Wet Snow Cover Mapping Algorithm Based on Multitemporal COSMO-SkyMed X-Band SAR Images

Abstract

Multitemporal COSMO-SkyMed (CSK) images are exploited to map wet snow cover in a mountainous area in South Tyrol by using a ratio and a probability of error (POE) approach. Free water in the snowpack attenuates the X-band synthetic aperture radar (SAR) signal and wet snow can be classified by comparing images acquired under wet snow and snow-free conditions. The three steps of the algorithms are: preprocessing of SAR data with particular attention on the potential of speckle filtering to improve the classification, classification of wet snow and postprocessing of the snow cover area (SCA) map. Furthermore, the choice of the snow-free reference and wet snow images on the classification threshold and the SCA is assessed as well as the influence of different landcover classes (blocky scree, grassland, forest). Thresholds to distinguish snow-covered and snow-free pixels are - 2.6 dB for grassland and rocks. To quantify the accuracy of the ratio method, POE maps are calculated. The advantage of the POE method is its independency from auxiliary information on snow cover and the possibility to limit the maximum error. SCA maps derived with a maximum POE of 25% and ratio SCA maps show good overall agreement with total SCA of 66.8% (ratio) and 65.6% (POE) on 26th April 2010. A comparison to SCA derived from Landsat 7 ETM+ reveals that total SCA is similar to SAR SCA when a NDSI threshold of 0.7 is applied, but only 86% of the pixels are detected as snow from both sensors at the same time.