Study On Distribution Pattern And Prediction Model Of Forest Fire In Ganzi Prefecture Of Sichuan Province

In recent years,with the global warming,many parts of the world have experienced extreme hot and dry weather.Abnormal climate change leads to frequent forest fires,which aggravates the deterioration of the earth’s ecological environment.A large number of facts show that climate change in recent years is closely related to human production and life.While modern industrialization brings good news to human beings,it also emits a lot of carbon dioxide and fluorine,which makes the greenhouse effect increasingly serious.How to effectively carry out the research of forest fire analysis and prediction and reduce the loss caused by forest fire to the greatest extent has become the focus of scholars at home and abroad.Ganzi Prefecture,located in the west of Sichuan Province,is rich in forest resources and diverse vegetation coverage.However,it is also a high incidence area of forest fires due to its unique geographical and climatic conditions.Based on the MODIS Fire Product data,terrain data,forest vegetation data and meteorological data of Ganzi Prefecture from2005 to 2019,this study deeply discussed the temporal and spatial distribution law of forest fire in Ganzi Prefecture by using mathematical statistics and spatial autocorrelation analysis methods on the basis of Arc GIS,SPSS,R language and other software,and superposed the forest fire data with terrain,vegetation,human and meteorological data To analyze the influence of different driving factors on forest fire.At the same time,binomial logistic regression,Poisson regression,negative binomial regression model and test theory are used to predict forest fire in Ganzi Prefecture from the annual scale and monthly scale,which provides scientific basis and theoretical reference for forest fire prediction and prediction,and has important research significance.The conclusions are as follows:(1)The results show that the interannual variation of the number of forest fires in Ganzi Prefecture in recent 15 years fluctuates on the whole.There are more forest fires in 2012 and 2013,and then it shows a slow downward trend with the increase of years.Spring and winter are the high incidence periods of forest fires in Ganzi Prefecture,and the number of fires from January to April is much greater than that in other months.From the perspective of geographical distribution,forest fires occur in all counties and cities of Ganzi Prefecture,and have a certain aggregation.Jiulong County and Kangding County are the areas with more occurrence times.Most of the forest fires in Ganzi Prefecture are caused by human factors,and roads and rivers can affect the distribution of forest fires to a certain extent.The forest area with an altitude of 4010m-4560 m is a high incidence area of forest fires,and the distribution of forest fires on each slope direction is relatively uniform,but the main occurrence is concentrated in the gentle slope section of 0-18 degrees.The main types of combustible vegetation of forest fire in Ganzi Prefecture are coniferous forest and shrub.Different climate environment also affects the risk of forest fire to a certain extent.(2)Based on the ratio of fire point and random non fire point of 1:1.5,binomial logistic regression model was used to model and analyze the forest fire occurrence results in Ganzi Prefecture with 0 and 1 in SPSS and 13 factors such as meteorological factors,terrain factors,vegetation factors and human factors,Elevation and population density have significant correlation with forest fire occurrence in Ganzi Prefecture,and meteorological factors are dominant,The accuracy of the model is 72.3%,and the prediction effect is good.According to the prediction results,the forest fire danger grade zoning of Ganzi Prefecture was realized.The results show that the northern part of Ganzi Prefecture is a low risk area,and the middle and high risk areas are mainly concentrated in the South and Southeast.(3)Considering that the occurrence of forest fire has the characteristics of aggregation,too many zero values of forest fire times in some counties and cities,and the overall discrete degree is large.Poisson regression,negative binomial regression,zero expansion Poisson regression and zero expansion negative binomial regression models were used to analyze the relationship between forest fire occurrence times and meteorological factors in Ganzi Prefecture on a monthly scale in R language,and Vuong test and AIC criterion were used to test the accuracy of the model.The results show that the zero expansion negative binomial model is better than the other three models,and it is the best prediction model of forest fire frequency in Ganzi Prefecture.