The importance of surface fluxes has been studied since so far as one of the dominant slow component of Earth climate system especially in mid-latitude summers. Initial soil state, and in particular the soil moisture content, has been addressed in a wide number of studies for climatic, meteorological and hydrological applications. A satellite data assimilation method has been developed, for the Regional Atmospheric Modelling System (RAMS), which incorporates satellite – observed heating rates in order to retrieve soil moisture based on the new generation of geostationary Meteosat Second Generation data, taking advantage of their enhanced spatial and temporal resolution. The method acts on the soil moisture in RAMS ground levels adjusting it, upward and downward, until the RAMS simulated surface heating rate is in close agreement with the satellite – observed one in each grid cell. The method simply carries out a forward integration of the Soil – Vegetation – Atmospheric RAMS component (the Land Ecosystem Atmosphere Feedback version 2 model, LEAF – 2) for a special assimilation period in order to adjust the model soil moisture according with the observed data. Iterations needed add just a small amount of time to be computed to the total simulation time, so it is possible a simple usage within an operational chain or seasonal / long runs too. The sub grid heterogeneity representation in LEAF – 2 guarantees an optimal physical description of surface latent heat and sensible heat flux, their respectively dominant regimes during wet or dry periods, and transition among them.