The feasibility of a radiance assimilation using a multi-layered snow physical model to estimate snow physical parameters is studied.The work is divided in five parts.The first two chapters are dedicated to the literature review. In the third chapter, experimental work was conducted in the alpine snow to estimate snow correlation (for microwave emission modelling) using near-infrared digital photography. We made microwave radiometric and near-infrared reflectance measurements of snow slabs under different experimental conditions. We used an empirical relation to link near-infrared reflectance of snow to the specific surface area (SSA), and converted the SSA into the correlation length. From the measurements of snow radiances at 21 and 35 GHz, we derived the microwave scattering coefficient by inverting two coupled radiative transfer models (RTM) (the sandwich and six-flux model).The correlation lengths found are in the same range as those determined in the literature using cold laboratory work.The technique shows great potential in the determination of the snow correlation length under field conditions. In the fourth chapter, the performance of the ensemble Kalman filter (EnKF) for snow water equivalent (SWE) estimation is assessed by assimilating synthetic microwave observations at Ground Based Microwave Radiometer (GBMR-7) frequencies (18.7, 23.8, 36.5, 89 vertical and horizontal polarization) into a snow physics model, CROCUS. CROCUS has a realistic stratigraphic and ice layer modelling scheme. This work builds on previous methods that used snow physics model with limited number of layers. Data assimilation methods require accurate predictions of the brightness temperature (Tb) emitted by the snowpack. It has been shown that the accuracy of RTMs is sensitive to the stratigraphic representation of the snowpack. However, as the stratigraphic fidelity increases, the number of layers increases, as does the number of state variables estimated in the assimilation. One goal of the present study is to investigate whether passive microwave measurements can be used in a radiance assimilation (RA) scheme to characterize a more realistic stratigraphy.The EnKF run was performed with an ensemble size of 20 using artificially biased meteorological forcing data.The snow model was given biased precipitation to represent systematic errors introduced in modelling, yet the EnKF was still able to recover the"true" value of SWE with a seasonally-integrated RMSE of only 1.2 cm (8.1%).The RA was also able to extract the grain size profile at much higher dimensionality which shows that the many-to-one problem of SWE-Tb relationship can be overcome by assimilation, even when the grain size profile varies constantly with depth.The last chapter was on the validation of the data assimilation system using a point-scale radiance observations from the CLPX-1 GBMR-7. We first predicted snow radiance by coupling the snow model CROCUS to the snow emission model (MEMLS). Significant improvement of Tb simulation was achieved for the late February window for all three frequencies.The range of the underestimation of the polarization difference is between 25% and 75%. We then assimilated all six channels measurements of the GBMR-7.The filter was able to accurately retrieve the SWE for periods of time when the Tb measurements were available.The results show that RA using EnKF with a multi-layered snow model can be used to determine snow physical parameters even with a biased precipitation forcing.