易秋香^,1,2,3,*, 刘英^1,2,3, 常存^1,2,3, 钟瑞森^1,2,31中国科学院新疆生态与地理研究所/荒漠与绿洲生态国家重点实验室, 新疆乌鲁木齐 830011
²新疆维吾尔自治区遥感与地理信息系统应用重点实验室, 新疆乌鲁木齐 830011
³中国科学院大学, 北京 100049

Estimation of cotton Car/Chla ratio by hyperspectral vegetation indices and partial least square regression

YI Qiu-Xiang^,1,2,3,*, LIU Ying^1,2,3, CHANG Cun^1,2,3, ZHONG Rui-Sen^{1,2,3 1}State Key Laboratory of Desert and Oasis Ecology/Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, Xinjiang, China
²Key Laboratory of GIS & RS Application Xinjiang Uygur Autonomous Region, Urumqi 830011, Xinjiang, China
³University of Chinese Academy of Sciences, Beijing 100049, China

通讯作者: * 易秋香, E-mail: yiqx@ms.xjb.ac.cn

收稿日期:2019-10-22接受日期:2020-03-24网络出版日期:2020-08-12

基金资助:

国家自然科学基金项目.41571428 国家自然科学基金项目.41871328

Received:2019-10-22Accepted:2020-03-24Online:2020-08-12

Fund supported:

National Natural Science Foundation of China.41571428 National Natural Science Foundation of China.41871328

摘要
类胡萝卜素(Car)与叶绿素a含量比值(Car/Chla)的变化与植被生长发育变化、环境胁迫及叶片衰老特征等密切相关, 可作为植被生理生态及物候的监测指标。不同植被类型和植被品种其色素变化随植被生长发育呈现出不同的变化特征。为了探究适用于干旱区棉花Car/Chla比值估算的光谱指数和估算方法, 本研究通过2011年和2012年连续2年的大面积田间试验, 获取了棉花不同生育期的叶片及冠层尺度光谱反射率及色素含量信息, 对多种光谱指数及偏最小二乘回归(Partial Least Square Regression, PLSR)用于Car/Chla比值和Car估算进行了探讨。对比表明, 基于光化学指数(Photochemical Reflectance Index, PRI)的线性和一元二次模型对Car/Chla比值和Car的估算精度最高, 由PRI-Car/Chla线性模型得到的叶片和冠层尺度的Car/Chla比值估算值与实测值之间的决定系数R²大于0.6, PRI-Car的R²大于0.36; 基于PLSR模型得到的Car/Chla比值估算值与实测值之间的拟合关系略优于基于PRI的估算模型, 由其得到的叶片及冠层尺度Car/Chla比值估算值与实测值之间的决定系数R²大于0.80, Car估算值与实测值之间R²大于0.73; 不论基于PRI还是基于PLSR方法, 对Car/Chla比值的估算精度均高于Car含量, 该结论进一步证实了Car/Chla比值遥感监测的可行性, 丰富了对棉花生长高温胁迫、养分胁迫等环境胁迫及病虫害等遥感监测的依据指标。
关键词： 类胡萝卜素/叶绿素a比值;类胡萝卜素;光化学指数;偏最小二乘回归;棉花

Abstract
Estimating the ratio between carotenoid to chlorophyll a (Car/Chla) provides an additional avenue for the assessment of physiology and phenology of plant growth and development. With the aim of assessing cotton Car/Chla ratio from hyperspectral reflectance, a wide range of carotenoid (Car) and chlorophyll a concentrations, and leaf and canopy reflectance at cotton different growth stages were measured. The performance of a variety of Car/Chla ratio related vegetation indices and partial least square regression (PLSR) for Car/Chla ratio and Car estimation were tested. Among all tested vegetation indices, PRI (Photochemical Reflectance Index) and linear PRI models had the most significant correlations with Car/Chla ratio and Car, and could accurately estimate, Car/Chla ratio (R²_{leaf level} = 0.69 and R²_{canopy level} = 0.67) and Car concentration (R²_{leaf level} = 0.44 and R²_{canopy level} = 0.36). The best estimation of the Car/Chla ratio and Car was provided by PLSR models with R² > 0.80 between the estimated and measured value for Car/Chla ratio and R²= 0.74 for Car. Both reflectance indices and PLSR method were more successful for the estimation of Car/Chla ratio than for that of Car concentration, indicating the promising potential of Car/Chla ratio as a powerful indicator using for plant status monitoring by remote sensing. Besides, accuracy test of models using validation dataset highlighted the remarkable performance of PLSR for Car/Chla (R²_{leaf level} = 0.87 and R²_{canopy level} = 0.84) and Car (R²_{leaf level} = 0.73 and R²_{canopy level} = 0.74) estimated by hyperspectral reflectance at both the leaf and canopy levels. The results further prove the remarkable performance of hyperspectral reflectance for the estimation of Car/Chla ratio, and enrich the parameters for monitoring high temperature stress, water deficit stress, and nutrient stress and pest diseases by remote sensing in cotton.
Keywords：Car/Chla ratio;carotenoid;PRI (Photochemical Reflectance Index);PLSR (Partial Least Square Regression);cotton


PDF (541KB)元数据多维度评价相关文章导出EndNote|Ris|Bibtex收藏本文
本文引用格式
易秋香, 刘英, 常存, 钟瑞森. 基于光谱指数和偏最小二乘的棉花类胡萝卜素/叶绿素a比值估算[J]. 作物学报, 2020, 46(8): 1266-1274. doi:10.3724/SP.J.1006.2020.94157
YI Qiu-Xiang, LIU Ying, CHANG Cun, ZHONG Rui-Sen. Estimation of cotton Car/Chla ratio by hyperspectral vegetation indices and partial least square regression[J]. Acta Agronomica Sinica, 2020, 46(8): 1266-1274. doi:10.3724/SP.J.1006.2020.94157


