Unsupervised post-fire assessment of burned areas with free and open multispectral data using OBIA
Résumé
"This paper presents a new usupervised workflow to assess the extension of the burning damages due to wild fire events. The workflow make use of open access medium and high resolution multispectral images from the Landsat-8 and Sentinel-2 satellites, together with open access low resolution products provided by the MODIS and the VIIRS missions.In particular, the proposed workflow considers strong variations in the Normalized Difference Vegetation Index (NDVI) between two observation dates in order to detect burning damages. The changes in NDVI not related to the wildfire effects are filtered thanks to the MODIS and VIIRS hotspot detections occurred during the same observation period. To this aim, an Object-Based Image Analysis (OBIA) approach is employed. In fact, the proposed false positive filtering method is based on object features such as {\it 'overlaps with'} and {\it 'distance to'} detected hotspots. Experimental results on the Vesuvius, Southern Italy, wildfires of July 2017, show the effectiveness of the proposed method."