浙江农林大学学报2025，Vol.42Issue(6)：1132-1141,10.DOI:10.11833/j.issn.2095-0756.20250395
基于Landsat影像的曲靖市主要针叶林地上生物量估算
Aboveground biomass estimation of major coniferous forests in Qujing City based on Landsat images
摘要
Abstract
[Objective]Forest biomass is a cruial indicator of the productivity and carbon storage capacity of forest ecosystems.Studing its dynamic changes can facilitate a profound understanding of the functions of forest ecosystems and offer a scientific basis for ecological management and carbon cycle research.[Method]Taking Qujing City of Yunnan Province as the study area,a remote sensing estimation model of aboveground biomass was constructed using data from one type of continuous clearing sample plots in every 5-year period from 1992 to 2017,in combination with DEM and Landsat 5 TM,Landsat 8 OLI images.For the main conifer species(Pinus yunnanensis,P.armandii and mixed conifer forests),five machine learning methods,namely GBRT,RF,XGBoost,PLS and CatBoost,were compared to systematically evaluate their fitting performance and estimation accuracy.The optimal model was selected through the comprehensive evaluation of R2,RMSE,rRMSE and MAE indicators,and used for the time series estimation of aboveground biomass of major coniferous forests in Qujing City from 1992 to 2022.[Result]The estimation performance of different models on the three dominant coniferous species varied:CatBoost model performed best in mixed coniferous forests(R2=0.90,RMSE=5.66),with strong nonlinear fitting ability;while GBRT model performed better in P.yunnanensis forests and P.armandii forests,with stronger estimation stability and explanatory power.[Conclusion]This study indicates that the performance of different machine learning models in forest aboveground biomass estimation varies according to tree species.Model selection should consider the growth characteristics of tree species and their response differences to remote sensing factors,and avoid uniform modelling strategies to improve the estimation accuracy and applicability.[Ch,3 fig.39 ref.].关键词
Landsat影像/森林地上生物量/机器学习/估算模型/曲靖市Key words
Landsat images/forest aboveground biomass/machine learning/estimation model/Qujing City分类
农业科技引用本文复制引用
WANG Tong,ZHANG Chao,ZHOU Hang..基于Landsat影像的曲靖市主要针叶林地上生物量估算[J].浙江农林大学学报,2025,42(6):1132-1141,10.基金项目
国家自然科学基金资助项目(32160405) （32160405）
