基于GEDI和Sentinel-2的内蒙古退耕还林地块树高估测

doi:10.11707/j.1001-7488.LYKX20240213

摘要/Abstract

摘要：

目的: 构建适宜于退耕还林地块的树高样本集，协同遥感数据与机器学习方法估测退耕还林地块树高，为新一轮退耕还林成效监测提供参考依据。方法: 为实现对内蒙古新一轮退耕还林地块树高的准确估测，本研究提出了一种优化的GEDI样本筛选方法，构建成一套适宜于退耕还林地块的高质量树高样本集；借助Sentinel-2中高空间分辨率遥感数据和地形数据，利用梯度提升树算法对退耕还林地块树高进行估测，并对内蒙古退耕还林地块的树高状况进行分析。结果: 基于GEDI和Sentinel-2机器学习模型，可以实现退耕还林乔木地块树高的准确估测，决定系数R²为0.73，估测精度EA为72%，均方根误差RMSE为1.82 m；GEDI样本的优化筛选能提升退耕还林地块树高估测精度，与未筛选的样本相比模型估测精度R²提高了0.32，RMSE降低了0.83 m，EA提升了13%；红边归一化植被指数、差值植被指数、海拔、坡度与坡向变量重要性较高，累计贡献度超过50%，由此证明植被指数与地形信息是树高估算的关键重要性因子。内蒙古退耕还林地块乔木树高区间为2.5~20 m，平均为5.5 m，其中53.51%分布在5~10 m。结论: 本研究所提出的GEDI样本优化筛选方法显著提高了退耕还林地块树高估测的精度，证明了针对退耕还林地块特点优化筛选的有效性。基于遥感数据与机器学习，本研究实现了内蒙古退耕还林乔木地块树高的估测，为退耕还林地块树高估测提供了可行方法。

关键词: GEDI, 内蒙古退耕还林, 树高, 梯度提升树, 可持续森林管理

Abstract:

Objective: A tree height sample set suitable for Grain for Green Project (GGP) stands was constructed, and the machine learning methods were integrated with remote sensing data to estimate tree height in GGP plots, in order to provide a reference basis for monitoring the effectiveness of the new round of GGP. Method: To accurately estimate the tree height of the new round of GGP stands in Inner Mongolia, this study proposed an optimized GEDI sample selection method, and constructed a high-quality tree height sample set suitable for the GGP stands in Inner Mongolia. With Sentinel-2 medium high spatial resolution remote sensing data and terrain data, the gradient boosting tree algorithm was applied to estimate the tree heights in the GGP stands, and analyze the tree height status in the GGP stands. Result: Based on the GEDI and Sentinel-2 machine learning model, the tree heights in the GGP stands were able to be accurately estimated with a coefficient of determination (R2) of 0.73, an estimation accuracy (EA) of 72%, and a root mean square error (RMSE) of 1.82 m. The optimized selection of GEDI samples improved the estimation accuracy for tree height in the GGP stands, with an increase in model estimation accuracy R2 by 0.32, a decrease in RMSE by 0.83 m, and an increase in EA by 13% compared to the unselected samples. The red-edge normalized vegetation index, difference vegetation index, and elevation, slope, and aspect variables are of high importance, with a cumulative contribution of over 50%, proving that vegetation indices and terrain information are key factors for tree height estimation. The distribution of tree height intervals in Inner Mongolia's GGP stands ranges from 2.5–20 m, with an average height of 5.5 m, mainly distributed in the 5–10 m range, which accounts for 53.51%. Conclusion: The GEDI sample optimization screening method proposed in this study can significantly improve the accuracy of tree height estimation in the GGP stands, demonstrating the effectiveness of tailoring the screening process to the specific characteristics of the GGP stands. Based on remote sensing data and machine learning, this study has achieved the estimation of tree height for the new round of GGP stands in Inner Mongolia, providing a feasible approach for tree height estimation in these regions.

