Mapping Savanna Wildfires in Southern Belize Using Sentinel-1 SAR and Object Based Image Analysis

A new approach for delineating burned areas in savannas is presented using a series of Sentinel-1. SAR images and object-based image analysis (OBIA). Using analysis-ready Sentinel-1 data increases the frequency of observation to several scenes per month, enabling more timely and precise detection of fire events during a fire season. Using an object-based segmentation, classified using backscatter indices calculated between consecutive image pairs, there was 87.6% spatial agreement with expert interpretation of cloud-free sentinel-2 imagery. Additionally, we map the timing and extent of individual burnt areas and their coalescence. We apply the method to produce more detailed and precise maps of the progression of fires during 2019 in Belize than previously available from passive multispectral sensors. We discuss the scope for extending the method and draw preliminary conclusions on the effectiveness of the temporal indices and OBIA for detecting burns in this landscape.