一种基于聚类分析的由林分因子估算地表可燃物载量的方法  

Forest surface fuel load estimation by stand factors based on cluster analysis method

作　　者：武军[1] 朱霁平[1] 贾敬蕊[1] 周建军[1] 张林鹤[1] 

机构地区：[1]中国科学技术大学火灾科学国家重点实验室,安徽合肥230026

出　　处：《火灾科学》2011年第2期99-104,共6页Fire Safety Science

基　　金：国家自然科学基金资助项目(30972380);国家科技部林业公益性行业科研专项重点项目200704027

摘　　要：地表可燃物载量的确定,是实现森林地表火蔓延预测的前提。已有火蔓延模型中,假定同一林分中可燃物分布是均匀的,不能准确描述可燃物载量的复杂空间分布。基于聚类分析方法,建立了一种根据易获取的几种林分因子来估算地表可燃物载量的方法。首先,对大兴安岭地区35块兴安落叶松林样地和21块樟子松林样地的树龄、郁闭度、胸径、树高等林分因子采用重心法进行系统聚类分析,分别将兴安落叶松林和樟子松林分为5类和7类。然后,计算得出每类可燃物的类中心,并以每类包含所有样地的可燃物载量的平均值来表示该类中心的载量,并由此建立可燃物载量与林分因子的对应关系。对聚类分析与线性回归预测模型从平均绝对误差AAD、标准误差SEE、模型预测稳定性指标SIE三方面进行对比,结果表明聚类分析模型要优于线性回归分析模型。Forest surface fuel load is one of the most important factors for forest fire spread prediction. In existing fire spread models, fuel load is usually assumed to be uniform in a region with the same forest type, although its spatial distribution is complex even in one kind of forest with different forest stand factors. In this article, a method for surface fuel load estimation by the forest stand factors is established based on the cluster analysis method. The stand factors used in cluster analysis include forest age, canopy density, average tree height and diameter at breast height which are all easy to be obtained. 35 Larix gmeli- nii plots and 21 Pinus sylvestris plots of Da Hinggan Mountains forests were used to carry on centroid cluster analysis and in re- sult they were divided into 5 and 7 clusters, respectively. The center of each cluster was calculated and its corresponding fuel load was represented by the average load of the plots grouped in this cluster. Finally, three statistical indicators, including the mean absolute error of the estimate, MAE, the standard error of the estimate, SEE, and the stable indicators of the estimate, SIE, were used to contrast the fitting errors of the cluster analysis method and the multiple linear regressions method. The re- sults show that the cluster analysis method is better than the latter one.

关 键 词：森林地表可燃物载量 林分因子 聚类分析 

分 类 号：S762[农业科学—森林保护学]