SYSTEMS AND TECHNOLOGIES FOR SUSTAINABLE SITE-SPECIFIC SOIL AND CROP MANAGEMENT
Location: Cropping Systems and Water Quality Research
Title: Estimating depth to argillic soil horizons using apparent electrical conductivity response functions
Submitted to: Symposium on Application of Geophysics to Engineering and Environmental Problems Proceedings
Publication Type: Proceedings
Publication Acceptance Date: January 9, 2009
Publication Date: March 30, 2009
Citation: Sudduth, K.A., Kitchen, N.R., Myers, D.B., Drummond, S.T. 2009. Estimating depth to argillic soil horizons using apparent electrical conductivity response functions. In Proc. 22nd Symposium on the Application of Geophysics to Engineering and Environmental Problems, March 29-April 2, 2009, Ft. Worth, Texas, CD-ROM.
Interpretive Summary: As farmers have become familiar with the technologies of precision agriculture, such as yield monitors and GPS, there has been increasing interest in being able to map soil variability within fields. One on-the-go soil sensing technology is soil apparent electrical conductivity (EC), for which there are several sensors commercially available. Each of the commercial sensors measures EC as a single reading averaged over a measurement depth, and this measurement depth is different for the different sensors. We wanted to find which sensor data would give the best picture of topsoil depth variability for the claypan soils of northern Missouri, because topsoil depth is an important factor in the productivity of these soils. We collected data with three commercial EC sensors on two fields in Missouri and related the data to topsoil depth with two different methods, one using statistical analysis and one using calculations based on the theoretical response functions of the sensors. We found that topsoil depth estimates by two methods were very similar and that the EC sensors with medium measurement depths gave the best results. We also found that better estimates, with a 30% reduction in error, could be obtained when combining data from multiple sensors in the statistical approach. We also plan to combine multiple sensor data with the theoretical approach in future research. These results will benefit users of EC instruments by providing methods to calculate soil variables from the EC data. The results will also benefit scientists and extension personnel who may need to recommend the best EC instrument to use in a particular situation.
Maps of apparent electrical conductivity (ECa) of the soil profile are widely used in precision agriculture practice and research. A number of ECa sensors are commercially available, each with a unique response function (i.e., the relative contribution of soil at each depth to the integrated ECa reading). In past research we estimated depth to an argillic horizon (i.e., topsoil depth, TD) on claypan soils by fitting empirical equations to ECa data from such sensors, either individually or in combination. The objective of this research was to take a more theoretical approach, estimating TD using a two-layer soil model incorporating data from different ECa sensors along with their published response functions. Data were obtained with three sensors having five different ECa depth-response functions (Veris 3150, Geonics EM38 vertical dipole mode, and DUALEM-2S) on two Missouri claypan-soil fields. Soil cores were also obtained in each field to provide calibration and validation TD data. Using a numerical optimization approach, response-function calibration models were developed for each ECa dataset. Best TD estimates (root mean square error of validation, RMSEv = 19 cm) were obtained with the shallower reading from the DUALEM-2S, while the least accurate estimates were obtained from the Veris datasets (RMSEv = 24-25 cm). Results from the model-based approach were very similar to those obtained by regressing TD on 1/ECa. For both approaches, most occurrences of high TD error were localized in one area of one field, possibly due to the presence of subsoil features of lower conductivity that violated the two-layer assumption. TD error in this area was reduced by including additional 1/ECa terms from multiple sensors in the regression approach. Inclusion of multiple ECa terms in the model inversion approach will be a subject of future investigation.