北京市基于哨兵2号数据估计主要林分因子联立模型的构建与应用

doi:10.13466/j.cnki.lczyyj.2025.03.013

摘要/Abstract

摘要： 为探索基于哨兵2号(Sentinel-2)数据构建主要林分因子预估模型,估计乔木林小班因子的现实可行性,利用北京市1 500个乔木林样地的地面实测数据和光谱特征参数,采用误差变量联立方程组方法,构建针叶林、硬阔林、软阔林3种主要森林类型的8项林分因子预估模型,包括平均胸径(D)、平均树高(H)、平均优势高(H_d),以及每公顷株数(N)、断面积(G)、蓄积量(V)、生物量(B)和碳储量(C);将全市乔木林小班按25 m×25 m网格单元提取光谱特征参数,利用所构建8项林分因子预估模型,完成对所有乔木林小班因子的估计。结果表明:1)对估计主要林分因子贡献最大的光谱特征参数是短波红外波段1反射率(B11)和短波红外波段2反射率(B12),其次是红边波段1反射率(B5)和比值植被指数(RVI);2)所构建的3种主要森林类型8项林分因子预估模型,自检和交叉检验的平均预估误差(E_MP)均在10%以内;3)根据模型反演得出的乔木林小班蓄积量累计值,与全市综合监测得到的森林蓄积量相比仅差-1.74%,在抽样调查允许误差范围内。所建北京市3种主要森林类型的8项林分因子预估模型,可用于对全市乔木林小班主要林分因子的估计;基于哨兵2号光谱特征参数构建的主要林分因子预估模型,其预估精度基本能满足森林资源调查监测的技术要求,可在生产实践中应用。

关键词: 光谱特征参数, 主要林分因子, 误差变量, 联立模型, 北京

Abstract:

In order to explore the feasibility of establishing main stand characteristics models based on Sentinel-2 data to estimate the factors of forest patches,the ground plot measured data and spectral characteristic metrics of 1 500 forest plots in Beijing were used to develop prediction models of three major forest types through the error-in-variable simultaneous equations.The models involve eight main stand characteristics,including mean DBH,mean height,dominant height,stem number,basal area,stock volume,biomass and carbon storage.Additionally,based on the spectral characteristic parameters extracted by 25 m×25 m grid cells within the forest patches in Beijing,the eight prediction models were used to estimate main stand characteristics of all forest patches.The results showed:1)The spectral characteristic metrics of Sentinel-2 that contributed the most to the estimation of main stand characteristics were B11 and B12(short wave infrared 1 and 2 band reflectance),followed by B5(red-edge 1 band reflectance)and RVI(Ratio Vegetation Index);2)The mean prediction errors(MPEs)of eight main stand characteristics models of three major forest types were less than 10%,either self-validation or cross-validation;3)The cumulative value of stock volume in all forest patches estimated by the volume models is only -1.74% lower than that obtained by the integrated monitoring of the municipality,which was within the allowable error range of sampling survey.The eight prediction models of three major forest types can be used to estimate the main stand characteristics of forest patches in Beijing;and the prediction accuracy of the main stand characteristics models based on spectral characteristic parameters of Sentinel-2 can almost meet the technical requirements of forest resource inventory and monitoring,then can be applied in practice.

Key words: spectral characteristic metrics, main stand characteristics, error-in-variable, simultaneous models, Beijing

中图分类号:

S771

曾伟生, 温雪香, 李骁尧, 谭炳香, 孙乡楠, 刘樯漪, 王甜. 北京市基于哨兵2号数据估计主要林分因子联立模型的构建与应用[J]. 林草资源研究, 2025,(3): 109-118.

ZENG Weisheng, WEN Xuexiang, LI Xiaoyao, TAN Bingxiang, SUN Xiangnan, LIU Qiangyi, WANG Tian. Establishment and application of simultaneous models for estimating main stand characteristics based on Sentinel-2 data in Beijing[J]. Forest and Grassland Resources Research, 2025,(3): 109-118.

