Detecting forest fire omission error based on data fusion at subpixel scale

Abstract

Forest disaster fires pose a serious threat to forest resources and people's lives and property, and may also cause ecological disasters and social crises. There is a phenomenon of forest fire omission error, due to combustibles incomplete combustion in the forest fires early stage, low ground temperature and weak infrared radiation energy, and influence of forest canopy shading. Smoke detection based on satellite imagery is imperative for forest fire detection. Therefore, when forest fire occurs, forest fire smoke detection combined with forest fire infrared radiation monitoring can greatly reduce the phenomenon of forest fire omission error. Meteorological satellite imagery is commonly used for near real-time forest fire monitoring, thanks to its high temporal resolution. However, due to the low spatial resolution of meteorological satellite imagery and a large number of mixed pixels, it cannot accurately locate forest fire, and we introduced the concept of subpixel mapping, which is studied in detail in the article (https://doi.org/10.3390/rs14102460) we published. In this paper, the Himawari-9 satellite imagery is used to detect forest fire smoke at subpixel scale based on the Modified Pixel Swapping Algorithm (MPSA) and detect forest fire infrared radiation at subpixel scale based on the Mixed-Pixel Unmixing integrated with Pixel-Swapping Algorithm (MPU-PSA), respectively, obtaining forest fire smoke detection and infrared radiation monitoring results, and average value of forest fire identification accuracy is 41.78 % and 83.96 %, respectively. The Threshold-Weighted Fusion (TWF) method is used to fuse subpixel scale forest fire smoke detection result and forest fire infrared radiation monitoring result, and average value of forest fire identification accuracy is increased from 83.96 % (forest fire infrared radiation monitoring alone) to 93.97 %. Results show that fusion of subpixel scale forest fire smoke detection result and forest fire infrared radiation monitoring result can greatly improve spatial positioning accuracy of forest fire monitoring and reduce phenomenon of forest fire omission error.

Keywords

• Subpixel scale
• Forest fire identification
• Data fusion
• Satellite remote sensing