中南林业科技大学学报

融合RFE-CV和Optuna-TPE超参数优化的机载LiDAR湿地松人工林碳储量估测模型

PDF下载

盛文蕾 ^1a , 温小荣 ^1b , 汪求来 ² , 叶金盛 ²

中南林业科技大学学报 | 林学 2026,46(8): 68-78

收起

中南林业科技大学学报 |林学 2026 , 46 (8) : 68 -78

融合RFE-CV和Optuna-TPE超参数优化的机载LiDAR湿地松人工林碳储量估测模型

全屏

盛文蕾^1a, 温小荣^1b, 汪求来², 叶金盛²

作者信息

^1a.南京林业大学　林草学院、水土保持学院，江苏　南京　210037

^1b.南京林业大学　南方现代林业协同创新中心，江苏　南京　210037

^2.广东省林业调查规划院，广东　广州　510520

通讯作者:

温小荣（njw9872@163.com），副教授，博士，硕士研究生导师。

作者简介:

盛文蕾（shengwenlei_1999@163.com），硕士研究生。

An integrated model for estimating carbon stock in airborne LiDAR-mapped wetland pine plantations combining RFE-CV and Optuna-TPE hyperparameter optimization

Wenlei Sheng^1a, Xiaorong Wen^1b, Qiulai Wang², Jinsheng Ye²

Affiliations

^1a.College of Forestry and Grassland, College of Soil and Water Conservation, Nanjing Forestry University, Nanjing 210037, Jiangsu, China

^1b.Co-Innovation Center for Sustainable Forestry in Southern China, Nanjing Forestry University, Nanjing 210037, Jiangsu, China

^2.Guangdong Forestry Survey and Planning Institute, Guangzhou 510520, Guangdong, China

出版时间: 2026-08-25 doi: 10.14067/j.cnki.1673-923x.2026.08.007

文章导航

摘要

收起

【目的】

融合机载LiDAR点云特征与超参数优化算法，构建精准的湿地松Pinus elliottii人工林碳储量估测模型，为大尺度碳汇监测提供科学依据，进而制定有助于实现“双碳”目标的森林经营决策。

【方法】

以广东省湿地松人工林为对象，基于机载LiDAR数据提取101个点云特征变量，采用递归特征消除结合交叉验证（RFE-CV）筛选最优特征组合，并引入基于TPE算法的Optuna超参数优化框架（以RMSE为目标函数）构建随机森林（RF）、K-近邻（KNN）和极端梯度提升（XGBoost）模型，模型性能通过R²、RMSE和MAE进行评估，利用SHAP值解析特征对湿地松碳储量预测的贡献机制。

【结果】

1）运用五折RFE-CV筛选出5个与湿地松碳储量估测紧密相关的最优特征变量（A_IH40、A_IH80、H₁₀、H₂₅、I_max），相较于使用全部变量构建模型，特征筛选显著提高了各模型的预测精度。其中，XGBoost模型效果最明显（R²从0.78提升至0.89，RMSE从0.63 t降至0.46 t，MAE从0.48 t降至0.41 t）。2）基于TPE算法的Optuna超参数优化框架效果显著优于随机搜索，XGBoost模型的RMSE值降低了2.31%，误差调整百分比（13.13%）达到随机搜索的3.57倍；RF和KNN模型的收敛迭代次数分别减少33.33%和49.18%。3）经Optuna-TPE优化后，各模型精度显著提升，XGBoost模型精度最优，R²从0.89显著提升至0.93。4）SHAP值表征不同模型的特征依赖性差异，A_IH40主导XGBoost模型预测（SHAP值为0.64），而H₂₅对RF模型更具关键性（SHAP值为0.28）。

【结论】

特征选择和基于TPE算法的Optuna超参数调优对提升湿地松碳储量估测模型性能至关重要。优化后的XGBoost模型结合最优特征组合，实现了对湿地松人工林碳储量的高精度估测。

关键词

Optuna-TPE超参数 / 碳储量 / 随机森林 / KNN / XGBoost / 贝叶斯优化

Abstract

收起

【Objective】

This study integrated airborne LiDAR point cloud features with a hyperparameter optimization algorithm to develop a precise carbon stock estimation model for Pinus elliottii plantations in wetlands. The aim is to provide a scientific basis for large-scale carbon sink monitoring, thereby supporting forest management decisions that align with the "dual carbon" goals.

