Combined Use of SAR and Optical Time Series Data for Near Real-Time Forest Disturbance Mapping

Tropical forest monitoring with EO is limited by two main factors: frequent cloud cover and rapid forest regrowth. Both can be overcome by using temporally dense optical image stacks and SAR imagery that is independent of cloud cover. We present a method making use of both SAR (Sentinel-1) and optical (Sentinel-2 and Landsat-8) time series data to map forest disturbances. An initial forest/non-forest map is calculated based on time-series of optical data. The initial forest/non-forest map is then updated based on the detected forest disturbances from SAR and optical data stacks which are merged based on the Bayes' theorem. The method was applied at a complex tropical forest site in Peru. Disturbance detection accuracies were computed for the S-1, optical only and combined approach. The combined approach shows the highest detection accuracies with 83.7 % for the area-based and 97.1 % for the plot-based validation. Our results argue in support of future near real-time multi-sensor tropical forest monitoring systems.