Comparison of multitemporal compositing methods for burnt area detection in Southeast Asian conditions

Abstract

This study investigated the usability of different multitemporal compositing methods for burnt area detection purposes in the humid tropical conditions of insular Southeast Asia, characterised by persisting cloud cover, varying fire-induced spectral changes and large amount of small burn scars. Six monthly composites of the Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance images (MOD09) were built using six different algorithms. The performance and usability of the compositing algorithms were evaluated using three criteria: separability between burnt and unburnt areas, homogeneity and average sensor zenith angle. Maximum surface temperature method (band 31, 11.0 μm) produced the most homogeneous composites with preference to close-to-nadir observations. However, these composites showed unexpectedly low separability between burnt and unburnt areas, mainly due to low spatial resolution of band 31 (1000m). Overall, taking into account the performance of the compositing methods and the large amount of small burnt areas in insular Southeast Asia, a minimum NIR (band 2, 0.86 μm) method, combined with pre-compositing cloud and cloud shadow removal, was seen as the most suitable compositing method for burnt area detection in this region. Its strengths were 250 m spatial resolution in pixel selection and high burnt/unburnt separability combined with reasonably good performance on homogeneity and average sensor zenith angle.