JXB Advance Access originally published online on September 21, 2006
Journal of Experimental Botany 2007 58(4):855-867; doi:10.1093/jxb/erl123
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Remote Sensing |
Hyperspectral remote sensing of plant pigments
Department of Geography, Lancaster University, Lancaster LA1 4YB, UK
* E-mail: alan.blackburn{at}lancaster.ac.uk
Received 15 May 2006; Accepted 17 July 2006
 |
Abstract |
|---|
 Top Abstract IntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
The dynamics of pigment concentrations are diagnostic of a range of plant physiological properties and processes. This paper appraises the developing technologies and analytical methods for quantifying pigments non-destructively and repeatedly across a range of spatial scales using hyperspectral remote sensing. Progress in deriving predictive relationships between various characteristics and transforms of hyperspectral reflectance data are evaluated and the roles of leaf and canopy radiative transfer models are reviewed. Requirements are identified for more extensive intercomparisons of different approaches and for further work on the strategies for interpreting canopy scale data. The paper examines the prospects for extending research to the wider range of pigments in addition to chlorophyll, testing emerging methods of hyperspectral analysis and exploring the fusion of hyperspectral and LIDAR remote sensing. In spite of these opportunities for further development and the refinement of techniques, current evidence of an expanding range of applications in the ecophysiological, environmental, agricultural, and forestry sciences highlights the growing value of hyperspectral remote sensing of plant pigments.
Key words: Anthocyanins, carotenoids, chlorophyll, hyperspectral, optical properties, pigments, radiative transfer, reflectance, remote sensing
|
Introduction |
|---|
|
TopAbstractIntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
The importance of plant pigments
Plant pigments are of tremendous significance in the biosphere. Indeed, it is argued that the chlorophylls are Earth's most important organic molecules as they are necessary for photosynthesis, the carotenoids are essential for plant and mammal survival through their photosynthetic and nutritional functions, while other pigment groups are key to the physiology of plants and the organisms with which they interact (Davies, 2004).
Within leaf chloroplasts the antenna pigments absorb solar radiation and transfer the energy to the reaction centre pigments which initiates photosynthesis (Richardson et al., 2002). The most important of these pigments are the chlorophylls (Chls; chlorophyll a and b) which are of interest in their own right, but, from a physiological perspective, Chls concentration is important for several reasons. Chls have a dominant control upon the amount of solar radiation that a leaf absorbs, therefore, foliar concentrations of Chls controls photosynthetic potential and, consequently, primary production. The molecular structure of the Chls incorporates a large proportion of total leaf nitrogen and several studies have found that foliar Chls concentration provides an accurate, indirect estimate of plant nutrient status (Filella et al., 1995; Moran et al., 2000). Chls generally decrease under stress and during senescence and the ratio of Chl a to Chl b changes with abiotic factors such as light (Fang et al., 1998), therefore, measurements of total Chl and Chl a and Chl b individually, can provide useful insights into plant–environment interactions (Richardson et al., 2002).
Carotenoids (Cars) are the second major group of plant pigments, composed of carotenes and xanthophylls. Cars can absorb incident radiation and contribute energy to photosynthesis as they are essential structural components of the photosynthetic antenna and reaction centre complexes (Bartley and Scolnik, 1995). The fraction of photosynthetically active radiation absorbed by a plant canopy (APAR) has been related to net primary productivity as a function of a light use efficiency (LUE) coefficient defining the carbon fixed per unit radiation intercepted (Landsberg et al., 1996). Such studies assume that the contribution of each pigment to the energetics of photosynthesis is equal, but this is an insufficient interpretation, as the concentration of Chl a is the limiting factor in the utilization of light for photosynthesis, because it receives energy absorbed by Chl b and Cars (Chappelle et al., 1992; Kim et al., 1994). Thus, the photosynthetic potential of two plants may differ even though their APAR is equal, depending upon the concentrations of individual pigments. Furthermore, when incident radiation exceeds that needed for photosynthesis, Cars that compose the xanthophyll cycle dissipate excess energy and protect the reaction centres. This is achieved by the conversion of the xanthophyll pigments from the epoxidized (violaxanthin) state to the de-epoxidized (antheraxanthin and zeaxanthin) state (Demming-Adams and Adams, 1996). Xanthophyll pigment composition is therefore closely related to the photosynthetic LUE. The dual photosynthetic and photoprotective roles of the carotenoids have also been evident in studies demonstrating two levels of regulation of carotenoid synthesis, the first dependent on chlorophyll synthesis and the second on LUE (Härtel and Grimm, 1998). Thus, while changes in Chls are indicative of stress and phenological stage, Cars concentration provides much complementary information on vegetation physiological status (Young and Britton, 1990).
