Classification of Mediterranean vegetation by TM and ancillary data for the evaluation of fire risk

Mediterranean vegetation is strongly subjected to the risk of wildfires, which can become a major cause of land degradation. The knowledge of the spatial variations of this risk is essential, therefore, for forest resource management. Relying on the fact that different vegetation types can be associated with different risk levels, a classification approach based on the use of Landsat Thematic Mapper (TM) scenes is currently proposed for the generation of maps related to fire risk. Hard and fuzzy classifications were tested for this purpose on Elba island (central Italy), taking into account the effects of the use of scenes from different periods (spring and summer) and of ancillary data. The fire risk images obtained were evaluated by comparison with the fire events that occurred on the island during the last decade. The results show that, while the acquisition period has only minor effects, classification accuracy is strongly dependent on the inclusion of ancillary data. Moreover, the fuzzy approach better exploits the information of the integrated datasets, producing maps which are temporally stable and highly indicative of the fire risk in the study area.