Accuracy of wet snow mapping using simulated Radarsat backscattering coefficients from observed snow cover characteristics
Wet snow cover mapping by means of airborne and spaceborne SAR is operational today and successfully applied in rugged high mountain terrain and in agricultural area. This paper proposes a numerical study to estimate the accuracy of wet snow mapping by using a radar backscattering model that simulates backscattering from a multi-layer snowpack for various snow cover conditions and for SAR parameters specific to Radarsat (C-HH). Field measurements carried out in numerous sites during the winters of 1994 to 1996 in several areas of Quebec (Canada) have allowed to choose some typical snow profiles and the corresponding snow/soil parameters. Results indicate that under the assumptions used in the model and the simulations, for the standard mode S1 of Radarsat (20 to 27.4) and in the case of wet snow cover with liquid water content of 1%, the optimum relative under-and over-estimation of wet snow pixels are of the order of 23.9% and 13.4%, respectively. For wet snow cover at 2%, the algorithm operates with a relative under-estimation of wet snow pixels around 8.5% and a relative over-estimation of the order of 1.7%. For wet snow with liquid water content of 4%, the relative under-and over-estimation of wet snow pixels is around 0.8% and 0.3%, respectively. They are negligible for wet snow with liquid water content higher than 4%. With the standard mode S7 of Radarsat (44.9 to 49.4), the wet snow mapping algorithm leads to a slightly lower performance than with the standard mode S1. The accuracy of the method for wet snow mapping demonstrates the high potential of SAR for snow monitoring. It is considered sufficient when the liquid water content of the snowpack is higher than 1% for actual snow conditions similar to those eight observed conditions used in this study.