Deforestation Monitoring Using Sentinel-L SAR Images in Humid Tropical Areas

Tropical forests are vulnerable to deforestation and various monitoring techniques have been developed based on remotely sensed data to map deforestation, but are facing multiple problems in the tropical areas. For instance, the techniques based optical data, which are widely used to monitor deforestation, face severe limitations in the humid tropical forest due to high cloud cover. Sentinel-l C-SAR dense time series can be used for a temporally more accurate monitoring. In this study, a change detection algorithm commonly used in the financial domain, the Cumulative Sum (CuSum) algorithm, was modified to be applied on time-series of Sentinel-l images in a forest concession of Democratic Republic of Congo (DRC) near Kisangani. The validation was made through the visual interpretation of PlanetScope OrthoScene images as in-situ data were missing. The results show a precision up to 0.75, an accuracy up to 0.95 and a kappa coefficient up to 0.40 for clear cut detection. The algorithm is able to detect forest degradation activities before the clear cuts.