Research On Multi-scale Model Of Grassland Fire Occurrence

Grassland is a type of vegetation with a large area of distribution on the earth.As one of the countries with the most abundant grassland resources in the world,China needs to pay more attention to grassland management.Fire is one of the most important disturbance factors in natural ecosystems.Once a fire occurs in grassland,it will cause huge losses to the grassland ecosystem and the life of people in pastoral areas.For example,a fire will affect the pastoral economy and endanger the human’s life in pastoral areas.The northeastern region of China’s Inner Mongolia Autonomous Region covers a relatively large area of grassland.Grassland fires occur common in this region.Nearly one-sixth of the land in the region suffers serious fire damage each year.Many domestic and foreign scholars have started a series of studies in different directions.The occurrence of grassland fires is such a complex issue that is affected by many driving factors.And the occurrence of grassland fires under different scales is affected by different factors.Therefore,this paper analyzes the occurrence of grassland fire from the two scales: large-scale and mesoscale.For the large-scale analysis,the four leagues in the eastern part of the Inner Mongolia Autonomous Region were selected as the research area.The analysis results were as follows:(1)There are strong correlations between the spatial distribution of grassland fire points and the four types of land use which are industrial and mining land,residential areas,water areas,cultivated land in the study area at the three points 2000,2005 and 2010。The closer to these land uses,that is,the denser the human activities,the more grassland fires occur.On a long-term scale,land use had a significant correlation with the occurrence of grassland fires each year.(2)The drying index is used to characterize the spatial distribution of dryness and humidity in the regional climate,so it was selected to analyse the relationship with the grassland fires in large scale.The results showed that: the correlation coefficients between the average dryness index and the number of grassland fires during the three periods of 2000-2002,2003-2007,and 2008-2012 in the study area were-0.541,-0.568,and-0.432,respectively,all of which were significantly negative correlated at the 0.01 level.(3)The degree of land use represents the development and utilization of land resources in the region.Therefore,the degree of land use was selected to analyze the relationship with the grassland fires at the large-scale.The results showed that the degree of land use in different periods had a significant impact on the occurrence of grassland fires.Among them,the R value of the linear fit between the land use degree in 2000 and the fire point in 2000-2002 was 0.828.The R value of the linear fit between the land use degree in 2005 and the fire points in 2003-2007 was 0.846.The R value of the linear fit between the land use degree 2010 and the fire points in 2008-2012 were 0.796,all of which were significantly positively correlated at the level of 0.01.(3)I calculated the degree of land use in 2015 and the average dryness index in 2013-2015,and counted the number of grassland fires occurring in different administrative regions from 2013 to 2015.According to the binary regression model established by the data from the above three time periods,2288 grassland fires occurred in the study area from 2013 to 2015,and between which there were 2,295 grassland fires actual occurred in the study area.The forecast error is Only 0.31%,the prediction error is small,so the regression model based on the degree of land use has strong predictability at the large-scale.For the purpose of mesoscale analysis,Hulunbuir,Inner Mongolia,was selected as the study area.The analysis results were as follows:(1)I calculated the prior probability of grassland fire occurrence in the study area,which was 0.0052.And then I calculated the weights,correlations,and standard deviations of correlations for each evidence layer,which is divided into binary thematic maps based on the Cs value.The bigger Cs is,the closer the relationship between the evidence factor and the occurrence of grassland fire,and the better the prediction ability of the evidence layer.According to the study results,the actual calculation the three selected evidence layers(combustibles,residential grid distance cost,and GDP)were divided into bivariate thematic maps separately,which is,Cs >1 was divided into one category,Cs<1 was divided into the other one category.(2)According to the bivariate thematic map of the evidence factor and the weight value,the minimum value of the posterior probability of the study area is.001893 and the maximum value is 0.087209.The regions with a posteriori probability of more than 0.015019 was classified into high fire risk areas,the regions with a posterior probability between 0.003865 and 0.015019 was classified as middle fire risk areas,and the regions with a posterior probability of less than 0.003865 was classified as low fire risk areas.High fire risk areas were mainly distributed in the east,the middle and north of the center of Hulunbuir.Middle fire risk areas were mainly distributed in most parts of the central and the north part of the east of Hulunbuir.While low fire risk areas were mainly distributed in the west of Hulunbuir,which was mainly located in the Hulun Buir Prairie.Due to the sparsely populated land in this region,there are few human activities and most of them are grassland use type,the risk of grassland fires on the HulunBuir prairie is relatively low.(3)The F value corresponding to the low fire risk area was calculated to be 0.36,the F value corresponding to the middle fire area was 1.49,the F value corresponding to the high fire area was 4.13,and the F value fully complied with the law of gradually increasing,I verified the overall spatial distribution of the fire points from 2013 to 2015,which is consistent with the model prediction results.The results showed that the fire risk distribution map and actual conditions obtained in this chapter is relatively consistent,so the model is in line with the actual situation and can provide support for local fire protection management.