基于ICESat-2/ATLAS数据的龙竹地上生物量估测  

Estimation of aboveground biomass of Dendrocalamus giganteus based on ICESat-2/ATLAS data

作　　者：吴再昆 周文武 舒清态[1] 夏翠芬 段丹丹 WU Zaikun;ZHOU Wenwu;SHU Qingtai;XIA Cuifen;DUAN Dandan(Southwest Forestry University,Kunming 650233,Yunnan,China;Guangyuan Forestry Workstation,Guangyuan 628000,Sichuan,China;Information Technology Research Center,Beijing Academy of Agriculture and Forestry Sciences,Beijing 100097,China)

机构地区：[1]西南林业大学,云南昆明650233 [2]广元市林业工作站,四川广元628000 [3]北京市农林科学院信息技术研究中心,北京100097

出　　处：《中南林业科技大学学报》2025年第1期48-57,共10页Journal of Central South University of Forestry & Technology

基　　金：“十四五”国家重点研发计划项目(2023YFD2201205)。

摘　　要：【目的】探究星载激光雷达数据对大型丛生竹类地上生物量(AGB)的反演潜力,以期利用星载激光雷达数据准确预测区域尺度丛生竹的地上生物量。【方法】以云南省新平县的龙竹为研究对象,以ICESat-2/ATLAS星载激光雷达为数据源,通过提取林地内的ATLAS光斑及其参数,使用空间插值技术实现光斑数据在空间上的连续分布。采用随机森林回归(RFR)、K-近邻法(KNN)、梯度提升回归树(GBRT)3种机器学习算法构建区域尺度的AGB预测模型,以实现研究区龙竹AGB的遥感估测。【结果】1)在提取的23个ATLAS参数指标中,经过皮尔逊(Pearson)相关性分析,dem_h、n_ca_photons、h_mean_canopy_abs等11个特征变量与龙竹AGB呈显著或极显著相关。2)采用KNN、RFR、GBRT 3种算法构建AGB估测模型,其决定系数(R^(2))分别为0.43、0.93、0.96,均方根误差(RMSE)分别为20.95、9.35、6.74 t·hm^(-2),总体预测精度(P)分别为68.68%、86.03%、89.92%。3)采用GBRT模型估测研究区龙竹的总AGB为1005714.91 t,平均AGB为69.13 t·hm^(-2)。【结论】基于龙竹林地内ICESat-2/ATLAS光斑的克里金插值结果,采用GBRT算法构建龙竹地上生物量模型可以有效预测研究区内龙竹的地上生物量,该法可为星载激光雷达估测丛生竹地上生物量提供一定的参考。【Objective】Explore the inversion potential of spaceborne lidar data for the aboveground biomass(AGB)of large sympodial bamboo species,with the aim of accurately predicting the aboveground biomass of sympodial bamboos at a regional scale using spaceborne lidar data.【Method】In this study,Dendrocalamus giganteus in Xinping County,Yunnan Province was taken as the research object,and ICESat-2/ATLAS spaceborne lidar was used as the data source.By extracting the ATLAS footprints and its parameters in the forest land,and using the spatial interpolation technology to realize the continuous distribution of the footprints data in space.Three machine learning algorithms,random forest regression(RFR),k-nearest neighbor(KNN)and gradient boosting regression tree(GBRT),were used to construct a regional-scale AGB prediction model to realize the AGB remote sensing estimation of Dendrocalamus giganteus in the study area.【Result】1)Among the 23 ATLAS parameters extracted,Pearson correlation analysis showed that 11 characteristic variables,such as dem_h,n_ca_photons,and h_mean_canopy_abs,were significantly or extremely significantly correlated with the AGB of Dendrocalamus giganteus.2)The coefficient of determination(R^(2))of AGB estimation models established by KNN,RFR and GBRT algorithms were 0.43,0.93 and 0.96,respectively,and the root mean square errors(RMSE)were 20.95,9.35 and 6.74 t·hm^(-2),respectively,and the overall prediction accuracy(P)were 68.68%,86.03%and 89.92%,respectively.3)The total AGB of Dendrocalamus giganteus in the study area was estimated to be 1005714.91 t by GBRT model,and the average AGB was 69.13 t·hm^(-2).【Conclusion】Based on the Kriging interpolation results of ICESat-2/ATLAS footprints within the Dendrocalamus giganteus forest,the construction of a Dendrocalamus giganteus aboveground biomass model using the GBRT algorithm can effectively predict the aboveground biomass of Dendrocalamus giganteus within the study area.This method can provide a certain reference for estimating the abovegro

关 键 词：ICESat-2/ATLAS 地上生物量 竹林 机器学习 空间插值 

分 类 号：S771.8[农业科学—森林工程]