植物生态学报, 2022, 46(10): 1251-1267 doi: 10.17521/cjpe.2021.0373

研究论文

基于Sentinel-2A数据的东北森林植物多样性监测方法研究

周楷玲1,2, 赵玉金,1,*, 白永飞,1,3,*

1中国科学院植物研究所植被与环境变化国家重点实验室, 北京 100093

2中国科学院大学生命科学学院, 北京 100049

3中国科学院大学资源与环境学院, 北京 100049

Study on forest plant diversity monitoring based on Sentinel-2A satellite data in northeast China

ZHOU Kai-Ling1,2, ZHAO Yu-Jin,1,*, BAI Yong-Fei,1,3,*

1State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinses Academy of Science, Beijing 100093, China

2College of Life Science, University of Chinese Academy of Sciences, Beijing 100049, China

3College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China

通讯作者: *(Zhao YJ,zhaoyj@ibcas.ac.cn;Bai YF,yfbai@ibcas.ac.cn)

编委: 苏艳军

责任编辑: 赵航

收稿日期: 2021-10-15   接受日期: 2022-01-14  

基金资助: 中国科学院战略性先导科技专项(A类)(XDA23080303)

Corresponding authors: *(Zhao YJ,zhaoyj@ibcas.ac.cn;Bai YF,yfbai@ibcas.ac.cn)

Received: 2021-10-15   Accepted: 2022-01-14  

Fund supported: Strategic Priority Research Program of Chinese Academy of Sciences(XDA23080303)

摘要

植物多样性监测是开展生物多样性评估, 制定生物多样性保护政策的基础。传统的森林植物多样性监测以实地调查为主, 难以快速获取森林植物多样性的空间分布及其动态变化信息。遥感技术的发展为评估区域尺度森林植物多样性提供了重要工具。该研究选取凉水、丰林和珲春3个国家级自然保护区, 利用Sentinel-2A卫星影像和野外实测数据, 探讨了基于像元和聚类的光谱多样性直接估算方法, 以及基于随机森林回归的森林植物多样性反演方法。研究结果表明: (1)在像元尺度, 基于凸包面积计算的光谱多样性指数对Shannon-Wiener多样性指数的估算精度(R2 = 0.74)优于基于变异系数的方法(R2 = 0.60); (2)基于像元的光谱多样性估算方法对Shannon-Wiener多样性指数的估算精度优于聚类分析方法(R2 = 0.59); (3)基于6个特征变量, 利用随机森林回归算法对Shannon-Wiener多样性指数的估算精度最高(R2 = 0.79); (4)上述方法均不能精确估算Simpson多样性指数和物种丰富度。研究发现基于Sentinel-2A卫星影像能较好地反演Shannon-Wiener多样性指数, 为下一步能在大尺度上进行森林植物多样性估算提供了参考和依据。

关键词： 森林植物多样性; Sentinel-2A; 光谱多样性; 聚类分析; 随机森林回归

Abstract

Aims Plant diversity monitoring is the basis of biodiversity assessment and developing conservation policy. Traditional forest plant diversity monitoring is mainly based on field surveys, which is difficult to quickly obtain the spatial distribution and dynamic change of forest plant diversity. The development of remote sensing technology provides an important tool for assessing forest plant diversity at the regional scale. In this study, we explored two methods of forest plant diversity estimation based on Sentinel-2A satellite images and field data in three selected national nature reserves (Liangshui, Fenglin, and Hunchun).

Methods We used two methods to estimate forest plant diversity: (1) Direct estimation based on spectral diversity at the pixel and cluster scales, respectively; (2) Indirect estimation based on random forest regression. The spectral diversity was calculated based on the coefficient of variation and convex hull area at the pixel scale, respectively. K-means clustering method was used for cluster analysis to calculate the spectral diversity between clusters. For the indirect estimation, we used 10-fold cross validation to select characteristic variables for later diversity calculation.

Important findings Our results showed that: (1) At the pixel scale, the estimation accuracy of Shannon-Wiener diversity index based on convex hull area (R2= 0.74) was better than that of coefficient of variation (R2= 0.60); (2) The pixel-based estimation accuracy of Shannon-Wiener diversity index outperformed clustering basis (R2= 0.59); (3) Based on six feature variables, the Shannon-Wiener diversity index was best estimated using the random forest regression algorithm (R2= 0.79); (4) Both the Simpson diversity index and species richness could not be accurately estimated by the above methods. Our findings indicate the capability of Sentinel-2A satellite images to estimate the Shannon-Wiener diversity index, providing reference and basis for forest plant diversity estimation at a large scale.

Keywords： forest plant diversity; Sentinel-2A; spectral diversity; cluster analysis; random forest regression

PDF (3293KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

引用本文

周楷玲, 赵玉金, 白永飞. 基于Sentinel-2A数据的东北森林植物多样性监测方法研究. 植物生态学报, 2022, 46(10): 1251-1267. DOI: 10.17521/cjpe.2021.0373

ZHOU Kai-Ling, ZHAO Yu-Jin, BAI Yong-Fei. Study on forest plant diversity monitoring based on Sentinel-2A satellite data in northeast China. Chinese Journal of Plant Ecology, 2022, 46(10): 1251-1267. DOI: 10.17521/cjpe.2021.0373

森林是陆地生态系统的重要组成部分, 占全球陆地面积的30%, 在维持生物多样性和生物圈功能方面发挥着重要作用(Pan et al., 2013; Harrison et al., 2014)。然而, 由于人类活动和气候变化的影响, 生物多样性正在以一种前所未有的速度丧失(Ceballos et al., 2015)。及时掌握森林生物多样性的现状、格局、变化趋势和受到的威胁, 是制定生物多样性保护政策和措施的前提。长久以来, 森林植物多样性监测依赖于大量的野外调查, 费时费力, 调查结果也受样地选择、调查方法、抽样调查的力度、参与调查人员的专业知识等诸多因素限制(Gotelli & Colwell, 2001; Graham & Hijmans, 2006)。同时, 野外调查集中在物种和样地水平, 难以在景观、区域乃至全球尺度上实现对森林植物多样性在时空上的连续监测(Kerr & Ostrovsky, 2003; Nicholson et al., 2009; Levrel et al., 2010; Pereira et al., 2010)。

近年来, 随着遥感技术的迅速发展, 使得在大尺度、多时空上评估生物多样性成为可能。遥感数据具有覆盖范围广、可持续性强以及可重复的特点(Nagendra, 2001; Duro et al., 2007), 有助于迅速地揭示大面积生物多样性丢失状况, 对大尺度、长时间序列的生物多样性评估至关重要(Turner, 2014)。生物多样性遥感监测方法可总结为直接法和间接法两种(Turner et al., 2003)。直接法是利用高空间分辨率和高光谱分辨率的卫星传感器直接识别物种、种群或群落。Ke等(2010)利用QuickBird高分辨率多光谱数据和激光雷达数据, 通过图像分割和基于目标的分类方法对森林物种进行识别, 发现两种数据均有较好的结果, 但是结合两种数据源能获取更高的精度, Kappa系数最高可达91.6%。Cross等(2019)利用WorldView-3卫星影像对热带森林里的6个森林物种进行了分类, 分类准确率高达85.37%。间接法是通过遥感数据反演与生物多样性密切相关的指标或参数, 再与野外实测数据构建统计模型以反演生物多样性。植被指数、植物生化组分和植被结构等都可以作为植物多样性研究的相关指标。归一化植被指数(NDVI)经常被用来指示区域的物种多样性(Fairbanks & Mcgwire, 2004; Feeley et al., 2005), 而增强型植被指数(EVI)则可以更好地在植被茂密的地区代替NDVI (Waring et al., 2006)。此外, 与光谱变异相关的叶片叶绿素、氮含量、色素、水分等生化组分的变化也可以用来估算森林物种多样性(Carlson et al., 2007)。

光谱变异假说是生物多样性直接遥感监测的主要理论基础, 即遥感图像光谱多样性与物种多样性有关。光谱多样性, 有时也被称为光学多样性(Palmer et al., 2002; Ustin & Gamon, 2010)、光谱异质性或光谱变异性(Rocchini et al., 2010), 表征光谱反射率的空间变化, 这种变化与植物多样性相关的植物光学性状的差异有关, 可以用来反映物种多样性(Gillespie et al., 2008; Nagendra & Rocchini, 2008)。基于光谱变异假说的光谱多样性与物种多样性的研究已经有很多, 在温带森林(Laliberté et al., 2020), 亚热带森林(Kalacska et al., 2007), 热带森林(Asner & Martin, 2009; Féret & Asner, 2014; Schäfer et al., 2016), 草原(Oldeland et al., 2010; Wang et al., 2016, 2018; Gholizadeh et al., 2020)均有报道。常用光谱多样性指数包括变异系数(CV)(Somers et al., 2015; Wang et al., 2016), 凸包体积(CHV)(Dahlin, 2016; Gholizadeh et al., 2018), 凸包面积(CHA)(Gholizadeh et al., 2018), 光谱角度制图(SAM)(Zhang et al., 2006; Gholizadeh et al., 2018), 光谱信息散度(SID)(Chang, 2000)等。

然而, 由于物种内的光谱变异会混淆物种间的光谱多样性, 进一步影响植物多样性的预测精度。聚类分析作为一种非监督分类绘制植物多样性图的有效方法, 可以基于聚类结果进一步计算光谱异质性, 能够在一定程度上减少物种内光谱变异(Medina et al., 2013; Féret & de Boissieu, 2020)。目前使用非监督分类方法的森林植物多样性研究还较少。Féret和de Boissieu (2020)基于Sentinel-2光谱影像, 利用K均值聚类生成光谱物种图, 再以此绘制了亚马孙某片森林的α和β多样性。也有研究基于分辨率更高的高光谱影像, 通过不同聚类方法反演植物多样性。Medina等(2013)基于AISA机载高光谱数据, 利用光谱方差估算物种丰富度, 发现基于层次聚类得到的Shannon-Wiener多样性指数与野外数据之间的相关性是可变的, 甚至在某些情况下是负的, 但在聚类后进行光谱解混可以显著改善。Zhao等(2018)利用机载高光谱和激光雷达数据对神农架自然保护区森林冠层的单木进行分割, 利用自适应模糊C均值聚类方法, 估算了该区域的树种丰富度和Shannon-Wiener多样性指数(R2 = 0.83, 均方根误差(RMSE) = 0.25)。

除了基于光谱多样性的方法, 基于机器学习的方法在森林植被分类和植物多样性监测中也得到了广泛的关注。随机森林(Breiman, 2001)是一种广泛被用于预测的非参数的机器学习算法, 它不假设数据的正态分布, 具有对噪声不敏感、分类更快更稳定的特点(Clark & Roberts, 2012; Miao et al., 2012), 是用于森林多样性建模的最优方法之一(Mallinis et al., 2020; Gyamfi-Ampadu et al., 2021)。随机森林里包括用于评估模型准确性的集成交叉验证, 能对重要变量进行排序, 并以非线性的方式处理交互数据。Erinjery等(2018)采用最大似然法和随机森林分类的方法, 基于Sentinel-2光谱波段及其衍生的NDVI和纹理、Sentinel-1 SAR波段及其纹理对高海拔热带雨林不同植被类型图进行了绘制, 分类精度可达75%以上。Mallinis等(2020)基于4种不同分辨率的卫星数据, 利用随机森林回归的方法对森林植物多样性进行了评估, 结果发现基于WorldView-2图像生成的模型精度最高, Shannon-Wiener多样性指数估算精度为0.44, Simpson多样性指数为0.37。

