基于时序遥感指数的中国竹林植被信息提取

doi:10.11707/j.1001-7488.LYKX20250100

摘要/Abstract

摘要：

目的: 竹林是中国特殊的森林类型，具有显著的生态、经济和社会价值。其光谱特征常与同纬度地区其他森林类型混淆。如何基于遥感技术精确提取中国的竹林分布，是个较大的挑战。本研究构建新的时序遥感指数，并结合随机森林算法评估其对竹林信息提取的贡献作用，从而得以提高竹林信息提取的精度，进而为竹林资源监测提供新的技术思路。方法: 首先基于目视解译，选择竹林、常绿林、落叶林、草地、建筑、裸地、水体和道路8种土地覆被类型的训练样点，利用2022—2023年的哨兵（Sentinel-2A）影像，分析竹林与其他覆被的类似光谱特征差异；在此基础上，创新性地构建能有效辨识竹林与其他林地光谱差异的3个单波段（R_c、RE1_c和SWIR_c）和两个多波段（MVI_c和NDWI_c）时序遥感指数；同时设计下面4种特征组合方案：原始波段+传统指数（FS1）、原始波段+传统指数+红边指数（FS2）、原始波段+传统指数+时序遥感指数（FS3）、原始波段+传统指数+红边指数+时序遥感指数（FS4），并利用随机森林分类算法比较FS1、FS2、FS3和FS4对竹林提取精度的影响，分析时序遥感指数在竹林提取中的重要性，并与《2021中国林草生态综合监测评价报告》的统计数据对比，以验证提取结果的准确性。结果: 在4种组合方案中，土地覆被分类总体精度排序表现为：FS4 > FS3 > FS2 > FS1，竹林的生产者精度和用户精度在FS4中也都是最大的，分别为0.95和0.85。方案组合结果比较表明：时序遥感指数的引入使竹林提取精度明显提升；FS4提取的竹林面积与统计数据具有更好的一致性，均方根误差从不使用时序遥感指数的FS2的17.53降低至7.46；构造的5个时序遥感指数都位于特征值重要性排序的前列，相对重要性均在75%以上，表明研发的时序遥感指数对竹林提取很有应用价值。结论: 引入时序遥感指数能够显著提高竹林提取的精度，基于多时相影像构建的时序遥感指数在竹林资源监测中具有很好的应用潜力。

关键词: 时序遥感指数, 竹林, 重要性排序, Google earth engine, 随机森林分类

Abstract:

Objective: Bamboo forest is a unique forest type in China with significant ecological, economic, and social values. Its spectral feature is often confused with those of other forest types in the same distribution region. It is thus challenging to accurately mapping bamboo forest distribution using remote sensing technology. This study aims to improve the accuracy of mapping bamboo forest by developing newly time-series remote sensing indices (TSI) and combining them with the random forest algorithm, and thus provide a new technical approach for bamboo forest resource monitoring. Method: Training samples for bamboo forest, evergreen forest, deciduous forest, grassland, building, bare land, water body and road were selected through visual interpretation. Based on the Sentinel-2A imagery from 2022–2023, spectral differences between bamboo forests and other cover types were firstly analyzed. Then three single-band (i.e. R_c, RE1_c, and SWIR_c) and two multi-band TSIs (MVI_c and NDWI_c) were innovatively developed to distinguish bamboo forests from other forest types, and four feature sets schemes were designed, namely original bands + traditional indices (FS1), original bands + traditional indices + red-edge indices (FS2), original bands + traditional indices + TSIs (FS3), and original bands + traditional indices + red-edge indices + TSIs (FS4). The random forest classification algorithm was used to compare the effect of FS1, FS2, FS3, and FS4 on the accuracy of mapping bamboo forest, and the importance of TSI in mapping bamboo forest distribution was analyzed. The area of bamboo forest derived from interpreted thematic map were validated against statistics from the 2021 China Forest and Grassland Ecological Comprehensive Monitoring and Evaluation Report. Result: In the four combination schemes, the overall accuracy ranking of land cover classification is as follows: FS4 > FS3 > FS2 > FS1. The producer and user accuracy of bamboo forests are also the highest in FS4, with values of 0.95 and 0.85, respectively. The comparison results of the combination of schemes show that the introduction of TSI significantly improves the accuracy of bamboo forest extraction. The bamboo forest area extracted by FS4 has a better consistency with statistical data, and the root mean square error has decreased from 17.53 in FS2 without using TSI to 7.46. The ranking of feature value importance shows that the five constructed TSIs are all at the top of the importance ranking, and their relative importance is above 75%, indicating that the developed TSIs have important contributions in bamboo forest extraction. Conclusion: The newly developed TSIs play a significant role in mapping bamboo forest distribution, effectively distinguishing bamboo forests from other forest types. The TSIs based on multi-temporal imagery pose a great application potential in forest cover classification.

