Identification of Peatland Burned Area based on Multiple Spectral Indices and Adaptive Thresholding in Central Kalimantan

Abstract

Nowadays, spectral index has become popular as a tool to identify fire-burned areas. However, the use of a single index may not be universally applicable to region with diverse landscape and vegetation as peatlands. Here, we propose to develop a procedure that integrates multiple spectral indices with an adaptive thresholding method to enhance the performance of burned area detection. We combined the Normalized Difference Vegetation Index (NDVI) and the Normalized Burn Ratio (NBR) using MODIS imagery from 2002 to 2022 to calculate  (Confirmed Burned Pixel) by filtering dNDVI and dNBR. The mean and standard deviation of  serve as inputs for image thresholding. We tested our approach in Sebangau peatland, Central Kalimantan, where fires occur annually. The results showed that the model performed well with overall accuracy > of 91%, indicating that the model is effective and reliable for identifying burned areas. The findings also revealed that the frequency of fire is below 2 times/year, with the southeastern is the most fire prone regions. Further, our findings provide an alternative approach for identifying burned areas in locations with diverse vegetation cover and different geographical regions.