How well do CMIP5 Climate Models Reproduce the Hydrologic Cycle of the Colorado River Basin?

The Colorado River, which is the primary source of water for nearly 40 million people in the arid Southwestern states of the United States, has been experiencing an extended drought since 2000, which has led to a significant reduction in water supply. As the water demands increase, one of the major challenges for water management in the region has been the quantification of uncertainties associated with streamflow predictions in the Colorado River Basin (CRB) under potential changes of future climate. Hence, testing the reliability of model predictions in the CRB is critical in addressing this challenge. In this study, we evaluated the performances of 17 General Circulation Models (GCMs) from the Coupled Model Intercomparison Project Phase Five (CMIP5) and 4 Regional Climate Models (RCMs) in reproducing the statistical properties of the hydrologic cycle in the CRB. We evaluated the water balance components at four nested sub-basins along with the inter-annual and intra-annual changes of precipitation (P), evaporation (E), runoff (R) and temperature (T) from 1979 to 2005. Most of the models captured the net water balance fairly well in the most-upstream basin but simulated a weak hydrological cycle in the evaporation channel at the downstream locations. The simulated monthly variability of P had different patterns, with correlation coefficients ranging from -0.6 to 0.8 depending on the sub-basin and the models from same parent institution clustering together. Apart from the most-upstream sub-basin where the models were mainly characterized by a negative seasonal bias in SON (of up to -50%), most of them had a positive bias in all seasons (of up to +260%) in the other three sub-basins. The models, however, captured the monthly variability of T well at all sites with small inter-model variabilities and a relatively similar range of bias (-7 °C to +5 °C) across all seasons. Mann-Kendall test was applied to the annual P and T time-series where majority of the models and all observed products displayed nonsignificant trends for annual P. In contrast, more than half of the models exhibited significant trend with annual T as the observations. The results of this work provide support when selecting climate models for impact studies required to develop policies and plan investments aimed at ensuring water sustainability in the CRB.