CAJ | 학술논문

目前国内外还没有对不同火险条件下草原火险时空发生概率的研究,而这方面研究对草原火灾管理对策和防火救助应急预案的制定具有重要意义.根据呼伦贝尔草原火灾统计月报表和相关气象、社会经济资料,利用Logistic回归模型建立草原火险预测模型,对草原火险进行了空间上的预测.结果表明,日平均风速、日降水量对草原火险影响较大. 以2005年所有火灾案例对草原火险预测模型进行检验,研究表明,该预测方法具有较高的可靠性,可为火灾管理和减灾决策的制定提供指导.
목전국내외환몰유대불동화험조건하초원화험시공발생개솔적연구,이저방면연구대초원화재관리대책화방화구조응급예안적제정구유중요의의.근거호륜패이초원화재통계월보표화상관기상、사회경제자료,이용Logistic회귀모형건립초원화험예측모형,대초원화험진행료공간상적예측.결과표명,일평균풍속、일강수량대초원화험영향교대. 이2005년소유화재안례대초원화험예측모형진행검험,연구표명,해예측방법구유교고적가고성,가위화재관리화감재결책적제정제공지도.
The present paper wants to introduce our research of the prairie fire hazard prediction based on the logistic regression simulation. As is known, prairie fire has been one of the fatal natural disasters that may influence the development of stockbreeding and the husbandry industry in China. However, there has not been enough research on the probability of prairie fire hazards and ways on how to reduce or avoid their occurrence either at home or abroad. On the other hand, so far as we know, there do exist lots of mathematical simulations that are likely to be available for such disaster prediction studies, such as the gray prediction model, the BP neural network model, the Bayes prediction model and multi-linear regression model, etc. though gray prediction model and BP neural network model are mainly used in time series prediction, whose range of dependent variables in multiple linear regression model is (-∞, +∞). In our research, we have taken the dependent variable as dichotomous variable, believing that such binary logistic regression models are fit for the prairie fire hazard research. In choosing such variables, we found that it is the human activities rather than those of natural fire that lead to such fires in accordance with the historical registration data in Hulunbeier on such fires. Therefore, we have chosen Hulunbeier grassland as a case study and the variable of population as our variable. In doing so, the key factors that affect the prairie fire hazard can be modeled by the logistic regression that employs daily grassland fire disaster statistics, related meteorological and economic data, and daily grassland fire hazard predicted. The results of our method show that the prairie fire hazard is highly affected by the average daily precipitation and average daily wind speed. The probability of grassland fire has been very high though the average monthly relative humidity and the average monthly precipitation is very low. The prediction of such fire hazards in May 1st, 2005 proves that the results of our prediction is highly reliable, and therefore the method can be taken a reference and guidance to managing the prairie fire hazards and mitigating the hazards and reducing the losses caused by it.