Update of land cover maps for watersheds of small reservoirs

The works presents the update of land cover maps through GIS and remote sensing techniques with the support of free or open-source products and data. The analysis was performed using a supervised classification approach, applying a Random Forest algorithm to Sentinel-2 images enhanced with various spectral indices. The procedure is developed in QGIS and Google earth engine environment (GEE). The study aims at increasing the knowledge of small watersheds characteristics, a fundamental aspect for the modelling and management of hydrological and erosion processes. Specifically, the land cover is computed in the watersheds of twenty small reservoirs in the upper Tiber river basin, enabling the identification of 7 different land classes: water, industrial and civil building, low density and intensive cultivation agricultural areas, forest, bare soil; the overall accuracy obtained was 92.1%. The land cover of the watersheds allows the categorisation of small lakes into macro-groups, linked to anthropic or natural areas, providing a first step for the management and environmental protection strategies to be adopted.