由于遥感数据的成本效益和区域可用性, 已经成为评估生态系统生物多样性的重要工具。尽管利用无人机、激光雷达、高分影像或高光谱可提高多样性的预测精度, 但其获取困难、计算复杂、花费相对较高且在大区域上很难实现。基于多光谱数据反演生物多样性在大区域植被和森林覆盖中有着更广泛的应用。Sentinel-2号卫星数据是目前可免费获取的最高空间分辨率的多光谱遥感数据, 也是唯一一个有3个红边带的卫星遥感图像, 它里面包含的叶绿素信息对植被的变化敏感, 对植被的健康监测有重要作用。本研究拟利用Sentinel-2A影像, 在研究区内分别测试基于像元和聚类分析的光谱多样性直接估算方法以及随机森林回归两种目前主流的反演森林植物多样性的方法, 以期发现Sentinel-2A数据在估算森林植物多样性方面的潜力, 为能在更大的尺度上进行森林植物多样性估算及相关研究提供可行的方法参考和依据, 为森林生态学研究和森林管理提供参考。

1 材料和方法

1.1 研究区概况

研究区选取东北区域涵盖针阔混交林、落叶阔叶林及落叶针叶林等不同森林类型的3个国家级自然保护区, 分别是凉水(128.80°-128.93° E, 47.12°- 47.24° N)、丰林(128.98°-129.25° E, 48.03°-48.20° N)和珲春(130.29°-131.24° E, 42.41°-43.47° N)国家级自然保护区, 3个保护区属同一片区, 地理位置接近, 气候土壤条件相似, 主要森林植被类型有一定差异, 以提升模型的可移植性(图1)。

图1

新窗口打开| 下载原图ZIP| 生成PPT

图1   东北森林研究区地理位置。A, 凉水国家级自然保护区。B, 丰林国家级自然保护区。C, 珲春国家级自然保护区。

Fig. 1   Location of the northeast forest study area. A, Liangshui National Nature Reserve. B, Fenglin National Nature Reserve. C, Hunchun National Nature Reserve.


凉水国家级自然保护区总面积为6 394 hm2。气候类型为典型的温带大陆性季风气候, 年平均气温-0.3 ℃, 年降水量676 mm, 积雪期130-150天。保护区内地形条件较为复杂, 属典型丘陵地区, 海拔275- 720 m。区内土壤为暗棕壤, 顶极森林群落类型为阔叶红松林, 主要树种有红松(Pinus koraiensis)、臭冷杉(Abies nephrolepis)、白桦(Betula platyphylla)、紫椴(Tilia amurensis)、红皮云杉(Picea koraiensis)、水曲柳(Fraxinus mandshurica)等, 保护区周围的人工林类型主要是落叶松(Larix gmelinii)、红松和樟子松(Pinus sylvestris var. mongolica)人工林。

丰林国家级自然保护区总面积为18 165 hm2, 海拔约为300-450 m, 最高海拔为688 m。大陆性季风气候, 年平均气温为-0.5 ℃, 年降水量640.5 mm, 主要集中在6-9月。丰林保护区的主要保护对象为原始阔叶红松林, 是全球原始红松林分布的中心地带, 是世界上最典型和完整的以红松为主的北温带针阔叶混交林生态系统。

珲春国家级自然保护区总面积为108 700 hm2, 保护区外围有因保护对象活动范围较大而设立的外围保护带, 面积为41 778 hm2。保护区气候类型为近海中温带海洋性季风气候, 年平均气温5.65 ℃。保护区南北跨越珲春全境, 山地主要分布于北、中和南部的部分地区, 北部全区最高点海拔达973 m, 南部最低点海拔5 m, 高差大。植被类型以阔叶林为主, 北部有少量针叶林和针阔混交林。主要保护树种为东北红豆杉(Taxus cuspidata)、红松、紫椴、黄檗(Phellodendron amurense)、水曲柳等。区内野生动物资源丰富, 其中东北虎(Panthera tigris ssp. altaica)、东北豹(Panthera pardus orientalis)、梅花鹿(Cervus nippon)等9种动物被列为国家I级重点保护野生动物。

自然保护区可分为核心区、缓冲区和实验区3个功能区。核心区是指保护区内未经或很少经过人为干扰的区域, 保护效果最好, 在核心区外围可划分出一定面积的允许从事科研活动的缓冲区, 在缓冲区周围的为实验区, 为人类活动和规划发展的区域(Ma et al., 2009; Xu et al., 2016)。合理的功能区划可以在促进生物多样性保护的同时为当地的经济文化发展作出贡献, 对整个自然保护区的长远发展有重要意义。本研究所选保护区的功能区划概况如图1所示。

1.2 数据及预处理

1.2.1 野外数据

于2019年7-8月在凉水、珲春、丰林国家级自然保护区内开展群落和物种调查, 然而, 在选取珲春自然保护区样点时由于保护区规划和安全问题, 研究点选在保护区北部东北虎豹国家公园。每个保护区各挑选2-4种具有代表性的林分类型, 每种林分类型下各设置3个30 m × 30 m样地, 共获得了20个调查样地, 记录样地4角的GPS坐标。在每个样地内进行每木检尺, 记录样方内出现的所有树种名称、树高、胸径(≥5 cm)、枝下高及冠幅等参数。同时记录样地土壤类型和郁闭度等信息, 样地基本情况如表1所示。其中, 1-5号样地位于丰林保护区内，是以红松、青楷槭(Acer tegmentosum)、紫椴、硕桦(Betula costata)等为主要树种的阔叶红松林。6-11号样地属于珲春保护区范围, 主要植被类型是以红松、紫椴、白桦、臭冷杉等为主要树种的针阔混交林。12-20号样地位于凉水保护区内, 12-14号样地是以臭冷杉、鱼鳞云杉(Picea jezoensis var. microsperma)为主的混交林, 伴生着白桦、花楷槭(Acer ukurunduense)、青楷槭、春榆(Ulmus davidiana var. japonica)等多种阔叶树种, 16-20号样地是以红松、鱼鳞云杉、白桦、水曲柳为主要树种的针阔混交林。

表1   东北森林样地调查信息表

Table 1  Sample plot survey information of northeast forest

样地
Sample plot		地理位置
Geographical position		郁闭度
Crown density		物种丰富度
Number of species		Shannon-Wiener多样性指数
Shannon-Wiener diversity index		Simpson多样性指数
Simpson diversity index
1129.19° E, 48.12° N0.8561.560.76
2129.19° E, 48.12° N0.75101.890.82
3129.19° E, 48.12° N0.8071.670.79
4129.18° E, 48.13° N0.85101.920.82
5129.18° E, 48.13° N0.75101.960.83
6131.11° E, 43.44° N0.70110.870.83
7131.11° E, 43.44° N0.60100.880.85
8131.11° E, 43.44° N0.6580.770.78
9131.07° E, 43.47° N0.6060.650.74
10131.07° E, 43.47° N0.7040.440.56
11131.07° E, 43.47° N0.6080.730.74
12128.90° E, 47.18° N0.6051.340.80
13128.89° E, 47.18° N0.7071.270.66
14128.89° E, 47.18° N0.7061.560.82
15128.89° E, 47.18° N0.6571.010.73
16128.89° E, 47.19° N0.6061.170.72
17128.89° E, 47.18° N0.6040.960.67
18128.86° E, 47.20° N0.70102.060.86
19128.86° E, 47.20° N0.70112.170.84
20128.86° E, 47.20° N0.7091.820.79

新窗口打开| 下载CSV


根据样方调查结果, 计算样方内胸径≥5 cm的树种的物种丰富度(S)、Shannon-Wiener多样性指数(H')和Simpson多样性指数(D)。物种丰富度用样方中的乔木物种数表示。计算公式如下:

 H′= –∑i=1nPilnPi
D = 1−∑i=1nPi2

式中, n为样方中的物种数, Pi为第i个物种的个体数占所有物种总个体数的比例(Pielou, 1966)。

1.2.2 遥感数据及数据预处理

同步于地面调查, 选取2019年7-8月的Sentinel- 2A卫星影像开展森林植物多样性遥感监测。该卫星携带多光谱成像仪(MSI), 轨道高度为786 km, 幅宽为290 km, 卫星的重访周期为10 d, 与Sentinel-2B互补, 重访周期缩短至5 d。影像覆盖了13个光谱波段, 其中包括4个可见光波段(蓝色、绿色、红色和近红外波段), 空间分辨率为10 m, 短波红外波段(SWIR1和SWIR2)、卷云波段(SWIR-Cirrus)和红边波段(红边波段1-4), 空间分辨率为20 m, 其余两个波段的分辨率为60 m。Sentinel-2A数据有3个波段数据在红边范围内, 这对植被健康的监测十分有效。

Sen2Cor (http://step.esa.int/main/snap-supported-plugins/sen2cor/)是Sentinel-2 Level 2A产品生成和格式化的工具集, 利用安装的Sen2Cor插件对下载的L1C级遥感影像进行辐射定标和大气校正, 大气校正后的影像去除了B10卷云波段。对于经过大气校正后的影像在SNAP (http://step.esa.int/main/download/snap-download/)中进行重采样, 分辨率设置为10 m, 最后将处理后的影像在ENVI软件中进行波段的调整、合并、镶嵌和裁剪等处理。对裁剪之后的影像进行辐射滤波, 通过选取NDVI的阈值大于0.3以去除土壤、建筑等非植被区域的影响, 设置近红外和蓝色光谱波段的阈值分别小于1 500和500以去掉阴影区域和云的影响。

1.3 研究方法

本研究测试了两种目前主流的森林植物多样性的反演方法, 包括基于原始影像像元和基于聚类分析的光谱多样性方法和基于随机森林的植物多样性回归的估算方法。基于像元的方法是通过计算变异系数和凸包面积两种光谱多样性指数以反演森林植物多样性。聚类分析主要采用K均值聚类和自定义模糊C均值聚类, 然而自定义模糊C均值聚类预测试结果并不理想, 因此本研究主要采用K均值聚类方法。随机森林是目前比较流行的机器学习方法, 本研究选取了与植物多样性相关的30个植被指数和12个特征波段, 通过随机森林回归估算森林植物多样性。考虑到样方的定位误差和边缘效应, 本研究选取5 × 5移动窗口(即50 m × 50 m)计算了光谱变异系数和光谱特征值。具体的流程如图2所示。

图2

新窗口打开| 下载原图ZIP| 生成PPT

图2   东北森林植物多样性研究方法流程图。NDVI, 归一化植被指数。

Fig. 2   Flow chart of the research methods for forest plant diversity in northeast China. NDVI, normalized difference vegetation index.


