广西师范大学学报（自然科学版） ›› 2026, Vol. 44 ›› Issue (4): 234-245.doi: 10.16088/j.issn.1001-6600.2025042601

• 农业科学 • 上一篇    下一篇

基于特征选择和集成机器学习算法的森林地上生物量估测

罗蜜1,2, 邓子蒨1, 赵学松2,3, 吕华权2,3, 莫晓峰1, 吴宇1, 周伟4*   

  1. 1.南宁师范大学 地理科学与规划学院, 广西 南宁 530001;
    2.自然资源部中国—东盟卫星遥感应用重点实验室,广西 南宁 530201;
    3.广西壮族自治区自然资源遥感院, 广西 南宁 530023;
    4.南宁师范大学 环境与生命科学学院, 广西 南宁 530001
  • 收稿日期:2025-04-26 修回日期:2025-05-19 出版日期:2026-07-05 发布日期:2026-07-01
  • 通讯作者: 周伟(1991—),男,山东淄博人,南宁师范大学教师,博士。E-mail: bsstzw@163.com
  • 基金资助:
    自然资源部中国—东盟卫星遥感应用重点实验室开放基金(KLCARS-2024-G06);广西科技基地和人才专项(桂科AD23026073);大学生创新创业训练计划(202410603025)

Forest aboveground biomass estimation based on feature selection and ensemblemachine learning algorithms

Luo Mi1,2, Deng Ziqian1, Zhao Xuesong2,3, Lü Huaquan2,3, Mo Xiaofeng1, Wu Yu1, ZhouWei4*   

  1. 1. School of Geography and Planning, Nanning Normal University, Nanning Guangxi 530001, China;
    2. KeyLaboratory of China-ASEAN Satellite Remote Sensing Applications, Ministry of Natural Resources of the People’s Republic of China, Nanning Guangxi 530201, China;
    3. Guangxi Zhuang Autonomous Region Remote Sensing Institute of Natural Resources, Nanning Guangxi 530023, China;
    4. School of Environment and Life Sciences, Nanning Normal University, Nanning Guangxi 530001, China
  • Received:2025-04-26 Revised:2025-05-19 Online:2026-07-05 Published:2026-07-01

PDF (PC)

1

摘要： 森林地上生物量(AGB)遥感估测中,特征变量的日益增多使得有效特征筛选成为提升精度的关键问题。本文以南宁市为研究区,基于哨兵2号遥感数据提取各波段光谱信息、纹理特征及地形因子等多元变量,分别采用逐步回归法、双变量相关法、随机森林法等3种特征选择方法筛选建模变量,并基于CatBoost和RF集成机器学习算法建立AGB估测模型。通过五折交叉验证评估模型性能,筛选最优模型组合实现研究区AGB空间制图。结果表明,在3种特征选择方法中,双变量相关法在马尾松林、桉树林、其他阔叶林中表现均为最佳,而在杉木林中,随机森林法表现最佳。其中,杉木林中随机森林特征筛选法结合RF算法为最优模型组合(R2=0.58,RMSE=8.53 Mg·hm-2);马尾松林中双变量相关法＋RF算法为最优组合(R2=0.51, RMSE=11.10 Mg·hm-2);桉树林中双变量相关法＋RF算法为最优模型组合(R2=0.56,RMSE=14.91 Mg·hm-2);其他阔叶林中双变量相关法＋RF算法为最优模型组合(R2=0.35,RMSE=40.55 Mg·hm-2)。研究表明,特征选择方法对模型的预测性能有显著影响,将特征选择方法和集成机器学习算法相结合,有利于提高AGB的估测精度。

