基于遥感数据和优化Blending算法的人工林地上生物量估算研究  

Estimation of Above-Ground Biomass of Planted Forest Based on Remote Sensing Data and Optimized Blending Algorithm

作　　者：辛守英 王晓红 焦琳琳 XIN Shouying;WANG Xiaohong;JIAO Linlin(College of Mining Engineering,North China University of Science and Technology,Tangshan 063210,Hebei,China;Hebei Geological Engineering Exploration Institute,Baoding 071051,Hebei,China;Hebei Industrial Technology Institute of Mine Ecological Remediation,Tangshan 063210,Hebei,China)

机构地区：[1]华北理工大学矿业工程学院,河北唐山063210 [2]河北省地质工程勘查院,河北保定071051 [3]河北省矿区生态修复产业技术研究院,河北唐山063210

出　　处：《西北林学院学报》2025年第2期207-219,共13页Journal of Northwest Forestry University

基　　金：中央引导地方科技发展资金项目(246Z5901G);唐山市科技局应用基础研究计划项目(20130202b);华北理工大学博士专项(BS201818);河北省自然科学基金青年基金项目(D2019209322,D2019209317)。

摘　　要：森林生物量是了解森林碳循环以及评估森林碳储量和健康状况的重要指标,准确估算塞罕坝林场主要优势树种人工林生物量对于评估区域森林碳循环及实现森林可持续发展具有重要指导意义。本研究以塞罕坝林场为研究区域,利用Sentinel-1雷达遥感数据、Sentinel-2光学遥感数据和DEM数据,分别提取雷达数据的雷达和纹理特征,光学数据的光谱、纹理、植被指数和缨帽变换特征,以及DEM数据的高程、坡度、坡向等地形特征,结合塞罕坝林场的森林二类调查数据,通过HPO-PCA-Blending集成估算模型实现塞罕坝林场主要优势树种人工林生物量的估算。结果表明:1)对Sentinel-1、Sentinel-2和DEM多源数据进行参数特征提取,得到234种参数特征;2)华北落叶松、白桦、樟子松、蒙古栎、云杉格网中心的森林生物量均近似正态分布规律;3)相比ANN、MLR、XGB、LSTM和KNN估算模型,HPO-PCA-Blending集成估算模型极大地提高了主要优势树种人工林生物量的估算精度,R^(2)均提高0.3以上。多源数据参数特征与HPO-PCA-Blending集成估算模型联合能够实现对塞罕坝林场人工林生物量的准确估算,可为塞罕坝林场乃至其他区域森林生物量的估算和森林资源的管理提供重要的方法和理论依据。Forest biomass is an important index for understanding forest carbon cycle and evaluating forest carbon storage and health status.Accurately estimating the biomass of the dominant tree species in the Saihanba Forest Farm is of great guiding significance for assessing regional forest carbon cycle and realizing sustainable forest development.This study took the Saihanba Forest Farm as the study area and used Sentinel-1 radar remote sensing data,Sentinel-2 optical remote sensing data,and DEM data to extract radar and texture features of radar data,spectrum,texture,vegetation index,tasseled cap features of optical data,and elevation,slope,slope aspect topographic features of DEM data,respectively.Combined with the forest management inventory data of the farm,the planted forest biomass of the main dominant tree species in the forest farm was estimated by using the HPO-PCA-Blending integrated estimation model.The results showed that 1)234 parameter features were extracted from Sentinel-1,Sentinel-2,and DEM multi-source data.2)The forest biomass of Larix gmelinii var.principis-rupprechtii,Betula platyphylla,Pinus sylvestris var.mongholica,Quercus mongolica,and Picea asperata in the grid center was in approximately normal distribution.3)Compared with ANN,MLR,XGB,LSTM,and KNN estimation models,the HPOPCA-Blending integrated estimation model greatly improved the estimation accuracy of the planted forest biomass of the main dominant tree species,and the R^(2) increased by more than 0.3.The combination of the multi-source data parameter features and the HPO-PCA-Blending integrated estimation model can achieve accurate estimation of the planted forest biomass in the Saihanba Forest Farm as well as some other regions,which provides an important method and theoretical basis for forest biomass estimation and forest resource management.

关 键 词：塞罕坝林场 人工林 森林生物量 机器学习 多源数据 

分 类 号：S718.556[农业科学—林学]