Geostatistical Modeling and Propagation of Uncertainty: Application to the Management of Agricultural Fields

1. Abstract This paper presents a geostatistical approach to model uncertainty about the spatial distribution of soil properties and to propagate it through a crop model, allowing for incorporation of uncertainty in the choice of management scenarios (no lime, single-rate liming and site-specific lime applications to acidic field soil). The methodology involved modeling the uncertainty about wheat yield, accounting for the local uncertainties about soil pH and lime requirement, and the uncertainties about crop model parameters used in the simulations. Indicator kriging (IK) was used together with stratified sampling of the probability distributions of variables and model parameters for the propagation of uncertainties through to the output yield and net profit maps. Comparison of the three scenarios showed that under the economic conditions of the analysis, the optimum was reached for a single-rate application of 3.5 Mg/Ha over the entire field instead of site-specific lime applications.