关键词: 特征选择, 随机森林, CatBoost, 森林地上生物量, 机器学习

Abstract: The increasing dimensionality of feature variables in remote sensing-based estimation of forest aboveground biomass (AGB) necessitates effective feature selection to enhance model accuracy. This study focuses on Nanning City as the research area, utilizing Sentinel-2 data as the remote sensing source. Spectral information from various bands, texture features, and additional factors such as elevation, slope, and aspect were extracted. Three feature selection methods including stepwise regression, bivariate correlation, and random forest were employed to identify modeling variables. Biomass estimation models were established based on CatBoost and random forest (RF) machine learning algorithms. Five-fold cross-validation was applied to evaluate model performance, and the best model was used to complete biomass mapping. The results indicated that among the three feature selection methods, the bivariate correlation method performed the best across three tree types: pine, eucalyptus, and broadleaf species. For Chinese fir forests, the random forest method showed superior performance. Specifically: For Chinese fir forests, the combination of the random forest feature selection method and the RF algorithm was optimal (R2=0.58, RMSE=8.53 Mg·hm-2). For Masson pine forests, the bivariate correlation method combined with the RF algorithm was the best choice (R2=0.51, RMSE=11.10 Mg·hm-2). For eucalyptus forests, the bivariate correlation method combined with the RF algorithm yielded the best results (R2=0.56, RMSE=14.91 Mg·hm-2). For broadleaf forests, the bivariate correlation method combined with the RF algorithm also proved optimal (R2=0.35, RMSE=40.55 Mg·hm-2). Feature selection methods have a significant impact on the predictive performance of models. Combining feature selection methods with ensemble machine learning algorithms is conducive to improving the estimation accuracy of AGB.

Key words: feature selection, random forest, CatBoost, forest aboveground biomass, machine learning