Key words: time-series remote sensing indices, bamboo forest, importance order, google earth engine, random forest classification

中图分类号:

S718.5

郝嘉珩,郭毅超,李浩,朱爱青,石雷. 基于时序遥感指数的中国竹林植被信息提取[J]. 林业科学, 2025, 61(9): 1-11.

Jiaheng Hao,Yichao Guo,Hao Li,Aiqing Zhu,Lei Shi. Vegetation Cover Extraction of Bamboo Forest in China Based on Time-Series Remote Sensing Indices[J]. Scientia Silvae Sinicae, 2025, 61(9): 1-11.

图/表 10

图1

图2

表1

表2

表3

图3

图4

图5

图6

图7

参考文献 0

	龚健雅. 人工智能时代测绘遥感技术的发展机遇与挑战. 武汉大学学报(信息科学版), 2018, 43 (12): 1788- 1796.
	Gong J Y. Development opportunities and challenges of surveying and remote sensing technology in the age of artificial intelligence. Geomatics and Information Science of Wuhan University, 2018, 43 (12): 1788- 1796.
	顾大形, 陈双林, 郑炜曼, 等. 竹子生态适应性研究综述. 竹子研究汇刊, 2010, 29 (1): 17- 23.
	Gu D X, Chen S L, Zheng W M, et al. Review of the ecologicol adaptability of bamboo. Journal of Bamboo Research, 2010, 29 (1): 17- 23.
	官凤英, 邓旺华, 范少辉. 毛竹林光谱特征及其与典型植被光谱差异分析. 北京林业大学学报, 2012, 34 (3): 31- 35.
	Guan F Y, Deng W H, Fan S H. Spectral characteristics of Phyllostachys pubescens forests and spectral differences from typical vegetation. Journal of Beijing Forestry University, 2012, 34 (3): 31- 35.
	国家林业和草原局. 2019. 中国森林资源报告(2014—2018). 北京: 中国林业出版社.
	National Forestry and Grassland Administration. 2019. China forest resources report (2014−2018). Beijing: China Forestry Publishing House. ［in Chinese］
	国家林业和草原局. 2023. 2021中国林草生态综合监测评价报告. 北京: 中国林业出版社.
	National Forestry and Grassland Administration. 2023. 2021 China forest and grassland ecological comprehensive monitoring and evaluation report. Beijing: China Forestry Publishing House. ［in Chinese］
	林兴稳, 闻建光, 吴胜标, 等. 地表反射率地形校正物理模型与效果评价方法研究进展. 遥感学报, 2020, 24 (8): 958- 974.
	Lin X W, Wen J G, Wu S B, et al. Advances in physical models and evaluation methods for topographic correction of surface reflectance. Journal of Remote Sensing, 2020, 24 (8): 958- 974.
	罗涟玲, 王修信, 农京辉, 等. 遥感图像林型纹理特征的ICA与SVM分类. 计算机工程与应用, 2012, 48 (13): 227- 229,239. doi: 10.3778/j.issn.1002-8331.2012.13.048
	Luo L L, Wang X X, Nong J H, et al. ICA and SVM classification of forest type texture features in remote sensing images. Computer Engineering and Applications, 2012, 48 (13): 227- 229,239. doi: 10.3778/j.issn.1002-8331.2012.13.048
	单博文, 李　祎, 石　雷, 等. 1960—2050年中国竹林区极端气候的时空变化. 林业科学, 2025, 61 (2): 50- 61. doi: 10.11707/j.1001-7488.LYKX20220728
	Shan B W, Li Y, Shi L, et al. Spatiotemporal change of the extreme climate in China’s bamboo forest during 1960—2050. Scientia Silvae Sinicae, 2025, 61 (2): 50- 61. doi: 10.11707/j.1001-7488.LYKX20220728
	孙晓艳, 杜华强, 韩　凝, 等. 面向对象多尺度分割的SPOT5影像毛竹林专题信息提取. 林业科学, 2013, 49 (10): 80- 87.
	Sun X Y, Du H Q, Han N, et al. Object-oriented multi-scale segmentation for Phyllostachys pubescens information extraction from SPOT5 imagery. Scientia Silvae Sinicae, 2013, 49 (10): 80- 87.
	田　宇. 2022. 基于生态环境效应约束的大庆市景观格局优化研究. 哈尔滨: 东北林业大学.
	Tian Y. 2022. Research on landscape pattern optimization of Daqing City based on ecological environment effect constraints. Harbin: Northeast Forestry University. ［in Chinese］
	王李娟, 孔钰如, 杨小冬, 等. 基于特征优选随机森林算法的农耕区土地利用分类. 农业工程学报, 2020, 36 (4): 244- 250. doi: 10.11975/j.issn.1002-6819.2020.04.029
	Wang L J, Kong Y R, Yang X D, et al. Land use classification in agricultural areas based on feature selection and random forest algorithm. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36 (4): 244- 250. doi: 10.11975/j.issn.1002-6819.2020.04.029
	严欣荣, 张美曼, 郑亚雄, 等. 基于Sentinel-2的丛生竹林信息提取方法比较及分布特征. 生态学杂志, 2020, 39 (3): 1056- 1066.
	Yan X R, Zhang M M, Zheng Y X, et al. Comparison of extraction methods and distribution characteristics of clumping bamboo forests based on Sentinel-2. Chinese Journal of Ecology, 2020, 39 (3): 1056- 1066.
