PDF version
for printing

© 2005 Plant Management Network.
Accepted for publication 28 July 2005. Published 28 September 2005.

Interpreting Yield Variability with Electrical Conductivity and Terrain Attributes across a Central Kentucky Landscape

B. G. Sears, Department of Plant and Soil Sciences, Robinson Station, University of Kentucky, Quicksand 41339-9081; and B. Mijatovic, T. G. Mueller, and R. I. Barnhisel, Department of Plant and Soil Sciences, University of Kentucky, Lexington 40546-0091

Corresponding author: T. G. Mueller. mueller@uky.edu

Sears, B. G., Mijatovic, B., Mueller, T. G., and Barnhisel R. I. 2005. Interpreting yield variability with electrical conductivity and terrain attributes across a Central Kentucky landscape. Online. Crop Management doi:10.1094/CM-2005-0928-01-RV.

Abstract

Interpretation of yield maps for management requires an adequate understanding of the factors that influence the variability of crop productivity. The objective of this study was to investigate the use of precision agriculture technologies for interpreting corn grain yield variability across a Central Kentucky landscape. Yields were measured in 2003 and 2004 on 75 plots distributed along four transects. Bulk soil electrical conductivity (EC), depth to bedrock, and topsoil clay content were also measured at each site. Digital elevation models (DEMs) were used to calculate terrain attributes. Slope, aspect, and curvature did not relate well with yield in either year. The specific catchment area and bulk soil EC were useful for explaining yield variability in 2004, when 8.6 inches of rain over eight-days resulted in the flooding of lowland areas early in the growing season. Multiple regression models accounted for 65% of the variability in 2004 but only 20% in 2003. Our findings suggest that specific catchment area and EC can be useful for understanding yield variability in years rain events cause flooding. Potentially, these tools could also be used to aid in artificial drainage and land-use decisions.

Introduction

For yield maps to be used as reliable tools for aiding with agronomic decisions, spatially varying soil and landscape factors that limit crop productivity should be adequately known (18). Terrain attributes and bulk soil electrical conductivity (EC) may provide to producers some of the necessary information that will allow them to better understand the factors governing yield variability. Digital elevation models (DEMs) are used to calculate terrain attributes. DEMs are datasets containing elevation values arranged in regular grids. Although they can be obtained on-line from the USGS (20), most yield variability studies have been conducted with DEMs generated by interpolating elevation data collected with survey grade GPS systems. Terrain attributes are related to crop productivity because landscape position drives processes such as erosion, deposition, infiltration, and overland flow (10).

Precipitation is perhaps the most important factor determining the relationship between yield and terrain attributes (10,12). Many studies have found statistically significant correlations between corn yield and simple terrain attributes such as elevation, slope, aspect, and curvature (4,6,7,10,11,12,17,19). However, complex terrain attributes may provide even more useful information about factors governing yield variability.

Upslope contributing area is an attribute derived by estimating water flow from upland areas. Kravchenko and Bullock (12) found that this attribute was weakly correlated with corn and soybean yield (-0.17 ≤ r ≤ 0.30) over four years and across eight Indiana and Illinois locations. Other terrain attributes have been studied in drier regions. A stronger linear relationship was found between wheat yield and specific catchment area (r² = 0.60) (19), where specific catchment area is the upslope contributing area normalized by the average flow width and is given as:

Specific Catchment Area =		Upslope Contributing Area		[1]
		Flow Width

Areas with small specific catchment areas would not be expected to have a great potential for flooding or erosion. Significant relationships have also been found between the topographic wetness index and winter barley (6) and winter wheat (4,19) yields. The topographic wetness index (1) is calculated from the expression:

Topographic Wetness Index =	ln		Specific Catchment Area		[2]
			tan β

where β is slope in degrees. Topographic wetness index values will be large for areas with large upslope areas that are relatively flat which tend to become water logged during the growing season. Another terrain attribute is the stream power index estimated as:

       Stream Power Index = Specific Catchment Area × tan β.                [3]

Areas with large stream power indices have a great potential for erosion.

Bulk soil electrical conductivity is the inverse of soil resistivity. It is often greater in eroded areas (5,8) because it tends to be positively correlated with clay content (21) and bulk density (8) and negatively correlated with topsoil thickness (3). Bulk soil EC is often greater in poorly drained soils (13) because it increases with soil moisture (9). A range of positive and negative correlations between EC and grain yield has been observed in multi-year and multi-location studies (11,13). Kravchenko et al. (13) observed negative correlations primarily in years with wet springs and positive correlations in years with dry springs. These reports suggest that the interrelationships between EC, terrain attributes, and yield are complex and must be understood before these precision agriculture variables can be adequately used for interpreting yield variability in corn production systems.

