The effect of vegetation type and density on X-band SAR backscatter after forest fires

Abstract

Various frequencies (e.g. visible light, infrared, microwaves) from remote sensing sensors can be used for active fire mapping, forest fire detection and fire emission assessment. However, little is known about the applicability of X-band SAR data for burned area detection. This paper presents a detailed SAR backscatter coefficient analysis and accuracy assessment with respect to CORINE 2006 land cover data. For this purpose five forest fires have been analyzed. Dry as well as wet acquisition conditions have been taken into account. The analysis demonstrated that largest differences in backscatter coefficients between pre- and post-fire conditions were linked to tall and dense vegetation types. Contrarily, scant vegetation was marked by lowest signal differences. High correlation coefficients have been obtained from regression analysis between vegetation indices and SAR backscatter changes. Moreover, a burned area classification algorithm with different thresholds for each vegetation type has been applied. The classification result illustrated that areas abundantly covered with vegetation showed classification accuracies of ~91 %, whereas sparse vegetation achieved ~5 % accuracies.