A Study On The Prediction Of Forest Fires In Liangshan Prefecture Under The Role Of Multiple Factors

Forest fires not only destroy the ecological environment,but also affect human productive activities.Strengthening forest protection therefore plays an important role in ecological calming and security.Located in southwestern Sichuan,Liangshan Prefecture is rich in forest resources and has one of the highest forest coverage rates in China.Because of its humid climate and lush vegetation,it is well suited to plantation activities.However,forest fires are frequent in the region,with large scale forest fires occurring every year,which have a significant impact on forest resources,the ecological environment and socio-economic aspects.Forest fire risk forecasting is therefore important for the protection of forest resources.Understanding the spatial and temporal distribution of forest fires in Liangshan Prefecture and investigating the influence of various factors on forest fires,analysing the occurrence of forest fires and predicting forest fire risk zones can effectively reduce the occurrence of forest fires in Liangshan Prefecture and provide theoretical references for forest fire prediction.This paper uses MODIS data from 2005 to 2020 to analyse the distribution of forest fires in time and space,and to understand the influence of meteorological factors on forest overfire area;analyses various factors affecting forest fires,constructs a forest fire prediction model for Liangshan Prefecture using a logistic-sti regression model and a geographically weighted logistic-sti regression model,and carries out forest fire risk class The following conclusions were drawn:(1)Forest fire points were extracted using MOD14A1,and non-forest fire points were removed by combining land use type data,resulting in 962 forest fire points.The spatial and temporal patterns of forest fires were investigated by extracting forest fire points from MOD14A1 data;the monthly forest fire area was extracted from MOD14A2,and a multiple linear regression model was established to investigate the relationship between forest fire area and meteorological factors.The results showed that as the number of meteorological factors increased,the correlation of the multiple linear regression model also became stronger.(2)The spatial and temporal distribution of forest fires in Liangshan Prefecture was investigated,and the changes in the number of forest fires were analysed in terms of years,seasons and months.The results show that from 2005 to 2020,the number of forest fires in Liangshan Prefecture fluctuated greatly,with 831 forest fires occurring in spring and winter,with January to May being the most frequent period;spatially,the distribution was uneven,with the highest number of forest fires occurring in Yangyuan and Dechang counties.(3)The values of the driving factors on the forest fire sites were extracted to analyse the influence of multiple factors on forest fires,and then the magnitude of the influence of the factors was determined by geographic probes as well as exploring the types of interactions between the factors.The results showed that the highest influence was 0.723 mean annual sunshine and the lowest was only 0.004 in slope direction;the interaction of any two factors was greater than the value of a single factor and greater than the sum of the two factor values.(4)A forest fire prediction model was constructed based on logisticsti regression and geographically weighted logisticsti regression models to predict forest fires in Liangshan.The factors affecting forest fires were subjected to multiple covariance diagnosis and variable screening to filter out factors that could be used for modelling,and finally the accuracy of the two models was compared to select a model more suitable for forest fire prediction in Liangshan.The results showed that the accuracy of the geographically weighted logistic regression model was better than that of the logistic regression model for all samples at 0.76 or above.(5)Based on the prediction results of the forest fire prediction model,a forest fire hazard zone map was established.The results showed that forest fires mainly occurred in the southwestern and southern regions,with fewer fires in the northeastern region.The distribution of forest fires in various districts and counties from 2005 to 2020 verified the reliability of the forest fire risk zoning in Liangshan Prefecture.