Joint assimilation of surface soil moisture and LAI observations into a land surface model

Land Surface Models (LSM) offer a description of land surface processes and set the lower boundary conditions for meteorological models. In particular, the accurate description of those surface variables which display a slow response in time, like root-zone soil moisture or vegetation biomass, is of great importance. Errors in their estimation yield significant inaccuracies in the estimation of heat and water fluxes in Numerical Weather Prediction (NWP) models. In the present study, the ISBA-A-gs LSM is used decoupled from the atmosphere. In this configuration, the model is able to simulate the vegetation growth, and consequently LAI. A simplified 1D-VAR assimilation method is applied to observed surface soil moisture and LAI observations of the SMOSREX site near Toulouse, in south-western France, from 2001 to 2004. This period includes severe droughts in 2003 and 2004. The data are jointly assimilated into ISBA-A-gs in order to analyse the root-zone soil moisture and the vegetation biomass. It is shown that the 1D-VAR improves the model results. The efficiency score of the model (Nash criterion) is increased from 0.79 to 0.86 for root-zone soil moisture and from 0.17 to 0.23 for vegetation biomass.