| In this study,Landsat OLI satellite remote sensing images were used to construct a random forest classification model using satellite remote sensing data,and the method of identifying the area of fire burnt areas and forest fire severity was studied and analyzed.At the same time,it was related to terrain factors such as altitude,slope,and aspect,as well as grasslands and forests.The vegetation type factors are combined to study the spatial distribution of the fire severity in the overfire area and the image of the vegetation type,and to use the multi-temporal vegetation index change time series to study the early recovery process of the vegetation in the overfire area.Conducted a study on the burning spots formed by the forest fires that occurred on May 2nd,2017 in the Daxinganling Forestry Bureau,Beidahe Forest Farm,Inner Mongolia,by collecting various types of Landsat 8 satellite remote sensing images derived from May to October each year from 2016 to 2019 Fire product data is used as the data source,DEM images of the research area are collected as the terrain factor data source,the threshold range of d NBR is obtained according to the K-means clustering algorithm,and the remote forest index such as NBR,NDVI,NDMI,EVI,SAVI is used to build a random forest classification model To identify the fire severity of the research fire,use the ENVI5.3software to analyze the altitude,slope,aspect and fire severity of the research area one by one,and use the supervision classification function in ENVI to divide the forest and grassland.Each land type is also analyzed one by one with the fire severity.Then,compare the restoration of vegetation index under different fire severity and different vegetation types between different indexes such as NDVI,EVI,NDMI and NBR,and finally select the index that can better reflect the characteristics of vegetation growth and recovery for early vegetation after fire Resume research,the results are as follows:(1)The classification model based on the random forest classifier has a good overall effect.The overall classification accuracy of forest fire severity reached 88.3%,and the KAPPA coefficient value was 0.857.According to the K-means clustering algorithm,four grades of d NBR are obtained:unburned,mild fire,moderate fire,and severe fire.The area of moderately burned area and lightly burnt area is larger,2888hm2 and 2590hm2 respectively,accounting for 40.2% and 36% of the total fired area,respectively.(2)Forest fire severity will be affected by terrain factors such as altitude,slope,aspect and vegetation types.The fire area is the largest in the area of 450 ? 550 m above sea level.The light forest fire severity is mainly distributed in the area of 450 ? 550 m above sea level,while the moderate andsevere fire severity is mainly distributed in the altitude of 550 ? 1000m;°)The fire area is the largest,and there is almost no light fire when the slope is above 15 °;the fire area on the Yangpo is larger than that on the shady slope.In terms of vegetation types,forests are mainly affected by moderate to severe fires,while grasslands are mainly affected by mild fires,with less damage.(3)In the early period after the fire,vegetation restoration has great differences under the effects of forest fire severity of different intensities and different vegetation types.At the same time,the recovery level of grasslands is much higher than that of forests.Most grasslands can quickly recover to 100%level in the year of the fire.However,due to the greater severity of the fire,the forests need to gradually recover to the level of 80% before the second year.The results of this study can provide a reference case for the evaluation of the fire severity of forest fires and the restoration of vegetation in burned areas. |
PDF Full Text Request