Forest loss (2018 - present) (CNES)

This map shows forest cover loss over Sentinel-1 acquisitions from January 2018 to present for Guyana, Suriname, Guyana, Gabon, Vietnam, Laos, Cambodia with a pixel size of 10 meters. The pixel values indicate the date of deforestation detection.

Forest loss maps are produced on the high computing performance CNES cluster. For example, 71,000 Sentinel-1 images have been processed to map forest loss from 2018 in Vietnam, Laos and Cambodia, covering approximately 1,230,000 km².

Forest loss detection method The detection method has been initially developed at CESBIO since 2015. The method has been constantly tested over various countries and improved, thanks to a fruitful collaboration between CESBIO, CNES and GlobEO, and to funding from ESA (SOFT project) then CNES (TropiSCO project). The method is based on the use of basemaps such as forest-non forest classifications, mangroves (Bunting et al., 2018), water surface areas (Pekel et al., 2016), and digital elevation models from SRTM. The forest-non forest mask from Potapov et al. (2019) has been used over Vietnam, Laos and Cambodia. The mask from the ECOSEO project has been used over Guyana, Suriname and French Guiana, and the mask from AGEOS over Gabon.

The two confidence indexes have been designed to answer various needs. The confidence index very high corresponds to the official validation results (false negatives of 10% and false positives of 0.9% respectively with a minimum mapping unit of 0.1 hectare in Vietnam, Laos and Cambodia). This level of confidence is associated to a lower rate of false positives, and as such is better adapted to operational monitoring with associated interventions in situ. The confidence high can be used for example for assessing relevant statistics.