	杨　丽, 李　祎, 单博文, 等. 竹林土壤碳储量的空间变异分析: 以四川长宁县为例. 生态学杂志, 2023, 42 (4): 854- 861.
	Yang L, Li Y, Shan B W, et al. Spatial variation of soil organic carbon of bamboo forest: a case study in Changning County, Sichuan Province. Chinese Journal of Ecology, 2023, 42 (4): 854- 861.
	姚茂林, 江　洪, 张丽玉. 基于Sentinel-2影像红边光谱指数与特征优选的竹林提取研究. 海南大学学报(自然科学版), 2022, 40 (4): 373- 381.
	Yao M L, Jiang H, Zhang L Y. Bamboo forest extraction based on red-edge spectral indices and feature selection from Sentinel-2 imagery. Journal of Hainan University (Natural Science Edition), 2022, 40 (4): 373- 381.
	郑容姝. 2003. 沿海沙地引种绿竹等竹子的抗盐抗旱机理研究. 福州: 福建农林大学.
	Zheng R S. 2003. Study on the salt and drought resistance mechanism of introducing bamboo such as green bamboo into coastal sandy land. Fuzhou: Fujian Agriculture And Forestry University. ［in Chinese］
	周芳纯. 1998. 竹林培育学. 北京: 中国林业出版社.
	Zhou F C. 1998. Bamboo forest cultivation. Beijing: China Forestry Publishing House. ［in Chinese］
	Abbas A M, Amanabo U H. Harnessing the economic and environmental benefits of bamboo in Nigeria. Alanya Academic Review, 2017, 1 (3): 39- 47.
	Leo B. Random forests. Machine Learning, 2001, 45 (1): 5- 32. doi: 10.1023/A:1010933404324
	Li J F, Wang L Y, Liu S Q, et al. An automatic cloud detection model for Sentinel-2 imagery based on Google Earth Engine. Remote Sensing Letters, 2022, 13 (2): 196- 206. doi: 10.1080/2150704X.2021.1988753
	Li L, Li N, Lu D, et al. Mapping moso bamboo forest and its on-year and off-year distribution in a subtropical region using time-series Sentinel-2 and Landsat 8 data. Remote Sensing of Environment, 2019, 231, 111265. doi: 10.1016/j.rse.2019.111265
	Li M N, Li C C, Jiang H, et al. 2016. Tracking bamboo dynamics in Zhejiang, China, using time-series of Landsat data from 1990 to 2014. International Journal of Remote Sensing 37(7): 1714−1729.
	Linderman M, Liu J, Qi J, et al. Using artificial neural networks to map the spatial distribution of understorey bamboo from remote sensing data. International Journal of Remote Sensing, 2004, 25 (9): 1685- 1700. doi: 10.1080/01431160310001598971
	Liu G, Li L, Gong H, et al. Multisource remote sensing imagery fusion scheme based on bidimensional empirical mode decomposition (BEMD) and its application to the extraction of bamboo forest. Remote Sensing, 2016, 9 (1): 19. doi: 10.3390/rs9010019
	McFeeters S K. The use of the normalized difference water index (NDWI) in the delineation of open water features. International Journal of Remote Sensing, 1996, 17 (7): 1425- 1432. doi: 10.1080/01431169608948714
	Nagler L P, Scott L R, Westenburg C, et al. Evapotranspiration on western U. S. rivers estimated using the enhanced vegetation index from MODIS and data from eddy covariance and Bowen ratio flux towers. Remote Sensing of Environment, 2005, 97 (3): 337- 351. doi: 10.1016/j.rse.2005.05.011
	Qi S, Bin Song, Liu C, et al. Bamboo forest mapping in China using the dense Landsat 8 image archive and Google Earth Engine. Remote Sensing, 2022, 14 (3): 762. doi: 10.3390/rs14030762
	Shoko C, Mutanga O. Examining the strength of the newly-launched Sentinel 2 MSI sensor in detecting and discriminating subtle differences between C3 and C4 grass species. ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 129, 32- 40. doi: 10.1016/j.isprsjprs.2017.04.016
	Takeuchi W, Yasuoka Y. Development of normalized vegetation, soil and water indices derived from satellite remote sensing data. Journal of the Japan Society of Photogrammetry and Remote Sensing, 2005, 43 (6): 7- 19.
	You S, Zheng Q, Lin Y, et al. Specific bamboo forest extraction and long-term dynamics as revealed by landsat time series stacks and Google Earth Engine. Remote Sensing, 2020, 12 (18): 3095. doi: 10.3390/rs12183095

