Forest fire detection in far east Russian region with NOAA AVHRR images

We report a study of forest fire detection around Yakutsk with NOAA AVHRR images. The purpose of the study is verifying that a method using a three-dimensional histogram has a higher accuracy for forest fire detection than a previous method using thresholds and verifying forest fire detection for an independent area. In our method, we used the forest fire information that we stored to train a three-dimensional histogram in far east Russia from 1996 to 2000. Our method showed better accuracy than the previous one in a different area from the training area. At non-fire points (e.g. high-brightness clouds) the previous method failed sometimes, but our method judged correctly. For this reason, our method can detect forest fire with better accuracy than the previous one and may be applied to other geographical areas.