The objective of this experiment was to examine the relationship between corn grain yield, terrain attributes, and soil electrical conductivity measurements across a Central Kentucky landscape.

Site Selection and Description

The study was conducted at the University of Kentucky Spindletop Research Farm. The corn plots were established in areas that had been in fescue for approximately 60 years. The experiment was conducted along transects indicated in Fig. 1 and 2. The centers of the plots are denoted by black circles with white halos. The sites were chosen in order to include a range of soil types, elevation, and slope variation that are common across Inner Bluegrass landscapes of Kentucky.

Fig. 1. The locations of experimental plots along Transects 1, 2, and 3 and aerial photograph draped over DEM (10× vertical exaggeration).

Fig. 2. The locations of experimental plots along Transect 4 and aerial photograph draped over DEM (10× vertical exaggeration).

Soil and Crop Management

Pioneer Hi-Bred 31G98 hybrid corn was planted on 2 June 2003 at approximately 33,000 seeds per acre in 30-inch rows. On 7 May 2004, Pioneer Hi-Bred 34G98RRBT hybrid was planted at approximately 28,000 seeds per acre in 30-inch rows. Herbicides and fertilizer were applied according to the University of Kentucky best management practices for no-till corn production. Corn was harvested for grain on 8 November 2003 and 10 November 2004 with a Massey Ferguson MF8 two-row plot combine harvester equipped with an AgLeader yield monitor system. Yield measurements at the ends of the transects were unreliable and were removed from the analyses.

Soil Electrical Conductivity Measurements

Bulk soil electrical conductivity measurements were collected with a Veris 3100 EC Mapping System (Veris Technologies, Salina, KS) and geo-referenced with a Trimble (Trimble, Sunnyvale, CA) Ag 132 DGPS receiver on April 15 of 2003. Measurements were collected along transects with multiple side-by-side passes. Shallow and deep EC data were collected simultaneously by measuring EC with two different coulter spacings. Average shallow and deep EC values for each experimental plot were obtained with GIS data extraction techniques. The ratio of deep to shallow EC was calculated as recommended by the manufacturer (21).

Soil Sampling and Particle Size Analysis

Depth to bedrock was measured by inserting a marked 60-inch tile probe into the ground approximately 1 m away from either side of the centers of each experimental area. From these same locations, two 3-ft soils cores were obtained with a 1 7/8-inch inside-diameter probe. Particle size analysis was conducted on the topsoil sample using a modified micro-pipette method (2,14) by the University of Kentucky Soil Testing Laboratory, Division of Regulatory Services. Topsoil included the surface A-horizon and transitional horizons (e.g., AE, AB, and EB) but no B-horizons.

Elevation and Terrain Analysis

Elevation measurements were obtained with two survey grade GPS receivers (Trimble AgGPS 132, Sunnyvale, CA). In addition, USGS elevation data were used for regions outside the GPS survey area. The ArcGIS Spatial Analyst software was used to create DEMs with the TOPOTORASTER command. Slope, plan curvature, profile curvature, upslope contributing area, and flow width were calculated with TapesG for Windows (University of Southern California, Los Angeles, CA). These terrain attributes were used to derive the specific catchment area (Equation 1), topographic wetness index (Equation 2), and stream power index (Equation 3).

Software programs use different conventions for reporting profile, plan, and tangential curvature values so readers should cautiously compare the results in this paper to those of others described in the literature. In this study, positive profile curvature values indicated convex slopes and greater erosion would generally be expected in these areas. Negative profile curvature values indicated concave shaped slopes. Greater deposition would be expected in these areas. Plan curvature is the rate of slope change along the horizontal plane and it indicates whether water would be expected to converge or diverge. Tangential curvature is the product of plan curvature and the sine of slope angle (15); therefore tangential curvature values are generally small in flat areas. Negative plan and tangential curvature values occur in valleys where water flow is expected to be convergent. Positive values are associated with divergent flow.

