Bias patterns and climate change signals in GCM-RCM model chains

Abstract

Abstract The assessment of regional climate change and the associated planning of adaptation and response strategies are often based on complex model chains. Typically these employ global and regional climate models (GCMs and RCMs), as well as one or several impact models. It is a common belief that the errors in such model chains behave approximately additive, thus the uncertainty should increase with each modeling step. If this hypothesis was true, the application of RCMs would not lead to any intrinsic improvement (beyond higher-resolution details) of the GCM results. Here, we investigate the bias patterns (offset during the historical period against observations) and climate change signals of two RCMs that have downscaled a comprehensive set of GCMs following the EURO-CORDEX framework. Our results show that the biases of the RCMs and GCMs are not additive and not independent. The two RCMs are systematically reducing the biases and modifying climate change signals of the driving GCMs, even on scales that are considered well resolved by the driving GCMs. The GCM projected summer warming at the end of the century is substantially reduced by both RCMs. These results are important, as the projected summer warming and its likely impact on the water cycle are among the most serious concerns regarding European climate change.