents at the surface. For these soils, EC values tended to be greater. Soils with less surface clay, deeper soils, soils with greater topsoil thickness, and soils that are deeper to fragipan tended to have lower EC values. Boundary-line analysis has been used to evaluate yield potential with EC as proposed by Kitchen et al. (1999). If yield potential were known, then areas in fields that yield under potential could be assessed for problems or sampled site specifically. The difference between yield potential and measured yield could be used as a management opportunity index. Specific methodologies are needed for assessing yield potential and management opportunity using bulk soil EC in Kentucky.

Multivariate relationships between soil and landscape properties and shallow EC were described with a 13-variable multiple-regression model (R² = 0.70; Table 5). Multicollinearity, which was not an issue in this study, occurs when strong relationships exist between independent variables and can be problematic for regression analysis, especially with regard to the estimation of regression coefficients. While significant correlations existed between some of the independent variables used in the models, the intercept-adjusted variance inflation factors were all < 3.0 and condition indices < 4.0. Regression parameters can be affected when variance inflation factors (Gunst and Mason, 1980) and condition indices (Belsley et al., 1980; SAS Inst., 2001) are > 10. Variances of the partial estimates for each factor category (i.e., soil texture, soil moisture, soil chemistry, soil temperature, depth to restrictive layer, and landscape) contributed to overall EC variability and thus were indicative of the relative importance of the factor categories (Table 5). Interactions between the factor categories may also contribute to overall EC variability; however, the low variance inflation factors and condition indices indicated that these interactions were small. Further, preliminary data analysis indicated that the inclusion of interaction terms in the regression analysis did not greatly enhance overall predictability. Consequently, they were not introduced into the model. We, therefore, conclude that the variances of the partial estimates provided a useful index for ranking the importance of the variables in this analysis.

After the soil texture and moisture variables, the next most important factor category was the soil chemistry category (BpH, BS, and Ca) (Table 3). The variance of the summed partial estimates for the soil chemistry factors was relatively low (4.2 mS² m^-2) compared with partial estimates for the soil texture variables (9.1 mS² m^-2) and moisture (7.8 mS² m^-2). Soil chemical properties may vary tremendously across agricultural fields, and it may not be possible to make accurate estimates with interpolation even at small sampling intervals (i.e., 30-m grids) (Mueller et al., 2001). If EC were to be used to primarily understand differences in soil texture, then variation in EC related to soil chemical variables would be considered noise. Fortunately, the relationship with soil chemical properties was not stronger.

Soil temperature was included in the model (Table 5) although within-field correlations between temperature and EC were generally fair to poor (Table 3), and the correlations across locations and dates were nonexistent (r = -0.08; Table 4). Soil temperature is known to affect soil EC variability (McNeill, 1980), but the variability of other factors (e.g., texture and clay) most likely masked the effect of soil temperature in the univariate analyses. However, in the multivariate analysis, the effect of other variables was taken into account, thereby resolving the effect of temperature on shallow EC. While soil temperature was not one of the more dominant factors affecting EC variability (variance of parameter estimate = 1.4 mS² m^-2; Table 5), contributions of temperature to EC variability were still important (the smallest contribution observed was 0.1 mS m^-1 for 0.9°C, the minimum soil temperature observed, and 5.0 mS m^-1 for 37°C, the maximum temperature observed).

The contribution of the topographic slope factor to predicted EC variability was smaller than expected (Table 5). For example, the variance of the partial estimates for slope² was only 0.3 mS² m^-2, and the partial contribution to EC was only 3.0 mS m^-1 for the largest slope observed (14.1%) in Shelby-2. The effect of slope on EC variability, however, was larger than indicated by the slope² model parameter. More severely sloping areas tended to be truncated soils (eroded) and had higher clay content than flatter regions of fields; therefore, the model took slope into account in three ways: the slope, depth to cambic or argillic horizon, and clay factors.

