学术
采用偏最小二乘回归方法估测森林郁闭度
杜晓明[1,2] 蔡体久[1] 琚存勇[1]
[1]东北林业大学林学院,哈尔滨150040 [2]国家林业局驻大兴安岭森林资源监督办公室,黑龙江加格达奇165000
摘 要:
以遥感数据与森林资源一类清查数据为基础，探讨了用Bootstrap方法筛选最优郁闭度估测变量，用偏最小二乘回归方法建立模型估测森林郁闭度的可行性．结果表明：无论是用所有变量构造的模型还是用所选最优变量构造的模型，郁闭度估测的相对偏差在5％左右．筛选出的最优变量与其他地区的研究结论差异很大，说明除了筛选方法，地带性植被和地形地貌的不同也会造成估测郁闭度最优变量的差异．[著者文摘]
关 键 词:
文章出处:
《应用生态学报》-2008年19卷2期 -273-277页
分 类 号:
文献标识码:
A
文章编号:
1001-9332（2008）02-0273-05
Estimation of forest canopy closure by using partial least square regression
DU Xiao-ming, CAI Ti-jiu, JU Cun-yong （ 1 College of Forestry, Northeast Forestry University, Harbin 150040, China; 2Supervision Office of Forest Resources Management in Great Xing＇ an Mountains, State Forestry Administration, Jagedaqi 165000, Heilongiiang, China ）
Abstract:
Based on remote sensing and forest resources inventory data, this paper approached the feasibility of using Bootstrap approach to select optimal variables and using partial least square （PLS） regression to build a model for estimating forest canopy closure. The results showed that whether using a model built with all variables or a model with the optimal variables selected by Bootstrap approach, the relative deviation in estimating forest canopy closure was about 5%. The optimal variables selected in this paper differed greatly with those in the studies for other areas, suggesting that besides selection method, zonal vegetation and terrain could also induce the differences of selected optimal variables for the estimation of forest canopy closure.[著者文摘]
Key words:
forest canopy closure ; remote sensing; terrain indices ; partial least square regression ; Bootstrap approach.
基金资助:
国家自然科学基金资助项目（30771713,30471424）.
更多>>免费文章
cqvip.com