Spatio-temporal Pattern And Risk Assessment Of Wildfire In Cross-border Area Of China,Mongolia And Russia

Based on MCD64A1 burned area products and MOD14A1\MYD14A1 thermal anomaly products,and combined with meteorological data,atmospheric circulation data,vegetation index data and other auxiliary data,this study uses spatial statistical analysis methods to analyze the spatio-temporal dynamic pattern of wildfire area and the number of fire spots in the cross-border area of China,Mongolia and Russia.By using Pearson correlation analysis and structural equation model analysis,this paper discusses the relationship between the burned area of wildfire and climatic factors,vegetation status and atmospheric circulation in cross-border area of China,Mongolia and Russia,the Russian far East,Mongolia and Inner Mongolia,respectively.Fourteen best fire risk factors are obtained by using random forest algorithm to select fire risk factors,and the importance order of fire risk factors is obtained by using the importance score of fire risk factors.Using the 14 fire risk factors selected finally,the random forest algorithm,PSO-ANFIS and GA-ANFIS are used to establish the wildfire risk assessment model in the cross-border area of China,Mongolia and Russia,which realizes the assessment and classification of the wildfire risk in the cross-border area of China,Mongolia and Russia.The results show that:(1)The wildfires in the cross-border areas between China,Mongolia and Russia mainly occurred in Orientale Province,Sukhbator Province and Kent Province in the northeast of Mongolia,Buryat Republic of Russia and Post-Baikal Border region,and relatively few wildfires occurred in Inner Mongolia of China.Wildfires occurred more than 2 times in the areas mainly concentrated in Orientale Province,Sukhbator and Post-Baikal Border region.From 2001 to 2017,the total burned area in the cross-border area between China,Mongolia and Russia was 4.6767 million km~2,and the total fire number was 62.56 million,and the number of fire points was basically consistent with the interannual change of the fire area,showing a slight downward trend.2003,2008,2011,2014 and 2015 were the most active years for wildfires.The overfire area and the number of fire spots in spring are much higher than those in other seasons.March,April,May,June and October are the months with the highest concentration of wildfires in the study area.(2)The total area and number of wildfires in the far East of Russia in recent 17 years are much higher than those in central and eastern Mongolia and Inner Mongolia Autonomous region of China.The area of wildfire in central and eastern Mongolia is much higher than that in Inner Mongolia Autonomous region,but the number of fire spots is slightly lower than that in Inner Mongolia.There are more fire spots in Inner Mongolia every year,but the overfire area is much lower,mainly because China’s Inner Mongolia region has perfect fire prevention institutions and policies,which can be put out at the first time when wildfires occur,thus reducing the spread of large-scale wildfires.Grassland fires and forest fires are the main types of wildfires in cross-border areas between China,Mongolia and Russia.From 2001 to 2017,the total overfire area of grassland fire accounted for more than half of the total overfire area of all landuse types,followed by forest fire,accounting for 41%.But the number of forest fires is nearly double that of grassland fires.(3)Using the Pearson correlation method,it can be concluded that there are some differences in the related factors of the wildfire overfire area in the whole Sino-Mongolian-Russian cross-border region,the Russian far East,the central and eastern Mongolia and Inner Mongolia Autonomous region in different months.In the Russian far East in the study area,precipitation is the main factor affecting the wildfire overfire area,and there is a significant negative correlation between the overfire area and precipitation.Based on the analysis of structural equation model,it is concluded that Arctic Oscillation,monthly mean temperature,normalized vegetation index,monthly mean temperature,monthly cumulative precipitation,Palmer drought index and North Atlantic Oscillation all significantly affect the overfire area of wildfire in the whole region.the overfire area in the Russian far East is mainly affected by monthly average temperature and monthly cumulative precipitation.The overfire area in central and eastern Mongolia is significantly affected by Arctic Oscillation,vegetation index,monthly cumulative precipitation,monthly mean temperature and North Atlantic Oscillation Index.In Inner Mongolia,China,the overfire area is only significantly affected by the Arctic Oscillation index North Atlantic Oscillation.(4)According to the order of the importance of the fire risk factors in the random forest model,the daily temperature difference has the greatest influence on the occurrence of wildfire,followed by the frequency of rain day,and the influence of slope,aspect and DMSP data is the least,and finally the 14 best fire risk factors are obtained.The fitting effect of random forest on training data set is the best,followed by PSO-ANFIS,but the generalization ability of PSO-ANFIS is stronger than that of random forest algorithm.GA-ANFIS performs poorly in both training data set and test data set,especially its generalization ability is much weaker than random forest model and PSO-ANFIS model.(5)The areas with high wildfire risk levels are mainly concentrated in the Oriental Province of Mongolia,the Post-Baikal Border region of Russia and the Republic of Buryat.There are a small number of scattered high-risk areas in the northeast of Inner Mongolia.The fire danger level in southwest Mongolia and the southwest of China’s Inner Mongolia Autonomous region is lower,and the fire insurance level in China in the border area of China,Mongolia and Russia is lower than that in the other two countries.