植被色素变化与植被胁迫、光合作用等密切相关。叶绿素(chlorophyll, Chl)及类胡萝卜素(carotenoid, Car)被认为是反映植被生理状况的良好指示器^[1]。随着遥感技术的发展, 植被色素含量诊断得到了进一步的发展, 其中, 叶绿素含量的监测已有很多研究, 并且建立了各类叶绿素含量估算模型^[2,3,4,5,6]。类胡萝卜素作为叶绿体中的第二大色素, 和叶绿素一样, 具有吸收、传递光能及光保护功能, 并且它可反映植被受胁迫情况^[7]。由于叶片中类胡萝卜素含量明显低于叶绿素含量, 并且两者显著相关^[8], 用于类胡萝卜素监测的敏感波段也同时受叶绿素含量影响, 这导致类胡萝卜素含量的估算较难^[9,10]。

对于大部分植被, 正是由于各类色素含量及其占比的变化, 导致叶片颜色变化^[11,12,13]。当植被受到高温胁迫、养分胁迫或在叶片老化过程中, 叶绿素和类胡萝卜素含量都会随之老化而降低, 但相比叶绿素含量的快速减少, 类胡萝卜素含量变化相对更稳定^[12,14-16], 这也就意味着类胡萝卜素与叶绿素含量比值可反映植被生长环境胁迫及叶片的老化特征, 如Sanger^[17]证实在衰老的叶片中类胡萝卜素要比叶绿素a更持久地存在; Pe?uelas等^[18,19]发现类胡萝卜素与叶绿素a (Chla)比值(Car/Chla)在老化和不健康的植被叶片中会增加, 而在健康的叶片中会降低。相比于利用单一叶绿素或类胡萝卜素含量进行植被生理状况监测, Car/Chl比值是一个无量纲值, 能够消除植被类型及品种之间类胡萝卜素和叶绿素含量本身的差异, 使得不同植被类型之间的色素变化情况更具有可比性, 并且, 相比单一叶绿素含量变化或类胡萝卜素含量变化能更灵敏地指示植被生理变化或衰老特征^[20], 可为植被生理生态状况监测提供更有效的监测指标。

目前对于类胡萝卜素与叶绿素含量比值的遥感监测已有部分研究。最早Pe?uelas等^[18]分析了光谱指数用于Car/Chla监测的可能性, 提出归一化色素指数(Normalized Difference Pigment Index, NDPI)与向日葵及水生植物的Car/Chla比值具有显著相关性; 随后Pe?uelas等^[21]又进一步获取了多种植被包括玉米、小麦、番茄、大豆、向日葵、甜菜、橡树、枫树以及肉质植物等的叶片光谱和色素信息, 探讨了

叶片光谱用于Car/Chla估算的最佳波段, 并构建了结构不敏感色素指数 (Structure Insensitive Pigment Index, SIPI), 认为该指数可削弱叶片表面及叶肉结构对Car/Chla估算的影响, 同时证实该指数与Car/Chla比值之间呈非线性相关, 当Car/Chla比值较小时(小于1), SIPI敏感性较低, 当Car/Chla比值大于10时, SIPI达到饱和; Merzlyak等^[16]发现多种植被(包括枫树、栗树、土豆以及锦紫苏)叶片老化及果实(包括柠檬和苹果)成熟过程中, 500 nm和680 nm附近的叶片光谱的差值(R₆₇₈-R₅₀₀)对叶片老化引起的色素含量变化最为敏感, 由此构建了用于类胡萝卜素与叶绿素比值估算的植被衰老指数(Plant Senescing Reflectance Index, PSRI), 通过PSRI与R₆₇₈/R₇₅₀之间的关系判定叶片老化和果实成熟阶段, 但不同植被在叶片老化和果实成熟过程中所展示的PSRI变化有所不同; Nakaji等^[22]和Garrity等^[23]分别证实日本落叶松及橡树和枫树的Car/Chl比值与光化学指数(Photochemical Reflectance Index, PRI)具有极显著相关性。此外, 也有研究利用辐射传输模型证实了光谱指数与Car/Chla比值之间的显著相关, 如Garrity等^[23]利用PROSPECT-5模型证实了PRI与Car/Chl之间显著相关, 决定系数R²=0.83, 但通过实测获取的两种针叶林植被的PRI和Car/Chl比值之间的判定系数仅为0.22, 研究者认为这与获取的叶片样本受干旱胁迫有关; Hernández-Clemente等^[24]的研究表明, PRI与叶绿素或类胡萝卜素之间的关系取决于PRI和Car/Chl比值之间的关系, 其利用PROSEPCT+DART模型模拟的针叶林PRI光谱指数与类胡萝卜素与叶绿素比值相关系数达到0.9。国内, 周贤峰^[20]利用实测的叶片和冠层光谱评估了一些常用于类胡萝卜素、叶绿素及两者比值估算的光谱指数, 用于玉米和冬小麦的Car/Chl比值估算的精度, 并构建了类胡萝卜素与叶绿素比值植被指数(Combined Carotenoid/Chlorophyll Ratio Index, CCRI)^[25], 也证实不同光谱指数对不同作物的Car/Chl比值估算精度与作物Car/Chl比值分布有关, 其中SIPI、PSRI以及PRI对玉米Car/Chl比值估算的精度要优于对冬小麦Car/Chl比值的估算, 主要因为玉米Car/Chl比值分布范围更广泛, 包含更大的变化。

由上可见, 已有研究中关于类胡萝卜素和叶绿素比值遥感监测所提出的光谱指数多针对林木植被展开, 并且都是基于研究对象的经验统计模型, 这类模型的估算精度受植被类型、植被生长环境、植被色素含量变化特征及植被本身光谱特征等多种因素影响, 无法直接应用于不同研究区域的不同作物, 同时对于作物Car/Chla比值遥感监测的研究目前还较少, 有待进一步探讨。本研究分析各类光谱指数用于棉花类胡萝卜素与叶绿素比值及类胡萝卜素的估算精度, 并与偏最小二乘方法对比, 旨在明确其估算的可行性, 丰富棉花生理生态遥感监测指标。

