Advantages of retrieving pigment content [μg/cm2] versus concentration [%] from canopy reflectance
Introduction
Terrestrial plants are vital for the production of oxygen and organic matter through photosynthesis. Photosynthesis is primarily controlled by pigments, which are important links to assess plant stress, plant functioning, biological cycles, and biosphere-atmosphere interactions (Nelson and Yocum 2006; Blackburn, 2006; Kattenborn et al., 2018). Photosynthesis is performed by chlorophylls and carotenoids. Carotenoids, together with anthocyanins, protect chlorophylls and other plant material from photodamage (excess and UV radiation). Anthocyanins are further important indicators for pathogen defence (Lev-Yadun and Gould 2008, Zarco-Tejada et al., 2018).
These pigments primarily affect the radiative transfer in the visible spectrum, where solar radiation is highest (400–700 nm), whereas incident radiation that is not absorbed by the canopy or the ground is scattered. These scattered remnants constitute the basis for quantifying pigments such as chlorophylls, carotenoids, or anthocyanins using optical remote sensing observations (Tucker and Sellers, 1986; Jacquemoud et al., 1996; Blackburn, 2006; Kattenborn et al., 2017; Zarco-Tejada et al., 2018). Commonly, pigments are quantified using two different metrics - either as pigment content, i.e. pigment mass per leaf area [μg/cm2] (hereafter referred as pigmentarea) or as pigment concentration, i.e. pigment mass per leaf dry mass [g/g or %] (hereafter referred as pigmentmass). Note that the terms content and concentrations are often used interchangeably, while here we use content for per-area and concentration for per-mass. The choice of quantification method in remote sensing appears to be inconclusive, as both metrics are frequently referred to in the relevant literature (e.g. Jacquemoud et al., 1996; Zarco-Tejada et al., 2001; Asner and Martin, 2009; Jetz et al., 2016). Here, we argue that quantifying pigmentmass with remote sensing is unsubstantial as 1) this measure does not explicitly reflect variation in pigments per se, but rather variation in leaf dry mass, 2) pigmentmass is less accurately retrieved than pigmentarea using optical remote sensing and 3) it is more difficult to scale-up pigmentmass to the canopy scale. We conclude that quantifying pigmentsarea is more appropriate in remote sensing due to its explicit relation to radiative transfer, enhanced scalability and as it is a more direct expression of plant stress and functioning.
Put simply, pigmentmass [%] is the ratio of pigmentarea [μg/cm2] and the Leaf Dry Mass per Area [g/cm2] (LMA):
Leaf dry mass is composed of carbohydrates (hemi-cellulose, cellulose, starch), proteins, lignin and waxes, and it generally reflects differences in leaf lifespan resulting from adaptations to environmental factors (Grime et al., 1997, Wright et al. 2004, Díaz et al., 2016). As evinced using global trait databases, LMA has a higher variance than leaf traits related to photosynthesis, e.g. leaf nitrogen content [mg/cm2] or photosynthetic capacity [μmol/m2/s] (see Wright et al. 2004; Osnas et al., 2013; Lloyd et al., 2013). This is critical as leaf resource investments (e.g. LMA) and leaf traits relating to photosynthesis are largely independent of one another (Osnas et al., 2013; Llyod et al. 2013; Osnas et al., 2018) and accordingly the division by LMA actually dominates the actual variation of pigments content.
Here we demonstrate these relationships for leaf pigments using a dataset comprising LMA, chlorophyllarea, carotenoidarea, and anthocyaninarea values from 45 herbaceous species retrieved in-situ (see supporting information for details). The coefficient of variation of LMA (38.4%) clearly exceeds that of chlorophyllarea (24.8%), carotenoidarea (15.0%), and anthocyaninarea (26.1%). Correspondingly, a principal component analysis (Fig. 1) of LMA, pigmentsarea and pigmentsmass reveals that pigmentsmass primarily reflect the LMA gradient (strong negative correlation). Gradients of pigmentsarea, in contrast, are largely orthogonal and thus uncorrelated with LMA. Thus, it can generally be expected that gradients of pigmentsmass predominantly mirror the variation in LMA, which in turn overshadows the actual variation of pigmentsarea.
As reported by previous authors, the retrieval of leaf constituents is more accurate for absolute contents per area than for concentration per mass (Grossman et al., 1996; Jacquemoud et al., 1996; Oppelt and Mauser, 2004). This can be explained by the radiative transfer mechanisms: Leaf constituents affect the reflectance properties of a plant canopy through absorption and scattering, whereas these effects increase with increasing contents of the respective constituent (e.g. pigments). The spectral signal is therefore determined by the absolute content of the constituent (e.g. pigmentsarea) and not by its concentration relative to LMA. In other words, concentrations (pigmentmass) cannot represent the absolute amount of matter interacting with electromagnetic radiation (also see Jacquemoud et al., 1996). For this reason, pigments in radiative transfer models are parametrized by specific absorption coefficients on an area basis. Pigmentmass is the ratio of pigmentarea to LMA, which further implies that remote sensing of pigmentmass (e.g. through statistical models) ideally requires the simultaneous consideration of spectral features corresponding to both pigments (in the visible range) and LMA (in the short wave infrared range), as illustrated using empirical canopy reflectance data in Fig. 2. However, the retrieval of LMA using optical canopy reflectance is commonly challenging, as the respective spectral features are overshadowed by water absorption (Jacquemoud et al., 1996; Homolová et al., 2013). Moreover, and in contrast to visible and near infrared wavelengths, the short-wave infrared information is generally affected by lower signal-to-noise ratios, increased spectral shifts, and increased calibration uncertainties (Cocks et al., 1998; Bachmann et al., 2015). Uncertainties in the retrieval of LMA spectral features propagate into errors of pigmentmass assessment. Thus, the retrieval of pigmentsmass is substantially impaired as it requires spectral information of the short wave infrared range (which is not always available) and the generally less accurate retrieval of the LMA variation. In contrast, the retrieval of pigmentsarea only relies on spectral features in the visible range (Fig. 2a).
