| 213 | 0 | 736 |
| 下载次数 | 被引频次 | 阅读次数 |
[目的]草地地上生物量(Aboveground Biomass,AGB)是表征草地生态系统结构与功能的重要指标。针对高寒草地AGB遥感反演中光学数据易饱和、微波数据易受土壤湿度与地表粗糙度干扰，以及传统机器学习模型生态学解释性不足等问题，探索多源遥感数据融合与可解释机器学习方法在青藏高原复杂地形区草地AGB估算中的适用性与优势。[方法]以青藏高原东南缘昌都地区为研究区，基于Sentinel-1微波与Sentinel-2光学遥感数据，结合地面样方调查数据，构建多源遥感驱动的草地AGB估算模型。采用Boruta特征选择方法筛选关键遥感变量，并分别利用极端梯度提升(eXtreme Gradient Boosting,XGB)、随机森林(Random Forest,RF)、支持向量回归(Support Vector Regression,SVR)和K近邻(K-Nearest Neighbors,KNN)4种机器学习算法开展对比建模。同时引入SHAP方法分析关键变量对AGB估算的贡献及其非线性响应特征。[结果]多源遥感融合模型的估算精度显著优于单一数据源模型，其中XGB模型表现最佳，预测决定系数R2达到0.87,RMSE和rRMSE分别为9.5 g·m~-2和13.57%;VH后向散射系数、NDVI和EVI是影响AGB估算的最关键变量，SHAP分析揭示其对不同生物量区间具有显著的非线性贡献特征；草地AGB空间分布明显受地形因子控制，低海拔和南向坡区域生物量较高，而高海拔及北向坡区域生物量显著偏低。[结论]光学与微波遥感数据的互补融合结合机器学习建模可有效提升高寒复杂地形区草地AGB反演精度，引入SHAP方法能够增强模型结果的生态学解释力。研究结果为青藏高原草地生物量遥感监测与生态资源管理提供可靠的技术路径和科学依据。
Abstract：[Objective]Grassland aboveground biomass(AGB) is a key indicator for characterizing the structure and functioning of grassland ecosystems. However, remote sensing-based AGB estimation in alpine grasslands is challenged by the saturation of optical data, the sensitivity of microwave data to soil moisture and surface roughness, and the limited ecological interpretability of conventional machine learning models. This study aims to explore the applicability and advantages of multi-source remote sensing data fusion combined with interpretable machine learning approaches for grassland AGB estimation in complex terrain of the QinghaiTibetan Plateau. [Methods]The study area was located in the Chamdo region on the southeastern margin of the Qinghai-Tibetan Plateau. Multi-source remote sensing-driven AGB estimation models were developed using Sentinel-1 microwave and Sentinel-2 optical data in conjunction with field plot measurements. Key remote sensing variables were selected using the Boruta feature selection algorithm, and four machine learning models, eXtreme Gradient Boosting(XGB), Random Forest(RF), Support Vector Regression(SVR), and K-Nearest Neighbors(KNN),were used for comparative analysis. In addition, SHapley Additive exPlanations(SHAP) were employed to quantify the contributions of key variables and reveal their nonlinear effects on AGB estimation.[Results]Multi-source data fusion models significantly outperformed single-source models, with the XGB model achieving the highest accuracy of R2 = 0.87,and RMSE and rRMSE of 9.5 g·m-2 and 13.57%, respectively; the VH backscatter coefficient, NDVI and EVI were identified as the most important predictors, and SHAP analysis revealed their significant nonlinear contributions across different biomass ranges; the spatial distribution of grassland AGB was strongly controlled by topographic factors, with higher biomass occurring at lower elevations and on south-facing slopes, while significantly lower biomass was observed at higher elevations and on north-facing slopes.[Conclusion]The integration of optical and microwave remote sensing data with machine learning techniques effectively improves grassland AGB estimation in alpine regions with complex terrain, while the incorporation of SHAP enhances the ecological interpretability of model results. This study provides a robust methodological framework and scientific support for grassland biomass monitoring and ecological resource management on the Qinghai-Tibetan Plateau.
[1]崔立晗,郑盛,徐敏.内蒙古森林和草地地上生物量遥感反演[J].地理科学,2024,44(12):2215-2224.CUI L H,ZHENG S,XU M.Remote sensing inversion of forest and grassland aboveground biomass in Inner Mongolia,China[J].Scientia Geographica Sinica,2024,44(12):2215-2224.