1 材料与方法

1.1 观测试验

于新疆维吾尔自治区北部石河子垦区148团、149团及150团(84°58'—86°24'E, 43°26'—45°20'N), 依据棉田连片面积, 2011及2012年分别选取4个种植面积大于50 hm²的地块作为试验样点布设田间观测点。在每个样点布设6~8个不等的观测点。分别于2011年6月9日(苗期)、7月14日(开花初期)、8月17日(花铃期)及2012年6月19日(现蕾期)、7月18日及(开花盛期) 8月1日(花铃期)试验观测。

1.2 数据获取

1.2.1 光谱测定 使用美国ASD (Analytical Spectral Device)公司的ASDFieldSpec Pro FRTM光谱仪测定光谱, 其波段值为350~2500 nm, 其中350~1000 nm光谱采样间隔为1.4 nm, 光谱分辨率为3 nm; 1001~2500 nm光谱采样间隔为2 nm, 光谱分辨率为10 nm。冠层光谱选择在无风无云或少云的天气测定, 测定时间为12:00-16:00。测量之前先开启光谱仪预热, 再次测量之前都先进行光谱仪的优化, 光谱测定前进行参考板反射率测定, 测量时探头垂直向下, 距棉花冠层垂直高度约0.7 m, 25°视场角。叶片光谱选取健康叶片利用叶片夹连接ASD光谱仪测定。冠层光谱和叶片光谱采样均以10条光谱为一采样光谱, 即每次记录10条光谱, 以其平均值作为该样点的光谱反射值。

1.2.2 色素含量测定 对应叶片光谱测定的位置, 剪取约0.2 g叶片置试管中, 加萃取液(丙酮:无水乙醇:蒸馏水=4.5:4.5:1.0)(v/v) 25 mL, 于室温下遮光静置至样品完全发白。然后采用722 s分光光度计分别测定663、645和440 nm处的光密度值(optical density, OD), 基于OD值计算叶绿素含量, 参照蒋德安等^[26]方法。

1.3 植被指数

考虑到Car/Chla比值与类胡萝卜素含量直接相关, 除已有的Car/Chla比值指数外, 还分析了各类研究中提出的类胡萝卜素光谱植被指数。Hernandez- Clemente等^[24]的研究中对用于类胡萝卜素及类胡萝卜素与叶绿素比值估算的植被指数进行了总结, 并构建了简单植被指数(R₅₁₅/R₅₇₀), 本文在利用其总结的植被指数外, 加入了Zhou等^[25]提出的CCRI, 共计19个植被指数。在对这些植被指数与棉花Car/ Chla比值及类胡萝卜素进行相关性分析的基础上, 筛选出在叶片及冠层尺度与Car/Chla比值及类胡萝卜素均达到极显著相关的植被指数进行进一步模型构建分析。采用的植被指数如表1。

Table 1
表1
表1所采用的光谱指数
Table 1Vegetation indices used for Car/Chla ratio and carotenoids estimation

指数 Index	全称 Full name	表达式 Formula	文献 Reference
PRI	Photochemical Reflectance Index	(R531-R570)/(R531+R570)	[27]
PRI*CI	Carotenoid/Chlorophyll Ratio Index	(R531-R570)/(R531+R570)*((R760/R700)-1))	[23]
PSRI	Plant senescence reflectance index	(R678-R500)/R750	[16]
R515/R570	Simple ratio vegetation index	R515/R570	[24]
CCRI	Combined carotenoid/chlorophyll ratio index	(R720-R521)*R705/(R750-R705)*R521	[25]

新窗口打开|下载CSV

1.4 偏最小二乘回归

偏最小二乘回归(partial least square regression, PLSR)是用于光谱数据分析的强有力方法, 该方法已被应用于各类研究^[28,29]。对于该方法的具体介绍可参考Geladi等^[30]和Wold等^[31]。本研究中, PLSR用于构建棉花叶片及冠层光谱反射率与Car/Chla比值及类胡萝卜素之间的线性模型, 采用波段400~ 1000 nm共计601个光谱反射率作为输入变量, 利用留一验证法(leave-one-out cross-validation, LOOCV)^[32]通过最小预测残差平方和(predicted residual sums of squares, PRESS)^[33]确定各估算模型的最佳潜变量(latent variables)个数。PLSR模型构建通过Matlab代码编程实现。

2 结果与分析

2.1 类胡萝卜素与叶绿素a比值(Car/Chla)及类胡萝卜素(Car)统计特征

将叶片和冠层尺度的Car/Chla比值及Car的实测数据随机分为检验样本和验证样本, 其中总数据的三分之二用于建模, 剩余的三分之一用于模型估算精度验证。建模样本和检验样本的Car/Chla比值及Car统计特征见表2。由表2可知, Car/Chla比值及Car的取值分布广泛, Car/Chla比值的最大最小值分别是0.109和0.495, Car含量最小3.97 μg cm^-2, 最大23.06 μg cm^-2。建模样本和检验样本统计特征相似, 说明样本随机分组合理。

Table 2
表2
表2用于模型构建和验证的Car/Chla比值以及Car的统计特征
Table 2Statistical information of the Car/Chla ratio and Car for calibration and validation datasets

尺度 Level	数据集 Dataset	样本数n	Car/Chla比值Car/Chla ratio				类胡萝卜素Car (μg cm-2)
			平均值Mean	取值范围 Range	标准偏差SD		平均值Mean	取值范围 Range	标准偏差SD
叶片尺度 Leaf level	总体All	141	0.286	0.109-0.495	0.081		13.49	3.97-23.06	3.99
	验证样本Calibration	94	0.283	0.109-0.478	0.081		13.08	5.76-22.83	3.72
	检验样本Validation	47	0.292	0.118-0.495	0.083		14.32	3.97-23.06	4.40
冠层尺度Canopy level	总体All	159	0.302	0.109-0.495	0.082		14.15	3.97-23.06	4.16
	验证样本Calibration	104	0.300	0.109-0.440	0.082		13.88	5.76-20.93	3.91
	检验样本Validation	55	0.306	0.118-0.495	0.082		14.55	3.97-23.06	4.51

