Selecting regional climate models based on their skill could give more credible precipitation projections over the complex Southeast Asia region

This study focuses on future seasonal changes in daily precipitation using Regional Climate Models (RCMs) from the Coordinated Regional Climate Downscaling Experiments-Southeast Asia ensemble (CORDEX-SEA). Projections using this RCM ensemble generally show a larger inter-model spread in winter than in summer, with higher significance and model agreement in summer over most land areas. We evaluate how well the RCMs simulate climatological precipitation using two skill metrics. To extract reliable projections, two sub-ensembles of 'better' and 'worse' performing models are selected and their respective projections compared. We find projected intensification of summer precipitation over northern SEA, which is robust across RCMs. On the contrary, in the southern part of SEA, the 'worse' ensemble projects a significant and widespread decrease in summer rainfall intensity whereas a slight intensification is projected by the 'better' ensemble. Further exploration of inter-model differences in future changes reveals that these are mainly explained by changes in moisture supply from large-scale sources (i.e., moisture convergence) with enhanced effects from local sources (i.e., evapotranspiration). The 'worse' models project greater changes in atmospheric circulation compared with the 'better' models, which can explain part of the uncertainty in projections for daily precipitation over the CORDEX-SEA domain. Hence, our findings might help assess more reliable projections over the SEA region by selecting models based on a two-step model evaluation: the ability of models to simulate historical daily precipitation and their performance in reproducing key physical processes of the regional climate.