Abstract:
Background The investigation of burned areas after forest fires is a key and difficult point in the field of forest fire research. Traditional manual on-site investigation methods, satellite remote sensing images, and drone technology, as common investigation methods, have their own drawbacks when used alone, which cannot solve the problem of quickly and accurately dividing the severity of large-scale forest fires. Therefore, taking the location of the Chongqing Jinyun Mountain fire in September 2022 as the experimental area, and combining the advantages of the above three methods, a "star machine ground" forest fire investigation system is constructed.
Methods Selecting mildly, moderatly, and severely fire plots, and ruler processing and mapping for each wood was conducted. Based on satellite remote sensing and unmanned aerial vehicle remote sensing images of the survey area, NDVI was calculated and spatially matched with the measured tree height, chest diameter, scorch height, and crown width of the plots. Grey correlation analysis was conducted to analyze the correlation between various indicators of the sample plots with different degrees of fire and NDVI values, and then NDVI values were extended to satellite images.
Results 1) According to ground survey data, the scorch height of trees in each fire plot increased with the increase of fire severity. The changes in the average values of tree height and diameter at breast height were consistent, with moderate fire being the largest, mild fire being the second, and severe fire being the smallest. The average change trend of crown diameter was different from these two items, with moderate fire being the largest, followed by severe fire, and mild fire being the smallest. Large fires mainly occur in the lower layers of the forest. 2) The relationship between the four indicators of the three plots and their NDVI was above 0.55. The scorch height, crown diameter, and crown diameter indicators of the moderately burned plots were moderately correlated with their NDVI. The correlation between the tree height, DBH, and the height of the lightly burned plot and their NDVI was above 0.8, indicating a strong correlation. The correlation degree of other indicators was between 0.6 and 0.8, indicating a strong correlation. 3) Using the NDVI threshold of drone multispectral images as the standard, the NDVI ranges for severe, moderate, and mild fires were determined to be 0-0.245 6, >0.245 6-0.347 1, and >0.347 1-0.690 0, respectively.
Conclusions Through the method of grey correlation analysis, the correlation between UAV NDVI and four ground survey data is strong, that is, the accuracy of UAV image NDVI is verified with ground survey data validation, indicating that NDVI can be used as the basis for forest fire grading. And using the NDVI threshold of each fire plot in the drone multispectral image as the standard to reclassify satellite remote sensing images of forest fire areas, it can achieve rapid and accurate classification of large-scale fire degree levels. On this basis, the soil type, dominant tree species and elevation data of different fire grades are analyzed, which is of great significance to the cause of fire and the restoration of ecological environment after forest fire.