新窗口打开|下载CSV

2.2 基于植被指数的估算模型构建

已有的研究证实Car/Chla比值及Car和光谱指数之间通常呈线性^[24,25]或一元二次的非线性相关^[23], 本研究构建了基于这2种拟合函数的估算模型。

线性模型为

(1)y = ax+b
一元二次模型为

(2)y = ax²⁺bx+c
其中, y表示Car/Chla比值或Car含量; x表示光谱指数。

基于各类植被指数构建的线性及一元二次估算模型的决定系数见表3。由表3可知, 对于Car/Chla比值及Car含量, 在叶片及冠层尺度, 均是基于PRI的一元二次模型具有最大决定系数, 并且不论是叶片尺度还是冠层尺度, Car/Chla比值的估算模型的决定系数总是大于Car含量的估算模型, PRI-Car/Chla模型在叶片尺度具有最大决定系数R²=0.732, 冠层尺度R²=0.681, 叶片尺度的Car/Chla估算模型精度略优于冠层尺度。进一步对比发现, 基于PRI的线性估算模型与一元二次模型的决定系数非常接近, 2种模型判定系数的差值均小于0.01。考虑到模型的简单性, 在估算精度相似的情况下, 本研究将采用表达形式更为简单的线性模型利用检验样本进行进一步精度检验, 在此基础上, 讨论利用叶片尺度的PROSPCT-D^[34]和冠层尺度的SAILH^[35]辐射传输模型模拟棉花叶片及冠层尺度的反射率, 进一步检验PRI用于棉花Car/Chla估算的有效性。

Table 3
表3
表3基于植被指数的Car/Chla比值及Car估算模型
Table 3Vegetation indices predictive models for Car/Chla ratio and Car

建模方法 Method	叶片尺度Leaf level			冠层尺度Canopy level
	比值Car/Chla ratio	类胡萝卜素Car		比值Car/Chla ratio	类胡萝卜素Car
PRI-Linear	y= -0.834x+0.338 R2=0.723	y= -30.58x+14.86 R2=0.541		y=1.343x+0.313 R2=0.670	y= -0.009x+0.135 R2=0.486
PRI-Polynomial	y=1.897x2-1.31x+0.348 R2=0.732	y=26.1x2-37.07x+14.9 R2=0.542		y=3.191x2-1.679x+0.309 R2=0.681	y=0.001x2-0.044x+0.349 R2=0.487
PRI*CI-Linear	y= -0.155x+0.328 R2=0.638	y= -5.35x +14.37 R2=0.426		y= -0.127x + 0.32 R2=0.519	y= -0.0718x+1.1565 R2=0.356
PRI*CI-Polynomial	y=0.104x2-0.293x+0.341 R2=0.681	y=5.38x2-12.49x+15.03 R2=0.488		y=0.073x2-0.248x+0.321 R2=0.627	y=0.01x2-0.34x+2.79 R2=0.431
PSRI-Linear	y=3.711x + 0.323 R2=0.203	y=155.2x + 14.47 R2=0.196		y=1.74x + 0.293 R2=0.157	y=0.0015x-0.0158 R2=0.093
PSRI-Polynomial	y= -180.3x2-0.66x+0.314 R2=0.257	y = -11414x2-122.7x+13.9 R2=0.314		y= -54.95x2+2.56x+0.31 R2=0.239	y= -0.0004x2 +0.01x-0.09 R2=0.244
CCRI-Linear	y=0.173x+0.079 R2=0.501	y=5.42x+6.38 R2=0.272		y=0.182x+0.082 R2=0.405	y=0.031x+0.758 R2=0.189
CCRI-Polynomial	y= -0.134x2+0.49x-0.097 R2=0.580	y= -7.09x2 +22.3x-2.89 R2=0.396		y= -0.364x2+0.976x-0.317 R2=0.526	y= -0.008x2 +0.239x-0.51 R2=0.417

新窗口打开|下载CSV

2.3 基于PLSR的估算模型构建

PLSR估算模型的最佳潜变量个数通过LOCCV获取最小预测残差平方和(PRESS)值来确定。潜变量是自变量信息的重新组合, 当潜变量个数对因变量解释性最强, 同时又最能概括自变量系统中的综合信息时, 为最佳潜变量个数, 通常用PRESS值来判定, 当PRESS值最小时, 所对应的潜变量个数即为模型最佳潜变量个数。叶片尺度及冠层尺度的Car/Chla比值及Car含量的PLSR模型的潜变量个数与PRESS之间的关系如图1所示。图中最小PRESS值所对应的潜变量个数用黑色实心圆点表示。由图1可知, 叶片及冠层尺度Car/ Chla比值的PLSR模型的最佳潜变量个数分别为19和16, Car含量的PLSR模型最佳潜变量个数分别为10和13。

图1

新窗口打开|下载原图ZIP|生成PPT
图1预测误差平方和(PRESS)与潜变量之间的变化关系

Fig. 1Relationships between the number of latent variables and the predicted residual sums of squares (PRESS)



在确定各估算模型最佳潜变量个数的基础上, 将叶片及冠层光谱反射率(400~1000 nm)和Car/ Chla比值及Car含量作为模型输入变量, 构建PLSR估算模型, 得到叶片及冠层尺度的Car/Chla比值和Car含量的估算值。基于建模样本的估算值与实测值之间的线性关系如图2所示。由图2可知, 在叶片和冠层尺度, Car/Chla比值及Car含量的估算值与实测值均高度拟合, 两者的拟合线与1:1线接近重叠, 并且相比Car含量, Car/Chla比值的估算精度更高, 叶片和冠层尺度其估算值与实测值的拟合决定系数R²均大于0.96, RMSE%约为5%, 小于Car含量的11%左右。

图2

新窗口打开|下载原图ZIP|生成PPT
图2基于建模样本的Car/Chla-PLSR和Car-PLSR估算模型的估算值与实测值之间的线性关系

Fig. 2Measured vs. estimated Car/Chla ratio and Car for PLSR models using calibration dataset