【Method】

Taking Pinus elliottii plantation in Guangdong Province as the research object, 101 point cloud feature variables were extracted based on airborne LiDAR data. The recursive feature elimination combined with cross-validation (RFE-CV) was used to screen the optimal feature combination, and the Optuna hyperparameter optimization framework based on the TPE algorithm (with RMSE as the objective function) was introduced to construct random forest (RF), k-nearest neighbor (KNN) and extreme gradient boosting (XGBoost) models. The model performance was evaluated by R², RMSE and MAE, and the SHAP value was used to analyze the contribution mechanism of features to the prediction of carbon storage of Pinus elliottii.

【Result】

1) Five optimal feature variables (A_IH40, A_IH80, H₁₀, H₂₅, I_max) closely related to the estimation of Pinus elliottii carbon storage were selected by 5-fold RFE-CV. Compared with the model using all variables, feature selection significantly improved the prediction accuracy of each model. Among them, the XGBoost model has the most obvious effect (R² increased from 0.78 to 0.89, RMSE decreased from 0.63 to 0.46 t, MAE decreased from 0.48 to 0.41 t); 2) The Optuna hyperparameter optimization framework based on the TPE algorithm is significantly better than random search: the RMSE value of the XGBoost model is reduced by 2.31%, and the error adjustment percentage (13.13%) is 3.57 times that of random search. The convergence iterations of the RF and KNN models are reduced by 33.33% and 49.18%, respectively; 3) After optimization by Optuna-TPE, the accuracy of each model has been significantly improved: the XGBoost model achieves the highest accuracy, with its R² significantly increasing from 0.89 to 0.93, thus becoming the optimal model; 4) SHAP value analysis revealed the difference of feature dependence in different slash pine models: A_IH40 dominated XGBoost prediction (SHAP = 0.64), while H₂₅ was more critical to the RF model (SHAP = 0.28).

【Conclusion】

This study showed that feature selection and Optuna hyperparameter tuning based on the TPE algorithm are crucial to improve the performance of the Pinus elliottii carbon storage estimation model. The optimized XGBoost model combined with the optimal feature combination realized the high-precision estimation of the carbon storage of Pinus elliottii plantations.

Key words

Optuna-TPE hyperparameter / carbon storage / random forest / KNN / XGBoost / Bayesian optimization

引用本文

盛文蕾, 温小荣, 汪求来, 叶金盛. 融合RFE-CV和Optuna-TPE超参数优化的机载LiDAR湿地松人工林碳储量估测模型. 中南林业科技大学学报, 2026 , 46 (8) : 68 -78 . DOI: 10.14067/j.cnki.1673-923x.2026.08.007

Wenlei Sheng, Xiaorong Wen, Qiulai Wang, Jinsheng Ye. An integrated model for estimating carbon stock in airborne LiDAR-mapped wetland pine plantations combining RFE-CV and Optuna-TPE hyperparameter optimization[J]. Journal of Central South University of Forestry & Technology, 2026 , 46 (8) : 68 -78 . DOI: 10.14067/j.cnki.1673-923x.2026.08.007

基金

收起

广东省林业科技创新项目(2021KJCX001)
国家重点研发计划项目(2016YFC0502704)
江苏省高校优势学科建设工程资助项目（PAPD）

参考文献

收起

文献

收起

参考文献引证文献

排序方式：

[1]

张邦文, 金苏蓉, 邵虹, 等. 基于小班尺度的乔木林碳储量和碳密度估算：以宁都县为例[J]. 南方林业科学, 2023, 51(4): 31-35.