中图分类号:  S718.5; TP181; TP79

[1] Fang J Y, Wang Z M. Forest biomass estimation at regional and global levels, with special reference toChina’s forest biomass[J]. Ecological Research, 2001, 16(3): 587-592. DOI: 10.1046/j.1440-1703.2001.00419.x.
[2] Jacon A D, Galvão L S, Dalagnol R, et al. Aboveground biomass estimates over Brazilian savannas using hyperspectral metrics and machine learning models: experiences with Hyperion/EO-1[J]. GIScience & Remote Sensing, 2021, 58(7): 1112-1129. DOI: 10.1080/15481603.2021.1969630.
[3] Mohite J, Sawant S, Pandit A, et al. Forest aboveground biomass estimation by GEDI and multi-source EO data fusion over Indian forest[J]. International Journal of Remote Sensing, 2024, 45(4): 1304-1338. DOI: 10.1080/01431161.
[4] Puliti S, Hauglin M, Breidenbach J, et al. Modelling above-ground biomass stock over Norway using national forest inventory data with ArcticDEM and Sentinel-2 data[J]. Remote Sensing of Environment, 2020, 236: 111501. DOI: 10.1016/j.rse.2019.111501.
[5] De Almeida C T, Galvão L S, De Oliveira Cruz E AragãO, Luiz Eduardo, et al. Combining LiDAR and hyperspectral data for aboveground biomass modeling in the Brazilian Amazon using different regression algorithms[J]. Remote Sensing of Environment, 2019, 232: 111323. DOI: 10.1016/j.rse.2019.111323.
[6] 李春梅, 张王菲, 李增元, 等. 基于多源数据的根河实验区生物量反演研究[J]. 北京林业大学学报, 2016, 38(3): 64-72.DOI: 10.13332/j.1000-1522.20150209.
[7] 韩宗涛, 江洪, 王威, 等. 基于多源遥感的森林地上生物量KNN-FIFS估测[J]. 林业科学, 2018, 54(9): 70-79.
[8] Luo M, Wang Y F, Xie Y H, et al. Combination of feature selection and CatBoost for prediction: the first application to the estimation of aboveground biomass[J]. Forests, 2021, 12(2): 216. DOI: 10.3390/f12020216.
[9] Chen Z L, Jia K, Xiao C C, et al. Leaf area index estimation algorithm for GF-5 hyperspectral data based on different feature selection and machine learning methods[J]. Remote Sensing, 2020, 12(13): 2110. DOI: 10.3390/rs12132110.
[10] 张少伟, 惠刚盈, 韩宗涛, 等. 基于光学多光谱与SAR遥感特征快速优化的大区域森林地上生物量估测[J]. 遥感技术与应用, 2019, 34(5): 925-938. DOI: 10.11873/j.issn.1004-0323.2019.5.0925.
[11] 孙钊, 谢运鸿, 王宝莹, 等. 基于无人机多维数据集的森林地上生物量估测模型研究[J]. 农业机械学报, 2024, 55(6): 186-195, 236.
[12] 丁家祺, 黄文丽, 刘迎春, 等. 基于机器学习和多源数据的湘西北森林地上生物量估测[J]. 林业科学, 2021, 57(10): 36-48.
[13] Jiang F G, Kutia M, Ma K S, et al. Estimating the aboveground biomass of coniferous forest in Northeast China using spectral variables, land surface temperature and soil moisture[J]. Science of The Total Environment, 2021, 785: 147335. DOI: 10.1016/j.scitotenv.2021.147335.
[14] Pelletier F, Cardille J A, Wulder M A, et al. Inter- and intra-year forest change detection and monitoring of aboveground biomass dynamics using Sentinel-2 and Landsat[J]. Remote Sensing of Environment, 2024, 301: 113931. DOI: 10.1016/j.rse.2023.113931.
[15] Li J N, Bao W, Wang X M, et al. Estimating aboveground biomass of boreal forests in northern China using multiple datasets[J]. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62: 4408410. DOI: 10.1109/TGRS.2024.3408316.
