Sensitivity of the hydrological cycle to increasing amounts of greenhouse gases and aerosols

The coupled atmosphere-ocean Climate Model of the Centre National de Recherches Meteorologiques (CNRM) has been used to run a time-dependent climate change experiment to study the impact of increasing amounts of greenhouse gases and aerosols on the simulated water cycle. This simulation has been initialised with the oceanic temperature and salinity profiles and the atmospheric trace gas concentrations observed in the 1950s, and has been carried out for 150 years after a 20-year spin-up. The simulated climate change has been analysed as the difference between two 30-year time slices: 1970-2000 and 2070-2100 respectively. The model achieves a reasonable simulation of present-day climate and simulates a general increase in precipitation throughout the twenty first century. The main exceptions are the subtropics, where the enhanced Hadley circulation has a drying impact, and the midlatitude continents, where the increased evaporation in spring and decreased moisture convergence in summer lead to a relative summer drying. Global and regional analyses suggest that the precipitation increase is generally limited by a decrease in the water vapour cycling rate and in the precipitation efficiency, which appear as key parameters of the simulated water cycle. In order to reduce the spread between climate scenarios, more efforts should be devoted to estimate these parameters from satellite observations and meteorological analyses, and their possible evolution over recent decades. In the present study, the impacts of global warming on the surface hydrology have been also investigated. The main findings are the amplification of the annual cycle of soil moisture in the mid-and-high latitudes, and the decrease in the Northern Hemisphere snow cover, at a rate that is consistent with recent satellite estimations and should increase during the twenty first century. The runoff simulated over the 1950-2100 period has been converted into river flow using a linear river routeing model. The trends simulated over recent decades are surprisingly consistent with the river flow measurements available from the Global Runoff Data Centre. These trends can differ from those estimated over the whole 150-year integration, thereby indicating that it is not safe to predict hydrological impacts just by extrapolating the trends found in the available observations. Our climate model seems likely to provide qualitative hydrological scenarios over large river basins, but it still shows serious biases in the simulation of present-day river flows. Regional hydrological projections remain a challenge for the global climate modelling community and downscaling techniques are still necessary for this purpose.