[2]高晓彤,徐敏,田海静,等.锡林郭勒盟草地地上生物量时空变化遥感诊断研究[J].生态与农村环境学报,2025,41(12):1543-1552.GAO X T,XU M,TIAN H J,et al.Remote sensing diagnosis of spatiotemporal variation in grassland above-ground biomass in Xilin gol league[J].Journal of Ecology and Rural Environment,2025,41(12):1543-1552.
[3]李文雄,靳瑰丽,刘文昊,等.基于无人机遥感的荒漠草地地上生物量反演研究[J].草地学报,2025,33(4):1258-1266.LI W X,JIN G L,LIU W H,et al.Research on aboveground biomass inversion of desert grassland based on UAV remote sensing[J].Acta Agrestia Sinica,2025,33(4):1258-1266.
[4]黄逸飞,谭炳香,陈振雄,等.协同主被动遥感数据的湖南典型草地地上生物量估测[J].陆地生态系统与保护学报,2024,4(6):32-46.HUANG Y F,TAN B X,CHEN Z X,et al.Above-ground biomass estimation of typical grassland in Hunan using active and passive remote sensing image[J].Terrestrial Ecosystem and Conservation,2024,4(6):32-46.
[5]王彩玲,王一鸣.机理模型与遥感数据同化的草地生物量估算相关研究[J].草业科学,2024,41(9):2104-2117.WANG C L,WANG Y M.A review of research on grassland biomass estimation based on remote sensing and mechanistic models[J].Pratacultural Science,2024,41(9):2104-2117.
[6]黄家兴,吴静,李纯斌,等.基于Sentinel-2和Landsat 8数据的天祝县草地地上生物量遥感反演[J].草地学报,2021,29(9):2023-2030.HUANG J X,WU J,LI C B,et al.Remote sensing retrieval of grassland above-ground biomass in Tianzhu County based on sentinel-2and landsat 8 data[J].Acta Agrestia Sinica,2021,29(9):2023-2030.
[7]FAZAKAS Z,NILSSON M,OLSSON H.Regional forest biomass and wood volume estimation using satellite data and ancillary data[J].Agricultural and Forest Meteorology,1999,98:417-425.
[8]胡仁杰,陈璇黎,陈金,等.MODIS NDVI饱和性对高寒草甸草地生物量遥感估测的影响——以青藏高原东缘为例[J].生态学报,2024,44(14):6357-6372.HU R J,CHEN X L,CHEN J,et al.MODIS NDVI saturation assessment of Alpine meadow grassland biomass estimation using remote sensing:A case study in the eastern edge of the Qinghai-Tibet Plateau[J].Acta Ecologica Sinica,2024,44(14):6357-6372.
[9]ZHAO Q X,YU S C,ZHAO F,et al.Comparison of machine learning algorithms for forest parameter estimations and application for forest quality assessments[J].Forest Ecology and Management,2019,434:224-234.
[10]DUBE T,MUTANGA O.The impact of integrating WorldView-2 sensor and environmental variables in estimating plantation forest species aboveground biomass and carbon stocks in uMgeni Catchment,South Africa[J].ISPRS Journal of Photogrammetry and Remote Sensing,2016,119:415-425.
[11]CHEN F L,ZHANG X F,ZHOU Q,et al.Combining Sentinel-1 and Sentinel-2 data for estimating grassland aboveground biomass in the Altay Mountains[J].ISPRS Journal of Photogrammetry and Remote Sensing,2021,172:132-151.
[12]MCROBERTS R E,TOMPPO E O,FINLEY A O,et al.Estimating areal means and variances of forest attributes using the k-Nearest Neighbors technique and satellite imagery[J].Remote Sensing of Environment,2007,111(4):466-480.
[13]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.HAN Z T,JIANG H,WANG W,et al.Forest above-ground biomass estimation using KNN-FIFS method based on multi-source remote sensing data[J].Scientia Silvae Sinicae,2018,54(9):70-79.
[14]赵露伟,范文义,聂永辉.应用双频合成孔径雷达(SAR)数据和干涉水云模型估算森林生物量[J].东北林业大学学报,2024,52(9):58-68.ZHAO L W,FAN W Y,NIE Y H.Estimating forest biomass using dual-frequency synthetic aperture radar(SAR)data and an interferometric water cloud model[J].Journal of Northeast Forestry University,2024,52(9):58-68.
[15]GAO T,ZHU J J,YAN Q L,et al.Mapping growing stock volume and biomass carbon storage of larch plantations in Northeast China with L-band ALOS PALSAR backscatter mosaics[J].International Journal of Remote Sensing,2018,39(22):7978-7997.
