Selective logging monitoring in Brazilian forests under management using REDD+AI data

Selective logging can be characterized as a sustainable forest management practice (legal) or as one of the main vectors of forest degradation (illegal), depending on its origin and intensity. Brazilian public forests under legal management contracts are operated by timber companies, but we can also observe patterns of potentially illegal selective logging outside the management units. REDD+AI platform combines deep learning with high spatial and temporal resolution images, representing an effective alternative for detecting human-induced disturbances. In this study, we evaluated the potential and limitations of the REDD+AI selective logging detection based on Planet NICFI imagery in national forests designated for management, monitored by the Brazilian Forest Service (SFB). Compared to SFB reference data, REDD+AI maps showed 67.8% average agreement (F1-score). Our analysis demonstrated the potential of this product for monitoring forest management, and also for potential illegal selective logging activities detection outside management units.