An outlier was identified along Transect 1. The specific catchment area, stream power index, and wetness index (Fig. 4) were large for this point. The upslope catchment area was determined to be 1.1 million ft² which was substantially larger than any other value. This version of TapesG for Windows software used a threshold (i.e., maximum cross-grading area) of 54,819 ft² (5000 m²) before the program switched from divergent (FD8) to stream channel flow (D8) flow algorithms (22). Ideally, this threshold should be set after testing various values and determining whether the terrain attributes accurately reflect field observations of channel flow during rain events. We observed that concentrated channel flow did not occur for the point in question during periods of heavy rainfall. Unfortunately, this parameter could not be adjusted with the version of TAPESG for Windows used in this study. Therefore, this plot was considered to be an outlier and excluded from the correlation and multiple regression analyses.

Statistical Analyses

Multiple stepwise linear regression was conducted using SAS (SAS Institute Inc., Cary, NC) PROC REG (SLENTRY = 0.50; SLSTAY = 0.15). Intercept adjusted variance inflation factors and condition indices were used to diagnose multicollinearity. The INFLUENCE option was used to calculate the RStudent statistic. This statistic was useful for identifying the outlier described in the previous section.

Impact of Erosion on Yield

Average grain yield was substantially greater in 2004 (193 bu/acre) than 2003 (169 bu/acre) because weather conditions were generally more favorable for plant growth in 2004 (Fig. 3). However, the year-to-year differences varied spatially (Fig. 4a). Across Transect 1, yields were similar for both years but values were substantially lower in 2003 than 2004 along Transects 2, 3, and 4. These differences likely occurred because the soils along these three transects were more eroded than in Transect 1. This was indicated by the generally more shallow depth to bedrock (blue symbols in Fig. 4f) and greater clay content (green symbols in Fig. 4f) in these areas. Eroded soils generally have less rooting volume than other areas so plants along these transects likely became more stressed in 2003. The soil moisture conditions were likely further exacerbated by high seeding rates used and late planting date this year.

Fig. 3. Cumulative precipitation during 2003 and 2004 compared to the 10 year average.

Fig. 4. Terrain attributes, EC, depth to rock, and clay content plotted against relative distance along transects.

Simple Terrain Attributes and Yield

Yield was positively correlated with elevation across transects in 2003 (Table 1), primarily because the low elevation portion of Transect 4 (red line in Fig. 4b) yielded much more poorly than any other areas (red line in Fig. 4a). The weak overall correlations (Table 1) were in agreement with those reported in most other studies (4,7,10,11,12,13,17).

Table 1. Correlations between grain yield in 2003 and 2004 with elevation, slope, and curvature.

   * denotes significance at .05 level.

 ** denotes significance at .01 level.

Slopes ranging between 1.1 to 7.9% tended to be substantially greater (red lines in Fig. 4b) in those areas with shallow soils (blue line in Fig. 4f). In both years, slope was poorly correlated with yield (Table 1). A wide range of correlation values have been reported in the literature with one study showing large negative correlations in dry years and weak positive values in wetter years (10). Curvature described the slope shape along the landscape gradient (Fig. 5c). The correlations between yield and curvature were generally weak although some values were statistically significant (Table 1). These results were consistent with other studies that have observed weak correlations during years with above average precipitation (7,10).

Fig. 5. Confidence intervals for 2003 yield prediction model (Table 4) and actual grain yield as a function of distance along the transects.

Complex Terrain Attributes and Yield

Correlations between grain yield and specific catchment area were high (Table 2). This occurred because yield at the beginning of Transect 4 was relatively low in 2004 (green line in Fig. 4a), and there was substantial flow into this area after 8.6 inches of precipitation occurred between May 25 and June 1 of 2004. Correlations between yield and topographic wetness or stream power indices were also highly significant but not as high as for specific catchment area. Flow from upslope areas was the major factor driving yield variability in 2004 across these transects. Some small-grain crop yield variability studies found similarly strong, but positive rather than negative correlations probably because they were conducted in drier areas where moisture was the major limiting factor (4,6,19).

Table 2. Correlations between grain yield in 2003 and 2004 with specific catchment area, topographic wetness index, and stream power index.

   * dentotes significance at .05 level.

 ** denotes significance at .01 level.

This may be the first study where these complex terrain attributes have been investigated for corn in a humid region across terrain with substantial relief. Kravchenko and Bullock (12) found weak correlations (-0.15 ≤ r ≤ 0.30) between yield and upslope contributing area in eight Midwestern fields in corn-soybean rotations. However, they studied landscapes that were relatively flat with average slope values ranging between 0.4 and 1.8% for the eight locations. In comparison, the average slope in this study was 4.3%.

