以云南省景谷县思茅松人工林为研究对象,以研究区2005年TM影像及2006年森林资源二类调查小班空间属性数据库为信息源,在前期建立思茅松单木生物量模型基础上,在ENVI下提取9个植被指数作为备选自变量,建立研究区思茅松人工林随机森林回归遥感估测模型。结果表明随机森林回归遥感估测模型的决定系数(R2)=0.97,均方根误差(RMSE)=4.97;模型的预估精度(P)=87.67%。利用已经训练好的随机森林估测模型,估测研究区思茅松人工林生物量为3 644 612.00t;单位面积生物量为59.90 t/hm2。研究结果可为其它典型森林类型生物量或碳储量估测提供案例分析。
The Simao Pine(Pinus kesiya var.langbianensis)plantations in Jinggu county are taken as the research object,and TM remote sensing image in 2005 and the forest resource inventory database for sub-compartment space attribute in 2006 as the data source.Based on the single tree biomass models,9 index as of vegetation were extracted under ENVI as the alternative variables.The estimation model of remote sensing random forest regression of the Simao pine plantations in the study area was established.The results are as followsR2=0.97,RMSE=4.97 and model estimation accuracy =87.76%.By using the estimation model of RF which has been trained,the predicted total biomass of Simao pine plantations was 3 644 612.00t in study area.The biomass of per-unit area was 59.90 t/hm2.The results provide a typical case analysis for estimation of the biomass and carbon stocks of other types of forests.
[1]IPCC.Good practice guidance for land-use change and forestry[R].Hayama,Japan:IPCC National Greenhouse Gas inventories programme,2003:1-295.
[2]戴小华,余世孝.遥感技术支持下的植被生产力与生物量研究进展[J].生态学杂志,2004(4):95-101.
[3]Fried M A,Davis F W,Michaelsen,et al.Scaling and uncertainty in the relationship between the NDVI and land surface biophysical variables:An analysis using a scene simulation model and data from FIFE[J].Remote Sensing,1995,54:233-246.
[4]袁野,李虎,刘玉峰.基于改进型B-P神经网络的西天山云杉林生物量估算[J].福建师范大学学报:自然科学版,2011,27(2):124-132 .
[5]曾明宇,陈振雄,刘庭威.基于ANN的森林蓄积量遥感估测研究[J].中南林业调查规划,2010,29(3):36-39.
[6]党承林,吴兆录.云南普洱地区思茅松林生物量[J].云南大学学报:自然科学版,1992,14(2): 119-127.
[7]胥辉,张会儒.林木生物量模型研究[M].昆明:云南科技出版社,2002.
[8]Bailey R L,Ware K D.Compatible Basal-Area Growth and Yield Model for Thinned and Unthinned Stand[J].Canadian Journal of Forest Research,1983,13:563-571.
[9]Candy S G.Growth and Yield Models for Pine in Tasmania [J].Forest Science,1989,19(1):112-133.
[10]Sullivan A D,Culture J L.A Simultaneous Growth and Yield Model for Loblolly Pine[J].Forest Science,1972,18(1):76-86.
[11]李江.思茅松幼龄人工林的生物量碳密度及其动态变化[D].北京:北京林业大学,2011:1-86.
[12]朱丽梅,胥辉.思茅松单木生物量模型研究[J].林业科技,2009,34(9):19-23.
[13]Huete S R.A Soil-adjusted Vegetation Index[J].Advances in Earth Science of Environment,1998,13(4):327-333.
[14]田庆久,闵祥军.植被指数研究进展[J].地球科学进展,1988(4):10-16.
[15]方匡南,吴见彬,朱建平,等.随机森林方法研究综述[J].统计与信息论坛,2011,26(3):32-38.
[16]Breiman L.Random Forest[J].Machine Learning,2001,45:5-32.
[17]李欣海.随机森林模型在分类与回归分析中的应用[J].应用昆虫学报,2013,50(4):1190-1197.
[18]Breiman L,Cutler A.Package random Forest[R/OL].(2013-08-12)[2015-12-26].http://cran.r-project.org/web/package/randomForest/randomForest.dbf.