Anthocyanins (Anths) are water-soluble flavonoids which are the third major group of pigments in leaves, but there is no unified explanation for their presence and function. For certain species or particular phenological stages Anths can become the dominant pigment group and they are a common component of all leaves. Anths can modify the light environment within a leaf and have the potential to regulate photosynthesis and limit photoinhibition and photobleaching (Barker et al., 1997) thereby having a photoprotective function (Steyn et al., 2002; Close and Beadle, 2003). Anths may act as osmotic adjusters offering plants resistance to freezing and drought stress (Chalker-Scott, 1999) and there is evidence that Anths have antioxidant properties that aid recovery after foliar injury (Gould et al., 2002). While the roles of Anths are currently the focus of some debate, it is clear that information on the dynamics of their concentrations is key to understanding the physiological reaction and resistance of plants to different environmental stress factors brought about by episodic events or seasonal fluctuations.
Variations in the relative concentrations of Chls, Cars, and Anths have recently been the subject of an emerging theory that proposes that the ‘bright’ foliar colours (i.e. low Chl, higher Cars and/or Anths) displayed by some plant species in autumn could be a result of coevolution between insects and trees (Archetti, 2000; Hamilton and Brown, 2001). It is proposed that the ‘bright’ colours are a signal of the defensive commitment of the plant which it uses to reduce parasite load. The coevolution theory may even be a general explanation for ‘bright’ colours of leaves, i.e. an explanation for inter- and intraspecific variations in the proportions of the major groups of pigments in leaves (Archetti and Brown, 2004). Whether or not this theory holds up to ongoing experimental scrutiny, variations in the absolute and relative concentrations of the different pigment groups have undoubted significance for plant physiological functioning and the interactions between flora and fauna (Bartley and Scolnik, 1995).
Given the demonstrable importance of plant pigments, information concerning the temporal dynamics and spatial variations of pigments can, from an applied perspective, provide key contributions to a wide range of scientific investigations and environmental/agricultural management endeavours. However, current capabilities for providing this information are limited. Traditional techniques for measuring foliar pigment concentrations involve extraction with a solvent and spectrophotometric analysis using standard procedures. This is possible because pigments have differing spectral absorption properties and simple combinations of absorbance values at a number of wavelengths can be used accurately to determine individual pigment concentrations from mixed extracts (Lichtenthaler, 1987). Pigment assay, particularly for individual carotenoids, can also be performed using high performance liquid chromatography (Dunn et al., 2004). However, these wet laboratory techniques are time- and labour-intensive and therefore expensive. Thus, for whole canopies, pigments must be quantified by extrapolation from a limited number of samples, which introduces inaccuracies. Moreover, the destructive nature of traditional methods can restrict the ability to monitor the temporal dynamics of pigments, particularly at the individual leaf scale. Faster, non-destructive measurements of total leaf Chl concentration can be obtained using hand-held field instruments such as the Minolta SPAD (Markwell et al., 1995), but there is still a requirement to extrapolate spatially from a restricted number of samples and measurements of Chl a and Chl b, Cars, or Anths cannot be made.
The potential for remote quantification of pigments
The spectral absorbance properties of pigments are manifest in the reflectance spectra of leaves and this offers the opportunity of using measurements of reflected radiation as a non-destructive method for quantifying pigments. A useful overview of the physical basis and applications of radiation measurements to enable remote sensing capabilities in plant ecophysiology is provided by Jones et al. (2003). Multispectral remote sensing systems such as the Landsat Thematic Mapper have been operational for several decades and spectral indices based on their broad wavebands have been developed for quantifying vegetation biophysical properties such as leaf area index (LAI). However, a number of studies have demonstrated that broad band multispectral data are inadequate for the remote sensing of vegetation biochemical properties and that narrow band (high spectral resolution; usually a bandwidth of 10 nm or less) hyperspectral data are required (Broge and Mortensen, 2002).