1.3.1 变异系数

光谱变异系数是α多样性的一种常用度量(Wang et al., 2016)。根据光谱变异假说, CV随着光谱变异性的增加而增加。本研究基于重采样后的Sentinel-2A卫星数据的12个波段计算了5 × 5移动窗口的变异系数。变异系数的计算公式如下:

CV=∑ λ=443  2190std(ρλ)mean(ρλ)n

式中, ρλ表示波长λ处的反射率,std(ρλ)和 mean(ρλ)表示所有像元波长λ处的标准差和平均值, n为波段数量。

1.3.2 凸包面积

凸包面积是一种新的光谱多样性指标(Deng et al., 2016; Gholizadeh et al., 2018)。CHA是样方L的平均光谱和该样方内第V个像元的光谱所组成的二维空间中的凸包面积。如果一个像元的平均光谱和该区域的平均光谱相似或与其高度相关, 这些波段在图里的位置会近乎在一条直线上, 凸包面积趋于0。样方平均CHA的表达式如下:

CHA¯L=∑V=1mCHA(RV,L,R¯L)m

式中, m为该样方中所有像元的数量,$\bar{R}_{L}$是样方L的平均光谱, RV,L是样方L内第V个像元的光谱。物种丰富度越高的样地, 与光谱的平均值偏差越大, CHA越大。计算CHA时需要将波段值进行归一化处理。

1.3.3 K均值聚类

K均值聚类是一种被广泛应用的非监督分类方法, 它的主要思想为: 将数据自动划分为K个组, 随机创建一个初始的聚类中心, 然后采用迭代的方法, 通过不断移动聚类中心不断尝试改进划分。具体的处理流程是: 首先选择K个中心点, 再将每个数据点按与聚类中心的距离重新划分到离其最近的中心点, 计算每个聚类中的点到该类中心点距离的平均值, 用平均值取代每个聚类中心, 再重复前两个步骤, 直到所有的观测值不再被分配或是达到最大的迭代次数。

本研究利用biodivMapR软件包, 基于Sentinel- 2A卫星影像的多光谱信息, 根据每个像元反射率对应的光学特征区分物种(Ustin & Gamon, 2010)。首先需要确定聚类窗口大小, 以窗口内能够计算多样性所需要的像元最少为原则, 综合考虑影像分辨率和现场调查的样地大小等因素, 通过测试3 × 3、5 × 5、7 × 7以及9 × 9等不同移动窗口(分辨率为10 m), 最终选取5 × 5的移动窗口为计算植物多样性最好的聚类窗口; 然后, 将预处理后的影像进行主成分分析, 选择2-6个能够突出植物属性、噪声较少的主成分, 再利用K均值聚类绘制波谱物种图, 最后基于每个光谱物种包含的像元数, 计算多样性指数。

1.3.4 随机森林回归

随机森林是Leo Breiman在2001年提出的一种基于分类和回归树的集成学习算法, 可以解释若干个自变量对因变量的作用。本研究使用随机森林回归算法构建了基于图像的森林多样性反演模型, 算法的基本流程是: (1)利用bootstrap重采样, 从原始样本n中随机抽取m个样本点, 得到新的训练集, 再利用每个训练集生成对应的CART回归树, 测试样本为每次重采样未被抽到的样本所构成的袋外数据(OOB); (2)随机抽取k个特征作为每个节点的分裂特征集, 再从k个特征中选择最优的分裂方式对该节点进行分裂; (3)每棵回归树自顶向下递归分枝, 直到满足分割终止条件, 让获得的CART回归树都能得到最大限度的生长; (4)生成的m棵回归树构成了随机森林回归模型, 每一棵回归树最终的预测结果为该样本点所到叶节点的均值, 最终的预测结果为所有CART回归树估算结果的均值。回归的精度评价采用OOB预测的残差均方(MSE)表示如下:

MSEOOB=n−1∑1n(yi−y^iOOB)2
RRF2=1−MSEOOBδ^y2

式中, yi表示袋外数据中因变量的实际值, y^i表示对袋外数据的预测值, δ^y2表示袋外数据预测值的方差, RRF2表示拟合优度。

随机特征的选取是基于影像的光谱信息以及衍生的植被指数。本研究采用了前面计算的两种光谱多样性指数、Sentinel-2A影像的12个波段信息和30个与森林多样性相关的植被指数作为特征波段, 相关植被指数的计算公式如表2所示。计算样地内的光谱变异系数, 以作为森林植被多样性的度量(Torresani et al., 2019; Mallinis et al., 2020)。

表2   植被指数计算公式

Table 2  Formula of calculating vegetation index

植被指数 Vegetation index计算公式 Calculate formula参考文献 Reference
TCARI3[(R699.19 - R668.98) - 0.2(R699.19 - R550.67)(R699.19/R668.98)]Kim et al., 1994
OSAVI(1 + 0.16)(R750 - R705)/(R750 + R705 + 0.16)Wu et al., 2008
OSAVI2(1 + 0.16)(R800 - R670)/(R800 + R670 + 0.16)Rondeaux et al., 1996
DATT(R850 - R710)/(R850 - R680)Datt, 1999
DATT2R850/R710Datt, 1999
Gitelson1/R700Gitelson et al., 1999
SR1R750/R700Gitelson & Merzlyak, 1997
SR2R700/R670McMurtrey III et al., 1994
SR3R730/R706Zarco-Tejada et al., 2003
SR4R675/R700Gitelson et al., 2003
MSIR1600/R819Hunt & Rock, 1989
NDII(R819 - R1649)/(R819 + R1649)Hardisky et al., 1983
CRI11/R510 - 1/R550Gitelson et al., 2002
CRI21/R510 - 1/R700Gitelson et al., 2002
ARI1/R550 - 1/R700Sims & Gamon, 2002
PSRI(R680 - R500)/R750Merzlyak et al., 1999
NDVI(R750.66 - R704.6)/(R750.66 + R704.6)Gitelson & Merzlyak, 1994
GNDVI(R783 - R560)/(R783 + R560)Rozenstein et al., 2019
TNDVI((R842 - R665)/(R842 + R665) + 0.5)^0.5Rozenstein et al., 2019
WDVIR842 - R665 × 0.5Rozenstein et al., 2019
NDI45(R705 - R665)/(R705 + R665)Delegido et al., 2011
SAVI(1 + L) × (R799.09 - R680.045)/(R799.09 + R680.045 + L) (L = 0.5)Huete, 1988
SAVI2R799.09/(R680.045+ b/a) (a = 0.969 1, b = 0.084 726)Major et al., 1990
ARVIRB = R680.045 - r(R444.5 - R680.045) (r = 1)
ARVI = (R799.09 - RB)/(R799.09 + RB)		Kaufman & Tanre, 1992
SARVIRB = R680.045 - r(R444.5 - R680.045) (r = 1, L = 0.5)
SARVI = (1 + L)(R799.09 - RB)/(R799.09 + RB + L)		Kaufman & Tanre, 1992
EVIG(R799.09 - R680.045)/(R799.09 + C1R680.045 - C2R444.5 + L)
(G = 2.5, C1 = 6, C2 = 7.5, L = 1)		Huete et al., 1997
IRECI(R783 - R665)/(R705/R740)Frampton et al., 2013
IPVIR842/(R842 + R665)Rozenstein et al., 2019
PSSRAR783/R665Rozenstein et al., 2019
RVIR842/R665Rozenstein et al., 2019

ARI, 花青素反射指数; ARVI, 耐大气植被指数; CRI, 类胡萝卜素反射指数; DATT, DATT植被指数; EVI, 增强型植被指数; Gitelson, Gitelson植被指数; GNDVI,绿色归一化差异植被指数; IPVI, 红外植被百分比指数; IRECI, 倒红边叶绿素指数; MSI, 水分胁迫指数; NDI45, 归一化差异指数; NDII, 归一化红外指数; NDVI, 归一化植被指数; OSAVI, 优化型土壤调节植被指数; PSRI, 植物衰老反射指数; PSSRA, 特征色素简单比值指数; RVI, 比值植被指数; SARVI, 土壤大气阻抗植被指数; SAVI, 土壤调节植被指数; SR, 比值植被指数; TCARI, 转换型叶绿素吸收植被指数; TNDVI, 转化后的归一化植被指数; WDVI, 加权差分植被指数。公式中的R及右下角数字代表位于该波长处的反射值。

ARI, anthocyanin reflectance index; ARVI, atmospherically resistant vegetation index; CRI, carotenoid reflectance index; DATT, DATT vegetation index; EVI, enhanced vegetation index; Gitelson, Gitelson vegetation index; GNDVI, green normalizad difference vegetation index; IRECI, inverted red-edge chlorophyll index; IPVI, infrared percentage vegetation index; MSI, moisture stress index; NDI45, normalized difference index 45; NDII, normalized difference infrared index; NDVI, vormalizad difference vegetation index; OSAVI, optimization soil-adjusted vegetation index; PSRI, plant senescence reflectance index; PSSRa, pigment specific simple ratio; RVI, ratio vegetation index; SARVI, soil atmospherically resistant vegetation index; SAVI, soil adjusted vegetation index; SR, simple ratio index; TCARI, transformed chlorophyll-absorbing vegetation index; TNDVI, transformed normalized difference vegetation index; WDVI, weighted difference vegetation index. R and the lower right number in the formula represent the reflection value at this wavelength.

新窗口打开| 下载CSV


利用44个变量初步构建随机森林回归模型, 对这些变量进行重要性评估。变量的重要性排序按“%IncMSE”的大小表示, “%IncMSE”即increase in mean squared error, 是基于每一棵决策树的袋外数据误差, 通过对每一个预测变量随机赋值, 如果袋外数据误差增幅越大, 其值被随机替换后模型估算的误差就越大, 说明该变量的重要性越高。但某些变量可能对多样性的估算作用很小, 或者变量间存在共线性, 从而影响回归结果。为进一步提高随机森林的估算精度, 减少噪声和冗余信息的干扰, 本研究基于十折交叉验证对变量进行筛选, 选取重要性高且使交叉验证误差最小的特征变量作为模型的输入变量(Genuer et al., 2010; Edwards et al., 2018)。基于所选择的变量构建最终的随机森林回归模型, 使用留一法验证模型精度。输入变量选择在R语言的“randomforest”包中实现, 留一法交叉验证在python中的sklearn库中实现。

2 结果

2.1 基于光谱多样性估算森林植物多样性

图3展示了基于全波段影像像元所计算的CV和CHA与实测的Shannon-Wiener多样性指数、Simpson多样性指数和物种丰富度的关系, 两种光谱多样性指数均能很好地估算Shannon-Wiener多样性指数, CHA对Shannon-Wiener多样性指数的估算精度(R2 = 0.74, RMSE = 0.07)高于CV对其的估算精度(R2 = 0.60, RMSE = 0.11)。但两种光谱多样性指数不能估算Simpson多样性指数和物种丰富度, CV和CHA与实测的Simpson多样性指数的估算精度均为0.12, 与实测的物种丰富度的精度分别为0.11和0.05, 结果均不显著。

图3

新窗口打开| 下载原图ZIP| 生成PPT

图3   基于原始波段的变异系数(CV)与凸包面积(CHA)与实测植物多样性(Shannon-Wiener多样性指数(H')、Simpson多样性指数(D)和物种丰富度(S))的关系。

Fig. 3   Relationship between coefficient of variation (CV) and convex hull area (CHA) based on original bands and measured plant diversity (Shannon-Wiener diversity index (H'), Simpson diversity index (D) and species richness (S)).


与基于像元的光谱多样性相比, 基于聚类的森林植物多样性估算结果略有降低(图4)。基于聚类分析可以较好地估算Shannon-Wiener指数(R2 = 0.59, RMSE = 0.12, p < 0.01), 但Simpson多样性指数与物种丰富度估算精度较差, 结果均不显著(Simpson多样性指数: R2 = 0.02, RMSE = 0.08; 物种丰富度: R2 = 0.12, RMSE = 7.94)。

图4

新窗口打开| 下载原图ZIP| 生成PPT

图4   基于聚类的植物多样性指数(Shannon-Wiener多样性指数(H')、Simpson多样性指数(D)和波谱物种数(S))与实测值的关系。

Fig. 4   Relationship between plant diversity index based on clustering (Shannon-Wiener diversity index (H'), Simpson diversity index (D) and the number of spectral species (S)) and measured value.