Zhang

B W

, Jin

S R

, Shao

, et al. Estimation of carbon storage and carbon density based on the subcompartment scale: a case study of Ningdu County, Jiangxi Province[J]. South China Forestry Science, 2023, 51(4): 31-35.

[2]

Koch

, Kaplan

J O

. Tropical forest restoration under future climate change[J]. Nature Climate Change, 2022, 12: 279-283.

[3]

Peduzzi

. The disaster risk, global change, and sustainability nexus[J]. Sustainability, 2019, 11: 957.

[4]

Auffhammer

, Gong

Y Z

. China's carbon emissions from fossil fuels and market-based opportunities for control[J]. Annual Review of Resource Economics, 2015, 7: 11-34.

[5]

王辽阔, 何介南. 湘北湿地松人工林生物量分配格局与碳汇潜力研究[J]. 湖南林业科技, 2025, 52(2): 10-18.

Wang

L K

, He

J N

. Study on biomass allocation pattern and carbon sink potential of Pinus elliottii plantation in northern Hunan[J]. Hunan Forestry Science & Technology, 2025, 52(2): 10-18.

[6]

Fei

, Chen

Y M

, Ke

, et al. Somatic embryogenesis of slash pine (Pinus elliottii Engelm.): initiation, maturation, germination and mycorrhization of regenerated plantlets[J]. Plant Cell, Tissue and Organ Culture (PCTOC), 2024, 157: 65.

[7]

张兴锐, 齐国辉, 陈利英, 等. 森林植被和土壤碳储量研究进展[J]. 青海农林科技, 2023(1): 56-61.

Zhang

X R

, Qi

G H

, Chen

L Y

, et al. Research progress of forest vegetation and soil carbon storage[J]. Science and Technology of Qinghai Agriculture and Forestry, 2023(1): 56-61.

[8]

Gibbs

H K

, Brown

, Niles

J O

, et al. Monitoring and estimating tropical forest carbon stocks: making REDD a reality[J]. Environmental Research Letters, 2007, 2: 045023.

[9]

Xie

X K

, Wang

Q L

, Dai

L M

, et al. Application of China's national forest continuous inventory database[J]. Environmental Management, 2011, 48: 1095-1106.

[10]

孙华, 龙依, 王储, 等. 森林储量遥感估测的不确定性研究进展[J]. 中南林业科技大学学报, 2026, 46(2): 1-12.

Sun

, Long

, Wang

, et al. Research progress on uncertainty in remote sensing-based forest reserves estimation[J]. Journal of Central South University of Forestry & Technology, 2026, 46(2): 1-12.

[11]

W F

, Cheng

Y F

, Luo

M Y

, et al. Progress and limitations in forest carbon stock estimation using remote sensing technologies: a comprehensive review[J]. Forests, 2025, 16: 449.

[12]

Gong

, Ge

W Y

, Guo

J Q

, et al. Satellite remote sensing of vegetation phenology: progress, challenges, and opportunities[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2024, 217: 149-164.

[13]

刘腾艳. 遥感结合过程模型的浙江省森林碳储量时空演变研究[D]. 杭州: 浙江农林大学, 2019: 2-3.

Liu

T Y

. Temporal-spatial evolution of forest carbon stock in Zhejiang based on remote sensing combined with process model[D]. Hangzhou: Zhejiang A & F University, 2019: 2-3.

[14]

陈明, 陈松, 王彬彬, 等. 星载激光雷达足印点质量对获取森林冠层高度的影响[J]. 中南林业科技大学学报, 2026, 46(3): 197-207.

Chen

, Chen

, Wang

B B

, et al. The impact of spaceborne LiDAR footprint quality on canopy height extraction[J]. Journal of Central South University of Forestry & Technology, 2026, 46(3): 197-207.