图/表 5

表1

表2

表3

3种主要森林类型的联立模型评价指标

森林类型	林分因子	自检结果						交叉检验结果
森林类型	林分因子	R²	E_SE	E_TR/%	E_AS/%	E_MP/%	E_MPS/%	R²	E_SE	E_TR/%	E_AS/%	E_MP/%	E_MPS/%
针叶林	D	0.266	3.09 cm	5.05	6.27	2.51	26.90	0.257	3.11 cm	5.04	6.37	2.53	27.15
	H	0.305	1.80 m	3.09	3.70	2.67	28.54	0.302	1.80 m	3.08	3.74	2.68	28.65
	H_d	0.332	2.13 m	2.43	2.85	2.65	28.58	0.328	2.13 m	2.40	2.90	2.66	28.67
	G	0.661	3.70 m²/hm²	1.48	5.57	3.91	39.23	0.649	3.77 m²/hm²	1.24	5.57	3.98	39.57
	N	0.282	402株/hm²	3.11	7.03	4.91	49.93	0.282	403株/hm²	2.98	6.95	4.91	50.23
	V	0.639	22.20 m³/hm²	-0.84	-1.02	5.16	48.13	0.623	22.70 m³/hm²	-1.20	-0.95	5.27	48.54
	B	0.494	26.80 t/hm²	-6.27	-0.63	5.37	45.99	0.469	27.50 t/hm²	-6.60	-0.57	5.50	46.45
	C	0.520	13.10 t/hm²	-6.12	-1.19	5.26	45.24	0.495	13.40 t/hm²	-6.45	-1.11	5.40	45.74
硬阔林	D	0.407	3.00 cm	1.28	1.52	1.59	18.34	0.399	3.02 cm	1.26	1.52	1.60	18.44
	H	0.321	1.95 m	0.89	0.50	1.65	18.47	0.316	1.96 m	0.84	0.48	1.66	18.56
	H_d	0.327	2.32 m	0.41	-0.06	1.64	18.55	0.322	2.33 m	0.37	-0.08	1.64	18.63
	G	0.606	3.88 m²/hm²	7.48	6.38	3.01	31.13	0.604	3.89 m²/hm²	7.47	6.51	3.02	31.33
	N	0.425	360株/hm²	8.75	11.99	3.57	38.34	0.418	362株/hm²	8.83	12.39	3.59	38.77
	V	0.576	23.70 m³/hm²	4.62	0.22	3.86	36.95	0.574	23.70 m³/hm²	4.61	0.32	3.86	37.12
	B	0.595	28.20 t/hm²	-1.61	-1.83	3.31	32.34	0.592	28.30 t/hm²	-1.59	-1.72	3.32	32.52
	C	0.592	13.60 t/hm²	-1.54	-1.72	3.32	32.59	0.589	13.60 t/hm²	-1.52	-1.60	3.34	32.76
软阔林	D	0.408	5.75 cm	4.72	5.97	3.46	24.33	0.399	5.80 cm	4.58	5.98	3.48	24.58
	H	0.554	3.25 m	0.43	-0.73	3.01	21.92	0.545	3.29 m	0.47	-0.61	3.04	22.13
	H_d	0.528	3.54 m	0.23	-0.81	2.89	20.99	0.517	3.58 m	0.26	-0.71	2.92	21.26
	G	0.521	4.56 m²/hm²	8.18	9.46	5.11	38.38	0.507	4.63 m²/hm²	7.59	9.38	5.19	38.86
	N	0.290	198株/hm²	5.35	8.30	5.72	41.49	0.263	201株/hm²	5.06	8.47	5.82	42.18
	V	0.482	45.90 m³/hm²	-5.74	-5.49	7.04	42.07	0.450	47.30 m³/hm²	-6.34	-5.50	7.26	42.21
	B	0.307	39.30 t/hm²	-9.45	-4.20	6.78	38.06	0.265	40.40 t/hm²	-9.96	-4.31	6.98	38.23
	C	0.275	19.00 t/hm²	-10.02	-4.24	6.87	38.05	0.232	19.60 t/hm²	-10.51	-4.35	7.07	38.26

