The evaluation of forest fire risk is an important issue in Mediterranean areas where the long arid season often creates favourable conditions for the occurrence of fires. In this Letter three indices related to this risk have been produced and compared for the western part of the Elba Island (Central Italy). The first index is based on the analysis of environmental information layers (topography, vegetation and soil type) within a Land Information System, while the other two are derived from a summer Landsat Thematic Mapper (TM) scene processed by unsupervised and supervised procedures, respectively. The results show the effectiveness of all these approaches, and, in particular, a greater accuracy of the supervised spectral index.