Anais do XX Simpósio Brasileiro de Sensoriamento Remoto

In this study, we present and evaluate the results of Collection 2 of DETER-R, a fully automated Near Real-Time forest disturbance detection system based on Sentinel-1 (S1) data. The main difference from Collection 1 is the change of the Minimum Mapping Unit from 1.0 to 0.4 ha. This change led to the detection of a higher number of smaller warnings, at the cost of including commission errors. During three months of tests, Collection 2 emitted 19,312 warnings, totaling 38,533.5 ha. It is estimated that, after a period of stabilization, around 10% of the number of warnings and 5% of the area do not correspond to real forest disturbance events. This was considered an acceptable compromise, which prompted the substitution of Collection 1 for Collection 2 in August 2022.