表3

图1

图2

参考文献 32

[1]	薛春泉, 陈振雄, 杨加志, 等. 省市县一体化森林碳储量估测技术体系:以广东省为例[J]. 林业资源管理, 2022(4):157-163.
[2]	曾伟生, 杨学云, 孙乡楠, 等. 森林资源调查监测中各级储量数据的一体化方法研究[J]. 林业资源管理, 2022(4):13-19.
[3]	庞勇, 李增元. 基于机载激光雷达的小兴安岭温带森林组分生物量反演[J]. 植物生态学报, 2012, 36(10):1095-1105.
[4]	曾伟生, 孙乡楠, 王六如, 等. 基于机载激光雷达数据估计林分蓄积量及平均高和断面积[J]. 林业资源管理, 2020(2):79-86.
[5]	曾伟生, 孙乡楠, 王六如, 等. 基于机载激光雷达数据的森林蓄积量模型研建[J]. 林业科学, 2021, 57(2):31-38.
[6]	袁钰娜, 彭道黎, 王威, 等. 利用机载激光雷达技术估测东北林区典型针叶林的蓄积量[J]. 应用生态学报, 2021, 32(3):836-844. doi: 10.13287/j.1001-9332.202103.001
[7]	李春干, 李振. 机载激光雷达大区域亚热带森林参数估测的普适性模型式[J]. 林业科学, 2021, 57(10):23-35.
[8]	ZOU Wentao, ZENG Weisheng, SUN Xiangnan. Simultaneous models for the estimation of main forest parameters based on airborne LiDAR data[J]. Forests, 2024, 15:775.
[9]	WANG Bing, JIA Kun, LIANG Shunlin, et al. Assessment of Sentinel-2 MSI spectral band reflectances for estimating fractional vegetation cover[J]. Remote Sensing, 2018, 10(12):1927. doi: 10.3390/rs10121927
[10]	REES W G, TOMANEY J, TUTUBALIA O, et al. Estimation of boreal forest growing stock volume in Russia from Sentinel-2 MSI and land cover classification[J]. Remote Sensing, 2021, 13(21):4483. doi: 10.3390/rs13214483
[11]	FANG Gengsheng, HE Xiaobing, WENG Yuhui, et al. Texture features derived from Sentinel-2 vegetation indices for estimating and mapping forest growing stock volume[J]. Remote Sensing, 2023, 15(11):2821. doi: 10.3390/rs15112821
[12]	龙依, 蒋馥根, 孙华, 等. 基于HLS数据的森林蓄积量遥感反演[J]. 森林与环境学报, 2021, 41(6):62-70.
[13]	吴胜义, 王义贵, 王飞, 等. 基于多距离度量kNN模型的森林蓄积量反演[J]. 中南林业科技大学学报, 2023, 43(2):15-23.
[14]	LIN Hui, ZHAO Wangguo, LONG Jiangping, et al. Mapping forest growing stem volume using novel feature evaluation criteria based on spectral saturation in planted Chinese for forest[J]. Remote Sensing, 2023, 15(2):402. doi: 10.3390/rs15020402
[15]	NAIK P, DALPONTE M, BRUZZONE L. Prediction of forest aboveground biomass using multitemporal multispectral remote sensing data[J]. Remote Sensing, 2021, 13(7):1282. doi: 10.3390/rs13071282
[16]	WAI P, SU Huiyi, LI Mingshi. Estimating aboveground biomass of two different forest types in Myanmar from Sentinel-2 data with machine learning and geostatistical algorithms[J]. Remote Sensing, 2022, 14(9):2146. doi: 10.3390/rs14092146
[17]	白嘎力, 萨如拉, 滑永春, 等. 基于哨兵2遥感影像的根河林区森林地上生物量估算[J]. 内蒙古林业调查设计, 2023, 46(2):25-30.
[18]	CHEN Xinyang, YANG Kemig, MA Jiang, et al. Aboveground biomass inversion based on object-oriented classification and pearson-mRMR-machine learning model[J]. Remote Sensing, 2024, 16(9):1537. doi: 10.3390/rs16091537
[19]	PEI Huiqing, OWARI T, TSUYUKI S, et al. Identifying spatial variation of carbon stock in a warm temperate forest in central Japan using Sentinrl-2 and digital elevation model data[J]. Remote Sensing, 2023, 15(8):1997. doi: 10.3390/rs15081997
[20]	王珠娜, 胡月, 张亚昊, 等. 基于遥感影像的森林质量评价方法研究[J]. 湖北林业科技, 2022, 51(4):37-42.
[21]	杨义炜, 高牧寒, 林腾. 基于哨兵2号卫星影像的建瓯市森林扰动变化监测研究[J]. 林业与生态科学, 2023, 38(4):106-113.
[22]	MORIN D, PLANELLS M, GUYON D, et al. Estimation and mapping of forest structure parameters from open access satellite images:development of a genetic method with a study case on coniferous plantation[J]. Remote Sensing, 2019, 11(11):1275. doi: 10.3390/rs11111275
[23]	LI Xinyu, ZHANG Meng, LONG Jiangping, et al. A novel method for estimating spatial distribution of forest above-ground biomass based on multispectral fusion data and ensemble learning algorithm[J]. Remote Sensing, 2021, 13(19):3910. doi: 10.3390/rs13193910
[24]	WU Xiangqian, SHEN Xin, ZHANG Zhengnan, et al. An advanced framework for multi-scale forest structural parameter estimations based on UAS-LiDAR and Sentinel-2 satellite imagery in forest plantations of northern China[J]. Remote Sensing, 2022, 14(13):2023. doi: 10.3390/rs14092023
[25]	SA Rula, FAN Wengyi. Estimation of forest parameters in boreal artificial coniferous forests using Landsat 8 and Sentinel-2A[J]. Remote Sensing, 2023, 15(14):3605. doi: 10.3390/rs15143605
[26]	邹文涛, 曾伟生, 孙乡楠. 基于机载激光雷达数据估计主要森林参数联立模型[J]. 中南林业调查规划, 2024, 43(4):32-38.
[27]	曾伟生, 唐守正. 立木生物量模型的优度评价和精度分析[J]. 林业科学, 2011, 47(11):106-113.
[28]	国家林业和草原局. 2021中国林草生态综合监测评价报告. 北京: 中国林业出版社, 2023:121.
[29]	国家林业局. 立木材积表:LY/T 1353—1999[S]. 北京: 中国标准出版社,1999.
[30]	全国森林资源标准化技术委员会. 主要树种立木生物量模型与碳计量参数:GB/T 43648—2024[S]. 北京: 中国标准出版社, 2024.
[31]	曾伟生, 唐守正. 非线性模型对数回归的偏差校正及与加权回归的对比分析[J]. 林业科学研究, 2011, 24(2):137-143.
[32]	国家林业局. 森林资源规划设计调查技术规程:GB/T 26424—2010[S]. 北京: 中国标准出版社, 2011.