[16] Luo H B, Qin S T, Li J, et al. High-density forest AGB estimation in tropical forest integrated with PolInSAR multidimensional features and optimized machine learning algorithms[J]. Ecological Indicators, 2024, 160: 111878. DOI: 10.1016/j.ecolind.2024.111878.
[17] Hu Y F, Nie Y H, Liu Z H, et al. Improving the potential of coniferous forest aboveground biomass estimation by integrating C- and L-band SAR data with feature selection and non-parametric model[J]. Remote Sensing, 2023, 15(17): 4194. DOI: 10.3390/rs15174194.
[18] Zhang T C, Long J P, Lin H, et al. A novel feature evaluation method in mapping forest AGB by fusing multiple evaluation metrics using PolSAR data[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21: 4006605. DOI: 10.1109/LGRS.2024.3378425.
[19] Li X J, Du H Q, Mao F J, et al. Estimation aboveground biomass in subtropical bamboo forests based on an interpretable machine learning framework[J]. Environmental Modelling & Software, 2024, 178: 106071. DOI: 10.1016/j.envsoft.2024.106071.
[20] Chen L Y, He A Q, Xu Z H, et al. Mapping aboveground biomass of Moso bamboo (Phyllostachys pubescens) forests under Pantana phyllostachysae Chao-induced stress using Sentinel-2 imagery[J]. Ecological Indicators, 2024, 158: 111564. DOI: 10.1016/j.ecolind.2024.111564.
[21] Waqas Khan P, Byun Y C, Lee S J, et al. Machine learning based hybrid system for imputation and efficient energy demand forecasting[J]. Energies, 2020, 13(11): 2681. DOI: 10.3390/en13112681.
[22] Yu X H, Ge H L, Lu D S, et al. Comparative study on variable selection approaches in establishment of remote sensing model for forest biomass estimation[J]. Remote Sensing, 2019, 11(12): 1437. DOI: 10.3390/rs11121437.
[23] GB/T 43648—2024主要树种立木生物量模型与碳计量参数[S].
[24] Cheng K, Chen Y L, Xiang T Y, et al. A 2020 forest age map for China with 30 m resolution[J]. Earth System Science Data, 2024, 16(2): 803-819. DOI: 10.5194/essd-16-803-2024.
[25] 胡天宇,苏艳军. 2021年10米分辨率中国草地覆盖度数据集[DS/OL]. Ⅵ.中国科学院植物科学数据中心(2024-09-06)[2025-04-23].https://www.plantplus.cn/doi/10.12282/plantadata.1607.CSTR:34735.11.PLANTDATA.1607.
[26] 鲁乐乐, 王震, 张雄清, 等. 基于贝叶斯接型平均法和逐步回归法构建杉木单木胸径生长模型[J]. 林业科学,2021, 57(9): 87-97. DOI: 10.11707/j.1001-7488.20210909.
[27] Wen Y Z, Yang A L, Kong X M, et al. A Bayesian-model-averaging Copula method for bivariate hydrologic correlation analysis[J]. Frontiers in Environmental Science, 2022, 9: 744462. DOI: 10.3389/fenvs.2021.744462.
[28] 吴立周, 王晓慧, 王志辉, 等. 基于随机森林法的农作物高光谱遥感识别[J]. 浙江农林大学学报, 2020, 37(1): 136-142. DOI: 10.11833/j.issn.2095-0756.2020.01.018.
[29] Veronika Dorogush A, Ershov V, Gulin A. CatBoost: gradient boosting with categorical features support[PP/OL]. arXiv (2018-10-24)[2025-04-26].https://doi.org/10.48550/arXiv.1810.11363.
[30] Breiman L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. DOI: 10.1023/A:1010933404324.
[31] Li Y C, Li C, Li M Y, et al. Influence of variable selection and forest type on forest aboveground biomass estimation using machine learning algorithms[J]. Forests, 2019, 10(12): 1073. DOI: 10.3390/f10121073.
[32] 段彩红,林辉,龙江平,等.结合改进模拟连续变化检测与分类算法的桉树年龄和蓄积量遥感估测[J].林业科学, 2025, 61(4): 46-55.DOI: 10.11707/j.1001-7488.LYKX20240753.
[33] 黄秋霞. 基于Landsat时间序列的广西桉树林识别及生物量估算[D]. 南宁: 南宁师范大学, 2022.