2.4 精度检验

进一步利用检验样本对基于PRI和PLSR的Car/Chla比值及Car含量估算模型进行精度验证。采用估算值与实测值之间线性拟合的决定系数R²、均方根误差RMSE、RMSE%以及估算值与实测值之间线性拟合的斜率(a)和截距(b)进行模型精度检验。决定系数R²以及斜率和截距可反映估算值与实测值之间变化趋势的吻合程度, 当R²越趋近1, 斜率越接近1, 截距越接近0时, 估算值和实测值之间吻合越好; RMSE%是RMSE值与实测平均值之间的百分比值, 它是无量纲值, 可用于单位不同的参数之间估算误差的比较, 直接体现模型估算误差。由表4可知, 叶片及冠层水平, 相比Car含量, Car/Chla比值的光谱指数模型及PLSR模型均具有更好的精度结果, 并且对于PLSR方法, 叶片尺度和冠层尺度Car/Chla比值估算的精度接近, 对于PRI, 叶片尺度光谱对Car/Chla比值估算的精度略优于冠层尺度。

Table 4
表4
表4基于检验样本的Car/Chla比值及Car含量估算模型精度检验结果
Table 4Cross-validation results for the Car/Chla and Car predictive models

尺度 Level	方法 Method	参数 Parameter	决定系数 R2	均方根误差RMSE	均方根误差 RMSE (%)	斜率 a	截距 b
叶片尺度Leaf level	PRI	Car/Chla	0.69	0.05	15.8	0.66	0.09
		Carotenoid	0.44	3.57	24.9	0.37	7.71
	PLSR	Car/Chla	0.87	0.03	10.7	0.83	0.04
		Carotenoid	0.73	2.39	16.7	0.67	4.02
冠层尺度Canopy level	PRI	Car/Chla	0.67	0.05	15.3	0.67	0.09
		Carotenoid	0.36	3.62	24.9	0.37	8.74
	PLSR	Car/Chla	0.84	0.03	10.1	0.84	0.05
		Carotenoid	0.74	2.19	14.9	0.72	3.87

新窗口打开|下载CSV

基于检验样本得到的Car/Chla比值及Car含量估算值和实测值之间的吻合趋势见图3。对比基于指数和基于PLSR方法的估算模型发现, 由PLSR模型得到的估算值与实测值之间拟合更好, 不论是Car/Chla比值还是Car含量, 利用PLSR方法得到的估算值都具有更大的斜率和更小的截距, 估算值与实测值更接近, 并且在叶片和冠层尺度Car/Chla比值的估算精度非常接近, 说明PLSR方法具有很好的稳定性。

图3

新窗口打开|下载原图ZIP|生成PPT
图3基于检验样本的Car/Chla比值及Car含量估算值与实测值之间的拟合关系

Fig. 3Measured vs. estimated Car/Chla ratio and Car using validation dataset

3 讨论

对多种光谱指数与Car/Chla比值及Car含量之间的相关分析表明,PRI指数在叶片和冠层尺度均是表现最佳的植被指数。本研究进一步利用叶片辐射传输模型PROSPECT-D模型和冠层SAILH模型模拟了棉花叶片及冠层尺度的光谱反射率,利用光谱模拟值分析了PRI对Car/Chl比值的估算能力,如图4所示,模拟结果也很好地证明了PRI与Car/Chl比值之间极显著线性及一元二次相关性。这一结论与Garrity等^[23]对落叶林Car/Chl比值的监测结果一致,该研究中落叶林模拟PRI值与Car/Chl比值之间的一元二次拟合关系的判定系数R²=0.99,与本研究接近。PRI是531 nm和570 nm处光谱反射率的归一化比值,最早用于叶黄素变化监测,但叶绿素和类胡萝卜素在531 nm处均有吸收特征,570 nm处只叶绿素有吸收特征,因此PRI与Car/Chl比值强烈相关^[9,36],很多研究都证实PRI是时间上从天尺度^[19,27]到月尺度^[37,38]和在叶片^[21,39-40]、冠层^[19,41-42]以及生态尺度^[43,44]监测光合作用的有效指数。本研究的结论也进一步证实了PRI在叶片尺度和冠层尺度,对干旱区作物Car/Chla比值估算的精度接近,凸显了该指数在不同光谱尺度的巨大应用价值。

图4

新窗口打开|下载原图ZIP|生成PPT
图4PROSPECT和SAILH模型模拟的叶片及冠层PRI指数与Car/Chl之间的拟合关系

Fig. 4Relationship of PROSPECT and SAILH simulated PRI with Car/Chl ratio



本研究结果证明不论是基于光谱指数方法还是基于偏最小二乘PLSR方法, 相比单一对Car含量的估算, 对Car/Chla比值的估算精度都更高, 这与Merzlyak等^[16]及Pe?uelas等^[21]证实的相比单一叶绿素或类胡萝卜素, 光谱指数对Car/Chl比值的估算更为有效的结论一致。基于PRI和PLSR方法的估算模型均能实现Car/Chla比值和Car含量的估算, 前者在叶片和冠层尺度的决定系数R²分别为0.69和0.67, 而后者为0.87和0.84, 可见偏最小二乘法对Car/Chla比值和Car含量的估算精度更好。但PRI是由2个与叶绿素含量和类胡萝卜素含量相关的敏感波段反射率值构成, PLSR方法的输入变量则是400~1000 nm波段范围内的所有光谱反射率, 对于波段分布有限且包含多种指数波段的多光谱卫星, 如Sentinel-2卫星数据, 通过构建光谱指数进行Car/Chla比值监测较为适用, 而当输入变量较多时, 如光谱分辨率较高的高光谱数据, 则可采用PLSR方法, PLSR包含的主成分分析计算, 能有效降低输入变量维度, 剔除冗余变量。

4 结论

(1)对比各类光谱指数, 光化学指数(PRI)与Car/ Chla比值及Car含量具有最显著相关性, 基于PRI指数的线性和一元二次非线性拟合模型对叶片和冠层尺度的Car/Chla比值和Car含量的估算具有最大决定系数; (2) PRI和偏最小二乘回归(PLSR)对Car/Chla比值的估算精度要优于对Car含量的估算, Car/Chla比值可作为棉花生理生态胁迫(如高温胁迫、养分胁迫和水分胁迫以及病虫害)遥感监测的有效参考指标; (3) PRI和PLSR是Car/ Chla比值和Car含量遥感监测的有效方法, 两者在叶片及冠层尺度对Car/Chla比值的估算精度接近, 对于叶片及冠层高光谱数据, PLSR模型估算能力略高于PRI模型。进一步研究, 期望利用Sentinel-2卫星数据在区域尺度上验证本研究结论, 并探讨Car/Chla比值用于棉花水分、养分胁迫及叶片衰老状况监测的特点。

