Multiple-depth electrical conductivity estimates of discrete-layer soil texture

Authors: Sudduth, Kenneth - Ken; Kitchen, Newell; Veum, Kristen; Vories, Earl; Drummond, Scott

Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 5/23/2019
Publication Date: 5/28/2019
Citation: Sudduth, K.A., Kitchen, N.R., Veum, K.S., Vories, E.D., Drummond, S.T. 2019. Multiple-depth electrical conductivity estimates of discrete-layer soil texture. In: Proc. 5th Global Workshop on Proximal Soil Sensing, May 28-31, 2019, Columbia, Missouri. p. 153-158.

Interpretive Summary: Bulk apparent soil electrical conductivity (ECa) is the most widely used type of soil sensing in precision agriculture. Soil ECa relates to multiple soil properties, including clay content, water content, and salt content (salinity), and farmers and others have found it useful for mapping variations within fields. What has been more difficult is developing numeric relationships between ECa readings and important soil properties. In part, this is because ECa readings represent the entire profile, while soil properties are most often determined for specific soil layers. In this research we investigated a mathematical approach known as model inversion to calculate individual layer ECa values for the same layers where soil property data were available. Input data was from ten fields in Missouri with a range of soils from mostly sandy to mostly clayey. The inversion process provided layer values that were strongly related to measured layer ECa at calibration points established in each field. Layer conductivity data successfully estimated soil texture for some, but not all, of the fields, and additional investigation is needed to understand why this was the case. If proven successful, the inversion approach could help provide soil datasets important for agronomic and environmental modelling, potentially benefiting producers, agribusiness entities, and the general public.

Technical Abstract: Bulk apparent soil electrical conductivity (ECa) is the most widely used soil sensing modality in precision agriculture. Soil ECa relates to multiple soil properties, including clay content, water content, and salt content (i.e., salinity). In general, ECa sensors respond to soil conductivity down to a sensor-dependent measurement depth, weighted by a nonlinear response function. Although multi-channel commercial ECa sensors give information about multiple depths, additional processing is required to derive conductivity for specific, discrete depths that can then be related to texture or other soil properties. Therefore, the purpose of this research was to evaluate the accuracy of discrete-layer soil texture estimates from multi-channel ECa sensing. Mobile ECa data were collected for ten Missouri, USA fields from 8 to 35 ha in size and containing a range of soil types, ranging from predominantly sand to predominantly clay. Profile ECa data as a function of depth were obtained at between 8 and 20 locations within each field using an instrumented penetrometer. Simultaneously, soil cores to 1 m depth were obtained at those same locations, segmented, and analyzed for texture. Commercial ECa inversion software was used to extract layer conductivities on 0.1-m increments. Those layer conductivities were validated against the measured penetrometer ECa. The good results obtained across all fields (R2 = 0.75) provided confidence in the accuracy of the inversion process. Inversion-based and measured conductivities were then related to the measured texture data. Results were not good for claypan-soil fields, where cations, especially Mg, were more strongly related to ECa than was clay. Better results were obtained when data were merged for all other fields, with measured ECa providing slightly better results (R2 = 0.84, RMSE = 5.7 % clay) than inversion-estimated layer ECa (R2 = 0.81, RMSE = 6.2 % clay). These findings show that it is possible to provide accurate mapped estimates of discrete-layer soil texture using multi-channel ECa sensors. These texture estimates can then be used as input data for spatially-explicit crop and environmental modelling.