RESUME / ABSTRACT  


Analyse 2D-Var simplifiée du l'humidité du sol à partir des données de température de surface pour l'initialisation de modèles numériques


by


Gianpaolo Balsamo

Météo France






The study concerns the analysis of soil moisture for the operational mesoscale model of Meteo-France ALADIN. A simplified variational technique (2D-Var) has been implemented to analyse soil moisture by assimilating screen-level observations of temperature and relative humidity. The linearized versions of the observation operator are obtained through finite differences from extra integrations of the numerical model with perturbed initial soil moisture states. Two fundamental assumptions of the methods (linearity of the observation operator and horizontal decoupling between grid points) are evaluated. The validation of the method is done with real observations using different time windows: a sequential assimilation cycle of 6-h allows a comparison with the operational optimal interpolation technique, while a 24-h window is considered to extend the temporal consistency of the assimilated observations in the analysis. Results from the 2D-Var analyses show realistic soil moisture retrievals when compared to an independent soil moisture estimation produced by an hydrological off-line system. The benefits of the 2D-Var analysis are also shown by an improvement of the forecast of screen-level variables.