2.2 基于随机森林回归估算森林植物多样性

基于已有的44个特征变量进行随机森林回归, 基于%IncMSE对输入变量的重要性程度进行排序, 重要性排序前30的变量如图5所示, 可以发现CV与CHA的重要性程度远高于其余变量。通过十折交叉验证对输入变量进行筛选, 验证误差随着特征变量数量的增加先下降再增加, 根据验证结果可知当保留6个变量时可以获得最理想的回归结果。用筛选出的特征变量构建最终的随机森林回归模型, 使用留一法对模型精度进行检验, 结果发现对Shannon- Wiener多样性指数的解释度最高(R2 = 0.79, RMSE = 0.06, p < 0.01), 对物种丰富度(R2 = 0.23, RMSE = 3.75)和Simpson多样性指数(R2 = 0.10, RMSE = 0.005)的解释度较低(图6)。

图5

新窗口打开| 下载原图ZIP| 生成PPT

图5   随机森林回归中重要性排序前30的变量。CV, 变异系数; CHA, 凸包面积; B4、B6、B8、B9分别表示Sentinel-2A的第4、6、8、9波段; 其余变量为植被指数, 具体含义见表2。

Fig. 5   Top 30 variables of importance in random forest regression. CV, coefficient of variation; CHA, convex hull area; B4, B6, B8 and B9 represent the fourth, sixth, eighth and ninth bands of Sentinel-2A, respectively; the other variables are vegetation indexes, with specific meanings shown in Table 2.


图6

新窗口打开| 下载原图ZIP| 生成PPT

图6   基于随机森林回归估算植物多样性指数(Shannon-Wiener多样性指数(H')、Simpson多样性指数(D)和物种丰富度(S))。

Fig. 6   Estimation of plant diversity index (Shannon-Wiener diversity index (H'), Simpson diversity index (D) and species richness(S)) based on random forest regression.


2.3 森林植物多样性成图

通过以上结果可以发现, 两种方法均能较好地估算Shannon-Wiener多样性指数, 精度为0.59-0.79。在分析比较之后, 采用随机森林回归的方法对研究区域的森林植物多样性进行区域成图(图7)。凉水自然保护区中间的核心区部分的主要林分类型为原始状态的阔叶红松林, 核心区的Shannon-Wiener多样性指数明显高于保护区南边的实验区以及保护区周围的缓冲区部分, 这可能是因为核心区多原始林, 人为干扰较少, 保护较好, 而保护区四周的缓冲区部分大多以次生林为主。丰林保护区的核心区和与其相连的缓冲区东部的Shannon-Wiener多样性值较高, 实验区的Shannon-Wiener多样性指数值较低, 在农田和建筑用地周围的多样性值最低。珲春自然保护区中部的实验区和南部的缓冲区部分有较多的农田、湖泊分布, 人类活动的影响使得其周围的Shannon-Wiener多样性指数值偏低, 北部的核心区和缓冲区部分多样性较高。整体来看, 丰林和凉水保护区的Shannon-Wiener多样性指数略高于珲春自然保护区, 前两者的森林类型主要是以红松为主的原始林和针阔混交林, 珲春保护区以阔叶林为主, 北部有少量针叶林和针阔混交林。

图7

新窗口打开| 下载原图ZIP| 生成PPT

图7   三个国家级自然保护区Shannon-Wiener多样性指数空间分布图。

Fig. 7   Spatial distribution of Shannon-Wiener diversity index in three national nature reserves.


3 讨论

3.1 光谱多样性估算森林植物多样性

3.1.1 像元尺度

本研究基于像元光谱反射率计算了变异系数和凸包面积两种光谱多样性指数, 结果发现两种指数均能很好地估算Shannon-Wiener多样性指数(p < 0.01)。这一结果与之前的研究(Lucas & Carter, 2008; Somers et al., 2015; Gholizadeh et al., 2018)一致, 表明CV和CHA是反映α多样性的有用指标。基于波段选择的凸包面积最早应用于草地生物多样性监测(Gholizadeh et al., 2018), 但在森林中还没有进一步的应用。此外, 本研究发现基于影像像元计算的CHA对实测的Shannon-Wiener多样性指数估算精度最高, 进一步证实了Gholizadeh的研究结论, 该研究发现随着分辨率降低, CHA优于包括CV在内的所有其他指标。一般来说, 空间分辨率的降低会产生混合像元, 进一步导致光谱异质性(Rocchini, 2007; Xi et al., 2019)。在所用影像都为中分辨率的情况下, CV的计算是用每个波段所有像元的变异系数之和除以波段数, 而CHA是先求出所有波段的变异再除以像元数。前者是整合所有波段值的空间变异, 可能反映所有物种或群落波谱变异的平均值, 而后者是每个波段空间变异的平均值, 可能反映所有物种或群落波谱变异的空间异质性, 更能体现物种或群落间的多样性。

但是, 研究发现两种光谱多样性指数均不能估算Simpson多样性指数和物种丰富度。这主要是由于遥感估算多样性是基于面上的信息, 物种所占的面积比例对影像光谱的影响更大, 而非物种的个数。Shannon-Wiener多样性指数考虑了物种的数量和每个物种的相对丰度用以衡量异质性(Pielou, 1966), 所以估算精度较高, 而Simpson多样性指数则是反映群落中少数物种的优势度(Fauvel et al., 2020)。此外, 多样性指数的估算比物种丰富度指数的估算更准确, 可能是因为丰富度仅基于物种的存在或缺失, 稀有物种的存在会影响物种丰富度, 但在影像上可能难以监测。在森林中稀有物种的冠幅占比较小, Sentinel-2A卫星的分辨率有限, 传感器很难检测到细微的单个物种变化, 而更多地指示像元内优势物种变异, 因此Shannon-Wiener多样性指数更适合区域尺度森林植物多样性的遥感估算(Oldeland et al., 2010)。

3.1.2 聚类尺度

聚类分析是一种根据数据相似性划分分组的无监督分类方法, 这种方法与训练数据收集和数据分布无关, 可以直接映射植物多样性而不单独辨别树种(Féret & Asner, 2014; Féret & de Boissieu, 2020)。聚类的方法很多, 本研究用到的是K均值聚类, 在尝试了3 × 3、5 × 5、7 × 7和9 × 9的聚类窗口后, 最终选定了聚类效果最好的5 × 5的聚类窗口。原因主要是受影像分辨率的限制, 更小的窗口导致聚类效果受影响, 无法聚类, 扩大窗口范围则不能与实地野外调查的样地大小相匹配, 估算结果误差较大。当所选的移动窗口为5 × 5时, 根据聚类结果计算得到的Shannon-Wiener多样性指数的估算结果最好(R2 = 0.58, p < 0.01), 7 × 7的移动窗口估算结果次之(R2 = 0.33, p < 0.01), 9 × 9的移动窗口效果最差(R2 = 0.26, p < 0.05)。通过聚类的方法获得的其他指数与实测值均无显著关系, 可能是因为通过聚类得到的物种数是基于光谱的波谱物种数, 并不是实际的物种数, 以此计算的结果与实测值有一定差异。

基于聚类方法的估算结果略低于基于像元光谱多样性方法, 这可能是受影像分辨率的限制。影像的分辨率对估算生物多样性有直接的影响, 分类的精度主要依赖于种间的光谱变异性(Zhang et al., 2006)。现有学者利用无人机和激光雷达数据, 将分辨率提高到单木水平, 通过单木分割并结合了植物叶片的生化组分以及结构参数等特征, 大大提高了分类精度(董文雪等, 2018; Zhao et al., 2018; 衣海燕等, 2020)。在预测试的时候本研究也尝试其他聚类方法, 自定义模糊C均值聚类可以根据聚类有效性函数自动获取聚类数目, 解决了在物种数量未知时的初始值选择的问题(Zhao et al., 2018; 衣海燕等, 2020), 可以较好地解决遥感信息的不确定性及多解性。但与前人的研究结果相比, 本研究使用该方法得到的结果精度并不高, 这也主要是受空间分辨率的限制, 不能从单木尺度进行聚类。

3.2 随机森林回归估算森林植物多样性

除了基于光谱多样性方法, 本研究也采用了机器学习的方法对森林植物多样性进行了估算, 结果发现基于所选特征指数的随机森林回归对Shannon- Wiener多样性指数有较好的反演结果(R2 = 0.79, p < 0.01), 而对其他多样性指数的估算效果较低。随机森林估算结果的准确性主要受两个方面因素的影响。一方面是受影像分辨率的影响, 影像分辨率不同对研究结果会产生一定影响。在草地生态系统中的大规模模拟研究表明, α多样性的估算精度随着空间分辨率的降低而下降, 最佳的估算精度是像元 大小与单个植株大小相接近时的精度(Wang et al., 2018)。这是因为遥感反映的是面上的信息, 利用遥感数据计算的不是物种间的光谱变异, 而是用像元代表的群落间的光谱变异间接指示物种间的变异,其中存在一定的误差, 而将分辨率提升到单木尺度会提高估算精度。另一方面受输入特征变量的影响, 本研究在44个输入变量中通过十折交叉验证对变量进行了筛选, 最后挑选了使估算结果最好的6个特征变量, 去除了变量间共线性和冗余信息的干扰。在变量挑选时发现输入的两种光谱多样性指数的重要性排序明显高于其余的特征变量, 说明这两种指数可以作为随机森林估算植物多样性的重要指标, 间接证明了前面直接利用两种指标进行植物多样性估算的可行性和合理性。但是, 重要性排序中的CV的重要性程度略高于CHA, 这可能是因为CHA与其他植被指数间存在一定相关性, 会对重要性结果产生一定影响。此外, 在森林生态系统类型中, 森林的空间分布和结构参数也会影响估算结果(Marceau et al., 1994; Gholizadeh et al., 2019), 即使采用相同分辨率的影像, 在不同地区或不同森林类型中结果可能存在一定差异(Mallinis et al., 2020; Gyamfi-Ampadu et al., 2021)。

3.3 方法的局限性

本研究目前只是在温带森林中开展, 因为气候地理条件差异, 如果要将此方法应用在不同地区的不同类型的森林中, 还需要进行进一步的验证和评估。国外相关学者基于Sentinel-2A影像, 以地中海地区的落叶阔叶林和针阔混交林为研究对象, 利用随机森林回归方法对森林植物多样性进行评估, 结果发现估算的Shannon-Wiener多样性指数精度在0.29-0.31之间, 而Simpson多样性指数的估算精度可达0.31-0.37 (Chrysafis et al., 2020; Mallinis et al., 2020)。本研究的结果与这些结论略有差异。本研究除了基于波段本身, 还加入了与植物多样性相关的指数作为随机森林的输入变量, 据此估算的Shannon-Wiener多样性指数精度较高, 但未发现光谱多样性与Simpson多样性指数有显著关系(R2 = 0.1), 这可能是受限于东北森林类型及其物种分布格局。今后的研究在应用到其他区域时, 可以在更大尺度上利用更多样方数据来进一步验证。

此外, 在所测试的两种方法中, 随机森林回归反演植物多样性的方法精度略高, 但该方法不能对超越训练数据集的范围进行估算, 而且它是一种黑箱的方法, 其内部的运行也无法控制。此外, 由于森林的复杂性和影像分辨率的限制, 目前的研究只是基于影像的光谱特征反演多样性, 而忽略了其他因素如叶片化学成分、植被结构参数、林下灌木或草本光谱的影响以及土壤反射率等的影响(Wang & Gamon, 2019), 方法上还有一定的局限性。如果要在全国范围内推广, 在后续研究中还需要考虑环境数据, 如地形、海拔、降水等环境因素的影响, 以进一步验证和完善方法, 为开展全国的森林健康评估提供较为全面可行的方法与参考。

