资源环境科技发展态势分析平台

The evaluation of regional climate models' (RCMs') ability to reproduce the present-day climate is critical to support their utility in impact studies under future climatic scenarios. This study evaluates the skill of an ensemble of state-of-the-art regional climate simulations from the European Coordinated Regional Downscaling Experiment initiative in reproducing the precipitation (P) properties in Sardinia, a Mediterranean island of 24,000km(2). The ensemble includes simulations at 0.44 degrees and 0.11 degrees grid spacing of the Historical experiment from 1950 to 2005. Precipitation records from a high-density network of gauges are used as reference data set. The interannual and seasonal climatology of P, presence of trend, and orographic effect are assessed at the original RCM grid spacings and different spatial scales of aggregation. Most models capture the observed positive relation between annual P and elevation, with better performance at 0.11 degrees spacing. The simulated spatial patterns of the climatological annual and seasonal means are well correlated with the observation at both spacings, but their spatial variability is overestimated. Positive and negative bias of up to 60% are found in the simulation of annual mean and interannual variability. While the majority of the models reproduce the phase of the seasonal cycle, they underestimate (overestimate) winter (summer) P. The RCMs exhibit different deficiencies in capturing the negative annual and seasonal observed trends. In general, models' skill degrades when analyses are conducted at smaller aggregation scales. Results of this study reveal insight on RCM performances in small-scale regions, which are often targeted by impact studies and have so far received less attention.