Global Sensitivity Analysis (GSA) Measures the Quality of Parameter Estimation. Case of Soil Parameter Estimation with a Crop Model

The behavior of crops can be accurately predicted when all the parameters of the crop model are well known, and assimilating data observed on crop status in the model is one way of estimating parameters. Nevertheless, the quality of the estimation depends on the sensitivity of model output variables to the parameters. In this paper, we quantify the link between the global sensitivity analysis (GSA) of the soil parameters of the mechanistic crop model STICS, and the ability to retrieve the true values of these parameters. The Global sensitivity indices were computed by a variance based method (Extended FAST) and the quality of parameter estimation (RRMSE) was computed with an importance sampling method based on Bayes theory (GLUE). Criteria based on GSA were built to link GSA indices with the quality of parameters estimation. The result shows that the higher the criteria, the better the quality of parameters estimation and GSA appeared to be useful to interpret and predict the performance of the estimation parameters process.