Object-based Vs Pixel-based Mapping of Fire Scars Using Multi-scale Satellite Data

Assessment and mapping of burned areas using satellite data has been implemented so far at local, regional and global scale, by employing various image analysis techniques, according to the spectral and spatial characteristics of the remotely sensed data. The spatial resolution of satellite data together with landscape configuration guides of the decision for the classification approach to be applied, since it determines the existence of a Highor a Low-resolution model of the scene. The recent availability of very high spatial resolution satellite data reconsiders the digital methods used for information extraction and thematic mapping. The strength of object-based analysis lies on the fact that individual pixels, composing real world objects, can be identified in the satellite imagery and classified using semantic and heuristic knowledge Object-based analysis of remotely sensed imagery has attracted attention in burned area mapping; a particular case of thematic mapping. However, some questions are posed regarding its applicability and performance at multiple scales. In this study, we discuss the strengths and the weaknesses of the object-based classification approach and compare it with a classical pixel-based for burned area mapping in a Mediterranean study area. Multi-sensor satellite data of multiple spectral and spatial resolutions from IKONOS, ASTER and MODIS sensors of the same burned area allowed us to evaluate the multi-scale performance of the object-based approach. Large scale true color aerial photography acquired shortly after the fire served for accuracy assessment of the multi-scale burned area maps.