With the advent of airborne [e.g. Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and HyMap] and, more recently, experimental spaceborne imaging spectrometers [e.g. Compact High Resolution Imaging Spectrometer (CHRIS) and Hyperion], with high spectral and radiometric resolutions and signal:noise ratios, there have been opportunities to acquire vegetation reflectance spectra and test methods for imaging plant pigment concentrations. Ustin et al. (2004) outline recent advances in the technology and applications of airborne and spaceborne imaging spectrometry for remote sensing of biophysical processes and properties. The forthcoming generation of operational spaceborne imaging spectrometer missions [e.g. Hyperspectral Environment and Resource Observer (HERO) and Environmental Mapping and Analysis Program (EnMAP)] will facilitate the development of a greater range of practical applications. At the same time as airborne and spaceborne technologies are developing, there is growing interest, particularly from the precision agriculture sector, in the development of field-based instruments that can be used to obtain plant spectra (Inoue and Peñuelas, 2001; Noh et al., 2005; Zillman et al., 2006). Such instruments can be mounted on tractors or mobile irrigation equipment and are able to obtain spectral data at a higher spatial resolution and temporal frequency and lower cost than airborne or spaceborne instruments. The maturation of sensor technology and growing operational deployment of ground-based, airborne and spaceborne instruments will greatly enhance the capabilities for routinely acquiring hyperspectral data and potentially, for quantifying plant pigments over a wide range of spatial scales, repeatedly.
Alongside developments in hyperspectral data acquisition there has been an increasing intensity of research focused on developing techniques for analysing plant spectra in order to quantify pigment concentrations. To extract pigment information the range of other factors which also influence vegetation reflectance spectra must first be taken into account. The internal structure of leaves, with large numbers of refractive discontinuities between cell walls and intercellular air spaces, scatters incident radiation and allows a large proportion to pass back through the upper epidermis to be observed as reflected radiation. Pigments, water, and other biochemicals absorb certain wavelengths of radiation which reduces reflectance in these regions. However, because of the overlapping absorption features of the pigments (Fig. 1) and other constituents, it may be difficult to relate reflectance at a single wavelength to the concentration of an individual pigment. Furthermore, leaf reflectance can vary independently of pigment concentrations due to differences in internal structure, surface characteristics (e.g. hairs/waxes) and moisture content. The reflectance spectrum of a whole canopy is subject to even more controlling factors, notably, effects of variations in number of leaf layers (LAI), orientation of leaves (leaf angle distribution; LAD), % canopy ground coverage, presence of non-leaf elements, areas of shadow and soil/litter surface reflectance. This range of factors, at leaf and canopy scales, obscures the relationships between spectral reflectance and concentrations of individual pigments.
|
Further issues arise because of the nature of hyperspectral data. While high spectral resolution is required in order to resolve the fine spectral differences between individual pigments, sensing in a large number of adjacent narrow wavebands can produce reflectance spectra containing considerable spectral autocorrelation. The autocorrelation causes redundancy within hyperspectral data sets, therefore it is necessary to employ appropriate techniques with which to characterize the main sources of spectral variability and to identify optimal wavebands that offer maximum informational content.
The following section reviews the progress in tackling this series of issues in order to quantify pigment concentrations from hyperspectral remotely-sensed data. A combination of research approaches have been used, incorporating experimental studies, usually where pigment concentrations are manipulated through stress treatments, and observational studies, often exploiting pigment variations due to natural phenological changes.
|
The development of hyperspectral analytical approaches |
|---|
|
TopAbstractIntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
Figure 2 displays the reflectance spectrum over the visible (400–700 nm) and near infrared (here 700–1000 nm) wavebands of a typical green leaf. This figure demonstrates the low reflectance across the visible, with a slight peak in the green (about 550 nm), with reflectance rising rapidly at the so-called ‘red edge’ leading to a plateau of high reflectance in the near-infrared, where pigments no longer absorb radiation. Several groups of spectral variables have been identified as being of value in characterizing the way in which reflectance spectra change as pigment concentrations vary and empirical relationships between these spectral variables and pigment concentrations have been defined. An alternative group of approaches for quantifying pigments from hyperspectral data are based on the numerical inversion of physically-based leaf and canopy radiative transfer models.
