Threshold-Variation Analysis of MOD14 Algorithm Model Using MODIS Datasets Over Tropical Area in Indonesia

Abstract

The Moderate Resolution Imaging Spectroradiometer (MODIS) which is equipped with thermal bands for measuring the temperature of the earth’s surface has been widely utilized in fire detection activities. Of the fire detection algorithms developed for MODIS the most commonly applied algorithm principle is the contextual algorithm. Such global algorithms for fire detection can be customized for local or regional areas to improve the accuracy of fire detection. For this reason there has been substantial research performed to apply and assess how accurate the global algorithm results are in certain local or regional areas. In principle, fire pixels detection is base on dichotomizing assessed imagery pixels values using certain threshold value. Determining the best threshold value leads to optimal detection results and minimum errors. This paper analyses the impact of variation of T4 and T threshold values on outputs from the MOD14 algorithm using MODIS data within different seasons in Indonesia. Analysis is conducted by analysis of the sensitivity of detected Fire Hotspots (FHS) to changes of T4 and T and also through statistical calculations. The results show that the sensitivity of the MODIS FHS algorithm is different between dry and wet seasons which indicates the need for different threshold values to be applied to optimise fire detection using MODIS within dry and wet seasons.