Calibrating Field-Specific Soil Parameters Using Remote Sensing and Several Years of Crop Yields

Currently, many precision agriculture recommendations are based on empirical models, for example when a nitrogen application rate is calculated from a drone or satellite image. In the future such recommendations will also use mechanistic models of crop growth. It is a challenge to calibrate these models to the specific conditions of each farm and each farm field, especially so for soil parameters. The aim of our work was to operationalize the estimation of soil parameters via inverse modelling, using remote sensing and yield monitor data. We calibrated the WOFOST model using 5 years of data from 10 commercial fields in The Netherlands on which potato was the main crop and where sugar beet, maize and wheat were grown in non-potato years. Starting values for soil parameters were taken from the national soil map. Model outcome proved especially sensitive to field capacity and depth of the soil. Inverse modelling resulted in improved accuracy of simulated LAI and crop yield. We report the increase in accuracy of the simulations and the change in soil parameters relative to their starting values.