More information is needed regarding the relationship between these terrain attributes and yield during dry years. It is also unclear how the wetness index relates to plant available soil moisture in Central Kentucky fields. Green and Erskine (4) reported poor correlations (r ≤ 0.24) between the wetness index and soil moisture measurements at eight different measurement times over a one-year period in eastern Colorado. In contrast, Moore et al. (16) found stronger relationships in New South Wales, Australia (0.31 ≤ r² ≤ 0.53) (16).

Soil EC and Yield

The correlations between EC and yield in 2003 and 2004 (Table 3) were significant because the low yielding area at the beginning of Transect 4 (Fig. 4a) had relatively high EC values. This was mainly due to excessively wet conditions early in the growing seasons. The EC values were substantially higher in this area because it was wetter than other areas at the time of sampling. These results were consistent with Kravchenko et al. (13) who found large negative correlations in wet years and poor correlations during years with normal precipitation.

Table 3. Correlations between grain yield in 2003 and 2004 with shallow and deep EC and the ratio of deep to shallow EC.

   * denotes significance at .05 level.

 ** denotes significance at .01 level.

The manufacturer of the EC sensor used in this study recommends that the ratio of deep to shallow EC be used to locate shallow soils (21). This ratio dropped considerably in areas (the blue lines in Fig. 4e) with a shallow depth to bedrock (blue lines in Fig. 4f) along Transects 2, 3, and 4. However, the EC ratio correlated poorly with yield (Table 3). This ratio may potentially be of greater value during drier cropping seasons.

Multiple Regression Analyses

The models explained approximately 20 and 65% of grain yield variability in 2003 and 2004, respectively (Table 4). These values were within the range reported by other investigators who have studied yield as a function of terrain attributes (10,11,12). Kitchen et al. (11) showed similar fits when models were developed with EC and simple terrain attributes (0.06 ≤ R² ≤ 0.61). Many studies have reported larger R² values, but most of these have included variables not measured with precision agricultural technologies (e.g., soil chemical properties).

Table 4. Multiple regression parameters for 2003 and 2004 (N = 75).

The 2003 model performed poorly across all transects (Fig. 5a). In most cases, the actual yield observations were outside of the 95% confidence interval. Yield predictions for the 2004 data were better associated with the actual data in 2004 (Fig. 6a). These results are consistent with the work of Kravchenko and Bullock (12) who found better fits in years with either extremely wet or dry conditions than in years with more normal precipitation.

Fig. 6. Confidence intervals for 2004 yield prediction model (Table 4) and actual grain yield as a function of distance along the transects.

Conclusions

Producers and crop consultants do not routinely examine relationships between grain yield, EC, and terrain attributes because these analyses are too time-consuming to be cost-effective for most management systems. However, the analytical procedures used in this study could potentially be automated by controlling GIS, terrain analysis, and statistical software applications with various computer programming languages. Then site-specific yield analysis results could be automatically provided to crop consultant or producer over the internet.

Our findings suggest that complex terrain attributes and EC measurements may be used to aid in the diagnosis of corn grain yield variability in years with above-average precipitation, particularly if periods of intensive rainfall result in water-logging conditions. If poor yields in wet years are associated with high EC values, stream power indices, topographic wetness indices, and specific catchment area values, flooding may have significantly reduced stands in these areas. Simple terrain attributes will have minimal utility in these cases. The data also suggested that the ratio of shallow to deep EC may be an indicator of soil depth and this should be investigated further. Yield relationships with soil depth were not consistent across our study area, so the ratio was poorly correlated with yield. However, the EC ratio may have a better relationship with yield during drier years.

For valid agronomic management interpretations, topographic indices and EC must be compared with multiple years of yield map data. If field regions with large specific catchment areas, topographic wetness index values, and EC values consistently have substantially lower yields, producers may choose to drain or remove these areas from production. This approach will be of greatest value to those managers who farm large acreages and who rely on others to conduct field operations throughout the growing season.

Acknowledgments

We acknowledge advice and assistance from Melissa Bridges, Paul Cornelius, Amanda Ferguson, Matt Gajdzic, Tasos Karathanasis, Ben Koostra, Adam Pike, Tim Phillips, Greg Schwab, Scott Shearer, Charlie Slack, David Williams, and Bill Witt. We are also grateful to Danna Reid, Dr. Frank Sikora, and the University of Kentucky Soil Testing Laboratory for analyzing the soil samples used in this study.

This work is contribution number 05-06-059 from the Kentucky Agricultural Experiment Station.

Literature Cited

1. Beven, K. J., and Kirkby, M. J. 1979. A physically-based, variable contributing area model of basin hydrology. Hydrol. Sci. Bull. 13:43-69.

