The Performance of CERES-Wheat Model in Wheat Planting Areas and Its Uncertainties

Crop models, coupling with climate data from climate models(GCMs, RCMs), are often employed to assess the impacts of climate change on crop production. However, there is a systematic mismatch of resolutions between climate models and crop models. Scaling up the crop model to regional scale is an appropriate method to resolve this problem. CERES Wheat crop model is used to simulate the wheat yields of 1981—2000 at 50 km×50 km grid scale. Performances of this simulation in wheat planting areas are evaluated based on the comparison of simulated yields to census values. The relative root mean square error(RMSE)between simulated and census yields for whole China is 27.9%, and the agreement index is 0.75. Of 2206 simulation units(50 km×50 km grid), 59.2% show relative RMSE less than 30%, in which 26.3% less than 15%. The performances differ among regions. Smallest bias occurs in agro-ecological zone 2(the largest wheat planting areas accounting for 39.9% of China's wheat planting area), with relative RMSE of 16.6% and D=0.68. To sum up, CERES Wheat crop model is able to produce reasonable results temporally and spatially. It can provide simulation information for policy making at macro scale despite existing uncertainties. The uncertainties of this regional simulation are ascribed to simplification and limitations of crop models, the aggregated inputs in wheat planting area, and errors in dataset etc, which need to be addressed in future.