Forest fire smoke analysis in far East Russia

Forest fires make large quantities of forest resources lost, and promote global warming. In Far East Russia, since forest fires occur every year, they have had a great influence on the earth environment. In this paper, we present a three-dimensional histogram method for detecting smoke of forest fires from NOAA AVHRR images. We used total 2,450 scenes from 2001 to 2003 and chose 378 scenes for experiments. The proposed method consists of the following two steps: create the three dimensional histogram from smoke pixels, detect smoke pixels from unknown images. We used 241 scenes of Zeya and Baikal regions to create the three dimensional histogram and 137 scenes of Sakhalin region for evaluation. There are ten combinations of the three dimensional histogram since the AVHRR has five channels. We confirmed that the proposed method can clearly distinguish smoke from clouds by comparing the smoke with the hot spots of forest fires. The combination of the channel-1, -2, and -4 is the best combination for detecting smoke pixels.