We hoped the multivariate analysis would provide information that would help us improve spatial estimates of soil properties. We considered whether it would be possible to enhance estimates of clay across locations and sampling dates if a suite of sensors were used to take into account some of the factors that affect EC (Table 5). Recent advances in GPS (Clark and Yao, 2000) have made it possible to obtain high-resolution DEMs from which slope is readily derived. On-the-go soil moisture sensors (Price et al., 1990) have not received much attention in the literature recently; however, at least one real-time commercial soil moisture sensor is available (Retrokool, Berkley CA). It is also feasible to measure soil temperature on the go, but we know of no commercially available real-time temperature sensors. Our attempts to improve estimates of clay content by including hypothetical sensor variables in the model (shallow EC, deep EC, soil moisture, soil temperature, and slope) and their transformations (lognormal and square) were not very successful. The final regression model to predict clay included shallow EC, soil temperature², and slope², and the R² value was 0.44, explaining only 4% more variability than EC alone (i.e., r² = 0.40) (Eq. [1]). This should be explored further using nonlinear models.

We next considered the spatial structure of soil EC maps. The RSV values were equal to or exceeded 50% in all cases (Table 6). In other words, most of the variation in the EC maps was attributable to factors other than measurement error and small-scale spatial variability that occurs 3 m. The RSV values were greater than 80% in more than half of all cases. The ranges of spatial correlation were large, excluding the map for the deep-EC measurements at the Trigg County site. However, the range and RSV values were difficult to interpret for models with two structures (not including the nugget structure). The next logical questions that must be addressed are whether the spatial structure of EC variability is adequate for management and at what scale must EC be measured to make EC maps of adequate quality. These are difficult questions and will require additional field experiments.

View larger version (31K): [in this window] [in a new window]   Fig. 6. Histograms for deep electrical conductivity (EC) at the Hardin location.

View larger version (83K): [in this window] [in a new window]   Fig. 7. Shallow electrical conductivity (EC) maps at the Hardin county location with data from four different dates and created with three mapping techniques: (a) point maps with equally spaced contour intervals, (b) point maps with contour intervals having the same number of points within each class, and (c) interpolated maps created with kriging (exponential semivariogram models) and with equally spaced contour intervals.

	CONCLUSIONS

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

	ACKNOWLEDGMENTS

 
This material is based on work supported by the Cooperative State Research, Education and Extension Service, USD, under Agreement no. 99-34408-7561 and the Kentucky Corn Growers Association. We are grateful to Mike Ellis, Wayne McAtee, and Charlie Stuecker for providing access to their farms to conduct this research. We greatly appreciate help in the field from the late Grant Thomas and advice from Ken Wells. We appreciate the review of this manuscript by Wilber Frye. Thanks to Danna Reid, Frank Sikora, and the staff of the University of Kentucky soil-testing lab for conducting the physical and chemical analyses for this study. We also gratefully acknowledge Blazan Mijatovic and Meng Fengxuan for their assistance at one of the Fayette County experimental sites. A special thanks to Rod Grusy for help with site selection.

	NOTES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Contrib. no. 01-06-61 from the Kentucky Agric. Exp. Stn., Lexington, KY.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 