Key words: GEDI, Grain for Green in Inner Mongolia, tree height, gradient boosting tree, sustainable forest management

中图分类号:

P237
S771.8

格根塔娜,月亮高可,李晓松,姬翠翠,王建和,沈通,王天璨. 基于GEDI和Sentinel-2的内蒙古退耕还林地块树高估测[J]. 林业科学, 2025, 61(3): 16-26.

Gentana Ge,Lianggaoke Yue,Xiaosong Li,Cuicui Ji,Jianhe Wang,Tong Shen,Tiancan Wang. Estimation of Tree Height in the Grain for Green Program Stands of Inner Mongolia Based on GEDI and Sentinel-2[J]. Scientia Silvae Sinicae, 2025, 61(3): 16-26.

图/表 10

表1

图1

图2

图3

图4

图5

图6

图7

图8

图9

参考文献 0

	陈宝瑞, 辛晓平, 朱玉霞, 等. 内蒙古荒漠化年际动态变化及与气候因子分析. 遥感信息, 2007, (6): 39- 44, 104. doi: 10.3969/j.issn.1000-3177.2007.06.010
	Chen B R, Xin X P, Zhu Y X, et al. Change and analysis of annual desertification and climate factors in Inner Mongolia using MODlS data. Remote Sensing Information, 2007, (6): 39- 44, 104. doi: 10.3969/j.issn.1000-3177.2007.06.010
	陈　浩, 权倍平, 李占斌, 等. 北洛河上游流域植被覆盖度时空变化及其地形分异特征. 水土保持研究, 2024, 31 (2): 219- 227, 239.
	Chen H, Quan B P, Li Z B, et al. Spatiotemporal changes of vegetation coverage and its topographic differentiation characteristics in the upper reaches of Beiluo river watershed. Research of Soil and Water Conservation, 2024, 31 (2): 219- 227, 239.
	顾润源, 周伟灿, 白美兰, 等. 气候变化对黄河内蒙古段凌汛期的影响. 中国沙漠, 2012, 32 (6): 1751- 1756.
	Gu R Y, Zhou W C, Bai M L, et al. Influence of climate change on ice slush period at Inner Mongolia section of Yellow River. Journal of Desert Research, 2012, 32 (6): 1751- 1756.
	阚鸿程, 张慧芳, 孟宝平, 等. 激光雷达飞行参数对低矮植被高度反演的影响——以大疆禅思L1为例. 草业科学, 2023, 40 (10): 2513- 2527. doi: 10.11829/j.issn.1001-0629.2022-0639
	Kan H C, Zhang H F, Meng B P, et al. LiDAR flight parameter influence on low vegetation height inversion: a case study of DJI Zenmuse L1. Pratacultural Science, 2023, 40 (10): 2513- 2527. doi: 10.11829/j.issn.1001-0629.2022-0639
	李明泽, 郭鸿郡, 范文义, 等. 基于GWR的大兴安岭森林立地质量遥感分析. 林业科学, 2017, 53 (6): 56- 66. doi: 10.11707/j.1001-7488.20170607
	Li M Z, Guo H J, Fan W Y, et al. Remote sensing analysis of forest site quality in Daxing’an Mountain based on GWR. Scientia Silvae Sinicae, 2017, 53 (6): 56- 66. doi: 10.11707/j.1001-7488.20170607
	李晓婷, 杨丽帆, 邹友峰, 等. 采煤胁迫下干旱区植被生长周期红边指数动态变化. 煤炭学报, 2021, 46 (5): 1508- 1520.
	Li X T, Yang L F, Zhou Y F, et al. Dynamic change of red edge vegetation index within a growth cycle in arid area under coal mining stress. Journal of China Coal Society, 2021, 46 (5): 1508- 1520.
	李晓英, 禹　熙, 王　方, 等. 西北干旱荒漠地区退耕还林还草工程综合效益评价. 水土保持研究, 2023, 30 (1): 216- 223, 232.
	Li X Y, Yu X, Wang F, et al. Comprehensive benefit evaluation on the project of converting farmland to forestland and grassland in arid desert area of northwest China. Research of Soil and Water Conservation, 2023, 30 (1): 216- 223, 232.
	美合日阿依·莫一丁, 买买提·沙吾提, 李金朝. 基于Sentinel-2时间序列数据及物候特征的棉花种植区提取. 干旱区地理, 2022, 45 (6): 1847- 1859.
	Moyidin M H R A Y, Sawuti M M T, Li J Z. Extraction of cotton planting area based on Sentinel-2 time series data and phenological characteristics. Arid Land Geography, 2022, 45 (6): 1847- 1859.
	苗百岭, 梁存柱, 韩　芳, 等. 内蒙古主要草原类型植物物候对气候波动的响应. 生态学报, 2016, 36 (23): 7689- 7701.
	Miao B L, Liang C Z, Han F, et al. Responses of phenology to climate change over the major grassland types. Acta Ecologica Sinica, 2016, 36 (23): 7689- 7701.
	石　希, 夏军强, 周美蓉, 等. 融合星载LiDAR系统GEDI数据与Sentinel-2影像的长江中游洲滩典型禾本科植物高度动态研究. 湖泊科学, 2024, 36 (2): 562- 574. doi: 10.18307/2024.0235
