Study On Forest Fire Risk Division Based On RS And GIS In Shexian County

As a global natural disaster,a large amount of forest resources destroyed by forest fire every year,which not only brings great property losses and casualties to human beings,but also seriously damages the ecological balance of nature,resulting in many environmental problems.Fire risk division is an important means to prevent fires and ensure the safety of forest resources.Based on Landsat 8 OLI remote sensing images,DEM data and survey data of 67 sample plots,the model of forest fuel loading was build based on machine learning algorithms,and the distribution characteristics of forest fuel loading was analyzed in Shexian County.Supported by RS and GIS technology,the forest fire risk evaluation index system was established by selecting seven indices such as vegetation,terrain and human factors based on forest burning ring theory.Fire risk level of each single pixel was determined using analytic hierarchy process and factor overlapping method,thus,to achieve forest fire division in Shexian county,and characteristic of the spatial distribution of forest fire risk was analyzed.The research results will play an important role in fire risk warning and provide scientific reference for local forestry departments to formulate forest fire management policies.The main findings of the study are as follows:(1)Python was used to analyze the correlation between remote sensing variables and fuel loading of sample plots.Six variables(i.e.,b1,RVI,PCA1,b6＿mean,b6＿vari and elevation)with the highest correlation were finally used as independent variables for model construction.Three non-parametric models based on Gradient boosting decision tree(GBDT),K-nearest neighbour algorithm(KNN)and Random forest(RF)were used to build the inversion model of the forest fuel loading in Shexian County.The coefficient of determination(R~2)and residual mean square error(RMSE)of each model was comparative analysised.Finally,Random forest algorithm with the highest model prediction accuracy(R~2=0.65,RMSE=50.59 t/hm2)was selected as the optimal inversion model.(2)Based on Random forest algorithm,forest fuel loading in Shexian county was inverted to determine the spatial distribution characteristics of 2021.The total forest fuel loading in Shexian County was 33.11 Mt,and the average fuel loading was 195.41 t/hm~2.The unit area fuel loading mainly concentrated in the range of 140-240 t/hm~2,covering an area of 132256.9hm~2 and accounting for 78%of the total fire risk area in the study area.The overall spatial distribution pattern of unit fuel loading was high in the east and low in the west.(3)Seven indicators(i.e.,fuel loading,fuel type,slope,aspect,altitude,road buffer zone and residential buffer zone)were chosen to construct the fire risk evaluation index system,and the weight of each indicator was determined by analytic hierarchy process.The weights of the seven indicators in sequence are 0.2214,0.2214,0.0802,0.0640,0.0257,0.1291 and0.2582,and the grade scores of each indicator was calculated.The forest fire risk index(FFR)of each pixel in the study area was calculated using the factor overlapping method and classified into four categories:low fire risk zone(0<FFR≤3),medium fire risk zone(3<FFR≤6),high fire risk zone(6<FFR≤8)and very high fire risk zone(FFR≥8).(4)A comprehensive forest fire risk division map of 2021 was drawn based on ArcGIS10.7.According to the classification of forest fire risk levels,the medium and high fire risk areas accounted for 97.33%of the total fire risk area,while the extremely high fire risk area accounted for 0.22%,covering an area of 382.11 hm~2.The high fire risk areas were concentrated in the northeast,while the fire risk level of the southern part was relatively low.An analysis of the differences between the fire risk levels in terms of different fire risk factors shows that the extremely high fire risk areas and high fire risk areas are mainly located in areas where the type of fuel type is flammable.