除了考虑与植物多样性的关系, 植物功能多样性也是生物多样性的重要组成部分, 它包括植物的化学、生理和形态结构的多样性, 反映了群落、景观、甚至大空间尺度内功能性状的变异性, 决定了生态系统的功能和稳定性(Tilman et al., 1997; Ruiz-Benito et al., 2014)。利用卫星遥感影像和野外实测功能性状的关系建模, 反演植物功能多样性对实现全球生物多样性快速监测具有重要意义(Jetz et al., 2016)。但由于计算植物功能多样性的可用区域很少且不可重复测量, 传统的计算植物功能多样性的方法不适合用于长期和大规模的监测。遥感反演功能多样性的原理是基于系统发育差异和资源限制导致的植物性状不同(Schweiger et al., 2018), 这些性状主要包括叶片氮浓度、叶片碳浓度、比叶面积、叶片干物质含量和叶面积等叶片性状, 以及株高、树冠横截面积和胸径等全株性状(Ma et al., 2019), 性状上的差异进而影响光谱反射率(Wang et al., 2018)。已经有很多研究基于机载高光谱数据绘制高空间分辨率的植物功能多样性空间分布图(Asner et al., 2017; Schneider et al., 2017; Zheng et al., 2021), 但在更大的区域范围内由于获取数据难以实现且成本昂贵, 研究还较少。Sentinel-2号卫星等多光谱空间遥感的发展有利于克服高空间分辨率和监测范围的问题(Rossi et al., 2020; Hauser et al., 2021)。目前相关的研究还较少且受很多因素影响, 比如野外数据和卫星影响时间不匹配, 物种的区域差异与方法的适用性等等, 这都有待进一步补充和验证, 可以考虑作为接下来的研究方向。

4 结论

本研究依托我国东北的3个森林生态系统国家级保护区, 基于遥感数据和野外调查数据, 探讨了基于像元和聚类分析的光谱多样性直接估算和基于随机森林回归的森林植物多样性反演。结果发现两种方法均能较好地估算研究区内Shannon-Wiener指数: (1)在像元尺度, 基于凸包面积计算的光谱多样性指数估算森林植物多样性精度(R2 = 0.74)优于基于变异系数方法(R2 = 0.60); (2)基于像元的光谱多样性估算方法优于聚类分析方法(R2 = 0.59); (3)利用随机森林回归算法估算精度(R2 = 0.79)最高。Sentinel-2号卫星影像是目前可以免费获取的最高分辨率的影像, 本研究为在大尺度上进行森林多样性估算提供了可行的方法和依据, 这对保护区森林管理和自然资源保护有重要意义。

参考文献

Asner GP, Martin RE. (2009).

Airborne spectranomics: mapping canopy chemical and taxonomic diversity in tropical forests

Frontiers in Ecology and the Environment, 7, 269-276.

DOI:10.1890/070152      URL     [本文引用: 1]

Asner GP, Martin RE, Knapp DE, Tupayachi R, Anderson CB, Sinca F, Vaughn NR, Llactayo W. (2017).

Airborne laser-guided imaging spectroscopy to map forest trait diversity and guide conservation

Science, 355, 385-389.

DOI:10.1126/science.aaj1987      PMID:28126815      [本文引用: 1]

Functional biogeography may bridge a gap between field-based biodiversity information and satellite-based Earth system studies, thereby supporting conservation plans to protect more species and their contributions to ecosystem functioning. We used airborne laser-guided imaging spectroscopy with environmental modeling to derive large-scale, multivariate forest canopy functional trait maps of the Peruvian Andes-to-Amazon biodiversity hotspot. Seven mapped canopy traits revealed functional variation in a geospatial pattern explained by geology, topography, hydrology, and climate. Clustering of canopy traits yielded a map of forest beta functional diversity for land-use analysis. Up to 53% of each mapped, functionally distinct forest presents an opportunity for new conservation action. Mapping functional diversity advances our understanding of the biosphere to conserve more biodiversity in the face of land use and climate change.Copyright © 2017, American Association for the Advancement of Science.

Breiman L. (2001).

Random forests

Machine Learning, 45, 5-32.

DOI:10.1023/A:1010933404324      URL     [本文引用: 1]

Carlson KM, Asner GP, Hughes RF, Ostertag R, Martin RE. (2007).

Hyperspectral remote sensing of canopy biodiversity in Hawaiian lowland rainforests

Ecosystems, 10, 536-549.

DOI:10.1007/s10021-007-9041-z      URL     [本文引用: 1]

Ceballos G, Ehrlich PR, Barnosky AD, García A, Pringle RM, Palmer TM. (2015).

Accelerated modern human-induced species losses: entering the sixth mass extinction

Science Advances, 1, e1400253. DOI: 10.1126/sciadv.1400253.

DOI:10.1126/sciadv.1400253      [本文引用: 1]

Chang CI. (2000).

An information-theoretic approach to spectral variability, similarity, and discrimination for hyperspectral image analysis

IEEE Transactions on Information Theory, 46, 1927-1932.

DOI:10.1109/18.857802      URL     [本文引用: 1]

Chrysafis I, Korakis G, Kyriazopoulos AP, Mallinis G. (2020).

Predicting tree species diversity using geodiversity and Sentinel-2 multi-seasonal spectral information

Sustainability, 12, 9250. DOI: 10.3390/su12219250.

DOI:10.3390/su12219250      [本文引用: 1]

Clark ML, Roberts DA. (2012).

Species-level differences in hyperspectral metrics among tropical rainforest trees as determined by a tree-based classifier

Remote Sensing, 4, 1820-1855.

DOI:10.3390/rs4061820      URL     [本文引用: 1]

Cross M, Scambos T, Pacifici F, Vargas-Ramirez O, Moreno- Sanchez R, Marshall W. (2019).

Classification of tropical forest tree species using meter-scale image data

Remote Sensing, 11, 1411. DOI: 10.3390/rs11121411.

DOI:10.3390/rs11121411      [本文引用: 1]

Dahlin KM. (2016).

Spectral diversity area relationships for assessing biodiversity in a wildland-agriculture matrix

Ecological Applications, 26, 2756-2766.

[本文引用: 1]

Datt B. (1999).

Visible/near infrared reflectance and chlorophyll content in Eucalyptus leaves

International Journal of Remote Sensing, 20, 2741-2759.

DOI:10.1080/014311699211778      URL     [本文引用: 2]

Delegido J, Verrelst J, Alonso L, Moreno J. (2011).

Evaluation of Sentinel-2 red-edge bands for empirical estimation of green LAI and chlorophyll content

Sensors, 11, 7063-7081.

DOI:10.3390/s110707063      PMID:22164004      [本文引用: 1]

ESA's upcoming satellite Sentinel-2 will provide Earth images of high spatial, spectral and temporal resolution and aims to ensure continuity for Landsat and SPOT observations. In comparison to the latter sensors, Sentinel-2 incorporates three new spectral bands in the red-edge region, which are centered at 705, 740 and 783 nm. This study addresses the importance of these new bands for the retrieval and monitoring of two important biophysical parameters: green leaf area index (LAI) and chlorophyll content (Ch). With data from several ESA field campaigns over agricultural sites (SPARC, AgriSAR, CEFLES2) we have evaluated the efficacy of two empirical methods that specifically make use of the new Sentinel-2 bands. First, it was shown that LAI can be derived from a generic normalized difference index (NDI) using hyperspectral data, with 674 nm with 712 nm as best performing bands. These bands are positioned closely to the Sentinel-2 B4 (665 nm) and the new red-edge B5 (705 nm) band. The method has been applied to simulated Sentinel-2 data. The resulting green LAI map was validated against field data of various crop types, thereby spanning a LAI between 0 and 6, and yielded a RMSE of 0.6. Second, the recently developed "Normalized Area Over reflectance Curve" (NAOC), an index that derives Ch from hyperspectral data, was studied on its compatibility with simulated Sentinel-2 data. This index integrates the reflectance curve between 643 and 795 nm, thereby including the new Sentinel-2 bands in the red-edge region. We found that these new bands significantly improve the accuracy of Ch estimation. Both methods emphasize the importance of red-edge bands for operational estimation of biophysical parameters from Sentinel-2.

Deng S, Katoh M, Yu X, Hyyppa J, Gao T. (2016).

Comparison of tree species classifications at the individual tree level by combining ALS data and RGB images using different algorithms

Remote Sensing, 8, 1034. DOI: 10.3390/rs8121034.

DOI:10.3390/rs8121034      [本文引用: 1]

Dong WX, Zeng Y, Zhao YJ, Zhao D, Zheng ZJ, Yi HY. (2018).

Forest species diversity mapping using airborne LiDAR and hyperspectral data

Journal of Remote Sensing, 22, 833-847.

[本文引用: 1]

[董文雪, 曾源, 赵玉金, 赵旦, 郑朝菊, 衣海燕 (2018).

机载激光雷达及高光谱的森林乔木物种多样性遥感监测

遥感学报, 22, 833-847.]

[本文引用: 1]

Duro DC, Coops NC, Wulder MA, Han T. (2007).

Development of a large area biodiversity monitoring system driven by remote sensing

Progress in Physical Geography: Earth and Environment, 31, 235-260.

DOI:10.1177/0309133307079054      URL     [本文引用: 1]

Edwards JA, Santos-Medellin CM, Liechty ZS, Nguyen B, Lurie E, Eason S, Phillips G, Sundaresan V. (2018).

Compositional shifts in root-associated bacterial and archaeal microbiota track the plant life cycle in field-grown rice

PLOS Biology, 16, e2003862. DOI: 10.1371/journal.pbio.2003862.

DOI:10.1371/journal.pbio.2003862      [本文引用: 1]

Erinjery JJ, Singh M, Kent R. (2018).

Mapping and assessment of vegetation types in the tropical rainforests of the Western Ghats using multispectral Sentinel-2 and SAR Sentinel-1 satellite imagery

Remote Sensing of Environment, 216, 345-354.

DOI:10.1016/j.rse.2018.07.006      URL     [本文引用: 1]

Fairbanks DHK, McGwire KC. (2004).

Patterns of floristic richness in vegetation communities of California: regional scale analysis with multi-temporal NDVI

Global Ecology and Biogeography, 13, 221-235.

DOI:10.1111/j.1466-822X.2004.00092.x      URL     [本文引用: 1]

Fauvel M, Lopes M, Dubo T, Rivers-Moore J, Frison PL, Gross N, Ouin A. (2020).

Prediction of plant diversity in grasslands using Sentinel-1 and -2 satellite image time series

Remote Sensing of Environment, 237, 111536. DOI: 10.1016/j.rse.2019.111536.

DOI:10.1016/j.rse.2019.111536      [本文引用: 1]

Feeley KJ, Gillespie TW, Terborgh JW. (2005).

The utility of spectral indices from Landsat ETM+ for measuring the structure and composition of tropical dry forests

Biotropica, 37, 508-519.

DOI:10.1111/j.1744-7429.2005.00069.x      URL     [本文引用: 1]

Féret JB, Asner GP. (2014).

Mapping tropical forest canopy diversity using high-fidelity imaging spectroscopy

Ecological Applications, 24, 1289-1296.