• Ahmed, S., and G. De Marsily. 1987. Comparison of geostatistical methods for estimating transmissivity using data on transmissivity and specific capacity. Water Resour. Res. 23:1717–1737.
• Barnhisel, R.I., M.J. Bitzer, J.H. Grove, and S.A. Shearer. 1996. Agronomic benefits of varying corn seed populations: Central Kentucky study. p. 957– 965. In P.C. Robert et al. (ed.) Precision agriculture. Proc. Int. Conf., 3rd, Minneapolis, MN. 23–26 June 1996. ASA, CSSA, and SSSA, Madison. WI.
• Belsley, D.A., E. Kuh, and R.E. Welsch. 1980. Regression diagnostics. John Wiley & Sons, New York.
• Clark, R.C., and H. Yao. 2000. Elevation of submeter and 2 to 5 meter accuracy GPS receivers to develop digital elevation models. Precis. Agric. 2(2):189–200.
• Cressie, N. 1993. Statistics for spatial data. Revised ed. John Wiley & Sons, New York.
• Damon, R.A., Jr., and W.R. Harvey. 1987. Experimental design, ANOVA, and regression. Harper & Row, New York.
• De Jong, E., A.K. Ballantyne, D.R. Cameron, and D.W.L. Read. 1979. Measurement of apparent electrical conductivity of soils by an electromagnetic induction probe to aid in salinity surveys. Soil Sci. Soc. Am. J. 43:810–812.[Abstract/Free Full Text]
• Doolittle, J.A., K.A. Sudduth, N.R. Kitchen, and S.J. Indorante. 1994. Estimating depths to claypans using electromagnetic induction methods. J. Soil Water Conserv. 49:572–575.
• Frye, W.W., L.W. Murdock, and R.L. Blevins. 1983. Corn yield–fragipan depth relations on a Zanesville soil. Soil Sci. Soc. Am. J. 47:1043–1045.[Abstract/Free Full Text]
• Goovaerts. P. 1997. Geostatistics for natural resource evaluation. Oxford Univ. Press, New York.
• Gunst, R.F., and R.L. Mason. 1980. Regression analysis and its application: A data-oriented approach. Marcel Dekker, New York.
• Hendrickx, J.M.H., B. Baerends, Z.I. Raza, M. Sadig, and M.A. Chaudhry. 1992. Soil salinity assessment by electrical induction of irrigated land. Soil Sci. Soc. Am. J. 56:1933–1941.[Abstract/Free Full Text]
• Isaaks, E., and R. Srivastava. 1989. An introduction to applied geostatistics. Oxford Univ. Press, New York.
• Kachanoski, R.G., E.G. Gregorich, and I.J. Van Wesenbeeck. 1988. Estimating spatial variations of soil water content using no contacting electromagnetic inductive methods. Can. J. Soil Sci. 68:715–722.
• Kitchen, N.R., K.A. Sudduth, and S.T. Drummond. 1996. Mapping of sand deposition from 1993 Midwest floods with electromagnetic induction measurements. J. Soil Water Conserv. 51(4):336–340.
• Kitchen, N.R., K.A. Sudduth, and S.T. Drummond. 1999. Soil electrical conductivity as a crop productivity measure for claypan soils. J. Prod. Agric. 12:607–617.
• McBride, R.A., A.M. Gordon, and S.C. Shrive. 1990. Estimating forest soil quality from terrain measurements of apparent electrical conductivity. Soil Sci. Soc. Am. J. 54:290–293.[Abstract/Free Full Text]
• McKenzie, R.C., R.J. George, S.A. Woods, M.E. Cannon, and D.L. Bennett. 1997. Use of the electromagnetic-induction meter (EM38) as a tool in managing salinisation. Hydrogeol. J. 5:37–50.
• McNeill, J.D. 1980. Electrical conductivity of soils and rocks. Tech. Note TN-5. Geonics Limited, Mississauga, ON, Canada.
• Mueller, T.G., and F.J. Pierce. 2003. Soil carbon maps: Enhancing spatial estimates with simple terrain attributes at multiple scales. Soil Sci. Soc. Am. J. 67:258–267.[Abstract/Free Full Text]
• Mueller, T.G., F.J. Pierce, O. Schabenberger, and D.D. Warncke. 2001. Map quality for site-specific management. Soil Sci. Soc. Am. J. 65:1547–1558.[Abstract/Free Full Text]
• Pannatier, Y. 1996. Variowin: Software for spatial data analysis in 2D. Springer, New York.
• Price, R.R., Ximing Huang, and L.D. Gaultney. 1990. Development of a soil moisture sensor. ASAE Paper 90–3555. ASAE, St. Joseph, MI.
• Rhoades, J.D., and D.L. Corwin. 1981. Determining soil electrical conductivity–depth relations using an inductive electromagnetic soil conductivity meter. Soil Sci. Soc. Am. J. 45:255–260.[Abstract/Free Full Text]
• Rhoades, J.D., and R.D. Ingvalson. 1971. Determining salinity in field soils with soil resistance measurements. Soil Sci. Soc. Am. Proc. 35:54–60.
• Rhoades, J.D., N.A. Manteghi, P.J. Shouse, and W.J. Alves. 1989. Soil electrical conductivity and soil salinity: New formulations and calibrations. Soil Sci. Soc. Am. J. 53:433–439.[Abstract/Free Full Text]
• Robertson, G.P., J.R. Crum, and B.G. Ellis. 1993. The spatial variability of soil resources following long-term disturbance. Oecologia 96:451–456.
• SAS Institute. 2001. SAS OnlineDoc: Version 8. SAS Inst., Cary, NC.
• Sheets, K.R., and J.M.H. Hendrickx. 1995. Noninvasive soil water content measurement using electromagnetic induction. Water Resour. Res. 2401–2409.
• Snedecor, W.G., and W.G. Cochran. 1967. Statistical methods. 6th ed. The Iowa State Univ. Press, Ames.
• Sparks, D.L. 1995. The chemistry of saline and sodic soils. Environmental soil chemistry. Academic Press, San Diego.
• Sudduth, K.A., S.T. Drummond, and N.R. Kitchen. 2001. Accuracy issues in electromagnetic induction sensing of soil electrical conductivity for precision agriculture. Comput. Electron. Agric. 31:239–264.
• Sudduth, K.A., N.R. Kitchen, and S.T. Drummond. 1999. Soil conductivity sensing on claypan soils: Comparison of electromagnetic induction and direct methods. p. 971–990. In P.C. Robert et al. (ed.) Precision agriculture. Proc. Int. Conf., 4th, St. Paul, MN. 19–22 July 1998. ASA, CSSA, and SSSA, Madison, WI.
• Veris Technologies. 2002. Veris Technologies [Online]. Available at http://www.veristech.com/ (verified 30 Jan. 2003). Veris Technol., Salina, KS.
• Williams, B.G., and D. Hoey. 1987. The use of electromagnetic induction to detect the spatial variability of the salt and clay contents of soil. Aust. J. Soil Res. 25:21–27.