参考文献 View Option 原文顺序
文献年度倒序
文中引用次数倒序
被引期刊影响因子

[1] Demmig-Adams B. Survey of thermal energy dissipation and pigment composition in sun and shade leaves
Plant Cell Physiol, 1998,39:474-482.
[本文引用: 1]

[2] Zarco-Tejada P J, Miller J R, Harron J, Hu B, Noland T L, Goel N. Needle chlorophyll content estimation through model inversion using hyperspectral data from boreal conifer forest canopies
Remote Sens Environ, 2004,89:189-199.
[本文引用: 1]

[3] Zhang Y, Chen J M, Miller J, Noland T. Needle chlorophyll content retrieval from airborne hyperspectral imagery
Remote Sens Environ, 2008,112:3234-3247.
[本文引用: 1]

[4] Hunt E R, Doraiswamy P C, McMurtrey J E, Daughtry C S T, Perry E M, Akhmedov B. A visible band index for remote sensing leaf chlorophyll content at the canopy scale
Int J Appl Earth OBS, 2013,21:103-112.
[本文引用: 1]

[5] Zarco-Tejada P J, Hornero A, Beck P S A, Kattenborn T, Kempeneers P, Hernández-Clemente R. Chlorophyll content estimation in an open-canopy conifer forest with Sentienl-2A and hyperspectral imagery in the context of forest decline
Remote Sens Environ, 2019,223:320-335.
URLPMID:31007289 [本文引用: 1]

[6] Gitelson A A, Chivkunova O B, Merzlyak M N. Non-destructive estimation of anthocyanins and chlorophylls in anthocyanic leaves
Am J Bot, 2009,96:1861-1868.
URLPMID:21622307 [本文引用: 1]

[7] Young A, Britton G. Carotenoids and stress. In: Alscher R G Jr, Cumming J R, eds. Stress Response in Plants: Adaptation and Acclimation Mechanisms. New York: Wiley-Liss, 1990. pp 87-112.
[本文引用: 1]

[8] Blackburn G A. Quantifying chlorophylls and caroteniods at leaf and canopy scales: an evaluation of some hyperspectral approaches
Remote Sens Environ, 1998,66:273-285.
[本文引用: 1]

[9] Sims D A, Gamon J A. Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages
Remote Sens Environ, 2002,81:337-354.
DOI:10.1016/S0034-4257(02)00010-XURL [本文引用: 2]

[10] Féret J B, Fran?ois C, Gitelson A, Gregory P A, Barry K M, Panigada C, Richardson A D, Jacquemoud S. Optimizing spectral indices and chemometric analysis of leaf chemical properties using radiative transfer modeling
Remote Sens Environ, 2011,115:2742-2750.
[本文引用: 1]

[11] Hendry G A F, Houghton J D, Brown S B. The degradation of chlorophyll: a biological enigma
New Phytol, 1987,107:255-302.
[本文引用: 1]

[12] Biswall B. Carotenoid catabolism during leaf senescence and its control by light
J Photochem Photobiol B: Biol, 1995,30:3-14.
[本文引用: 2]

[13] Buchanan-Wollastin V. The molecular biology of leaf senescence
J Exp Bot, 1998,49:181-199.
[本文引用: 1]

[14] Gitelson A A, Merzlyak M N. Spectral reflectance changes associate with autumn senescence of Aesculus hippocastanum L. and Acer platanoides L. leaves. Spectral features and relation to chlorophyll estimation
J Plant Physiol, 1994,143:286-292.
[本文引用: 1]

[15] Gitelson A A, Merzlyak M N. Quantitative estimation of chlorophyll-a using reflectance spectra: experiments with autumn chestnut and maple leaves
J Photochem Photobiol B: Biol, 1994,22:247-252.


[16] Merzlyak M N, Gitelson A A, Chivkunova O B, Rakitin V Y. Non-destructive optical detection of pigment changes during leaf senescence and fruit ripening
Physiol Plant, 1999,106:135-141.
[本文引用: 3]

[17] Sanger J E. Quantitative investigations of leaf pigments from their inception in buds through autumn coloration to decomposition in falling leaves
Ecology, 1972,52:1075-1089.
[本文引用: 1]

[18] Pe?uelas J, Field C, Griffin K, Gamon J. Assessing community type, plant biomass, pigment composition and photosynthetic efficiency of aquatic vegetation from spectral reflectance
Remote Sens Environ, 1993,46:1-25.
[本文引用: 2]

[19] Pe?uelas J, Gamon J, Freeden A, Merino J, Field C. Reflectance indices associated with physiological changes in nitrogen and water limited sunflower leaves
Remote Sens Environ, 1994,48:135-146.
DOI:10.1016/0034-4257(94)90136-8URL [本文引用: 3]

[20] 周贤峰. 色素含量比值进行作物氮素营养状况诊断方法研究
中国科学院大学博士学位论文, 北京, 2017.
[本文引用: 2]

Zhou X F. Research on the Methods of Crop Nitrogen Status Diagnosis Based on Carotenoid and Chlorophyll Ratio Values
PhD Dissertation of University of Chinese Academy of Sciences, Beijing, China, 2017 (in Chinese with English abstract).
[本文引用: 2]

[21] Pe?uelas J, Baret F, Filella I. Semi-empirical indices to assess carotenoids/chlorophyll a ratio from leaf spectral reflectance
Photosynthetica, 1995,31:221-230.
[本文引用: 3]

[22] Nakaji T, Oguma H, Fujinuma Y. Seasonal changes in the relationship between photochemical reflectance index and photosynthetic light use efficiency of Japanese larch needles
Int J Remote Sens, 2006,27:493-509.
[本文引用: 1]