DOI:10.1890/13-1824.1      URL     [本文引用: 2]

Féret JB, de Boissieu F (2020).

biodivMapR: an R package for α- and β-diversity mapping using remotely sensed images

Methods in Ecology and Evolution, 11, 64-70.

DOI:10.1111/2041-210X.13310      URL     [本文引用: 2]

Frampton WJ, Dash J, Watmough G, Milton EJ. (2013).

Evaluating the capabilities of Sentinel-2 for quantitative estimation of biophysical variables in vegetation

ISPRS Journal of Photogrammetry and Remote Sensing, 82, 83-92.

DOI:10.1016/j.isprsjprs.2013.04.007      URL     [本文引用: 1]

Genuer R, Poggi JM, Tuleau-Malot C. (2010).

Variable selection using Random forests

Pattern Recognition Letters, 31, 2225-2236.

DOI:10.1016/j.patrec.2010.03.014      URL     [本文引用: 1]

Gholizadeh H, Gamon JA, Helzer CJ, Cavender-Bares J. (2020).

Multi-temporal assessment of grassland α- and β-diversity using hyperspectral imaging

Ecological Applications, 30, e02145. DOI: 10.1002/eap.2145.

DOI:10.1002/eap.2145      [本文引用: 1]

Gholizadeh H, Gamon JA, Townsend PA, Zygielbaum AI, Helzer CJ, Hmimina GY, Yu R, Moore RM, Schweiger AK, Cavender-Bares J. (2019).

Detecting prairie biodiversity with airborne remote sensing

Remote Sensing of Environment, 221, 38-49.

DOI:10.1016/j.rse.2018.10.037      [本文引用: 1]

This study assessed the application of airborne imaging spectroscopy to monitor alpha-diversity in restored grassland plots. The study site was located within the Central Platte River ecosystem, south of Wood River in Central Nebraska, USA, and consisted of two sets of plots (young and old). Exotic species had recently invaded the old plots, confounding the original study design, while the young plots did not have significant invasion by weeds, reflecting the original study design. We used spectral variation (i.e. spectral diversity, expressed as the coefficient of variation) as a proxy for alpha-diversity (expressed as species richness and Shannon index). Airborne data collected at two flight altitudes and two flight directions tested the validity of "spectral diversity-alpha-diversity" relationship at different sampling scales and flight directions. Our results showed a strong relationship between spectral diversity and alpha-diversity in young, non-invaded plots exhibiting strong differences in alpha-diversity. However, in the old, invaded plots, the spectral diversity-alpha-diversity relationship was non-significant. Factors likely contributing to this failure in the old plots included the spatial mismatch between airborne and field-based sampling, the convergence in diversity levels over time, and the unique reflectance signatures of the invasive species related to their different structural and phenological properties. Unlike previous airborne studies in manipulated experimental prairie plots, but similar to results in more natural settings, the strong spectral diversity-alpha-diversity relationship in the young plots remained even at the spatial resolution of 1 m, demonstrating the potential of airborne remote sensing to assess diversity patterns in prairie grasslands. These findings demonstrate the importance of experimental remote sensing in evaluating spectral diversity, and provide insight for the development of operational airborne methods to assess biodiversity.

Gholizadeh H, Gamon JA, Zygielbaum AI, Wang R, Schweiger AK, Cavender-Bares J. (2018).

Remote sensing of biodiv improve ersity: soil correction and data dimension reduction methods assessment of α-diversity (species richness) in prairie ecosystems

Remote Sensing of Environment, 206, 240-253.

DOI:10.1016/j.rse.2017.12.014      URL     [本文引用: 6]

Gillespie TW, Foody GM, Rocchini D, Giorgi AP, Saatchi S. (2008).

Measuring and modelling biodiversity from space

Progress in Physical Geography: Earth and Environment, 32, 203-221.

DOI:10.1177/0309133308093606      URL     [本文引用: 1]

Gitelson A, Merzlyak MN. (1994).

Quantitative estimation of chlorophyll-a using reflectance spectra: experiments with autumn chestnut and maple leaves

Journal of Photochemistry and Photobiology B: Biology, 22, 247-252.

DOI:10.1016/1011-1344(93)06963-4      URL     [本文引用: 1]

Gitelson AA, Buschmann C, Lichtenthaler HK. (1999).

The chlorophyll fluorescence ratio F735/F700 as an accurate measure of the chlorophyll content in plants

Remote Sensing of Environment, 69, 296-302.

DOI:10.1016/S0034-4257(99)00023-1      URL     [本文引用: 1]

Gitelson AA, Gritz Y, Merzlyak MN. (2003).

Relationships between leaf chlorophyll content and spectral reflectance and algorithms for non-destructive chlorophyll assessment in higher plant leaves

Journal of Plant Physiology, 160, 271-282.

DOI:10.1078/0176-1617-00887      PMID:12749084      [本文引用: 1]

Leaf chlorophyll content provides valuable information about physiological status of plants. Reflectance measurement makes it possible to quickly and non-destructively assess, in situ, the chlorophyll content in leaves. Our objective was to investigate the spectral behavior of the relationship between reflectance and chlorophyll content and to develop a technique for non-destructive chlorophyll estimation in leaves with a wide range of pigment content and composition using reflectance in a few broad spectral bands. Spectral reflectance of maple, chestnut, wild vine and beech leaves in a wide range of pigment content and composition was investigated. It was shown that reciprocal reflectance (R lambda)-1 in the spectral range lambda from 520 to 550 nm and 695 to 705 nm related closely to the total chlorophyll content in leaves of all species. Subtraction of near infra-red reciprocal reflectance, (RNIR)-1, from (R lambda)-1 made index [(R lambda)(-1)-(RNIR)-1] linearly proportional to the total chlorophyll content in spectral ranges lambda from 525 to 555 nm and from 695 to 725 nm with coefficient of determination r2 > 0.94. To adjust for differences in leaf structure, the product of the latter index and NIR reflectance [(R lambda)(-1)-(RNIR)-1]*(RNIR) was used; this further increased the accuracy of the chlorophyll estimation in the range lambda from 520 to 585 nm and from 695 to 740 nm. Two independent data sets were used to validate the developed algorithms. The root mean square error of the chlorophyll prediction did not exceed 50 mumol/m2 in leaves with total chlorophyll ranged from 1 to 830 mumol/m2.

Gitelson AA, Merzlyak MN. (1997).

Remote estimation of chlorophyll content in higher plant leaves

International Journal of Remote Sensing, 18, 2691-2697.

DOI:10.1080/014311697217558      URL     [本文引用: 1]

Gitelson AA, Zur Y, Chivkunova OB, Merzlyak MN. (2002).

Assessing carotenoid content in plant leaves with reflectance spectroscopy

Photochemistry and Photobiology, 75, 272-281.

DOI:10.1562/0031-8655(2002)075<0272:accipl>2.0.co;2      PMID:11950093      [本文引用: 2]

Spectral reflectance of maple, chestnut and beech leaves in a wide range of pigment content and composition was investigated to devise a nondestructive technique for total carotenoid (Car) content estimation in higher plant leaves. Reciprocal reflectance in the range 510 to 550 nm was found to be closely related to the total pigment content in leaves. The sensitivity of reciprocal reflectance to Car content was maximal in a spectral range around 510 nm; however, chlorophylls (Chl) also affect reflectance in this spectral range. To remove the Chl effect on the reciprocal reflectance at 510 nm, a reciprocal reflectance at either 550 or 700 nm was used, which was linearly proportional to the Chl content. Indices for nondestructive estimation of Car content in leaves were devised and validated. Reflectances in three spectral bands, 510+/-5 nm, either 550+/-15 nm or 700+/-7.5 nm and the near infrared range above 750 nm are sufficient to estimate total Car content in plant leaves nondestructively with a root mean square error of less than 1.75 nmol/cm2.

Gotelli NJ, Colwell RK. (2001).

Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness

Ecology Letters, 4, 379-391.

DOI:10.1046/j.1461-0248.2001.00230.x      URL     [本文引用: 1]

Graham CH, Hijmans RJ. (2006).

A comparison of methods for mapping species ranges and species richness

Global Ecology and Biogeography, 15, 578-587.

DOI:10.1111/j.1466-8238.2006.00257.x      URL     [本文引用: 1]

Gyamfi-Ampadu E, Gebreslasie M, Mendoza-Ponce A. (2021).

Evaluating multi-sensors spectral and spatial resolutions for tree species diversity prediction

Remote Sensing, 13, 1033. DOI: 10.3390/rs13051033.

DOI:10.3390/rs13051033      [本文引用: 2]

Hardisky M, Klemas V, Smart RM. (1983).

The influence of soil-salinity, growth form, and leaf moisture on the spectral radiance of Spartina alterniflora canopies

Photogrammetric Engineering and Remote Sensing, 48, 77-84.

[本文引用: 1]

Harrison PA, Berry PM, Simpson G, Haslett JR, Blicharska M, Bucur M, Dunford R, Egoh B, Garcia-Llorente M, Geamănă N, Geertsema W, Lommelen E, Meiresonne L, Turkelboom F. (2014).

Linkages between biodiversity attributes and ecosystem services: a systematic review

Ecosystem Services, 9, 191-203.

DOI:10.1016/j.ecoser.2014.05.006      URL     [本文引用: 1]

Hauser LT, Féret JB, An Binh N, van der Windt N, Sil AF, Timmermans J, Soudzilovskaia NA, van Bodegom PM. (2021).

Towards scalable estimation of plant functional diversity from Sentinel-2: in-situ validation in a heterogeneous (semi-)natural landscape

Remote Sensing of Environment, 262, 112505. DOI: 10.1016/j.rse.2021.112505.

DOI:10.1016/j.rse.2021.112505      [本文引用: 1]

Huete AR. (1988).

A soil-adjusted vegetation index (SAVI)

Remote Sensing of Environment, 25, 295-309.

DOI:10.1016/0034-4257(88)90106-X      URL     [本文引用: 1]

Huete AR, Liu HQ, Batchily K, Vanleeuwen W. (1997).

A comparison of vegetation indices over a global set of TM images for EOS-MODIS

Remote Sensing of Environment, 59, 440-451.

DOI:10.1016/S0034-4257(96)00112-5      URL     [本文引用: 1]

Hunt Jr ER, Rock BN. (1989).

Detection of changes in leaf water content using Near- and Middle-Infrared reflectances

Remote Sensing of Environment, 30, 43-54.

DOI:10.1016/0034-4257(89)90046-1      URL     [本文引用: 1]

Jetz W, Cavender-Bares J, Pavlick R, Schimel D, Davis FW, Asner GP, Guralnick R, Kattge J, Latimer AM, Moorcroft P, Schaepman ME, Schildhauer MP, Schneider FD, Schrodt F, Stahl U, Ustin SL. (2016).

Monitoring plant functional diversity from space

Nature Plants, 2, 16024. DOI: 10.1038/NPLANTS.2016.24.

DOI:10.1038/NPLANTS.2016.24      [本文引用: 1]

Kalacska M, Sanchez-Azofeifa GA, Rivard B, Caelli T, White HP, Calvo-Alvarado JC. (2007).

Ecological fingerprinting of ecosystem succession: estimating secondary tropical dry forest structure and diversity using imaging spectroscopy

Remote Sensing of Environment, 108, 82-96.

DOI:10.1016/j.rse.2006.11.007      URL     [本文引用: 1]

Kaufman YJ, Tanre D. (1992).

Atmospherically resistant vegetation index (ARVI) for EOS-MODIS

IEEE Transactions on Geoscience and Remote Sensing, 30, 261-270.

