Towards a fully automatic processing chain for operationally mapping burned areas countrywide exploiting Sentinel-2 imagery

Burned area mapping is essential for quantifying the environmental impact of wildfires, for compiling statistics, and for designing effective short-to mid-term impact mitigation measures. The Sentinel-2 satellites are providing an unparalleled wealth of high-resolution remotely sensed information with a short revisit cycle, which is ideal for mapping burned areas both accurately and timely. However, the high detail and volume of the information provided actually encumbers the automation of the mapping process, at least for the level of automation required to map systematically wildfires on a national level. This paper presents a preliminary methodology for mapping burned areas using Sentinel-2 data, which aims to eliminate user interaction and achieve mapping accuracy that is acceptable for operational use. It follows an object-based image analysis (OBIA) approach, whereby the initial image is segmented into a set of adjacent and non-overlapping small regions (objects). The most unambiguous of them are labeled automatically through a set of empirical rules that combine information extracted from both a pre-fire Sentinel-2 image and a post-fire one. The burned area is finally delineated following a supervised learning approach, whereby a Support Vector Machine (SVM) is trained using the labeled objects and subsequently applied to the whole image considering a set of optimally selected object-level features. Preliminary results on a set of recent large wildfires in Greece indicate that the proposed methodology constitutes a solid basis for fully automating the burned area mapping process.