[34] Zhang Y Z, Ma J, Liang S L, et al. A stacking ensemble algorithm for improving the biases of forest aboveground biomass estimations from multiple remotely sensed datasets[J]. GIScience & Remote Sensing, 2022, 59(1): 234-249. DOI: 10.1080/15481603.2021.2023842.
[35] Zhai W G, Li C C, Fei S P, et al. CatBoost algorithm for estimating maize above-ground biomass using unmanned aerial vehicle-based multi-source sensor data and SPAD values[J]. Computers and Electronics in Agriculture, 2023, 214: 108306. DOI: 10.1016/j.compag.2023.108306.
[36] 陈宸.基于多平台激光雷达的温带森林地上生物量反演研究[D].济南:山东师范大学, 2024.
[1] 杨丽莎, 林钏, 张文齐, 崔换峰, 王艳霞. 气候变化情景下云南省茶适生区空间模拟研究[J]. 广西师范大学学报（自然科学版）, 2025, 43(1): 110-120.
[2] 张天祥, 王艳霞, 张雪珂, 林钏, 周汝良. 基于样本优化和机器学习的地质灾害气象风险预报模型研究——以云南省怒江州为例[J]. 广西师范大学学报（自然科学版）, 2025, 43(1): 201-215.
[3] 周鹏飞, 刘琳, 杨晋帆, 刘婷婷, 魏贵宇, 周汝良. 基于机器学习的云南省蝙蝠空间分布及其危害分析[J]. 广西师范大学学报（自然科学版）, 2023, 41(6): 139-149.
[4] 郭永昊, 刘储睿, 孙珍. 应用代谢组学技术预测啤酒发酵过程中的腐败菌[J]. 广西师范大学学报（自然科学版）, 2023, 41(5): 161-170.
[5] 陈昆, 曲大义, 王少杰, 王其坤. 基于二次分解和融合多特征的短时交通流量组合预测模型[J]. 广西师范大学学报（自然科学版）, 2023, 41(4): 33-46.
[6] 杨烁祯, 张珑, 王建华, 张恒远. 声音事件检测综述[J]. 广西师范大学学报（自然科学版）, 2023, 41(2): 1-18.
[7] 卓文花, 李丽香, 盘远方, 姜勇, 林红玲, 何雁, 方耀成, 梁士楚, 苏权. 桂林丘陵区不同坡位对灌木物种多样性和土壤因子的影响[J]. 广西师范大学学报（自然科学版）, 2023, 41(1): 192-199.
[8] 杨迪, 方扬鑫, 周彦. 基于MEB和SVM方法的新类别分类研究[J]. 广西师范大学学报（自然科学版）, 2022, 40(1): 57-67.
[9] 路凯峰, 杨溢龙, 李智. 一种基于BERT和DPCNN的Web服务分类方法[J]. 广西师范大学学报（自然科学版）, 2021, 39(6): 87-98.
[10] 张永生, 朱文焌, 史若琪, 杜振华, 张瑞, 王志. 基于可信度的Android恶意代码多模型协同检测方法[J]. 广西师范大学学报（自然科学版）, 2020, 38(2): 19-28.
[11] 林越, 刘廷章, 王哲河. 具有两类上限条件的虚拟样本生成数量优化[J]. 广西师范大学学报（自然科学版）, 2019, 37(1): 142-148.
[12] 韦振汉, 宋树祥, 夏海英. 基于随机森林的锂离子电池荷电状态估算[J]. 广西师范大学学报（自然科学版）, 2018, 36(4): 27-33.
[13] 许远静, 胡维平. 基于随机森林的不同程度病态嗓音识别[J]. 广西师范大学学报（自然科学版）, 2018, 36(4): 34-41.
[14] 张仁津, 唐翠芳, 刘彬. 基于人工神经网络游戏程序的研究和设计[J]. 广西师范大学学报（自然科学版）, 2011, 29(2): 119-124.
[15] 袁鼎荣, 谢扬才, 陆广泉, 刘星. 一种新的基于软集合理论的文本分类方法[J]. 广西师范大学学报（自然科学版）, 2011, 29(1): 129-132.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
[1] 田晟, 赵凯龙, 苗佳霖. 基于改进YOLO11n模型的自动驾驶道路交通检测算法研究[J]. 广西师范大学学报（自然科学版）, 2026, 44(1): 1 -9 .
[2] 唐程华, 易见兵, 吴欣, 熊文武, 王敬永. 跨域少样本图像语义分割方法综述[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 1 -27 .
[3] 田晟, 谢华林, 陈东. 基于改进深度强化学习的燃料电池汽车能量管理策略[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 28 -45 .
[4] 张旭, 刘迪迪. 基于TD3算法的电动汽车智能充/放电调度策略[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 46 -55 .
[5] 闫远洋, 谢丽蓉, 张龙军, 任娟, 黄晨晨, 胡超. 基于多目标优化的超短期风电功率预测模型[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 56 -70 .
[6] 陶振卓, 韦笃取. 参数未知永磁同步电机的自适应混沌同步控制[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 71 -78 .
[7] 王成龙, 宋强, 李文峰, 张仕民. PAM-DETR: 基于改进RT-DETR的医用手套小目标缺陷检测算法[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 79 -95 .
[8] 韦吴杰, 陈庆锋. 基于视图解耦与反事实增强的公平图学习[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 96 -106 .
[9] 曹锋, 吴澍康, 朱伟臻, 易见兵. 基于多属性决策的矛盾体分离式评估方法[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 107 -120 .
[10] 罗珵, 黄文韬, 何东平, 张越. 一类具有十参数的复三次多项式微分系统的弱持续中心问题[J]. 广西师范大学学报（自然科学版）, 2026, 44(4): 121 -129 .
版权所有 © 广西师范大学学报（自然科学版）编辑部
地址：广西桂林市三里店育才路15号　邮编：541004
电话：0773-5857325　E-mail: gxsdzkb@mailbox.gxnu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发