This article has been cited by other articles:

				M. S. Grigera, R. A. Drijber, K. M. Eskridge, and B. J. Wienhold Soil Microbial Biomass Relationships with Organic Matter Fractions in a Nebraska Corn Field Mapped using Apparent Electrical Conductivity Soil Sci. Soc. Am. J., August 3, 2006; 70(5): 1480 - 1488. [Abstract] [Full Text] [PDF]

				J. A. Terra, J. N. Shaw, D. W. Reeves, R. L. Raper, E. van Santen, E. B. Schwab, and P. L. Mask Soil Management and Landscape Variability Affects Field-Scale Cotton Productivity Soil Sci. Soc. Am. J., December 2, 2005; 70(1): 98 - 107. [Abstract] [Full Text] [PDF]

				W. K. Jung, N. R. Kitchen, K. A. Sudduth, R. J. Kremer, and P. P. Motavalli Relationship of Apparent Soil Electrical Conductivity to Claypan Soil Properties Soil Sci. Soc. Am. J., May 6, 2005; 69(3): 883 - 892. [Abstract] [Full Text] [PDF]

				J. A. Guretzky, K. J. Moore, C. L. Burras, and E. C. Brummer Distribution of Legumes along Gradients of Slope and Soil Electrical Conductivity in Pastures Agron. J., March 1, 2004; 96(2): 547 - 555. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (16) Citing Articles via Google Scholar Google Scholar Articles by Mueller, T. G. Articles by Barnhisel, R. I. Search for Related Content PubMed Articles by Mueller, T. G. Articles by Barnhisel, R. I. Agricola Articles by Mueller, T. G. Articles by Barnhisel, R. I. Related Collections Spatial Distribution Statistics Geostatistics Spatial Variability