In this research, the characterization of snow cover is made from data collected in September, February and March of 2002 and 2003, during Cold-land Processes Field Experiments project of the NASA. These data include snow and forests characteristic measurements, meteorological conditions, digital elevation model (DEM) and polarimetric multifrequency SAR data (C, L and P bands) acquired from AIRSAR-POLSAR airborne sensor. These data will be used to analyze multifrequency polarimetric techniques to characterize snow cover over forested areas (open area, sparse coniferous forest, and dense coniferous forest). Different techniques have been developed to detect wet snow over different forested areas. The methodology of wet snow detection developed by Rott and Nagler (1995) is first analyzed. The best result is obtained in HH polarization (13% for the sparse coniferous forest site and 25% for the dense coniferous forest site). C-band data in circular polarizations improves these results, but the errors remain high (22% for the sparse coniferous forest site and 13% for the dense coniferous forest site). The use of [sigma][omicronn] ratio in dB [sigma][omicronn][subscript LHH] /[sigma][omicronn][subscript CHH], [sigma][omicronn][subscript LHV]/[sigma][omicronn] [subscript CHH], [sigma][omicronn][subscript LHV] /[sigma][omicronn][subscript CHV] and [sigma][omicronn][subscript LVV] /[sigma][omicronn][subscript CHH] allows to detect wet snow ([less-than or equal to] 13% errors) for both the open area and the dense coniferous forest sites. However, with this technique, higher errors ([greater-than or equal to] 16%) are obtained for the sparse coniferous forest site. The analysis of polarimetric signatures in the three bands shows that their shapes vary according to snow conditions (wet or dry) and forest densities. The pedestal height of polarimetric signatures in P band allows to apply a thresholding approach to discriminate between snow conditions (wet or dry). The error matrix generated from polarimetric signature techniques applied to snow pit measurements shows error higher than 6%. For the characterization of snow condition, target decomposition theorems show promising results. For the three bands, the Freeman-Durden and Cloude-Pottier decompositions allow to understand scattering mechanisms of snow-covered-forested areas. Also, a thresholding approach applied to volume scattering power of the Freeman-Durden decomposition in C band as well as to entropy parameter together with angle [alpha] value of Cloude-Pottier decomposition shows abilities to detect wet snow over forested areas. The technique using the volume scattered power shows detection errors higher than 16%. No classification error is obtained in the error matrix generated from entropy values over the snow pits. The analysis of backscattering coefficients as a function of forest density (open area, sparse coniferous forest and dense coniferous forest) shows variations in the signal as a function of frequency, polarization, density and forest structures as well as with ground conditions (snow-free, dry snow, wet snow). Three radar vegetation indexes (IVR, IVRD[subscript HH] and IVRD[subscript VV]) are analyzed. The IVR index in C and L bands, as well as the IVRD[subscript VV] index in L band are sensitive to forest density. The volume scattered power of the Freeman-Durden decomposition also allows to characterize forest density in C, L and P bands.In order to partially reduce the effect of forested area on the backscattering of a snow cover, image difference between the C-band backscattering coefficient (HH polarization) and the C-band volume scattered power in wet snow condition is performed. The error matrix generated over the snow pit shows that a threshold of 1.5 dB applied to the image difference leads to errors less than 6%. The obtained results clearly show the utility of multifrequency, multipolarisation and polarimetric SAR data for wet snow detection over different forested areas.