Crop models are indispensable to assess crop production under climate change, and they need evaluation preferably on long-term experiments. The certainty of estimates depends on the detail of input data and selection of an appropriate model. Aggregation may allow the use of simpler models, but it is unknown how this increases the prediction uncertainty. The purpose of our study was to assess the efficiency of a meteorological dataset from regional and local sources using two different models simulating sugar beet growth. Experimental data (1961-1992) originated from the Static Fertilization Experiment in Bad Lauchstädt, Germany. Regional weather (temperature, radiation and rainfall) was retrieved from the European database. Biomass of sugar beet was simulated with a relatively simple single-season model (BB) and a more complex mechanistic multi-season model (CA). Simulations compared best to de-trended observations using CA with local daily weather data (R²=0.55). Using model CA with regional weather data decreased the prediction certainty to that for BB (R²=0.45 and 0.46). The complex model (CA) overestimated biomass (+1.4 t/ha) whereas the simple model (BB) under-estimated observed biomass above 21 t /ha (−3.8 t/ha). An estimate of initial soil water content proved essential for the application of the BB model under continental climate. The results suggest that the predictive capacity of simple and complex models for crop growth is similar when only regional input data are available.