Analysis of the contribution of radar satellite images for the snow cover estimation

Snow dynamics is a key hydrological process in alpine catchments. The snow accumulation formed during the winter feeds the dams during melting and so the snow quantification is important for dams managing. Snow data obtained from optical images (MODIS product) can be used to improve the simulation of the water flow using an hydrological model (MORDOR-TS, Le Lay, 2018). However, when there are clouds, this data cannot give any information.

To overcome this difficulty, this study presents an additional snow detection method using Synthetic Aperture Radar (SAR) data. The SAR images analysed come from Sentinel-1 (C-band) acquired in IW mode with a resolution of 5m by 20m. These images are obtained under two different polarizations (VV and VH). Before analyzing the SAR images, several pre-treatments such as despeckling, radiometric calibration, coregistration and layover detection are carried out.

The study is conducted around two snow gauges located at high altitude (2275m and 2685m) in the Guil catchment during an accumulation-melting cycle (September 2018-June 2019).

Two types of snow detection methods are used. The first one is a wet snow detection method (Nagler et al., 2016) that compares the analyzed image with a reference image. It allows to determine in a binary format if there is snow or not. The second one is a dry snow detection method (Lievens et al., 2019) which performs a comparison between the two polarizations of the analyzed image and determines a proportional snow depth.

The results were compared to the snow gauges data. Both methods appear to be complementary. Moreover, the time series obtained with snow dry detection method follows the tendency of snow gauges data during cold periods. Spatially over an area of 1600m by 1000m, the complementarity of the two methods can be seen once again. Despite this complementarity, a little presence of misdetection are observed at the resolution of the S1-images. However, when averaged to the resolution of MODIS data (500m by 500m), the detection results are consistent with the ground truth data. 

In the end, this study shows that we can efficiently detect snow with SAR images thanks to two complementary methods. Thus, SAR images add information about the snow cover up to the point of even estimating the snow depth with higher resolution than optical images. 

Le Lay M., Rouhier L., Garavaglia F., Hendrickx F., Monteil C., Le Moine N., and Ribstein P. (2018) Use of snow data in a hydrological distributed model: different approaches for improving model realism, EGU General Assembly 2018, Vienna, Austria.

Lievens, M. D., Marshall, H.-P., Reichle, R. H., Brucker, L., Brangers, I.,de Rosnay, P., Dumont, M., Girotto, M., Immerzeel, W. W., Jonas, T., Kim, E. J., Koch, I.,Marty, C., Saloranta, T., Schöber, J., and Lannoy, G. J. D. (2019). Snow depth variability inthe Northern Hemisphere mountains observed from space.nature communications.

Nagler, H. R., Ripper, E., Bippus, G., and Hetzenecker, M. (2016). Advan-cements for Snowmelt Monitoring by Means of Sentinel-1 SAR.remote sensing.