[23] Garrity S R, Eitel J U H, Vierling L A. Disentangling the relationships between plant pigments and the photochemical reflectance index reveals a new approach for remote estimation of carotenoid content
Remote Sens Environ, 2011,115:628-635.
[本文引用: 4]

[24] Hernández-Clemente R, Navarro-Cerrillo R M, Zarco-Tejada P J. Carotenoid content estimation in a heterogeneous conifer forest using narrow-band indices and PROSPECT + DART simulations
Remote Sens Environ, 2012,127:298-315.
[本文引用: 3]

[25] Zhou X F, Huang W J, Zhang J C, Kong W P, Casa R, Huang Y B. A novel combined spectral index for estimating the ratio of carotenoid to chlorophyll content to monitor crop physiological and phonological status
Int J Appl Earth OBS, 2019,76:128-142.
[本文引用: 3]

[26] 蒋德安, 朱诚. 植物生理学实验指导. 成都: 成都科技大学出版社, 1999. pp 20-23.
[本文引用: 1]

Jiang D A, Zhu C. Guide for Plant Physiology Experiment. Chengdu: Chengdu University of Science and Technology Press, 1999. pp 20-23(in Chinese).
[本文引用: 1]

[27] Gamon J A, Pe?uelas J, Field C B. A narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency
Remote Sens Environ, 1992,41:35-44.
[本文引用: 1]

[28] Yi Q X, Jiapaer G L, Chen J M, Bao A M, Wang F M. Different units of measurement of carotenoids estimation in cotton using hyperspectral indices and partial least square regression
ISPRS J Photogramm, 2014,91:72-74.
[本文引用: 1]

[29] 依尔夏提·阿不来提, 买买提·沙吾提, 白灯莎·买买提艾力, 安申群, 马春玥. 基于随机森林法的棉花叶片叶绿素含量估算
作物学报, 2019,45:81-90.
[本文引用: 1]

Ershat A, Mamat S, Baidengsha M, An S Q, Ma C Y. Estimation of leaf chlorophyll content in cotton based on the random forest approach
Acta Agron Sin, 2019,45:81-90 (in Chinese with English abstract).
[本文引用: 1]

[30] Geladi P, Kowalski B R. Partial least-squares regression: a tutorial
Analy Chimica Acta, 1986,185:1-17.
[本文引用: 1]

[31] Wold S, Sj?str?m M, Eriksson L. PLS-regression: a basic tool of chemometrics
Chemometr Intell Lab, 2001,58:109-130.
[本文引用: 1]

[32] Shao J. Linear model selection by cross-validation
J Am Stat Assoc, 1993,88:486-494.
[本文引用: 1]

[33] Chen S, Hong X, Harris C J, Sharkey P M. Sparse modeling using orthogonal forward regression with PRESS statistic and regularization
IEEE Trans Syst Man Cybern B Cybern, 2004,34:898-911.
DOI:10.1109/tsmcb.2003.817107URLPMID:15376838 [本文引用: 1]
The paper introduces an efficient construction algorithm for obtaining sparse linear-in-the-weights regression models based on an approach of directly optimizing model generalization capability. This is achieved by utilizing the delete-1 cross validation concept and the associated leave-one-out test error also known as the predicted residual sums of squares (PRESS) statistic, without resorting to any other validation data set for model evaluation in the model construction process. Computational efficiency is ensured using an orthogonal forward regression, but the algorithm incrementally minimizes the PRESS statistic instead of the usual sum of the squared training errors. A local regularization method can naturally be incorporated into the model selection procedure to further enforce model sparsity. The proposed algorithm is fully automatic, and the user is not required to specify any criterion to terminate the model construction procedure. Comparisons with some of the existing state-of-art modeling methods are given, and several examples are included to demonstrate the ability of the proposed algorithm to effectively construct sparse models that generalize well.

[34] Féret J B, Gitelson A, Noble S D, Jacquemoud S. PROSPECT-D: towards modeling leaf optical properties through a complete lifecycle
Remote Sens Environ, 2017,193:204-215.
[本文引用: 1]

[35] Kuusk A. The angular distribution of reflectance and vegetation indices in barley and clover canopies
Remote Sens Environ, 1991,37:143-151.
[本文引用: 1]

[36] Filella I, Porcar-Castell A, Munné-Bosch S, Back J, Garbulsky M F, Pe?uelas J. PRI assessment of long-term changes in carotenoids/chlorophyll ratio and short-term changes in de-epoxidation state of the xanthophyll cycle
Int J Remote Sens, 2009,30:4443-4455.
[本文引用: 1]

[37] Stylinski C D, Gamon J A, Oechel W C. Seasonal patterns of reflectance indices, carotenoid pigments and photosynthesis of evergreen chaparral species
Oecologia, 2002,131:366-374.
DOI:10.1007/s00442-002-0905-9URLPMID:28547708 [本文引用: 1]
This study examined the ability of the Photochemical Reflectance Index (PRI) to track seasonal variations in carotenoid pigments and photosynthetic activity of mature evergreen chaparral shrubs. Our results confirm that PRI scales with photosystem two (PSII) photochemical efficiency across species and seasons, as demonstrated by PRI's strong correlation with de-epoxidized (photoprotective) xanthophyll cycle pigment levels (normalized to chlorophyll) and with the chlorophyll fluorescence index, DeltaF/Fm'. PRI and carotenoid pigment levels (de-epoxidized xanthophyll cycle pigments normalized to chlorophyll or total carotenoid pigments normalized to chlorophyll) were correlated with seasonal fluctuations in midday net CO2 uptake of top-canopy leaves. By contrast, chlorophyll levels (as measured by the Chlorophyll Index) were not as strongly linked to photosynthetic activity, particularly when all species were considered together. Likewise, the Normalized Difference Vegetation Index (NDVI, an index of canopy greenness) did not correlate with net CO2 uptake. Canopy NDVI also did not correlate with canopy PRI, demonstrating that these indices were largely independent over the temporal and spatial scales of this study. Together, these patterns provide evidence for coordinated regulation of carotenoid pigments, PSII electron transport, and carboxylation across seasons and indicate that physiological adjustments are more important than structural ones in modifying CO2-fixation capacity during periods of photosynthetic down-regulation for these evergreen species. The strong correlation between PRI of whole canopies and PRI of top-canopy leaves suggests that the canopy can be treated as a

