High-density proximal soil sensing data and topographic derivatives to characterise field variability

Abstract

Proximal soil sensing platforms can provide high-density yet affordable sensor data to describe agricultural field variability. The availability of such data, along with recent advances in analysis methods, allows for the optimisation of model errors and a determination of their spatial variability. Most current sensors measure in-field parameters indirectly, rather than directly linking them to agronomic properties relevant to crop growth. Uncertainty analysis for predicting soil properties is an emerging challenge in precision agricultural practice. High-density soil sensor data and their capacity to contribute to the prediction of soil properties were investigated. An assessment of model accuracy was made by comparing model outputs to validation data points. High-accuracy topography and apparent soil electrical conductivity (EC a ) mapped with either DUALEM-21S or RTK GNSS sensors were used to characterise field-scale soil variability at 12 field sites in Ontario. Lab analyses of six soil properties [pH; buffer pH (BpH); Soil Organic Matter (SOM); Phosphorus (P); Potassium (K); and Cation Exchange Capacity (CEC)] were undertaken to characterise soil variability across the fields. DUALEM sensor variables were co-linear to one another. The topographic variables of slope and topographic wetness index, along with the remainder of the sensor variables, were key inputs to the prediction model. High Pearson's correlation coefficients ( r ≥ 0.60) indicated a strong correlation between topographic parameters and shallow EC a (PRP1: 0–0.5 m) sensor variables, and field-measured soil properties, allowing accurate predictions of several chemical properties ( i.e ., SOM, P, and CEC) at different locations. Among the 12 agricultural fields, two fields presented well-patterned data, resulting in the lowest prediction errors. Drawing on topographic variables provided promising predictions of field SOM and CEC. This highlights the powerful potential of proximal soil sensing technologies to define the site-specific crop production environment in terms of terrain and physical characterisation of the soil. The integration of conceptually different sensors allows for better prediction of certain soil properties than a single measurement approach. • Integration of high-density sensor measurements to predict soil properties. • The model error compared to the North American Proficiency Testing results. • Topographic variables provided promising predictions for soil organic matter and CEC. • Proximal soil sensing technologies allow mapping for a site-specific soil and crop management.