[16]王雁鹤,邢英梅.SAR与多光谱协同反演的三江源高寒森林碳储量优化[J].中国环境科学,2025,45(10):5682-5694.WANG Y H,XING Y M.Optimized modeling of Alpine forest carbon storage in the Sanjiangyuan Nature Reserve via SAR and multispectral synergistic inversion[J].China Environmental Science,2025,45(10):5682-5694.
[17]SOLBERG S,NÆSSET E,GOBAKKEN T,et al.Forest biomass change estimated from height change in interferometric SAR height models[J].Carbon Balance and Management,2014,9(1):5.
[18]JAYATHUNGA S,OWARI T,TSUYUKI S.The use of fixed–wing UAV photogrammetry with LiDAR DTM to estimate merchantable volume and carbon stock in living biomass over a mixed conifer–broadleaf forest[J].International Journal of Applied Earth Observation and Geoinformation,2018,73:767-777.
[19]曹健,刘立忠,李耀威,等.基于机器学习的合金钢拉伸性能预测及SHAP特征分析[J].材料与冶金学报,2025,24(4):393-402.CAO J,LIU L Z,LI Y W,et al.Prediction of tensile properties of alloy steel based on machine learning and SHAP feature analysis[J].Journal of Materials and Metallurgy,2025,24(4):393-402.
[20]王婷,周伟,肖洁芸,等.基于遥感数据和机器学习算法的草地地上生物量估算研究[J].冰川冻土,2023,45(2):753-762.WANG T,ZHOU W,XIAO J Y,et al.Estimating the grassland aboveground biomass based on remote sensing data and machine learning algorithm[J].Journal of Glaciology and Geocryology,2023,45(2):753-762.
[21]宋柯馨,蒋馥根,胡宗达,等.西藏自治区草地地上生物量遥感反演研究[J].生态学报,2023,43(13):5600-5613.SONG K X,JIANG F G,HU Z D,et al.Remote sensing inversion of above-ground biomass of grassland in the Tibet Autonomous Region[J].Acta Ecologica Sinica,2023,43(13):5600-5613.
[22]赵娅冰,彭道黎,郭发苗,等.基于特征选择和机器学习的森林蓄积量估算[J].北京林业大学学报,2025,47(4):155-167.ZHAO Y B,PENG D L,GUO F M,et al.Estimating forest stock volume based on feature selection and machine learning[J].Journal of Beijing Forestry University,2025,47(4):155-167.
[23]蒋晋豫,王海燕,张馨之,等.基于XGBoost算法的森林生物量多源遥感反演[J].西北林学院学报,2025,40(2):198-206,219.JIANG J Y,WANG H Y,ZHANG X Z,et al.Inversion of forest biomass with multi-source remote sensing data based on XGBoost algorithm[J].Journal of Northwest Forestry University,2025,40(2):198-206,219.
[24]熊向阳,杨小周,赵银超,等.基于超参数优化随机森林算法的森林生物量遥感反演[J].中南林业科技大学学报,2024,44(5):102-111.XIONG X Y,YANG X Z,ZHAO Y C,et al.Remote sensing inversion of forest biomass based on hyperparametric optimized random forests algorithm[J].Journal of Central South University of Forestry&Technology,2024,44(5):102-111.
[25]郝君,吕康婷,胡天祺,等.基于机器学习的红树林生物量遥感反演研究[J].林草资源研究,2024(1):65-72.HAO J,LYU K T,HU T Q,et al.Remote sensing inversion of mangrove biomass based on machine learning[J].Forest and Grassland Resources Research,2024(1):65-72.
[26]熊珂,邢元军,和晓风,等.基于KNN算法的森林地上生物量遥感估测[J].林草资源研究,2024(3):106-112.XIONG K,XING Y J,HE X F,et al.Remote sensing estimation of forest AGB based on KNN algorithm[J].Forest and Grassland Resources Research,2024(3):106-112.
基本信息:
中图分类号:TP181;S812
引用信息:
[1]彭泰来,李佳,吴后建,等.基于多源哨兵数据与可解释机器学习的青藏高原南部草地生物量估算[J].沈阳农业大学学报,2026,57(01):113-122.
基金信息:
国家林业和草原局中南调查规划院2023年度院立项科技创新项目（2023010）; 国家林业和草原局中南调查规划院2025年度院立项科技创新项目（2025011）; 西藏自治区科技计划项目（XZ202501ZY0074）; 湖南省自然科学基金面上项目（2021JJ31158）
2026-02-15
2026-02-15