Being a relative concentration, pigmentmass is generally an inconclusive metric: high pigmentmass can result from either high pigmentarea and intermediate LMA or intermediate pigmentarea and low LMA. It is therefore possible for two leaves or plant canopies to have equivalent pigmentmass, but differ greatly in pigmentarea and LMA. Accordingly, pigmentmass does not explicitly indicate if a plant canopy actually has low pigment content, e.g. due to stress or its inherent plant functional properties (compare Fig. 3).
This ambiguity similarly limits the scalability to the canopy level, which is pigment content per canopy surface area [g/m2] (hereafter referred as pigmentcanopy). Pigmentcanopy relates to the absolute photosynthesis of a vegetated area and is thus directly relevant for assessing productivity or atmosphere-biosphere interactions (De Pury and Farquhar, 1997; Peng et al., 2011). Here, we seek to demonstrate the limited scalability of pigmentmass using a straightforward approach, i.e. upscaling leaf constituents to the canopy scale by incorporating Leaf Area Index [m2/m2] (LAI). LAI is a proxy for the total foliage area within the canopy area and can be retrieved from remote sensing data with acceptable accuracy (Zarco-Tejada et al., 2001; Myneni et al., 2002; Schlerf et al., 2005). In case of pigmentarea, upscaling to pigmentcanopy merely requires a multiplication with LAI (Eq. (2)). In contrast, scaling pigmentmass to pigmentcanopy requires prior knowledge on the absolute foliage mass in the entire canopy surface area, i.e. the product of LAI and the LMA (Eq. (3)).
However, as described in section 2, the quantification of LMA requires SWIR information and is generally limited using canopy reflectance (compare Homolová et al., 2013). Thus, scaling pigmentmass to the canopy requires additional information on the dry weight of the foliage (LMA) and may be negatively affected by error propagation of the LMA estimates.
Section snippets
Discussion and concluding remarks
For monitoring vegetation photosynthesis and physiological status, from the above arguments, we strongly advocate a focus on pigment content per area, rather than pigment mass concentration. Most studies currently reporting on pigmentmass (see supplementary information Table S-2) do so without a precise justification on why they quantify pigments as concentration. We assume that the frequent use of pigmentmass may primarily be adopted from plant ecology, where leaf nutrients (e.g. nitrogen or
Acknowledgements
The project was funded by the German Aerospace Center (DLR) on behalf of the Federal Ministry of Economics and Technology (BMWi), FKZ50EE 1347. We would like to thank all employees of the botanical garden of the Karlsruher Institute for Technology (KIT), especially Peter Nick and Christine Beier, for their generous support. We further want to thank Kyle Kovach, Carsten Dormann and Pieter Beck for very helpful comments on the manuscript.
References (28)
- et al.
Critique of stepwise multiple linear regression for the extraction of leaf biochemistry information from leaf reflectance data
Remote Sens. Environ.
(1996) - et al.
Review of optical-based remote sensing for plant trait mapping
Ecol. Complex.
(2013) - et al.
Estimating leaf biochemistry using the PROSPECT leaf optical properties model
Remote Sens. Environ.
(1996) - et al.
Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data
Remote Sens. Environ.
(2002) - et al.
Remote estimation of gross primary production in maize and support for a new paradigm based on total crop chlorophyll content
Remote Sens. Environ.
(2011) - et al.
Remote sensing of forest biophysical variables using HyMap imaging spectrometer data
Remote Sens. Environ.
(2005) - et al.
Airborne spectranomics: mapping canopy chemical and taxonomic diversity in tropical forests
Front. Ecol. Environ.
(2009) - et al.
Estimating the influence of spectral and radiometric calibration uncertainties on EnMAP data products-examples for ground reflectance retrieval and vegetation indices
Remote Sens.
(2015) Hyperspectral remote sensing of plant pigments
J. Exp. Bot.
(2006)- et al.
The HyMapTM airborne hyperspectral sensor: the system, calibration and performance
A Handbook of Protocols for Standardised and Easy Measurement of Plant Functional Traits Worldwide
Simple scaling of photosynthesis from leaves to canopies without the errors of big-leaf models
Plant Cell Environ.
The global spectrum of plant form and function
Nature
Integrated screening validates primary axes of specialisation in plants
Oikos
Cited by (41)
Advancing our understanding of plant diversity-biological invasion relationships using imaging spectroscopy
2024, Remote Sensing of EnvironmentRetrieval of crop biophysical-biochemical variables from airborne AVIRIS-NG data using hybrid inversion of PROSAIL-D
2024, Advances in Space ResearchMapping foliar photosynthetic capacity in sub-tropical and tropical forests with UAS-based imaging spectroscopy: Scaling from leaf to canopy
2023, Remote Sensing of EnvironmentFrom spectra to plant functional traits: Transferable multi-trait models from heterogeneous and sparse data
2023, Remote Sensing of EnvironmentComparison of the performance of Multi-source Three-dimensional structural data in the application of monitoring maize lodging
2023, Computers and Electronics in Agriculture