Assessing model performance of climate extremes in China: an intercomparison between CMIP5 and CMIP3

Abstract

In this study, we present a brief analysis of the performances of global climate models participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5) in simulating climate extreme events in China and compare the results with those of the previous model generation (CMIP3). The primary focus of this analysis is the climate mean and variability of each extreme index. Results show that the CMIP5 models are generally able to capture the mean climate extremes and trends compared with a new gridded observational dataset. The model spread for some extreme indices is reduced in CMIP5 when compared with CMIP3. Furthermore, the models generally show higher skills in simulating the temperature-based indices than the precipitation-based indices in terms of means and linear trends. Results from six reanalyses further reveal large uncertainties for these indices and it is difficult to say which reanalysis is better for comparison with the simulations of all indices. Based on the relative errors of the climatology, the model evaluation varies considerably from one index to another. However, some models appear to perform substantially better than the others when the average of all indices is considered for each model, and the median ensembles outperform the individual models in terms of all the extreme indices and their means. Additionally, a relationship is observed between the improved simulation of the climate mean and the improved performance of its variability, although this improvement is limited to particular models.