Study On Fire Occurrence In Eastern Inner Mongolia Based On GIS

As one of the important disturbance factors of ecosystem,fire has an important impact on forest and grassland ecosystems.Forest and grassland resources in eastern Inner Mongolia are rich and they are important ecological barriers in China.They play extremely important roles in maintaining the balance of the northern ecosystem.This area is also one of the most serious areas of forest and grassland fires in China.Studying the temporal and spatial differentiation of fire in eastern Inner Mongolia and its impact mechanism can provide data support for the long-term prediction of future fire occurrence,and provide scientific basis for local scientific and reasonable fire prevention work and resource management.In this paper,a burned area extraction algorithm based on MODIS time series data is proposed in eastern Inner Mongolia from 2001 to 2015.Based on this result,a variety of GIS methods were used to analyze the annual,seasonal and monthly variation of the number of fires and the area of fire in eastern Inner Mongolia during the past 15 years,and to explore the spatial distribution characteristics such as the correlation and heterogeneity of fire occurrence.Single factor analysis and geographic detector methods were used to reveal the main factors affecting fire,such as climate,topography,combustibles and human disturbance.Finally,a geographically weighted logistic regression model is constructed to analyze the spatial action of the main influencing factors of fire occurrence and predict it.The main conclusions of this study are as follows:(1)Based on MODIS surface reflectance data and MODIS vegetation cover data,it is possible to extract burned area at regional scale.In this paper,the forest and grassland fires in eastern Inner Mongolia are taken as an example to verify the accuracy of the fire extraction algorithm based on MODIS time series data.The results show that the overall accuracy of the algorithm is about 80%,and the algorithm has high reliability.(2)Between 2001 and 2015,the eastern part of Inner Mongolia was seriously affected by the fire,with an average annual burning area of 230,000 hectares.The fluctuations between years are large,and the frequency of fire rises first and then rises and then rises.The most serious year was in 2003,with over 500,000 hectares of fire,the year with the least impact from the fire was in 2007,and the area of fire was only 20,000 hectares.The fires are mainly distributed in the agriculture and forestry interlaced zone on the east side of Daxing'anling and Hulunbuir grassland.(3)Geodetector can identify the nonlinear relationship between the number of fire occurrences and the impact factor,and quantify the contribution of each impact factor.The influencing factors that have a significant contribution to the occurrence of forest fires are annual average wind speed,annual average temperature,slope,NDVI and altitude;the factors that contribute significantly to the occurrence of grassland fires are annual average temperature,distance from cultivated land,altitude,NDVI and vegetation coverage.Annual average temperature,NDVI and altitude are important factors affecting the occurrence of fire in eastern Inner Mongolia.(4)The results of geographically weighted logistic regression model show that the annual average temperature and NDVI in the eastern Inner Mongolia have significant spatial changes in the impact on forest fires.The spatial variation of altitude,slope and wind speed is not significant.Forest fires are likely to occur at higher annual average temperatures,lower altitudes,and gentle slope area.The influence of NDVI and distance from cultivated land on grassland fire showed significant spatial changes,and the spatial variation of annual average temperature,vegetation coverage and altitude was not significant.Grassland fires are prone to high NDVI and high vegetation cover areas with low altitude and close distance from cultivated land.Using the prediction ability of the fire data test model from 2016 to 2018,the predicted AUC values are 0.877 and 0.808,respectively,and the model fitting effect is better.