Feasibility of Landsat TM imagery for burned area mapping in forests (Case study: Qouri Qaleh area in Kermanshah Province)

According the available statistics, the average annual rate of approximately 700 ha of forested areas are subject to severe degradation caused by wildfires in Kermanshah province located in western Iran. However, the heterogeneous spatial distribution of fires is mainly driven by factors such as the occurrence of harsh mountainous areas, which is coupled with the accessibility problems that notably constrain the ability of administrative units to achieve reliable quantitative information e.g. on the size of the burned areas and the level of damages. The satellite imagery provide useful information on various spatiotemporal scales which enable a more realistic assessment of consequences of forest fire. Therefore, this study aims at testing the performance of LANDSAT TM data for burned area mapping across a selected test site in Zagros Mountains. To this aim, the required imagery was obtained for 1389 (2010) over Qoori Qaleh area of Kermanshah Province. A set of digital image processing methods were applied, including band ratioing, principal component analysis and analysis of separability amongst the burned area classes. The resulted dataset was further classified by a supervised Maximum Likelihood approach. The results showed the TM band 4 as well as a composite of bands 3, 4, 7 and the first principal components of bands 1-3 and 4-7 yielded an overall accuracy of 90 and 92.75% and kappa coefficients of 0.8 and 0.86, respectively. Thus the applied dataset is concluded to be appropriate for burned area mapping in such areas.