SE France is the area most affected by wildfires and there is a variation of the fire occurrence and of the fire size within this region. So for better fire prevention on a local scale, it is necessary to better understand this variation of wildfires in space.Using the forest fires database of the French Forest Office, the fire density (1960-2011) was investigated at local scale in the département Bouches du Rhône. The municipalities of the study area were divided into 5 clusters according to descriptive variables (climate, land cover, WUI) using hierarchical cluster analysis. Statistical analyses showed that the mean fire density differed significantly according to the territory. Co-inertia and spatial analyses analyses were performed on each of the 5 clusters taking into account climate conditions, topography, land cover, WUI, network and population/housing densities as explanatory variables and 3 classes of fire density as dependent variables.According to both types of analyses, high fire density was mainly linked to high proportions of WUI except in cluster 4, of wildland vegetation in clusters 1 and 5 and of agricultural area only in cluster 4. Cool and wet climate conditions and steep slope were linked to high fire density in clusters 1 and 2 (for the slope only) and high population/housing density was an important driving factor in clusters 3 and 4. Network density (especially minor roads) was linked to high fire density only in cluster 3. Low fire density was mainly linked to high proportion of agricultural area except in clusters 3 and 4 in which relief climate conditions and high elevation as well as steep slope (only in cluster 4) were the main factors that mitigated fire density. This work showed that, at the local scale, the fire density driving factors varied within the study area. Their identification could improve fire prevention as they can better be targeted. This method can be extended to other regions regardless of the ecosystems.