Abstract:

Two-year series of monthly solutions of separated continental liquid water (surface water, soil moisture and groundwater) and snow mass changes have been computed from the CSR/GFZ geoids (04/2002-07/2004) at the resolution of ~660 km (deg. 25-30), using a robust inverse method of separation that combines GRACE observations with a priori covariances/errors of available global models (Ramillien et al., 2004; Ramillien et al., 2005). At the regional scale of a large drainage basin, these Land Waters (LW) solutions were used to compute time-series of water mass (i.e., Amazon) and snow mass (in high-latitude boreal regions), expressed in terms of equivalent-water volume variations. By resolution of the water mass balance equation, global maps of the evapotranspiration (ET) rate were derived by combining from the LW solutions and precipitation and runoff data from models. The validation of our method consists of comparing our estimates of water/snow mass changes with: (1) the results obtained using other previous approaches (i.e., the averaging filtering proposed by Wahr et al., (1998); 1-D ET rate profile for the Mississippi basin computed by Rodell et al., (2004)); (2) the monthly mean forecasts of different global models (WGHM, LaD, GLDAS). Further analyses reveal good agreement between the GRACE-derived LW solutions and model outputs in particular regionally and at the seasonal scale.