Comparison of Three Different Random Forest Approaches to Retrieve Daily High-Resolution Snow Cover Maps from MODIS and Sentinel-2 in a Mountain Area, Gran Paradiso National Park (NW Alps)

In the Alpine environment, snow plays a key role in many processes involving ecosystems, biogeochemical cycles, and human wellbeing. Due to inaccessibility of mountain areas high spatial temporal heterogeneity snowpack, satellite spatio-temporal data without gaps offer unique opportunity monitor on fine scale. this study, we present random forest approach within three different workflows combine MODIS Sentinel-2 products retrieve daily gap-free cover maps at 20 m resolution. The differ terms type ingested and, consequently, used. required inputs are MODIS/Terra Snow Cover Daily L3 Global dataset 500 m, automatically retrieved through recently developed revised-Let It workflow, from which selected are, alternatively, Extent (SCE) map or Normalized Difference Index (NDSI) map, Digital Elevation Model (DEM) consistent resolution with imagery. algorithm is based two steps, first fill second downscale obtain time series. workflow applied case study Gran Paradiso National Park. proposed represents attempt use purpose extracting parameters snow. validation was achieved by comparison both an independent assess accuracy, including snowline elevation prediction, algorithm’s performance topographic conditions, in-situ collected meteorological stations, focus seasonal phenology parameters. Results show that all approaches provide robust series (overall accuracies A1 = 93.4%, A2 A3 92.6% against Sentinel-2, 93.1%, 93.7%, 93.6% weather stations), but requires about one fifth computational resources needed for other two. fully automatic input readily globally available, promises be easily reproducible high-resolution series, crucial understanding snow-driven scale, such as vegetation dynamics after snowmelt.