Uncertainties in the GSWP-2 precipitation forcing and their impacts on regional and global hydrological simulations

The Global Soil Wetness Project (GSWP) is an international initiative aimed at producing global data sets of soil wetness and energy and water fluxes by driving land surface models with state-of-the-art 1° by 1° atmospheric forcing and land surface parameters. It also provides a unique opportunity to develop and test land surface parameterizations at the global scale, using multi-year off-line simulations that are not affected by the systematic errors found in atmospheric models. Nevertheless, the accuracy and reliability of the 10?year GSWP-2 atmospheric forcing remain questionable. A first comparison using the high-resolution Rhône-AGGregation (Rhône-AGG) database reveals that the baseline GSWP-2 precipitation forcing is drastically overestimated over the Rhône river basin. Hydrological simulations driven with each dataset and using the ISBA land surface model and the MODCOU river routing model are also compared. The simulated river discharges are validated against a dense network of river gauges and are generally less realistic when using the GSWP-2 instead of the Rhône-AGG precipitation forcing. Secondly, the GSWP-2 precipitation forcing is compared with three alternative data sets (GPCP-2, CRU-2, CMAP) at the global scale. Moreover, the results of a global sensitivity study to the precipitation forcing conducted with six land surface models are shown. The TRIP river routing model is used to convert daily runoff from all models into river discharges, which are compared at 80 gauging stations distributed over the globe. In agreement with the regional evaluation, the results reveal that the baseline GSWP-2 precipitation forcing is generally overestimated over the mid and high latitudes, which implies systematic errors in the simulated discharges. This study reveals that the empirical wind corrections applied to the GSWP-2 precipitation forcing are exaggerated, whereas the GPCP satellite adjustments seem to be useful for simulating realistic annual mean river discharges over the East Siberian river basins.