Surface albedo is a key variable for controlling the energy budget in the atmospheric boundary layer. In this regard, an accuracy assessment about 3% is required for surface albedo in the numerical weather prediction (NWP ) and climate modelling, which justifies the effort for performing atmospheric and anisotropic corrections on satellite radiometry. Actually, since climate models are partitioning the surface albedo into soil and vegetation albedos separately, a clear understanding of their respective variations is searched. While vegetation albedo varies mostly on a seasonal basis, day-to-day variations of soil albedo are mostly caused by the precipitations. The present study aims at quantifying first the magnitude of near surface soil moisture (SM) on soil albedo derived from MSG/SEVIRI observations. Then, persistency of the soil albedo with time is achieved using a Kalman filter and ECOCLIMAP as back-up. The variations of soil albedo with SM are well modelled in the literature using an exponential function. This relationship is convinced by comparing 4km SEVIRI data with field measurements of SM over SMOSMANIA stations located in south-western France, which encompasses various soil textures. Time series of SM and SEVIRI-derived soil albedos using a new approach are found in good agreement for the cases of reference of vegetated-spare areas. Results of comparison with ASCAT SM products at 8km will also be shown. Further, a prognostic surface albedo is built based on variations of SM and LAI and assuming a static vegetation albedo. An assimilation of the prescribed surface albedo is performed within the SURFEX environment over France by using the forcing SAFRAN and analysed soil moisture during 2009-2010 as inputs. Results in this context of a modification of the surface albedo at short-term scale will be discussed.