Spatial resolution of precipitation and radiation: The effect on regional crop yield forecasts

This paper explores the effect of uncertainty in precipitation and radiation on crop simulation results at local (50 x 50 km grids) and regional scale (NUTS1 regions) and on the crop yield forecasts for Germany and France. Two experiments were carried out where crop yields for winter-wheat and grain maize were simulated using the crop growth monitoring system (CGMS) for the year 2000 with different precipitation and radiation inputs. The first experiment used precipitation and radiation inputs interpolated from weather stations while the second experiment used accurate precipitation and radiation inputs derived from the European Land Data Assimilation System (ELDAS). The differences between the simulated water-limited yields of the two experiments demonstrate that uncertainty in precipitation and radiation translates into a considerable uncertainty in crop yield at the level of 50 x 50 km grids. This uncertainty strongly decreases when simulation results are spatially aggregated to NUTS I regions. European Statistical Office (EUROSTAT) yield statistics and CGMS model output for grain maize over the period 1990-1999 were used to develop yield forecasting equations for France and Germany. These equations were applied to the simulation results of both experiments. We concluded that uncertainty in radiation and precipitation in CGMS has little influence on the CGMS yield forecast at national level. Finally, the effect of averaging of precipitation and radiation was evaluated by comparing CGMS simulation results at 10 x 10 km level with results at 50 x 50 km level. We concluded that the CGMS grid size of 50 x 50 km is an appropriate resolution because the distributed simulation results at 10 x 10 km scale almost linearly with the results at 50 x 50 km obtained using averaged rainfall and radiation. Improvements of the system should therefore focus on providing average unbiased estimates of weather variables at 50 x 50 level, rather then increasing the grid resolution. (c) 2005 Elsevier B.V. All rights reserved.