	Shi X, Xia J Q, Zhou M R, et al. Integrating GEDI and Sentinel-2 data for mapping height dynamics of floodplain representative Poaceae vegetation in the Middle Yangtze River. Journal of Lake Sciences, 2024, 36 (2): 562- 574. doi: 10.18307/2024.0235
	王爱娟. 2015. ICESat-GLAS大光斑LiDAR波形模拟及森林最大冠层高度反演研究. 哈尔滨: 东北林业大学.
	Wang A J. 2015. ICESat-GLAS large-footprint LiDAR waveform simulation and study on sorest maximum canopy height estimation. Harbin: Northeast Forestry University. ［in Chinese］
	王　兵, 牛　香, 郭　珂. 国家退耕还林工程生态监测体系区划布局研究. 陆地生态系统与保护学报, 2022, 2 (1): 57- 72. doi: 10.12356/j.2096-8884.2021-0017
	Wang B, Niu X, Guo K. The layout of the ecological monitoring system of the Gain for Green Project in China. Terrestrial Ecosystem and Conservation, 2022, 2 (1): 57- 72. doi: 10.12356/j.2096-8884.2021-0017
	吴东昊. 2023. 荒漠植物高光谱特性与叶绿素含量预测建模优化研究. 呼和浩特: 内蒙古农业大学.
	Wu D H. 2023. Prediction modeling and optimization of chlorophyll content based on Hyperspectral information of desert plants. Hohhot: Inner Mongolia Agricultural University. ［in Chinese］
	吴　英, 张万幸, 张丽琼, 等. 基于DEM的地形与植被分布关联分析. 东北林业大学学报, 2012, 40 (11): 96- 98. doi: 10.3969/j.issn.1000-5382.2012.11.023
	Wu Y, Zhang W X, Zhang L Q, et al. Analysis of correlation between terrain and forest spatial distribution based on DEM. Journal of Northeast Forestry University, 2012, 40 (11): 96- 98. doi: 10.3969/j.issn.1000-5382.2012.11.023
	徐晋涛, 陶　然, 徐志刚. 退耕还林: 成本有效性、结构调整效应与经济可持续性——基于西部三省农户调查的实证分析. 经济学(季刊), 2004, (4): 139- 162.
	Xu J T, Tao R, Xu Z G. Sloping land conversion program: cost-effectiveness, structural effect and economic sustainability. China Economic Quarterly, 2004, (4): 139- 162.
	叶静芸, 吴　波, 贾晓红, 等. 极干旱区稀疏荒漠植被地上生物量遥感估算. 干旱区地理, 2022, 45 (2): 478- 487. doi: 10.12118/j.issn.10006060.2021.177
	Ye J Y, Wu B, Jia X H, et al. Estimation of aboveground biomass of sparse desert vegetation based on remote sensing techniques in hyper-arid area. Arid Land Geography, 2022, 45 (2): 478- 487. doi: 10.12118/j.issn.10006060.2021.177
	袁鸷慧, 聂　胜, 张合兵, 等. GEDI地面高程和森林冠层高度的精度评价与影响分析. 遥感技术与应用, 2022, 37 (5): 1056- 1070.
	Yuan Z H, Ni S, Zhang H B, et al. Accuracy evaluation and impact analysis of GEDI ground elevation and canopy height. Remote Sensing Technology and Application, 2022, 37 (5): 1056- 1070.
	张　超, 高　晶, 赵艳丽. 基于GIS内蒙古荒漠草原气候变化分析. 草业科学, 2014, 31 (12): 2212- 2220. doi: 10.11829/j.issn.1001-0629.2014-0067
	Zhang C, Gao J, Zhao Y L. Climate changes analysis in the Inner Mongolia desert grassland based on GlS. Pratacultural Science, 2014, 31 (12): 2212- 2220. doi: 10.11829/j.issn.1001-0629.2014-0067
	张增磊, 吕　达, 赵　灯, 等. 多种数值插值算法插值性能分析及优化. 科学技术创新, 2021, (36): 8- 12. doi: 10.3969/j.issn.1673-1328.2021.36.003
	Zhang Z L, Lv D, Zhao D, et al. Analysis and optimization of interpolation performance of a variety of numerical interpolation algorithms. Scientific and Technological Innovation, 2021, (36): 8- 12. doi: 10.3969/j.issn.1673-1328.2021.36.003
	韩明辉, 邢艳秋, 李国元, 等. GEDI不同算法组数据反演森林最大冠层高度和生物量精度比较. 中南林业科技大学学报, 2022, 42 (10): 72- 82.
	Zhao M H, Xing Y Q, Li G Y, et al. Comparison of the accuracy of the maximum canopy height and biomass inversion of the data of different GEDI algorithm groups. Journal of Central South University of Forestry & Technology, 2022, 42 (10): 72- 82.
	郑　颖, 刘华民, 刘东伟, 等. 内蒙古湿地空间分布格局及动态变化研究. 环境科学与技术, 2016, 39 (12): 1- 6,16.
	Zheng Y, Liu H M, Liu D W, et al. Spatial distribution pattern and dynamic change of wetlands in Inner Mongolia. Environmental Science & Technology, 2016, 39 (12): 1- 6,16.
