Improved Simulation of Regional Climate by Global Models with Higher Resolution: Skill Scores Correlated with Grid Length

Abstract

he current generation of climate models, as represented by phase 5 of the Coupled Model Intercomparison Project (CMIP5), has previously been assessed as having more skill in simulating the observed climate than the previous ensemble from phase 3 of CMIP (CMIP3). Furthermore, the skill of models in reproducing seasonal means of precipitation, temperature, and pressure from two observational datasets, quantified by the nondimensional Arcsin?Mielke skill score, appeared to be influenced by model resolution. The analysis is extended to 42 CMIP5 and 24 CMIP3 models. For the combined skill scores for six continents, averaged over the three variables and four seasons, the correlation with model grid length in the 66-model ensemble is ?0.73. Focusing on the comparison with ERA-Interim data at higher resolution and with greater regional detail, correlations are nearly as strong for scores over the ocean domain as for land. For the global domain (excluding the Antarctic cap), the correlation of the overall skill score with grid length is ?0.61, and it is nearly as strong for each variable. For most tests the improved averaged score of CMIP5 models relative to those from CMIP3 is largely consistent with their increased resolution. However, the improvement for precipitation and the correlations with length are both smaller if rmse is used as a metric. They are smaller again using the GPCP observational data, as the regional detail from a high-resolution model can lead to larger differences when compared to relatively smooth observational fields.