Extraction And Application Of Forest Fire Information Based On Multi Source Remote Sensing Imagery

In recent years,the frequent occurrence of forest fires around the world not only posed a great threat to terrestrial ecosystems,but also had a serious impact on people’s lives.Satellite remote sensing technology,which has the advantages of high update frequency,wide monitoring range and high efficiency,has become an effective mean of real-time monitoring,dynamic tracking and post-disaster assessment of forest fires.To address the shortcomings of the current fire point monitoring algorithm and fire point monitoring products based on single remote sensing data,such as low spatial and temporal resolution,low accuracy and poor adaptability,this paper takes the "4?20Mianning County forest fire" as an example and conducts a study on the extraction method and application of fire point location and fire traces in forest fire information with high spatial and temporal resolution based on multi-source remote sensing data.The main work of this paper is as follows:(1)To address the spatial and temporal limitations of fire point information extraction caused by the single data in traditional remote sensing fire point monitoring,the method of fire point information extraction by integrating multiple sources of remote sensing data is proposed to realize the spatial and temporal optimization of fire point information extraction.The high temporal resolution of GF-4 satellite is used to achieve real-time monitoring of fire point information during daytime;the high spatial resolution of Sentinel-2 satellite data is used to extract small fire point information to compensate for the omission of small fire points extracted from GF-4 data;and Sentinel-3 satellite is used to fill the gap of fire point monitoring at night.(2)Based on the two shortwave infrared channels of the Sentinel-2 satellite,we propose the Normalized Short Wave Infrared Difference Index(NSWID)and study the method of extracting fire point information based on NSWID.Experimental results show that this method has unique advantages in extracting location information of small and medium-sized fires,and can improve the resolution of fire information to 10 m.(3)Based on the advantages of Support Vector Machine(SVM)in extracting the extent of forest fire trails,SVM was applied to the extraction of fire trails from Sentinel-2 data,and the results were more than 80% consistent with the fire trails extracted by traditional remote sensing indices(NDVI,NBR,BAI),indicating that the SVM algorithm has good potential for application in the extraction of forest fire trails.(4)This paper proposes a method to extract fire traces based on the coherence of radar images,which is based on the strong penetration of microwave remote sensing to clouds and smoke.Based on the Sentinel-1 satellite radar image,the coherence coefficient of the radar image after the occurrence of fire is used to extract the fire traces.The experimental results show that the high coherence of the post-fire radar images can be used to extract the fire traces effectively,and the method can be used as an effective supplement to optical remote sensing.(5)The forest fire simulation software Flam Map can better predict the development of forest fires in the short term,but the simulation accuracy gradually decreases with time due to the influence of random factors such as flying fires.In this paper,the Flam Map forest fire development model for the "4?20 Mianning County Forest Fire" was modified using high spatial and temporal resolution forest fire information extracted from multi-source remote sensing imagery,which effectively improved the fire development accuracy of the "4?20 Mianning County Forest Fire".The accuracy of the fire development prediction of the "4?20 Mianning County Forest Fire" has been effectively improved.