This paper focuses on determining ignition probabilities of forest fires in Wildland-Urban Interface (WUI) areas to more effectively develop prevention plans. Multivariate Logistic Regression methodology was used to identify the most important biophysical and human variables to explain the emergence of ignition points, incorporating spatial analysis from Remote Sensing and Geographical Information Systems (GIS) data. To test this model we used two representative Wildland-Urban Interface landscapes in a Mediterranean environment, located in Catalonia (northeast Spain): an example of dispersed housing in a forested area associated to metropolitan processes and an agro-forestry mosaic connected with tourism development. For a better understanding a temporal comparison has been made, analyzing data from 1990s and from 2000s. Results show differences in the explicative models; in the former study area, high ignition probabilities are associated to human activity, mainly distance to urban areas and road networks, whereas in the latter they are related with land-use (scrubland and coniferous forest) and mean maximum temperatures. As a consequence, prevention tasks seem to be less difficult in the more metropolitan study area because the spatial model is further disperse in the agro-forestry mosaic. Finally, temporal analysis indicates that both areas were more prone to forest fires in the most recent decade than in the 1990s. © 2011 Elsevier Ltd.
- GIS and remote sensing data
- Ignition points
- Multivariate logistic regression
- Wildland-urban interface