Contextual algorithm adapted for NOAA-AVHRR fire detection in Indonesia

Abstract

A widely cited contextual algorithm for fire pixel detection was applied to NOAA-AVHRR data of Indonesia acquired during the severe fire season of 1997. The contextual algorithm tended to underestimate the number of fire pixels. Many fire pixels which could be identified visually in AVHRR band 3 imagery were missing in the output of the contextual algorithm. Through experimentation with the data. the authors were able to developed two alterations to the original algorithm to enhance the accuracy of fire detection in the Indonesian context. The enhanced algorithm was made capable of detecting fire pixels that the original algorithm failed to identify.