Spatial And Temporal Evolution Analysis And Prediction Of Vegetation Fire Point In Inner Mongolia Autonomous Region

Vegetation is an important part of terrestrial ecosystems,and it is also an area where fires occur frequently.Vegetation fires are unique earth system disturbances affecting the coupled biosphere,with enormous impacts on ecology and human society.As an important ecological barrier in northern my country,Inner Mongolia is very beneficial to maintain the balance of the northern ecosystem.The ecological environment here is fragile,and it is one of the most serious areas for vegetation fires in our country.To study the temporal and spatial evolution law and influence mechanism of vegetation fire point in Inner Mongolia,to provide data support for fire point prediction,has important practical significance for avoiding fire and protecting vegetation ecosystem.This thesis uses VIIRS(Visible infrared Imaging Radiometer)active fire data,supplemented by natural factors and human activity factors,and uses methods such as gravity center shift,mean clustering,and spatial correlation to analyze the spatiotemporal dynamic evolution of fire points.Regression coefficients and geographic detector methods were used to reveal the correlation and action mechanism between climate,topography,vegetation and other factors and the fire point.Logistic regression and random forest algorithm were used to establish a fire point prediction model,and the results of the model were evaluated.conclusion as below:(1)From 2012 to 2021,the fire in the study area was the most serious in 2014 and the smallest in 2012.Spring and autumn are the seasons with high incidence of fire,especially March-April and October.The fire points mainly occurred in the eastern and central-southern parts of the study area,and the land types were mainly concentrated in grassland and farmland.The inter-annual shift of the center of gravity is not obvious,located in the Xilin Gol League,while the inter-monthly shift of the center of gravity is concentrated in the central cities,showing two closed circles.The fire points in the study area showed three clustering types: "spring-dominated","autumn-dominated" and "no obvious seasonal characteristics" due to different regions.The overall level of spatial autocorrelation in the fire point is low,and the aggregation is reduced after 2019.(2)The vegetation fire point in Inner Mongolia is jointly affected by multiple factors.The detection result of the factor intensity has the strongest explanatory power for the occurrence of the fire point,and the non-meteorological factors such as elevation,slope and slope aspect have less influence on the fire point.The interaction of driving factors is mainly manifested as nonlinear enhancement.The results of the logistic regression coefficient analysis showed that the surface air temperature,air pressure and sunshine hours had a great influence on the fire point.(3)Logistic regression and random forest were used to build fire point prediction models respectively.The AUC value of the logistic regression model was greater than0.89,and the prediction accuracy was 86.8%-87.5%.The AUC value of the random forest model is greater than 0.99,and the prediction accuracy is 96.0%-97.4%.In terms of prediction accuracy,random forest algorithm is more suitable for vegetation fire point prediction in Inner Mongolia.