特征类别 Feature category	特征名称 Feature name	计算公式 Calculation formula
传统指数 Traditional index	归一化植被指数 Normalized difference vegetation index （NDVI）	$ \mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}=\dfrac{\mathrm{N}\mathrm{I}\mathrm{R}-R}{\mathrm{N}\mathrm{I}\mathrm{R}+R} $
	归一化水体指数 Normalized difference water index （NDWI）	$ \mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}=\dfrac{G-\mathrm{N}\mathrm{I}\mathrm{R}}{G+\mathrm{N}\mathrm{I}\mathrm{R}} $
	增强植被指数 Enhanced vegetation index （EVI）	$ \mathrm{E}\mathrm{V}\mathrm{I}=\dfrac{2.5(\mathrm{N}\mathrm{I}\mathrm{R}-R)}{\mathrm{N}\mathrm{I}\mathrm{R}+6R-7.5B+1} $

红边指数 Red-edge index	红边比值植被指数 Red-edge ratio vegetation index （RVI）	$ \mathrm{R}\mathrm{V}\mathrm{I}=\dfrac{\mathrm{N}\mathrm{I}\mathrm{R}}{\mathrm{R}\mathrm{E}1} $
	红边综合效应指数 Red-edge integrated vegetation index （MVI）	$ \mathrm{M}\mathrm{V}\mathrm{I}=\dfrac{\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{E}1-\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}2}{\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{E}1+\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}2} $
	红边竹林指数1 Red-edge bamboo forest index （BI1）	$ \mathrm{B}\mathrm{I}1=\dfrac{\mathrm{R}\mathrm{V}{\mathrm{I}}_{\mathrm{r}\mathrm{e}1}-\mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}}{\mathrm{R}\mathrm{V}{\mathrm{I}}_{\mathrm{r}\mathrm{e}1}+\mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}} $
	红边竹林指数2 Red-edge bamboo forest index （BI2）	$ \mathrm{B}\mathrm{I}2=\dfrac{\mathrm{M}\mathrm{V}{\mathrm{I}}_{\mathrm{r}\mathrm{e}1}-\mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}}{\mathrm{M}\mathrm{V}{\mathrm{I}}_{\mathrm{r}\mathrm{e}1}+\mathrm{N}\mathrm{D}\mathrm{W}\mathrm{I}} $

编号 No.	特征组合 Feature set	特征变量个数 Number of feature variables
FS1	原始波段+传统指数 Original bands + traditional indices	6+3
FS2	原始波段+传统指数+红边指数 Original bands + traditional indices + red-edge indices	6+3+4
FS3	原始波段+传统指数+时序遥感指数 Original bands + traditional indices + time-series indices	6+3+5
FS4	原始波段+传统指数+红边指数+时序遥感指数 Original bands + traditional indices + red-edge indices + time-series indices	6+3+4+5

