论文部分内容阅读
林火分布模型是在较大区域上描述林火空间分布的强有力工具,并可以确定影响林火分布的控制因子。本研究基于黑龙江省1996—2006年的历史火烧记录数据,分别采用广义线性模型和最大熵模型分析了地形、人类活动和土地覆被类型等环境控制因子对黑龙江省林火空间分布的影响,并比较了模型预测精度、评价环境变量重要性及预测火点概率分布图等。结果表明:两个模型的预测精度达中等水平,而最大熵模型的预测精度要略高于广义线性模型。总体而言,与人类活动相关的变量是林火分布模型最佳的环境变量,地形变量次之。尽管两个模型在预测精度和环境变量重要性方面都有很大的相似性,但最大熵模型产生的火点概率图空间格局与广义线性模型产生的明显不同。本研究说明,为了更加精确地确定森林火灾发生的热点地区,应该采用不同模型进行比较,或者有选择性地进行组合以产生综合的预测结果,从而为森林防火工作提供更加合理高效的建议。
The forest fire distribution model is a powerful tool to describe the spatial distribution of forest fire over a large area and to determine the control factors that affect the distribution of forest fires. Based on historical records of fire records from 1996 to 2006 in Heilongjiang Province, this study analyzed the influence of environmental control factors such as topography, human activities and land cover types on the spatial distribution of forest fires in Heilongjiang Province by using generalized linear model and maximum entropy model respectively The accuracy of model prediction, the importance of evaluating environmental variables and the probability distribution of predicted fire points are compared. The results show that the prediction accuracy of the two models is at medium level, while that of the maximum entropy model is slightly higher than that of the generalized linear model. In general, the variables related to human activities are the best environmental variables for the forest fire distribution model, followed by the topographic variables. Although the two models have great similarities in prediction accuracy and environmental variables, the spatial pattern of the fire probability plot generated by the maximum entropy model is obviously different from that produced by the generalized linear model. This study shows that in order to determine more precisely the hot spots where forest fires occur, different models should be used for comparison or selective combination to produce comprehensive forecast results, so as to provide more reasonable and efficient suggestions for forest fire prevention work.