Detection And Characterization of Forest Harvesting In Piedmont Through Sentinel-2 Imagery: A Methodological Proposal

Abstract

This study evaluated the effectiveness of Sentinel-2 (S2) as a tool for early detection and estimation of forest harvesting in the Piemonte Region, which can be used by the regional forest administration. The priority was the detection, at the regional scale, of annual forest cover changes with the following goals: i) mapping of irregular (in respect of the regional Forestry Regulation) forest cuts; ii) quantification of the intensity of the silvicultural interventions. Results are expected to support forest police controls. The proposed procedure is based on a supervised classification approach based on Random Forest algorithm. Accuracy of harvested areas detection proved to be high (overall accuracy 98%). Characterization of the occurred forest cuts was obtained computingthe local coefficient of variationof the normalized difference vegetation index (NDVI) after harvesting, that showed to be a good predictor of forest harvesting intensity.