	Atmani F, Bookhagen B, Smith T. Measuring vegetation heights and their seasonal changes in the western namibian savanna using spaceborne lidars. Remote Sensing, 2022, 14 (12): 2928. doi: 10.3390/rs14122928
	Clevers J G P W, Gitelson A A. Remote estimation of crop and grass chlorophyll and nitrogen content using red-edge bands on Sentinel-2 and-3. International Journal of Applied Earth Observation and Geoinformation, 2013, 23, 344- 351. doi: 10.1016/j.jag.2012.10.008
	Dubayah R, Blair J B, Goetz S, et al. The global ecosystem dynamics investigation: high-resolution laser ranging of the earth's forests and topography. Science of Remote Sensing, 2020, 1, 100002. doi: 10.1016/j.srs.2020.100002
	Duncanson L, Kellner J R, Armston J, et al. Aboveground biomass density models for NASA's global ecosystem dynamics investigation (GEDI) lidar mission. Remote Sensing of Environment, 2022, 270, 112845. doi: 10.1016/j.rse.2021.112845
	Gaumont-Guay D, Black T A, Barr A G, et al. Eight years of forest-floor CO₂ exchange in a boreal black spruce forest: Spatial integration and long-term temporal trends. Agricultural & Forest Meteorology, 2014, 184, 25- 35.
	Lang N, Jetz W, Schindler K, et al. A high-resolution canopy height model of the Earth. Nature Ecology & Evolution, 2023, 7 (11): 1778- 1789.
	Li W, Niu Z, Shang R, et al. High-resolution mapping of forest canopy height using machine learning by coupling ICESat-2 LiDAR with Sentinel-1, Sentinel-2 and Landsat-8 data. International Journal of Applied Earth Observation and Geoinformation, 2020, 92, 102163. doi: 10.1016/j.jag.2020.102163
	Li X X, Wessels K, Armston J, et al. First validation of GEDI canopy heights in African savannas. Remote Sensing of Environment, 2023, 285, 113402. doi: 10.1016/j.rse.2022.113402
	Liu X Q, Su Y J, Hu T Y, et al. Neural network guided interpolation for mapping canopy height of China’s forests by integrating GEDI and ICESat-2 data. Remote Sensing of Environment, 2022, 269, 112844. doi: 10.1016/j.rse.2021.112844
	Ni X L, Zhou Y K, Cao C X, et al. Mapping forest canopy height over continental China using multi-source remote sensing data. Remote Sensing, 2015, 7 (7): 8436- 8452. doi: 10.3390/rs70708436
	Potapov P, Li X Y, Hernandez-Serna A, et al. Mapping global forest canopy height through integration of GEDI and Landsat data. Remote Sensing of Environment, 2021, 253, 112165. doi: 10.1016/j.rse.2020.112165
	Simard M, Pinto N, Fisher J B, et al. 2011. Mapping forest canopy height globally with spaceborne lidar. Journal of Geophysical Research: Biogeosciences, 116: G04021.
	Zhu X X, Wang C, Nie S, et al. Mapping forest height using photon-counting LiDAR data and Landsat 8 OLI data: a case study in Virginia and North Carolina, USA. Ecological Indicators, 2020, 114, 106287. doi: 10.1016/j.ecolind.2020.106287
	Zandler H, Brenning A, Samimi C. Quantifying dwarf shrub biomass in an arid environment: comparing empirical methods in a high dimensional setting. Remote Sensing of Environment, 2015, 158, 140- 155. doi: 10.1016/j.rse.2014.11.007
	Zhang X, Liu L Y, Chen X D, et al. GLC_FCS30: global land-cover product with fine classification system at 30m using time-series Landsat imagery. Earth System Science Data, 2021, 13 (6): 2753- 2776. doi: 10.5194/essd-13-2753-2021
	Zhao T, Mu X H, Song W J, et al. Mapping spatially seamless fractional vegetation cover over China at a 30-m resolution and semimonthly intervals in 2010-2020 based on Google Earth Engine. Journal of Remote Sensing, 2023, 3, 0101. doi: 10.34133/remotesensing.0101

