Mapping snow cover from daily Collection 6 MODIS products over Austria

Abstract Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover maps have been successfully applied in regional snow cover mapping and hydrological modelling in many regions. The new MODIS products (i.e. Collection 6) provide information about the Normalized Difference Snow Index (NDSI) instead of simple binary information about snow cover of the former products. The objective of this study is to investigate the sensitivity and accuracy of different NDSI thresholds used for snow cover mapping and to compare the results with former snow cover classification based on a fixed NDSI threshold (NDSI=0.4). The accuracy is tested for both Aqua (MYD10A1) and Terra (MOD10A1) daily snow cover products by using daily snow depth observations at 665 climate stations in Austria in the period 2002 - 2014. The results show the overall classification accuracy over 665 climate stations to be larger than 97% (97.4% for Terra and 97.6% for Aqua) in Austria. The best NDSI thresholds fitted to individual stations can differ from a fixed threshold (NDSI=0.4) and are sensitive to the snow depth threshold indicating snow cover at the ground. The thresholds vary seasonally, decrease with increasing elevation and are lower in forested than open land cover settings. We found that the thresholds fitted to different elevation and land cover classes improve regional snow cover mapping by 3-10% in forested regions above 900m a.s.l. in January to March.