Study Of The Prediction For Forest Fire-burned Area In Daxing'an Mountains Region

Daxing'an mountains region is one of forest fire-prone areas in China.Forest fires have caused tremendous damage on forest resources,and the prediction of forest fires has been the focus of study of forest fires.Knowing the relevant rule of development of forest fires is an important dependent of carrying out fighting and protecting fires scientifically and effectively. The study on the prediction of forest burnt area is of great significance in prediction of forest fires of Daxing'an mountains.This paper analyzed the incidence of forest fires in Daxing'an mountains region over the past 20 years and concluded the rules and characteristics of forest fires;by analyzing and processing historical data of forest fires and corresponding meteorological data,this paper established the combination forecasting model of forest burnt area using principal component regression analysis,further tested and verified the model;at the same time,forecasting forest burnt area in use of fuzzy clustering analysis.This paper established the combination forecasting model of forest burnt area using principal component regression analysis,by selecting the average speed,relative humidity, average temperature and rainfall,etc.ten independent variables,and the logarithm value of fire areas as a dependent variable.Next the model was verified,the correct rate is 70.73%,which shows that the model can be used to predict real-time forest burnt area.Using the forecasting method of fuzzy clustering analysis on predicting forest burnt area in Daxing'an mountains region and establishing the fuzzy relation matrix and fuzzy clustering transitive closure of samples of historical data,the history of samples were divided into eight categories.And in accordance with features of in various types of meteorological,determine the attribution of the samples under test category,next to predict forest burnt area.The result indicates the correct rate of predication is 80.58%,the average relative error is 15.72%and the efficiency is better than one of linear models.