变量类型 Variable type	变量名称 Variable name	描述 Description	公式 Formula
光谱信息 Spectral information	B₂、B₃、B₄、B₅、B₆、B₇、B₈、B₁₁、B₁₂	Sentinel-2影像波段信息 Sentinel-2 image band information
	NDVI	归一化植被指数 Normalized difference vegetation index	$\dfrac{{{B_8} - {{\rm B}_4}}}{{{B_8} + {B_4}}}$
	RENDVI	红边归一化植被指数 Red edge normalized difference vegetation index	$\dfrac{{{B_6} - {B_5}}}{{{B_6} + {B_5}}}$
	NDWI	归一化差值水体指数 Normalized difference water index	$\dfrac{{{B_3} - {B_8}}}{{{B_3} + {B_8}}}$
	EVI	增强植被指数 Enhanced vegetation index	$\dfrac{{2.5 \times \left( {{B_8} - {B_4}} \right)}}{{{B_8} + 6 \times {B_4} - 7.5 \times {B_2} + 1}}$
	RVI	比值植被指数 Ratio vegetation index	$\dfrac{{{B_8}}}{{{B_4}}}$
植被指数 Vegetation index	DVI	差值植被指数 Difference vegetation index	${B_8} - {B_4}$
	VDVI	差异性植被指数 Visible-band difference vegetation index	$\dfrac{{2 \times {B_3} - {B_4} - {B_2}}}{{2 \times {B_3} + B{}_4 + {B_2}}}$
	SAVI	土壤调节植被指数 Soil adjusted vegetation index	$\dfrac{{1.5 \times \left( {{B_8} - {B_4}} \right)}}{{{B_8} + {B_4} + 0.5}}$
	MSAVI	改良土壤调节植被指数 Modified soil adjusted vegetation index	$\dfrac{{\left[ {2 \times {B_8} + 1 - \sqrt {{{\left( {2 \times {B_8} + 1} \right)}^2} - 8 \times \left( {{B_8} - {B_4}} \right)} } \right]}}{2}$
	NBR	归一化燃烧率指数 Normalized burn ratio index	$ \dfrac{{{B_8} - {B_{12}}}}{{{B_8} + {B_{12}}}} $
	NDMI	归一化差值水分指数 Normalized difference moisture index	$ \dfrac{{{B_8} - {B_{11}}}}{{{B_8} + {B_{11}}}} $
	DEM	从DEM中提取高程信息 Extract elevation information from the DEM
地形信息 Terrain information	Aspect	从DEM中提取坡向信息 Extract aspect information from the DEM
	Slope	从DEM中提取坡度信息 Extract slope information from the DEM