[15]

王可月, 王轶夫, 陈馨, 等. 基于集成学习算法和Optuna调优的江西省森林碳储量遥感估测[J]. 生态学报, 2025, 45(2): 685-700.

Wang

K Y

, Wang

Y F

, Chen

, et al. Remote sensing estimation of forest carbon storage in Jiangxi Province based on ensemble learning algorithm and Optuna tuning[J]. Acta Ecologica Sinica, 2025, 45(2): 685-700.

[16]

Jiang

F G

, Li

M D

, Chen

S W

. Adaptive Gaussian-PSO XGBoost model for Alpine forests aboveground biomass estimation using spaceborne PolSAR and LiDAR data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025, 18: 10157-10171.

[17]

Cheng

F Y

, Ou

G L

, Wang

, et al. Remote sensing estimation of forest carbon stock based on machine learning algorithms[J]. Forests, 2024, 15: 681.

[18]

Luo

, 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: 216.

[19]

Masek

J G

, Collatz

G J

. Estimating forest carbon fluxes in a disturbed southeastern landscape: integration of remote sensing, forest inventory, and biogeochemical modeling[J]. Journal of Geophysical Research: Biogeosciences, 2006, 111: G01006.

[20]

高宇航. 基于Landsat 8和机器学习的高峰林场桉树地上碳储量模型研究[D]. 北京: 北京林业大学, 2022: 51-53.

Gao

Y H

. Study on aboveground carbon storage model of Eucalyptus in Gaofeng forest farm based on Landsat 8 and machine learning[D]. Beijing: Beijing Forestry University, 2022: 51-53.

[21]

Bergstra

, Bengio

. Random search for hyper-parameter optimization[J]. Journal of Machine Learning Research, 2012, 13: 281-305.

[22]

Bergstra

, Bardenet

, Bengio

, et al. Algorithms for hyper-parameter optimization[C]//Proceedings of the 25th International Conference on Neural Information Processing Systems, ACM, 2011: 2546-2554.

[23]

李海霞, 宋丹蕾, 孔佳宁, 等. 传统机器学习模型的超参数优化技术评估[J]. 计算机科学, 2024, 51(8): 242-255.

H X

, Song

D L

, Kong

J N

, et al. Evaluation of hyperparameter optimization techniques for traditional machine learning models[J]. Computer Science, 2024, 51(8): 242-255.

[24]

T C

. Optuna hyperparameters optimization for discriminating life-threatening ventricular arrhythmia problem[D]. New York: State University of New York at Buffalo, 2023: 1-10.

[25]

Lin

, Zhang

, et al. Optimized LightGBM power fingerprint identification based on entropy features[J]. Entropy, 2022, 24: 1558.

[26]

Hyndman

R J

, Koehler

A B

. Another look at measures of forecast accuracy[J]. International Journal of Forecasting, 2006, 22(4): 679-688.

[27]

Poorazimy

, Shataee

, Mcroberts

R E

, et al. Integrating airborne laser scanning data, space-borne radar data and digital aerial imagery to estimate aboveground carbon stock in Hyrcanian forests, Iran[J]. Remote Sensing of Environment, 2020, 240: 111669.

[28]

Sokolova

, Lapalme

. A systematic analysis of performance measures for classification tasks[J]. Information Processing & Management, 2009, 45: 427-437.

[29]

GB/T 43648—2024主要树种立木生物量模型与碳计量参数[S].

GB/T 43648-2024 Tree biomass models and related parameters to carbon accounting for major tree species[S].

[30]

, Yin

, Zhang

X X

, et al. Tree-level carbon stock estimations across diverse species using multi-source remote sensing integration[J]. Computers and Electronics in Agriculture, 2025, 231: 109904.

[31]

Sarkar

, Sagan

, Bhadra

, et al. Soybean seed composition prediction from standing crops using Planet Scope satellite imagery and machine learning[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 204: 257-274.

