An efficient detection method of forest fires from MODIS and Landsat satellite images

Forest fires could cause significant damages in environment, human society and economy. These possibilities lead forest fires to a worldwide issue. Furthermore, damages from forest fires give rise to secondary damage such as floods. Thus, a rapid analysis on forest fire should take effect. However, it is considerably dangerous to access to forest fire area and almost impossible to get the field data measured by person. Also, because forest fires are quickly spread, real-time respond is surely required. Thus, satellite images, which have advantages not only on accessibility but also on safety, are very useful to measure forest fires quickly. Therefore, the main objective of this study is to detect forest fire in the forest area using Moderate Resolution Imaging Spectroradiometer (MODIS) images. To detect forest fire areas from MODIS images, an efficient fire detection algorithm is used. The algorithm uses brightness temperatures extracted from the channel 21 and 22 (4μm) and the channel 31 (11 μm) of the MODIS instrument. Forest fire is detected when the brightness temperature derived from these channels is higher than 360K in daytime, 330K at night time. To prevent false detection, pixels for T₄ < 15K or T₄ - T₁₁< 10K are removed at daytime. With this efficient algorithm, we detected forest fire and square the area of the active forest fire using the MODIS images. In addition, to remedy MODIS's lower resolution, 1km × 1km, we extract forest fire areas using Normalized Burn Ratio (NBR) from Landsat images which have 30m × 30m resolution. The results show that we can observe forest fire area fairly using MODIS image and calculate delicate fire area using pre- and post-fire Landsat images. This process could help have more accurate results of fire detection area.