统计特征	平均胸径 (D)/cm	平均高 (H)/m	优势高 (H_d)/m	株数 (N)/(株/hm²)	断面积 (G)/(m²/hm²)	蓄积量 (V)/(m³/hm²)	生物量 (B)/(t/hm²)	碳储量 (C)/(t/hm²)
平均数	14.0	8.8	10.3	771	10.8	59.0	68.8	33.2
最小值	5.0	1.9	2.0	15	0.1	0.1	0.1	0.1
最大值	44.8	27.4	28.1	4 110	43.9	435.4	338.6	163.7
标准差	6.1	3.9	4.4	598	7.8	57.5	54.9	26.5
变动系数/%	43.6	44.2	42.4	77.6	72.8	97.4	79.7	79.7

森林类型	林分因子	解释变量	联立模型(9)的参数估计值
森林类型	林分因子	解释变量	a₀~g₀	a₁~f₁	a₂~d₂	a₃~d₃
针叶林 (n=400)	D	X₂	2.262	-1.219
	H	X₃	0.756 1	-1.283
	G	X₂、X₃	8.059	-13.98	-12.15
	V	X₂、X₃	10.48	-14.67	-16.17
	H_d		-0.562 4	1.285
	B		0.437 5	1.218
	C		0.497 2
硬阔林 (n=800)	D	X₂、X₃、X₄	1.576	-3.223	1.388	0.321 0
	H	X₂、X₄	1.419	-1.218	0.213 5
	G	X₂、X₃、X₄	4.066	-5.223	-10.52	0.497 0
	V	X₂、X₃、X₄	6.424	-8.478	-10.69	0.551 5
	H_d		-0.092 6	1.214
	B		1.424	1.441
	C		0.479 6
软阔林 (n=300)	D	X₁、X₂、X₃	1.532	-1.023	-5.612	4.050
	H	X₁、X₂、X₃	0.595 5	-1.658	-3.862	2.948
	G	X₁、X₂、X₃	7.779	-18.83	-20.04	17.42
	V	X₁、X₂、X₃	11.99	-25.59	-31.71	28.86
	H_d		1.272	1.032
	B		6.004	0.849 6
	C		0.480 8