A Comparative Study of Precipitation and Evaporation between CMIP3 and CMIP5 Climate Model Ensembles in Semiarid Regions

Abstract

he twentieth-century climatology and twenty-first-century trend in precipitation P, evaporation E, and P ? E for selected semiarid U.S. Southwest and Mediterranean regions are compared between ensembles from phases 3 and 5 of the Coupled Model Intercomparison Project (CMIP3 and CMIP5). The twentieth-century simulations are validated with precipitation from observation and evaporation from reanalysis. It is found that the Special Report on Emissions Scenarios (SRES) A1B simulations in CMIP3 and the simulations with representative concentration pathways (RCPs) 4.5 and 8.5 in CMIP5 produce qualitatively similar seasonal cycles of the twenty-first-century trend in P ? E for both semiarid regions. For the southwestern United States, it is characterized by a strong drying trend in spring, a weak moistening trend in summer, a weak drying trend in winter, and an overall drying trend for the annual mean. For the Mediterranean region, a drying trend is simulated for all seasons with an October maximum and July minimum. The consistency between CMIP3 and CMIP5 scenarios indicates that the simulated trend is robust; however, while the trend in P ? E is negative in spring for the southwestern United States for all CMIP ensembles, CMIP3 predicts a strongly negative trend in P and minor negative trend in E whereas both CMIP5 scenarios predict a nearly zero trend in P and positive trend in E. For the twentieth-century simulations, the P, E, and P ? E of the two model ensembles are statistically indistinguishable for most seasons. This ?stagnation? of the simulated climatology from CMIP3 to CMIP5 implies that the hydroclimatic variable biases have not decreased in the newer generation of models. Notably, over the southwestern United States the CMIP3 models produce too much precipitation in the cold season. This bias remains almost unchanged in CMIP5.