Fire Spectral Analysis for Detection and Estimation of Fuel Area Using Satellite Images

Every year hundreds of fires are occurring in our country; So that thousands hectare of forests, pastures and vegetables are being destroyed. Fires, in addition to economic damages cause environmental effects. In spite of effective factors in fire occurrence, monitoring, controlling and evaluating damages are of great importance. Frequent fire occurrence in different regions of Iran in the past few years cause natural resources researchers to conduct new research about forests in order to prevent from occurring and expanding fire or to differ its time. After launching Remote Sensing satellites, collecting information from earths surface become possible and by developing these technology, monitoring natural arena in short time become possible. Monitoring and evaluating damages are some of the Remote Sensing applications. In this research, different satellite images for detecting and determining fired region in the Khour palms in the Esfahan province were used. MODIS sensor recorded a fire on 1/8/2008, which the enhanced fire detection algorithm specified location of this fire. In the proposed fire detection algorithm, water and cloud mask is applied to eliminate thecorresponding pixels and then an appropriate threshold is applied to identify the potential fire regions. Afterwards, the pixels which have brightness temperature, spectral reflectance and temperature difference values smaller than 293 K in band 22, 0.3 in band 2 and 10 K between bands 22 and 31 respectively, will be labeled as potential fire pixels. Then using different thresholds and tests, fire pixels are extracted. The results indicate that based on global scale, using threshold value of 310 K for band 22 is not useful for detection of small and cold fires. Thus an optimized threshold has been used. Also, for evaluating damages and determining the area of burnt regions, a Landsat ETM+ satellite image was used which maximum likelihood classification is performed to identify the burned area. The achieved results show that about 9 hectare of the palms are burnt due to the occurred fire.