Identification of burned areas in the Liguria region using Landsat and QuickBird images. The case study of Monte Fasce.

The aim of this study is to test and compare different remote sensing techniques to define burned areas in Liguria. The test area is the Monte Fasce site, affected by a huge fire in September 2009. The work is based on the Landsat TM and the QuickBird images acquired before and after the event. We considered bands, PCA, texture analysis and several spectral indices reported in literature. The indices were compared empirically and using two algorithms (ROI separability and the software SEATH) to find the most suited ones to detect the burned zones. Once the base data have been characterised, the burned area was extracted using different methods: thresholds, decision trees, the Maximum Likelihood classification, the ENVI and RHSEG segmentation and the Change Detection technique. The maps' accuracy of the areas covered by fire was estimated by comparing the satellite data with those taken on the ground by the Forest Service and the ones provided by a visual analysis of the post-event QuickBird image. The best results were obtained with the multitemporal technique computing the pre- and postimage difference: the Landsat data give an overall error of 22.75% applying a multithreshold technique with the indices NDVI, NBR and NBRT; the QuickBird data show an error of 22.8% using the NDVI index. Future improvements should envisage a methodology to reduce the error and a thorough analysis on a range of burned areas in the Liguria region.