Detecting burnt severity and vegetation regrowth classes using a change vector analysis approach: a case study in the southern part of Sumatra, Indonesia.

This study describes the development of burn severity and vegetation regrowth classes using vegetation (NDVI) and bareland (NDBI) indices-based change vector analysis (VI-CVA) with a case study on the fire event that occurred at the Berbak National Park, Jambi Province, in 2015. The main objective was to determine the type and the severity level of change due to fire or vegetation regrowth, as summarised in CVA magnitude and direction images. The vegetation and bareland indices were derived from Landsat medium-resolution images to detect the degree of change caused by the forest fires. The study found that severity and vegetation regrowth could be classified into five classes: unburnt, very low, low, and moderate severity burn classes and a moderate regrowth class from bare land to oil palm plantation, and unburnt. It was also found that the performance of this CVA approach was superior to the delta normalized burn ratio (dNBR) method as indicated by its ability to detect five post-fire severity classes with 87.7% overall accuracy compared with dNBR, which detected four post-fire severity classes with 66.9% overall accuracy. This paper reports that burn severity and vegetation regrowth can be classified well using change vector analysis (CVA). CVA using an NDVI and NDBI approach was superior to the dNBR method as indicated by the accuracy difference between both methods. [ABSTRACT FROM AUTHOR]