Unsupervised Burned Area Estimation through Satellite Tiles: A multimodal approach by means of image segmentation over remote sensing imagery

Climate change is increasing the number and the magnitude of wildfires, which become every year more severe. An accurate delineation of burned areas, which is often done through time consuming and inaccurate manual approaches, is of paramount importance to estimate the economic impact of such events. In this paper we introduce Burned Area Estimation through satellite tiles (BAE), an unsupervised algorithm that couples image processing techniques and an unsupervised neural network to automatically delineate the burned areas of wildfires from satellite imagery. We show its capabilities by performing an evaluation over past wildfires across European and non-European countries.