Develop algorithm to detect active fire and burned area in Landsat imagery

Abstract:

The massive wildfires occurred in this past two year, especially the 2018 California wildfires and 2019 Australian bushfire. These two-wildfire case occurred in different time period of summer season, California is in Northern Hemisphere, the summer begins mid-June opposite with the Australia which in Southern Hemisphere, the summer begins mid-December. Note these massive wildfire case time period occurs less than a year, the probability of wildfire occurred two time period a year in global coverage area. In other hand the fire also happens due to human disturbances such as traditional farming of slash-and-burn, most of fire cases occurred in developing country. Thus, it is crucial to detect the active fire and burned area for further research and investigation. Here, we develop an algorithm to detect active fire and burned area by using multispectral bands of Landsat images. To detect the active fire and burned area we include some variable such as Landsat SWIR 2 band, thermal band and LST (land surface temperature), also for burned scar variable based on Near Infrared band, burn index and NDVI (Normalized Difference Vegetation Index) value and the buffer area from active fire. More than 100 Landsat images sample are chosen in global coverage, to test our algorithm. The final product of our algorithm is a masking image file of active fire and burned area in TIF file, the product can be used for further research and study about the wildfires. To validate the results, we compare spatial relationship of our product with MODIS M6 (MODIS Collection 6 NRT Hotspot) product, it shows high correlation coefficients more than 0.75. It is concluded that the algorithm can be used as a good indicator in determining the active fire and burned area.