[1]	笪明慧,冉松松,傅彦榕,邹秉章,王思荣,万晓华,余再鹏,黄志群. 邻体树种多样性对杉木叶片氮与钾、钙、镁计量比的影响[J]. 林业科学, 2024, 60(6): 37-43.
[2]	徐奇刚,雷相东,郑宇,胡兴国,雷渊才,何潇. 基于Richards方程的冷杉树高曲线深度神经网络激活函数[J]. 林业科学, 2023, 59(10): 50-56.
[3]	黄宏超,谢栋博,段光爽,张状,张海江,符利勇. 华北落叶松和白桦半参数树高曲线模型[J]. 林业科学, 2022, 58(10): 101-110.
[4]	王彬,田相林,曹田健. 油松幼树树高生长预测的不确定性贝叶斯分析[J]. 林业科学, 2020, 56(11): 73-86.
[5]	毛学刚, 姚瑶, 范文义. 基于Landsat长时间序列的森林扰动参数提取与树高估算[J]. 林业科学, 2019, 55(3): 79-87.
[6]	王冬至, 张冬燕, 李永宁, 张志东, 李大勇, 黄选瑞. 基于贝叶斯法的针阔混交林树高与胸径混合效应模型[J]. 林业科学, 2019, 55(11): 85-94.
[7]	耿林, 李明泽, 范文义, 王斌. 基于机载LiDAR的单木结构参数及林分有效冠的提取[J]. 林业科学, 2018, 54(7): 62-72.
[8]	尤号田, 邢艳秋, 彭涛, 丁建华. 机载LiDAR航带旁向重叠对针叶林结构参数估测的影响[J]. 林业科学, 2018, 54(6): 109-118.
[9]	娄明华, 张会儒, 雷相东, 李春明, 臧颢. 基于空间自相关的天然蒙古栎阔叶混交林林木胸径-树高模型[J]. 林业科学, 2017, 53(6): 67-76.
[10]	刘鲁霞, 庞勇, 李增元. 基于地基激光雷达的亚热带森林单木胸径与树高提取[J]. 林业科学, 2016, 52(2): 26-37.
[11]	王冬至, 张冬燕, 张志东, 黄选瑞. 基于非线性混合模型的针阔混交林树高与胸径关系[J]. 林业科学, 2016, 52(1): 30-36.
[12]	祖笑锋, 倪成才, Gorden Nigh, 覃先林. 基于混合效应模型及EBLUP预测美国黄松林分优势木树高生长过程[J]. 林业科学, 2015, 51(3): 25-33.
[13]	张雄清, 张建国, 段爱国. 基于贝叶斯法估计杉木人工林树高生长模型[J]. 林业科学, 2014, 50(3): 69-75.
[14]	李海奎;法蕾. 基于分级的全国主要树种树高-胸径曲线模型[J]. 林业科学, 2011, 47(10): 83-90.
[15]	于颖;范文义;李明泽;杨曦光. 利用大光斑激光雷达数据估测树高和生物量[J]. 林业科学, 2010, 46(9): 84-87.