火灾科学
火災科學
화재과학
FIRE SAFETY SCIENCE
2011年
2期
99-104
,共6页
武军%朱霁平%贾敬蕊%周建军%张林鹤
武軍%硃霽平%賈敬蕊%週建軍%張林鶴
무군%주제평%가경예%주건군%장림학
森林地表可燃物载量%林分因子%聚类分析
森林地錶可燃物載量%林分因子%聚類分析
삼임지표가연물재량%림분인자%취류분석
Forest surface fuel Load%Forest stand factors%Cluster analysis
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SIF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。
地錶可燃物載量的確定,是實現森林地錶火蔓延預測的前提。已有火蔓延模型中,假定同一林分中可燃物分佈是均勻的,不能準確描述可燃物載量的複雜空間分佈。基于聚類分析方法,建立瞭一種根據易穫取的幾種林分因子來估算地錶可燃物載量的方法。首先,對大興安嶺地區35塊興安落葉鬆林樣地和21塊樟子鬆林樣地的樹齡、鬱閉度、胸徑、樹高等林分因子採用重心法進行繫統聚類分析,分彆將興安落葉鬆林和樟子鬆林分為5類和7類。然後,計算得齣每類可燃物的類中心,併以每類包含所有樣地的可燃物載量的平均值來錶示該類中心的載量,併由此建立可燃物載量與林分因子的對應關繫。對聚類分析與線性迴歸預測模型從平均絕對誤差AAD、標準誤差SEE、模型預測穩定性指標SIF三方麵進行對比,結果錶明聚類分析模型要優于線性迴歸分析模型。
지표가연물재량적학정,시실현삼임지표화만연예측적전제。이유화만연모형중,가정동일림분중가연물분포시균균적,불능준학묘술가연물재량적복잡공간분포。기우취류분석방법,건립료일충근거역획취적궤충림분인자래고산지표가연물재량적방법。수선,대대흥안령지구35괴흥안락협송림양지화21괴장자송림양지적수령、욱폐도、흉경、수고등림분인자채용중심법진행계통취류분석,분별장흥안락협송림화장자송림분위5류화7류。연후,계산득출매류가연물적류중심,병이매류포함소유양지적가연물재량적평균치래표시해류중심적재량,병유차건립가연물재량여림분인자적대응관계。대취류분석여선성회귀예측모형종평균절대오차AAD、표준오차SEE、모형예측은정성지표SIF삼방면진행대비,결과표명취류분석모형요우우선성회귀분석모형。
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.