Bayesian methods for updating crop-model predictions, applications for predicting biomass and grain protein content

Abstract

Crop models can be used for predicting crop quality and for optimizing agricultural practices, but the errors of prediction of these models are often important. The objective of this paper is to describe several methods for improving the accuracy of crop-model predictions with real-time measurements. First, we present a simple linear dynamic crop model simulating winter-wheat biomass production and we show how the Kalman-filter method can be used for updating model predictions. Second, we describe a general framework for updating complex nonlinear dynamic crop models. Finally, we present a case study with a nonlinear crop model predicting winter-wheat grain quality. The results presented in this paper show that Bayesian methods are useful for improving the reliability of the information provided to stakeholders by crop models but that additional research is required for implementing these methods with complex models