2. Burt, R., Reinsch, T. G., and Miller, W. P. 1993. A micro-pipette method for water dispersible clay. Commun. Soil Sci. and Plant Anal. 24:2531-2544.

3. Doolittle, J. A., Sudduth, K. A., Kitchen, N. R., and Indorante, S. J. 1994. Estimating depths to claypans using electromagnetic induction methods. J. of soil Water Conservation. 49:572-575.

4. Green, T. R., and Erskine, R. H. 2004. Measurement, scaling, and topographic analysis of spatial crop yield and soil water content. Hydrol. Process. 18:1447-1465.

5. Hartsock, N. J., Mueller, T. G., Karathanasis, A. D., and Cornelius, P. L. 2005. Interpreting soil electrical conductivity and terrain attribute variability with soil surveys. Precision Agriculture. 6:53-72.

6. Heckrath, G., Djurhuus, J., Quine, T. A., Van Oost, K., Govers, G., and Zang, Y. 2005. Tillage erosion and its effect on soil properties and crop yield in Denmark. J. Environ. Qual. 34:312-324.

7. Jaynes, D. B., Kaspar, T. C., Colvin, T. S., and James, D. E. 2003. Cluster analysis of spatiotemporal corn yield patterns in an Iowa Field. Agron. J. 95:574-586.

8. Johnson, C. K., Mortensen, D. A., Wienhold, B. J., Shanahan, J. F., and Doran, J. W. 2003. Site-specific management zones based on soil electrical conductivity in a semiarid cropping system. Agron. J. 95:330-315.

9. Kaschanoski, R. G., Gregorich, E. G., and Van Wesenbeeck, I. J. 1988. Estimating spatial variations of soil water content using no contacting electromagnetic induction methods. Can. J. Soil Sci. 68: 715-722.

10. Kaspar, T. C., Colvin, T. S., Jaynes, D. B., Karlen, D. L., James, D. E., Meek, D. W., Pulido, D., and Butler, H. 2003. Relationship between six years of corn yields and terrain attributes. Precision Agriculture 4:87-101.

11. Kitchen, N. R., Drummond, S. T., Lund, E. D., Sudduth, K. A., and Buchleiter., G. W. 2003. Soil electrical conductivity and topography related to yield for three contrasting soil-crop systems. Agron. J. 95:483-495.

12. Kravchenko, A. N., and Bullock, D. G. 2000. Correlation of corn and soybean grain yield with topographic and soil properties. Agron. J. 92:75-83.

13. Kravchenko, A. N., Thelen, K. D., Bullock, D. G., and Miller, N. R. 2003. Relationship among crop grain yield, topography, and soil electrical conductivity studied with cross-correlograms. Agron. J. 95:1132-1139.

14. Miller, W. P., and Miller, D. M. 1987. A micro-pipette method for soil mechanical analysis. Comm Soil S. 18:1-15.

15. Mitasova, H., and Hofierka, J. 1993. Interpolation by regularized spline with tension, II: application to terrain modeling and surface geometry analysis. Math Geol. 25:657-669.

16. Moore, I. D., Burch, G. J., and Mackenzie, D. H. 1988. Topographic effects on the distribution of surface soil water and the location of ephemeral gullies. Trans. Amer. Sco. Agric. Engrs. 31:1098-1107.

17. Ping, J. L., Green, C. J., Bronson, K. F., Zartman, R. E., and Dobermann, A. 2004. Identification of relationships between cotton yield, quality, and soil properties. Agron. J. 96:1588-1597.

18. Plant, R. E., Mermer, A., Pettygrove, G. S., Vayssieres, M. P., Young, J. A., Miller, R. O., Jackson, L. F., Denison, R. F., and Phelps, K. 1999. Factors underlying grain yield spatial variability in three irrigated wheat fields. T. ASAE. 42:1187-1202.

19. Si, B. S., and Farrell, R. E. 2004. Scale dependent relationship between wheat yield and topographic indicies: a wavelet approach. Soil Sci. Soc. Am. J. 68:577-587.

20. USGS. Seamless data distribution system. Online. Nat. Center Earth Resources Observ. Sci. (EROS).

21. Veris Technologies, Inc. 2004. Frequently asked questions about soil electrical conductivity (EC). Online. Salina, Kansas.

22. Wilson, J. P., and Gallant, J. C. 2000. Terrain Analysis: Principles and Applications. John Wiley & Sons, Inc. New York.