DOI:10.1109/36.134076      URL     [本文引用: 2]

Ke YH, Quackenbush LJ, Im J. (2010).

Synergistic use of QuickBird multispectral imagery and LIDAR data for object-based forest species classification

Remote Sensing of Environment, 114, 1141-1154.

DOI:10.1016/j.rse.2010.01.002      URL     [本文引用: 1]

Kerr JT, Ostrovsky M. (2003).

From space to species: ecological applications for remote sensing

Trends in Ecology & Evolution, 18, 299-305.

DOI:10.1016/S0169-5347(03)00071-5      URL     [本文引用: 1]

Kim MS, Daughtry CST, Chappelle EW, McMurtrey JE, Walthall CL. (1994).

The use of high spectral resolution bands for estimating absorbed photosynthetically active radiation (a Par)

//Proceedings of the 6th International Symposium on Physical Measurements and Signatures in Remote Sensing. Val dʼIsère, France. 299-306.

[本文引用: 1]

Laliberté E, Schweiger AK, Legendre P. (2020).

Partitioning plant spectral diversity into alpha and beta components

Ecology Letters, 23, 370-380.

DOI:10.1111/ele.13429      PMID:31773839      [本文引用: 1]

Plant spectral diversity - how plants differentially interact with solar radiation - is an integrator of plant chemical, structural, and taxonomic diversity that can be remotely sensed. We propose to measure spectral diversity as spectral variance, which allows the partitioning of the spectral diversity of a region, called spectral gamma (γ) diversity, into additive alpha (α; within communities) and beta (β; among communities) components. Our method calculates the contributions of individual bands or spectral features to spectral γ-, β-, and α-diversity, as well as the contributions of individual plant communities to spectral diversity. We present two case studies illustrating how our approach can identify 'hotspots' of spectral α-diversity within a region, and discover spectrally unique areas that contribute strongly to β-diversity. Partitioning spectral diversity and mapping its spatial components has many applications for conservation since high local diversity and distinctiveness in composition are two key criteria used to determine the ecological value of ecosystems.© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.

Levrel H, Fontaine B, Henry PY, Jiguet F, Julliard R, Kerbiriou C, Couvet D. (2010).

Balancing state and volunteer investment in biodiversity monitoring for the implementation of CBD indicators: a French example

Ecological Economics, 69, 1580-1586.

DOI:10.1016/j.ecolecon.2010.03.001      URL     [本文引用: 1]

Lucas KL, Carter GA. (2008).

The use of hyperspectral remote sensing to assess vascular plant species richness on Horn Island, Mississippi

Remote Sensing of Environment, 112, 3908-3915.

DOI:10.1016/j.rse.2008.06.009      URL     [本文引用: 1]

Ma X, Mahecha MD, Migliavacca M, van der Plas F, Benavides R, Ratcliffe S, Kattge J, Richter R, Musavi T, Baeten L, Barnoaiea I, Bohn FJ, Bouriaud O, Bussotti F, Coppi A, et al. (2019).

Inferring plant functional diversity from space: the potential of Sentinel-2

Remote Sensing of Environment, 233, 111368. DOI: 10.1016/j.rse.2019.111368.

DOI:10.1016/j.rse.2019.111368      [本文引用: 1]

Ma Z, Li B, Li WJ, Han NY, Chen JK, Watkinson AR. (2009).

Conflicts between biodiversity conservation and development in a biosphere reserve

Journal of Applied Ecology, 46, 527-535.

DOI:10.1111/j.1365-2664.2008.01528.x      URL     [本文引用: 1]

Major DJ, Baret F, Guyot G. (1990).

A ratio vegetation index adjusted for soil brightness

International Journal of Remote Sensing, 11, 727-740.

DOI:10.1080/01431169008955053      URL     [本文引用: 1]

Mallinis G, Chrysafis I, Korakis G, Pana E, Kyriazopoulos AP. (2020).

A random forest modelling procedure for a multi-sensor assessment of tree species diversity

Remote Sensing, 12, 1210. DOI: 10.3390/rs12071210.

DOI:10.3390/rs12071210      [本文引用: 5]

Marceau DJ, Gratton DJ, Fournier RA, Fortin JP. (1994).

Remote sensing and the measurement of geographical entities in a forested environment. 2. The optimal spatial resolution

Remote Sensing of Environment, 49, 105-117.

DOI:10.1016/0034-4257(94)90047-7      URL     [本文引用: 1]

McMurtrey III JE, Chappelle EW, Kim MS, Meisinger JJ, Corp LA. (1994).

Distinguishing nitrogen fertilization levels in field corn (Zea mays L.) with actively induced fluorescence and passive reflectance measurements

Remote Sensing of Environment, 47, 36-44.

DOI:10.1016/0034-4257(94)90125-2      URL     [本文引用: 1]

Medina O, Manian V, Chinea JD. (2013).

Biodiversity assessment using hierarchical agglomerative clustering and spectral unmixing over hyperspectral images

Sensors, 13, 13949-13959.

DOI:10.3390/s131013949      PMID:24132230      [本文引用: 2]

Hyperspectral images represent an important source of information to assess ecosystem biodiversity. In particular, plant species richness is a primary indicator of biodiversity. This paper uses spectral variance to predict vegetation richness, known as Spectral Variation Hypothesis. Hierarchical agglomerative clustering is our primary tool to retrieve clusters whose Shannon entropy should reflect species richness on a given zone. However, in a high spectral mixing scenario, an additional unmixing step, just before entropy computation, is required; cluster centroids are enough for the unmixing process. Entropies computed using the proposed method correlate well with the ones calculated directly from synthetic and field data.

Merzlyak MN, Gitelson AA, Chivkunova OB, Rakitin VY. (1999).

Non-destructive optical detection of pigment changes during leaf senescence and fruit ripening

Physiologia Plantarum, 106, 135-141.

DOI:10.1034/j.1399-3054.1999.106119.x      URL     [本文引用: 1]

Miao X, Heaton JS, Zheng S, Charlet DA, Liu H. (2012).

Applying tree-based ensemble algorithms to the classification of ecological zones using multi-temporal multi- source remote-sensing data

International Journal of Remote Sensing, 33, 1823-1849.

DOI:10.1080/01431161.2011.602651      URL     [本文引用: 1]

Nagendra H. (2001).

Using remote sensing to assess biodiversity

International Journal of Remote Sensing, 22, 2377-2400.

DOI:10.1080/01431160117096      URL     [本文引用: 1]

Nagendra H, Rocchini D. (2008).

High resolution satellite imagery for tropical biodiversity studies: the devil is in the detail

Biodiversity and Conservation, 17, 3431-3442.

DOI:10.1007/s10531-008-9479-0      URL     [本文引用: 1]

Nicholson E, Mace GM, Armsworth PR, Atkinson G, Buckle S, Clements T, Ewers RM, Fa JE, Gardner TA, Gibbons J, Grenyer R, Metcalfe R, Mourato S, Muûls M, Osborn D, et al. (2009).

Priority research areas for ecosystem services in a changing world

Journal of Applied Ecology, 46, 1139-1144.

DOI:10.1111/j.1365-2664.2009.01716.x      URL     [本文引用: 1]

Oldeland J, Wesuls D, Rocchini D, Schmidt M, Jürgens N. (2010).

Does using species abundance data improve estimates of species diversity from remotely sensed spectral heterogeneity?

Ecological Indicators, 10, 390-396.

DOI:10.1016/j.ecolind.2009.07.012      URL     [本文引用: 2]

Palmer MW, Earls PG, Hoagland BW, White PS, Wohlgemuth T. (2002).

Quantitative tools for perfecting species lists

Environmetrics, 13, 121-137.

DOI:10.1002/env.516      URL     [本文引用: 1]

Pan Y, Birdsey RA, Phillips OL, Jackson RB. (2013).

The structure, distribution, and biomass of the worldʼs forests

Annual Review of Ecology, Evolution, and Systematics, 44, 593-622.

DOI:10.1146/annurev-ecolsys-110512-135914      URL     [本文引用: 1]

Pereira HM, Belnap J, Brummitt N, Collen B, Ding H, Gonzalez- Espinosa M, Gregory RD, Honrado J, Jongman RHG, Julliard R, McRae L, Proenca V, Rodrigues P, Opige M, Rodriguez JP, et al. (2010).

Global biodiversity monitoring

Frontiers in Ecology and the Environment, 8, 459-460.

DOI:10.1890/10.WB.23      URL     [本文引用: 1]

Pielou EC. (1966).

The measurement of diversity in different types of biological collections

Journal of Theoretical Biology, 13, 131-144.

DOI:10.1016/0022-5193(66)90013-0      URL     [本文引用: 2]

Rocchini D. (2007).

Effects of spatial and spectral resolution in estimating ecosystem α-diversity by satellite imagery

Remote Sensing of Environment, 111, 423-434.

DOI:10.1016/j.rse.2007.03.018      URL     [本文引用: 1]

Rocchini D, Balkenhol N, Carter GA, Foody GM, Gillespie TW, He KS, Kark S, Levin N, Lucas K, Luoto M, Nagendra H, Oldeland J, Ricotta C, Southworth J, Neteler M. (2010).

Remotely sensed spectral heterogeneity as a proxy of species diversity: recent advances and open challenges

Ecological Informatics, 5, 318-329.

DOI:10.1016/j.ecoinf.2010.06.001      URL     [本文引用: 1]

Rondeaux G, Steven M, Baret F. (1996).

Optimization of soil- adjusted vegetation indices

Remote Sensing of Environment, 55, 95-107.

DOI:10.1016/0034-4257(95)00186-7      URL     [本文引用: 1]

Rossi C, Kneubüler M, Schütz M, Schaepman ME, Haller RM, Risch AC. (2020).

From local to regional: functional diversity in differently managed alpine grasslands

Remote Sensing of Environment, 236, 111415. DOI: 10.1016/j.rse.2019.111415.

DOI:10.1016/j.rse.2019.111415      [本文引用: 1]

Rozenstein O, Haymann N, Kaplan G, Tanny J. (2019).

Validation of the cotton crop coefficient estimation model based on Sentinel-2 imagery and eddy covariance measurements

Agricultural Water Management, 223, 105715. DOI: 10.1016/j.agwat.2019.105715.

DOI:10.1016/j.agwat.2019.105715      [本文引用: 6]

Ruiz-Benito P, Gómez-Aparicio L, Paquette A, Messier C, Kattge J, Zavala MA. (2014).

Diversity increases carbon storage and tree productivity in Spanish forests

Global Ecology and Biogeography, 23, 311-322.

DOI:10.1111/geb.12126      URL     [本文引用: 1]

Schäfer E, Heiskanen J, Heikinheimo V, Pellikka P. (2016).

Mapping tree species diversity of a tropical montane forest by unsupervised clustering of airborne imaging spectroscopy data

Ecological Indicators, 64, 49-58.

DOI:10.1016/j.ecolind.2015.12.026      URL     [本文引用: 1]

Schneider FD, Morsdorf F, Schmid B, Petchey OL, Hueni A, Schimel DS, Schaepman ME. (2017).

Mapping functional diversity from remotely sensed morphological and physiological forest traits

Nature Communications, 8, 1441. DOI: 10.1038/s41467-017-01530-3.

DOI:10.1038/s41467-017-01530-3      [本文引用: 1]

Schweiger AK, Cavender-Bares J, Townsend PA, Hobbie SE, Madritch MD, Wang R, Tilman D, Gamon JA. (2018).

Plant spectral diversity integrates functional and phylogenetic components of biodiversity and predicts ecosystem function

Nature Ecology & Evolution, 2, 976-982.

