The effect of bias correction and climate model resolution on wheat simulations forced with a regional climate model ensemble

Regional climate model (RCM) simulations are often used with agricultural models to assess the impact of climate change on agriculture. This study assesses the suitability of climate data sets from an RCM ensemble for forcing simulations of wheat yields for New South Wales, Australia, performed using the Agricultural Production Systems sIMulator (APSIM). Differences in yields between APSIM simulations forced with RCM output for the 1990-2009 period and APSIM simulations forced with observations are examined, as are simulated changes in yields between 1990-2009 and 2060-2079. The RCM ensemble downscales four global climate models (GCMs) using three different RCMs, firstly to a horizontal resolution of approximately 50 km, and then to approximately 10 km. All of the RCM simulations, 50 and 10 km, are able to simulate the spatial pattern of yields across the study region. However, some simulations have biases in yields. The largest of these are due to biases in rainfall during the growing season inherited from the GCMs. If the RCM output is bias-corrected, the largest positive yield biases are reduced because the underlying biases in growing season rainfall are reduced. Simulated future changes in yields are affected because there is a non-linear relationship between simulated yields and growing season rainfall. Downscaling to 10 km, rather than 50 km, is only beneficial when bias correction is used. The bias-correction technique used does not eliminate all biases in growing season rainfall and increases some of the smaller biases. Nonetheless, because of the potentially large effect of the rainfall biases on simulated future yield changes, this study supports the use of bias correction in assessments of the impacts of climate change that use RCM output to force APSIM. Indeed, our findings suggest that bias correction may be necessary to obtain reliable future changes in other outputs of other biophysical models that respond non-linearly to climate inputs.