Downscaling Extreme Precipitation from CMIP5 Simulations Using Historical Analogs

Abstract An approach for downscaling daily precipitation extremes using historical analogs is applied to simulations from phase 5 of the Coupled Model Intercomparison Project (CMIP5). The method employs a multistep procedure in which the occurrence of extreme precipitation on a given target day is determined on the basis of the probability of extreme precipitation on that day’s closest historical analogs. If extreme precipitation is expected, daily precipitation observations associated with the historical analogs are used to approximate precipitation amounts on the target day. By applying the analog method to historical simulations, the ability of the CMIP5 models to simulate synoptic weather patterns associated with extreme precipitation is assessed. Differences between downscaled and observed precipitation extremes are investigated by comparing the precipitation frequency distributions for a subset of rarely selected extreme analog days with those for all observed days with extreme precipitation. A supplemental composite analysis of the synoptic weather patterns on these rarely selected analog days is utilized to elucidate the meteorological factors that contribute to such discrepancies. Overall, the analog method as applied to CMIP5 simulations yields realistic estimates of historical precipitation extremes, with return-period precipitation biases that are comparable in magnitude to those obtained from dynamically downscaled simulations. The analysis of rarely selected analog days reveals that precipitation amounts on these days are generally larger than precipitation amounts on all days with extreme precipitation, leading to an underestimation of return-period precipitation amounts at many stations. Furthermore, the synoptic composite analysis reveals that tropical cyclones are a common feature on these rarely selected analog days.