[32]

Y Y

, He

H S

, Hawbaker

T J

, et al. Evaluating k-nearest neighbor (kNN) Imputation models for species-level aboveground forest biomass mapping in northeast China[J]. Remote Sensing, 2019, 11: 2005.

[33]

Chen

T Q

, Guestrin

. XGBoost: a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, ACM, 2016: 785-794.

[34]

Štrumbelj

, Kononenko

. Explaining prediction models and individual predictions with feature contributions[J]. Knowledge and Information Systems, 2014, 41: 647-665.

[35]

穆喜云, 刘清旺, 庞勇, 等. 基于机载激光雷达的森林地上碳储量估测[J]. 东北林业大学学报, 2016, 44(11): 52-56.

X Y

, Liu

Q W

, Pang

, et al. Forest aboveground carbon storage using RF algorithmic model and airborne LiDAR data[J]. Journal of Northeast Forestry University, 2016, 44(11): 52-56.

[36]

Chinembiri

T S

, Mutanga

, Dube

. Carbon stock prediction in managed forest ecosystems using Bayesian and frequentist geostatistical techniques and new generation remote sensing metrics[J]. Remote Sensing, 2023, 15: 1649.

[37]

Wang

Z Y

, Zhang

, Li

F L

, et al. Regional mangrove vegetation carbon stocks predicted integrating UAV-LiDAR and satellite data[J]. Journal of Environmental Management, 2024, 368: 122101.

[38]

傅志强. 基于机器学习的林地碳储量遥感反演[D]. 南京: 南京信息工程大学, 2024: 29-41.

Z Q

. Remote sensing inversion of forest land carbon storage based on machine learning[D]. Nanjing: Nanjing University of Information Science & Technology, 2024: 29-41.

[39]

Chen

, Xu

, Zhang

, et al. Cross-scale mapping of above-ground biomass and shrub dominance by integrating UAV and satellite data in temperate grassland[J]. Remote Sensing of Environment, 2024, 304: 114024.

2026年第46卷第8期

PDF下载

引用本文

BibTeX

文章信息

doi: 10.14067/j.cnki.1673-923x.2026.08.007

接收时间：2025-07-15
首发时间：2026-10-30
出版时间：2026-08-25

补充材料

相关文章

文章信息

作者

出版历史

收稿日期：2025-07-15
修回日期：2025-08-21

基金

广东省林业科技创新项目(2021KJCX001)

国家重点研发计划项目(2016YFC0502704)

江苏省高校优势学科建设工程资助项目（PAPD）

作者信息

^1a.南京林业大学　林草学院、水土保持学院，江苏　南京　210037

^1b.南京林业大学　南方现代林业协同创新中心，江苏　南京　210037

^2.广东省林业调查规划院，广东　广州　510520

通讯作者:

温小荣（njw9872@163.com），副教授，博士，硕士研究生导师。

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/znlykj/CN/10.14067/j.cnki.1673-923x.2026.08.007

分享至

全文二维码

扫描看全文

引用本文

BibTeX

本文的引用情况

2种不同金属材料的力学参数

科 Family	属数 Number of genus	种数 Number of species	占总种数比例 Percentage of total species (%)	属 Genus	种数 Number of species	占总种数比例 Percentage of total species (%)
鹅膏菌科Amanitaceae	2	11	5.26	鹅膏菌属 Amanita	10	4.78
小菇科 Mycenaceae	2	12	5.74	丝盖伞属 Inocybe	5	2.39
多孔菌科 Polyporaceae	8	14	6.70	蜡蘑属 Laccaria	5	2.39
红菇科 Russulaceae	3	23	11.00	小皮伞属 Marasmius	6	2.87
				小菇属 Mycena	11	5.26
				光柄菇属 Pluteus	5	2.39
				红菇属 Russula	17	8.13
				栓菌属 Trametes	5	2.39

关闭全屏

BibTeX
EndNote
RefWorks
TxT