Study On Occurring Space And Forecasting Model Of Forest Fires In Zhejiang Province

Based on forest fire statistical data of Zhejiang province during1991-2012, forest meteorological elements and basic geographic data, and with SPSS, ArcGIS, R_project and Maxent software for technical support, the temporal and spatial distribution pattern of forest fires in Zhejiang Province were studied by using of periodogram analysis, spatial autocorrelation analysis and mathematical statistics methods. And taking Longquan and Jinyun County as an example, some issues, such as forecasting model of forest fires and the relationship between forest fire and weather and some spatial factors are discussed in detail by using of model simulation and testing theories of international frontier. The results were great significance to the prediction and management of forest fire and can supply a scientific thesis foundation for forest fire’s forecast.The results showed that there were mainly small and medium-scale forest fires in Zhejiang Province, the frequency of which is proportional to the extent of damage, and these fires were mostly caused by person; Due to different fuel combustion characteristic of vegetation types, the severity of forest fire disasters was different. According to forest vegetation types, the number of forest fires from more to less was, in turn, coniferous forest, mixed needle leaf and bamboo forests and thickets; Annual forest fires times varied significant, which did not show obvious trend and periodicity, but average area of forest fire was a clear upward trend; Seasonal changes of forest fires was obvious,the peak period of forest fires in Zhejiang Province was from January to April, which accounted for80.08%of the total number of forest fires; On the spatial distribution, forest fires presented an overall increasing trend from north to south, the southern mountain of Zhejiang Province were relatively serious forest fire hazard areas, east to yongjia county, lishui city on the west, north to jinyun county, south county area was particularly serious.The traditional ordinary least squares (OLS) regression method, Poisson and negative binomial (NB) models were used to describe the relationship between forest fire occurrence and weather factors in Zhejiang province for1991-2012using R-Project statistic software. Negative binomial model was the best model to reflect the relationships between forest fire occurrence and weather factors, which showed the monthly mean wind speed and the monthly mean temperature were two important factors affecting forest fires occurred in Zhejiang Province.Taking Longquan City as an example, the forecasting model of forest fires of the meteorological factors as variables was discussed. We choosed Poisson, Zero-inflated Poisson(ZIP), negative binomial(NB) and Zero-inflated negative binomial(ZINB) models to model and analysised the relationship between forest fire and weather faetors, the results showed, zero-inflated negative binomial (ZINB) model was superior to the other3models. Therefore, the best prediction model by forest fires in Longquan city was ZINB model, its expression is: E(Y)=exp(11.3516-0.1552DMRH-0.1115DMT)and logit(p)=0.6899+1.4002DMWS-0.2333DMTBased on forest fire data and the factors controlling the forest fire occurrence from1991to2012in Jinyun County, we used Maxent model to quantitatively, intuitively predict the potential occurrence of forest fires in Jinyun County. The results showed that the maximum entropy model exhibited moderate level, which was the effective model to describe the spatial distribution pattern of of forest fire. The key environmental control factors of forest fires in Jinyun County was the density distribution pattern of fire point, followed by non combustible, forest and slope, aspect has minimal impact.