Accuracy, robustness and behavior of the STICS soil–crop model for plant, water and nitrogen outputs: Evaluation over a wide range of agro-environmental conditions in France

Soil-crop models are increasingly used as predictive tools to assess yield and environmental impacts of agriculture in a growing diversity of contexts. They are however seldom evaluated at a given time over a wide domain of use. We tested here the performances of the STICS model (v8.2.2) with its standard set of parameters over a dataset covering 15 crops and a wide range of agropedoclimatic conditions in France. Model results showed a good overall accuracy, with little bias. Relative RMSE was larger for soil nitrate (49%) than for plant biomass (35%) and nitrogen (33%) and smallest for soil water (10%). Trends induced by contrasted environmental conditions and management practices were well reproduced. Finally, limited dependency of model errors on crops or environments indicated a satisfactory robustness. Such performances make STICS a valuable tool for studying the effects of changes in agro-ecosystems over the domain explored. STICS v8.2.2 soil-crop model was evaluated over a large and varied dataset using its standard set of parameters.Level of accuracy is 10-50% for plant, soil water and nitrate outputs.Model reproduces well trends arising from contrasted agro-environmental conditions.Errors are weakly dependent on the agro-environmental conditions tested.Model accuracy and robustness is considered good for scenario testing and large scale use within the conditions tested here.