Evaluation of CMIP climate model hydrological output for the Mississippi River Basin using GRACE satellite observations

We use measurements of terrestrial water storage (TWS) inferred from Gravity Recovery and Climate Experiment (GRACE) satellite observations to evaluate the hydrological output composite-averaged over the Mississippi River Basin (MSRB) and over the ten-year period 2003-2012 from a subset of GCMs from the World Climate Research Programme's Coupled Model Intercomparison Project Phase 5 (CMIP5) and Phase 3 (CMIP3). We then investigate mid-20th to 21st century hydrological trends over the MSRB projected by the CMIP5 models. Improvements were found in CMIP5 simulations of the annual cycle of composite TWS, estimated as the sum of modeled depth-integrated soil moisture and snow water, over those of CMIP3 when compared with the GRACE composite TWS cycle. These improvements coincide with higher horizontal resolution and changes in hydrological parameterizations applied in most of the CMIP5 GCMs compared to earlier CMIP3 versions. Simulated values of composite hydrological budget terms among CMIP5 models, however, are not improved overall, with some models exhibiting increased precipitation and others decreased runoff from CMIP3 to CMIP5 to values outside long-term observed ranges. Since the effect of both increased precipitation and decreased runoff is to increase infiltration and soil water retention, the composite TWS annual cycles from these CMIP5 models, whose earlier CMIP3 simulations in some cases highly underestimated TWS annual cycle amplitudes compared to GRACE, now better agree with GRACE. In spite of the improved prediction of the composite TWS annual cycle, multi-decadal hydrological trends for the MSRB produced by the CMIP5 models vary. A consensus for future decreasing soil moisture is found among models, but with varied responses in magnitude as well as in direction of annual precipitation, evapotranspiration and runoff trends. Overall, GRACE data appear highly useful for evaluating GCM hydrological predictions over large river basins, and a longer time period of these data as more retrievals become available should help to evaluate GCM hydrological output on a multi-decadal time scale. (C) 2014 Elsevier B.V. All rights reserved.