[本文引用: 1]

Sims DA, Gamon JA. (2002).

Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages

Remote Sensing of Environment, 81, 337-354.

DOI:10.1016/S0034-4257(02)00010-X      URL     [本文引用: 1]

Somers B, Asner GP, Martin RE, Anderson CB, Knapp DE, Wright SJ, van de Kerchove R (2015).

Mesoscale assessment of changes in tropical tree species richness across a bioclimatic gradient in Panama using airborne imaging spectroscopy

Remote Sensing of Environment, 167, 111-120.

DOI:10.1016/j.rse.2015.04.016      URL     [本文引用: 2]

Tilman D, Knops J, Wedin D, Reich P, Ritchie M, Siemann E. (1997).

The influence of functional diversity and composition on ecosystem processes

Science, 277, 1300-1302.

DOI:10.1126/science.277.5330.1300      URL     [本文引用: 1]

Torresani M, Rocchini D, Sonnenschein R, Zebisch M, Marcantonio M, Ricotta C, Tonon G. (2019).

Estimating tree species diversity from space in an alpine conifer forest: the Raoʼs Q diversity index meets the spectral variation hypothesis

Ecological Informatics, 52, 26-34.

DOI:10.1016/j.ecoinf.2019.04.001      [本文引用: 1]

Forests cover about 30% of the Earth surface, they are among the most biodiverse terrestrial ecosystems and represent the bulk of many ecological processes and services. The assessment of biodiversity is an important and essential goal to achieve but it can results difficult, time consuming and expensive when based on field data. Remote sensing covers large areas and provides consistent quality and standardized data, which can be used to estimate species diversity. One method to estimate species diversity from remote sensing data is based on the Spectral Variation Hypothesis (SVH), which assumes that the higher the spectral variation of an image, the higher the environmental heterogeneity and the species diversity of the considered area. SVH has been tested using different spectral heterogeneity (SH) indices and measures, recently the Rao's Q index has been proposed as a new spectral variation measure to be applied to remote sensing data. In this paper, we tested the SVH in an alpine coniferous forest to estimate tree species diversity. We evaluated the performance of the Rao's Q index and compared it with another widely used SH index, the Coefficient of Variation (CV), validating them against values of Shannon's H (used as species diversity index) derived from in-situ collected data. A NDVI time-series (for 2016 and 2017) derived from the Sentinel-2A and 2B and Landsat 8 OLI satellites has been used to test the effect of the spatial grain of both the sensors and to understand the seasonality of the SVH. The results showed that the SVH is season and sensor dependent. For both years and satellites, the relation between Rao's Q and field data reached the highest R-2 between June and July, decreasing towards winter and spring similarly to the NDVI time-series. This relationship could be given because, when NDVI reaches its highest values, it is able to capture small variation in reflectance of different leaf traits typical of specific trees. The relation between field and spectral diversity reached a value of R-2 = 0.70 (2017) and R-2 = 0.48 (2016) for Sentinel-2 and of R-2 = 0.42 (2017) and R-2 = 0.47 (2016) for Landsat 8. CV showed the same NDVI temporal tendency. However, the relation between field-derived Shannon's H and CV was on average lower than that we found for Rao's Q. This research underlined the goodness of the Rao's Q index, the relevance of the NDVI in the study of the SVH and the importance of the multi-temporal approach.

Turner W. (2014).

Sensing biodiversity

Science, 346, 301-302.

DOI:10.1126/science.1256014      PMID:25324372      [本文引用: 1]

Turner W, Spector S, Gardiner N, Fladeland M, Sterling E, Steininger M. (2003).

Remote sensing for biodiversity science and conservation

Trends in Ecology & Evolution, 18, 306-314.

DOI:10.1016/S0169-5347(03)00070-3      URL     [本文引用: 1]

Ustin SL, Gamon JA. (2010).

Remote sensing of plant functional types

New Phytologist, 186, 795-816.

DOI:10.1111/j.1469-8137.2010.03284.x      PMID:20569415      [本文引用: 2]

Conceptually, plant functional types represent a classification scheme between species and broad vegetation types. Historically, these were based on physiological, structural and/or phenological properties, whereas recently, they have reflected plant responses to resources or environmental conditions. Often, an underlying assumption, based on an economic analogy, is that the functional role of vegetation can be identified by linked sets of morphological and physiological traits constrained by resources, based on the hypothesis of functional convergence. Using these concepts, ecologists have defined a variety of functional traits that are often context dependent, and the diversity of proposed traits demonstrates the lack of agreement on universal categories. Historically, remotely sensed data have been interpreted in ways that parallel these observations, often focused on the categorization of vegetation into discrete types, often dependent on the sampling scale. At the same time, current thinking in both ecology and remote sensing has moved towards viewing vegetation as a continuum rather than as discrete classes. The capabilities of new remote sensing instruments have led us to propose a new concept of optically distinguishable functional types ('optical types') as a unique way to address the scale dependence of this problem. This would ensure more direct relationships between ecological information and remote sensing observations.

Wang R, Gamon JA. (2019).

Remote sensing of terrestrial plant biodiversity

Remote Sensing of Environment, 231, 111128. DOI: 10.1016/j.rse.2019.111218.

DOI:10.1016/j.rse.2019.111218      [本文引用: 1]

Wang R, Gamon JA, Emmerton CA, Li H, Nestola E, Pastorello GZ, Menzer O. (2016).

Integrated analysis of productivity and biodiversity in a southern Alberta prairie

Remote Sensing, 8, 214. DOI: 10.3390/rs8030214.

DOI:10.3390/rs8030214      [本文引用: 3]

Wang R, Gamon JA, Schweiger AK, Cavender-Bares J, Townsend PA, Zygielbaum AI, Kothari S. (2018).

Influence of species richness, evenness, and composition on optical diversity: a simulation study

Remote Sensing of Environment, 211, 218-228.

DOI:10.1016/j.rse.2018.04.010      URL     [本文引用: 3]

Waring RH, Coops NC, Fan W, Nightingale JM. (2006).

MODIS enhanced vegetation index predicts tree species richness across forested ecoregions in the contiguous USA.

Remote Sensing of Environment, 103, 218-226.

DOI:10.1016/j.rse.2006.05.007      URL     [本文引用: 1]

Wu CY, Niu Z, Tang Q, Huang WJ. (2008).

Estimating chlorophyll content from hyperspectral vegetation indices: modeling and validation

Agricultural and Forest Meteorology, 148, 1230-1241.

DOI:10.1016/j.agrformet.2008.03.005      URL     [本文引用: 1]

Xi YB, Ren CY, Wang ZM, Wei SQ, Bai JL, Zhang B, Xiang HX, Chen L. (2019).

Mapping tree species composition using OHS-1 hyperspectral data and deep learning algorithms in Changbai Mountains, northeast China

Forests, 10, 818. DOI: 10.3390/f10090818.

DOI:10.3390/f10090818      [本文引用: 1]

Xu W, Li X, Pimm SL, Hull V, Zhang J, Zhang L, Xiao Y, Zheng H, Ouyang Z. (2016).

The effectiveness of the zoning of Chinaʼs protected areas

Biological Conservation, 204, 231-236.

DOI:10.1016/j.biocon.2016.10.028      URL     [本文引用: 1]

Yi HY, Zeng Y, Zhao YJ, Zheng ZJ, Xiong J, Zhao D. (2020).

Forest species diversity mapping based on clustering algorithm

Chinese Journal of Plant Ecology, 44, 598-615.

DOI:10.17521/cjpe.2019.0347      URL     [本文引用: 2]

[衣海燕, 曾源, 赵玉金, 郑朝菊, 熊杰, 赵旦 (2020).

利用聚类算法监测森林乔木物种多样性

植物生态学报, 44, 598-615.]

DOI:10.17521/cjpe.2019.0347      [本文引用: 2]

本研究基于机载LiDAR和高光谱数据, 从森林物种叶片的生理化学源头探寻生化特征与光谱特征的内在关联, 探讨生化多样性、光谱多样性与物种多样性之间的响应机制, 选择最优植被指数并结合最优结构参数, 通过聚类方法构建森林物种多样性遥感估算模型, 在古田山自然保护区开展森林乔木物种多样性监测。研究结果表明: (1)从16种叶片生化组分中, 筛选出叶绿素a、叶绿素b、类胡萝卜素、叶片含水量、比叶面积、纤维素、木质素、氮、磷和碳可通过偏最小二乘法用叶片光谱有效模拟(R2 = 0.60 – 0.79, p &lt; 0.01), 并选择有效的植被指数TCARI/OSAVI、CRI、WBI、RVI、PRI和CCCI表征相应的最优生化组分; (2)基于机载LiDAR数据利用结合形态学冠层控制的分水岭算法可获得高精度单木分离结果(R2 = 0.77, RMSE = 16.48), 同时采用逐步回归方法从常用的森林结构参数中选取了树高和偏度作为最优结构参数(R2 = 0.32, p &lt; 0.01); (3)基于6个最优植被指数和2个最优结构参数, 以20 m × 20 m为窗口通过自适应模糊C均值方法进行聚类, 实现了研究区森林乔木物种丰富度(Richness, R2 = 0.56, RMSE = 1.81)和多样性指数Shannon-Wiener (R2 = 0.83, RMSE = 0.22)与Simpson (R2 = 0.85, RMSE = 0.09)的成图。本研究在冠层尺度上获取了与物种多样性相关的生化、光谱和结构参数, 将单木个体作为最小单元, 利用聚类算法直接估算物种类别差异, 无需判定具体的树种属性, 是利用遥感数据进行区域尺度森林物种多样性监测与成图的实践, 可为亚热带地区常绿阔叶林的森林物种多样性监测提供借鉴。

Zarco-Tejada PJ, Pushnik JC, Dobrowski S, Ustin SL. (2003).

Steady-state chlorophyll a fluorescence detection from canopy derivative reflectance and double-peak red-edge effects

Remote Sensing of Environment, 84, 283-294.

DOI:10.1016/S0034-4257(02)00113-X      URL     [本文引用: 1]

Zhang J, Rivard B, Sánchez-Azofeifa A, Castro-Esau K. (2006).

Intra- and inter-class spectral variability of tropical tree species at La Selva, Costa Rica: implications for species identification using HYDICE imagery

Remote Sensing of Environment, 105, 129-141.

DOI:10.1016/j.rse.2006.06.010      URL     [本文引用: 2]

Zhao YJ, Zeng Y, Zheng ZJ, Dong WX, Zhao D, Wu BF, Zhao QJ. (2018).

Forest species diversity mapping using airborne LiDAR and hyperspectral data in a subtropical forest in China

Remote Sensing of Environment, 213, 104-114.

DOI:10.1016/j.rse.2018.05.014      URL     [本文引用: 3]

Zheng Z, Zeng Y, Schneider FD, Zhao Y, Zhao D, Schmid D, Schaepman ME, Morsdorf F. (2021).

Mapping functional diversity using individual tree-based morphological and physiological traits in a subtropical forest

Remote Sensing of Environment, 252, 112170. DOI: 10.1016/j.rse.2020.112170.

DOI:10.1016/j.rse.2020.112170      [本文引用: 1]

/

〈 〉

Copyright © 2022 版权所有 《植物生态学报》编辑部
地址: 北京香山南辛村20号, 邮编: 100093
Tel.: 010-62836134, 62836138;　Fax: 010-82599431;　E-mail: apes@ibcas.ac.cn, cjpe@ibcas.ac.cn
备案号: 京ICP备16067583号-19