分类组合 Feature set	总体精度 Overall accuracy	Kappa系数 Kappa coefficient	竹林 Bamboo forest		常绿林 Evergreen forest		落叶林 Deciduous forest		非林地 Non forest land
分类组合 Feature set	总体精度 Overall accuracy	Kappa系数 Kappa coefficient	PA	UA	PA	UA	PA	UA	PA	UA
FS1	0.75	0.73	0.72	0.61	0.60	0.68	0.68	0.77	0.87	0.86
FS2	0.81	0.77	0.79	0.75	0.74	0.80	0.82	0.77	0.89	0.92
FS 3	0.87	0.80	0.89	0.83	0.79	0.83	0.85	0.88	0.91	0.93
FS4	0.92	0.88	0.95	0.85	0.84	0.92	0.92	0.88	0.91	0.97

[1]	张翱,李文婷,王天祥,武耀星,雷刚,漆良华. 毛竹林土壤易氧化有机碳区域分异及影响因素[J]. 林业科学, 2024, 60(6): 1-12.
[2]	王春玲,史锴源,庞勇,蒙诗栎. 基于变化更新的中东欧国家森林覆盖制图[J]. 林业科学, 2024, 60(5): 116-126.
[3]	郑亚雄,范少辉,张璇,周潇,官凤英. 带状采伐毛竹林生产力变化规律[J]. 林业科学, 2023, 59(2): 22-29.
[4]	刘彩霞,陈俊辉,秦华,梁辰飞,徐秋芳. 有机无机肥长期配施对毛竹林土壤固碳和固氮微生物的影响[J]. 林业科学, 2022, 58(7): 82-92.
[5]	侯志康,曾松伟,莫路锋,周宇峰. 基于GA-BP神经网络的雷竹林CO₂浓度反演[J]. 林业科学, 2022, 58(2): 42-48.
[6]	张玮,何玉友,郭子武,汪舍平,陈双林. 失管毛竹林演替过程中乔木树种群落结构和多样性特征[J]. 林业科学, 2022, 58(12): 12-20.
[7]	胡文杰,庞宏东,胡兴宜,黄发新,杨佳伟,徐丽君,龚苗. 竹林密度和施肥种类对幕阜山区毛竹笋产量和品质及土壤理化性质的影响[J]. 林业科学, 2021, 57(12): 32-42.
[8]	王安可,毕毓芳,温星,王玉魁,蔡函江. 4种芳香植物精油对竹林病原真菌的抗菌性[J]. 林业科学, 2020, 56(6): 59-67.
[9]	熊雨露,周宇峰,李平衡,童亮,周国模,施拥军,杜华强. 毛竹林竹鞭生长特征和空间结构的探地雷达无损探测[J]. 林业科学, 2020, 56(12): 19-27.
[10]	李阳光, 杜华强, 毛方杰, 李雪建, 崔璐, 韩凝, 徐小军. 基于时序Landsat数据的浙江省竹林信息提取及时空演变[J]. 林业科学, 2019, 55(3): 88-96.
[11]	陈亮, 周国模, 杜华强, 刘玉莉, 毛方杰, 徐小军, 李雪建, 崔璐, 李阳光, 朱迪恩. 基于随机森林模型的毛竹林CO₂通量模拟及其影响因子[J]. 林业科学, 2018, 54(8): 1-12.
[12]	杨水生, 徐秀英, 石道金. 劳动力老龄化对竹林地利用效率的影响——基于中介效应的实证研究[J]. 林业科学, 2018, 54(10): 132-142.
[13]	杜满义, 封焕英, 范少辉, 苏文会, 毛超, 唐晓鹿, 刘广路. 闽西毛竹林不同施肥处理下土壤有机碳含量垂直分布与季节动态[J]. 林业科学, 2017, 53(3): 12-20.
[14]	李翀, 周国模, 施拥军, 周宇峰, 徐林, 范叶青, 沈振明, 李少虹, 吕玉龙. 不同经营措施对毛竹林生态系统净碳汇能力的影响[J]. 林业科学, 2017, 53(2): 1-9.
[15]	高国龙, 杜华强, 韩凝, 徐小军, 孙少波, 李雪建. 基于特征优选的面向对象毛竹林分布信息提取[J]. 林业科学, 2016, 52(9): 77-85.