Evaluation of crop yield simulations of an eco-hydrological model at different scales for Germany

Abstract

A prerequisite for integrated crop model applications is the evaluation at the desired spatial and temporal scale. Here, we analysed the eco-hydrological model SWIM simulating crop yields. Historic simulations for winter wheat and silage maize from 1991 to 2010 were used to examine the model performance at the county level in reproducing the county statistics for crop yields. The focus laid on the replication of mean yield levels and interannual crop yield variability. Simulations of silage maize performed better than simulations of winter wheat with R2-values for interannual yield variability of 0.72 and 0.26 respectively at the national level. In particular, silage maize showed a tendency to perform better in areas of lower soil water availability. The reasons for the clear superiority of silage maize were supposedly the short growing season, the lower susceptibility to pests and diseases and, hence, the direct translation of water stress into yield reductions. This signal was less evident for winter wheat and was additionally superposed of climate induced biotic and abiotic stresses – primarily originating in the cold season - which were not implemented in the model. Overall, the simulation bias seemed to originate rather from unconsidered processes than from uncertainties of input data or in model parameterisation.