[38] Garbulsky M F, Pe?uelas J, Papale D, Filella I. Remote estimation of carbon dioxide uptake of terrestrial ecosystems
Global Change Biol, 2008,14:2860-2867.
[本文引用: 1]

[39] Gamon J A, Serrano L, Surfus J S. The photochemical reflectance index: an optical indicator of photosynthesis radiation use efficiency across species, functional types and nutrient levels
Oecologia, 1997,112:492-501.
DOI:10.1007/s004420050337URLPMID:28307626 [本文引用: 1]
The photochemical reflectance index (PRI), derived from narrow-band reflectance at 531 and 570 nm, was explored as an indicator of photosynthetic radiation use efficiency for 20 species representing three functional types: annual, deciduous perennial, and evergreen perennial. Across species, top-canopy leaves in full sun at midday exhibited a strong correlation between PRI and DeltaF/Fm', a fluorescence-based index of photosystem II (PSII) photochemical efficiency. PRI was also significantly correlated with both net CO2 uptake and radiation use efficiency measured by gas exchange. When species were examined by functional type, evergreens exhibited significantly reduced midday photosynthetic rates relative to annual and deciduous species. This midday reduction was associated with reduced radiation use efficiency, detectable as reduced net CO2 uptake, PRI, and DeltaF/Fm' values, and increased levels of the photoprotective xanthophyll cycle pigment zeaxanthin. For each functional type, nutrient deficiency led to reductions in both PRI and DeltaF/Fm' relative to fertilized controls. Laboratory experiments exposing leaves to diurnal courses of radiation and simulated midday stomatal closure demonstrated that PRI changed rapidly with both irradiance and leaf physiological state. In these studies, PRI was closely correlated with both DeltaF/Fm' and radiation use efficiency determined from gas exchange at all but the lowest light levels. Examination of the difference spectra upon exposure to increasing light levels revealed that the 531 nm Delta reflectance signal was composed of two spectral components. At low irradiance, this signal was dominated by a 545-nm component, which was not closely related to radiation use efficiency. At progressively higher light levels above 100 mumol m(-2) s(-1), the 531-nm signal was increasingly dominated by a 526-nm component, which was correlated with light use efficiency and with the conversion of the xanthophyll pigment violaxanthin to antheraxanthin and zeaxanthin. Further consideration of the two components composing the 531-nm signal could lead to an index of photosynthetic function applicable over a wide range of illumination. The results of this study support the use of PRI as an interspecific index of photosynthetic radiation use efficiency for leaves and canopies in full sun, but not across wide ranges in illumination from deep shade to full sun. The discovery of a consistent relationship between PRI and photosynthetic radiation use efficiency for top-canopy leaves across species, functional types, and nutrient treatments suggests that relative photosynthetic rates could be derived with the

[40] Richardson A D, Berlyn G P. Spectral reflectance and photosynthetic properties of Betula papyrifera(Betulaceae) leaves along an elevational gradient on Mt. Mansfield, Vermont, USA
Am J Bot, 2002,89:88-94.
DOI:10.3732/ajb.89.1.88URLPMID:21669715 [本文引用: 1]
We studied relationships between spectral reflectance and photosynthesis of mountain paper birch, Betula papyrifera var. cordifolia (Regel) Fern., leaves from three different elevations on Mt. Mansfield (summit elevation 1339 m above sea level) in the Green Mountains of Vermont, USA. The different reflectance indices we used all suggested progressively increasing stress with increasing elevation. The photochemical reflectance index (PRI) indicated lower photosynthetic radiation use efficiency at higher elevations, the red edge position (lambdaRE) indicated lower chlorophyll concentrations at higher elevations, and the structure-independent pigment index (SIPI) indicated a higher carotenoid : chlorophyll a ratio at higher elevations. The rate of change in these indices with changes in elevation was much higher than we have observed in our studies of red spruce and balsam fir reflectance along a similar elevational gradient; we take this to be an indicator of the greater susceptibility of paper birch to elevation-related stressors compared to the very stress-tolerant conifers. At all light levels, photosynthesis decreased with increasing elevation; this pattern was most noticeable in the light-saturated rate of photosynthesis (A(sat)), which was nearly twice as high in low-elevation leaves (17.0 +/- 1.0 mumol.m(-2).s(-1)) than in high-elevation leaves. The quantum yield of photosynthesis (Phi) exhibited a similar trend. Furthermore, the highest elevation leaves showed a much sharper transition from the light-limited to the light-saturated part of the light response curve than did the lowest elevation leaves. The photochemical reflectance index was highly correlated with A(sat) (r(2) = 0.99) and Phi (r(2) = 0.96). In addition to contributing to our knowledge of the ecophysiology of paper birch along a steep environmental gradient, these results are further evidence of the usefulness of reflectance measures for the rapid and noninvasive detection of plant stress, especially when used in conjunction with direct measurements of photosynthesis.

[41] Filella I, Amaro T, Araus J L, Pe?uelas J. Relationship between photosynthetic radiation-use efficiency of barley canopies and the photochemical reflectance index (PRI)
Physiol Plant, 1996,96:211-216.
[本文引用: 1]

[42] Trotter G M, Whitehead D, Pinkney E J. The photochemical reflectance index as a measure of photosynthetic light use efficiency for plants with varying foliar nitrogen contents
Int J Remote Sens, 2002,6:1207-1212.
[本文引用: 1]

[43] Nichol C J, Huemmrich K F, Black T A, Jarvis P G, Walthall C L, Grace J, Hall F G. Remote sensing of photosynthetic-light-use efficiency of boreal forest
Agric For Meteorol, 2000,101:131-142.
DOI:10.1016/S0168-1923(99)00167-7URL [本文引用: 1]

[44] Pe?uelas J, Inoue Y. Reflectance assessment of canopy CO₂ uptake
Int J Remote Sens, 2000,21:3353-3356.
DOI:10.1080/014311600750019958URL [本文引用: 1]