|
Reflectance in individual narrow wavebands
While there is little agreement on the optimal wavelengths, there is good evidence from modelling studies that, at wavelengths where absorption coefficients of pigments are low, reflectance is more sensitive to high concentrations, while spectral regions with high absorption are more sensitive to lower pigment concentrations (Jacquemoud and Baret, 1990; Yamada and Fujimara, 1991). This is confirmed by empirical evidence which has demonstrated that reflectance at wavelengths corresponding to the lower and upper flanks of the major chlorophyll absorption feature in the red region (i.e. approximately 530–630 nm and a narrower band around 700 nm, respectively) is most sensitive to pigment concentrations over the normal range found in most leaves (Lichtenthaler et al., 1996; Mariotti et al., 1996; Carter and Knapp, 2001; Buscaglia and Varco, 2002; Carter and Spiering, 2002; Zhao et al., 2003) and canopies (Fillella et al., 1995; O'Neill et al., 2002). In studies incorporating low concentrations of pigments, as found in early immature and later senescent leaves and canopies with low leaf area and canopy cover, reflectance at wavelengths corresponding the centre of the major absorption features are most sensitive to pigment concentrations (Blackburn 1998a, b; Sari et al., 2005).
Reflectance in multiple narrow bands
In response to difficulties in relating reflectance in individual narrow bands to pigment concentration due to variations in the multiple controlling factors on reflectance at leaf and canopy scales (see section entitled ‘The potential for remote quantification of pigments’) and the observed non-linearity of relationships between pigments and reflectance, many analytical approaches have used reflectance in multiple narrow bands. Most so-called ‘spectral pigment indices’ have employed ratios of narrow bands within areas of the spectrum that are sensitive to pigments and those areas not sensitive and/or related to some other control on reflectance. Thus, pigment indices have been proposed as a means of solving the problems of the overlapping absorption spectra of different pigments (Chappelle et al., 1992) and the effects of leaf structure, leaf surface interactions, and canopy structure (Peñuelas et al., 1995). Most pigment indices have been derived for Chl quantification and are based on ratios of narrow bands in the visible and near-infrared (Schepers et al., 1996; Blackburn 1998a, b), while some use ratios of narrow bands only in the visible (Filella et al., 1995), in the red edge region (Vogelman et al., 1993), or in the near-infrared shoulder and red edge regions (Gitelson and Merzlyak, 1997). Most of these indices employ reflectance in two narrow wavebands, although some use three (Chappelle et al., 1992); most relate to total Chl, while some differentiate Chl a and Chl b (Datt, 1998); the majority have been developed at the leaf scale. le Maire et al. (2004) provide a comprehensive listing of the Chl spectral indices published until 2002.
Recent work has tended to focus on improving the robustness and generality of Chl indices by testing and refinement over a range of species and physiological conditions. Interestingly, much of this work has resulted in the incorporation of more narrow wavebands, usually chosen from the near-infrared, red edge, and visible regions. Indices using three bands have been found to be generally applicable at the leaf scale (Sims and Gamon, 2002; Gitelson et al., 2003; le Maire et al., 2004), while at the canopy scale both three (Dash and Curran, 2004; Gitelson et al., 2005) and four band indices (Thenkabail et al., 2002) have been developed. The merits of exploiting reflectance in several narrow bands are further demonstrated in studies that have attempted to quantify leaf and canopy Chl concentrations by subjecting multiple bands and, in some cases, the whole spectrum to principal components (Yao and Tian, 2003) and factor analysis (Coops et al., 2002), artificial neural networks (Tumbo et al., 2002; Chen et al., 2006) and stepwise multiple regression (O'Neill et al., 2002; Osborne et al., 2002). However, it has been found, when working on a single crop type, that a normalized difference ratio employing just two narrow wavebands was superior to a multiple regression model using five wavebands (Goel et al., 2003) or a partial least squares regression model based on the whole spectrum (Hansen and Schjoerring, 2003). Indeed, other work using single species has confirmed the efficiency of ratios of reflectance in two narrow wavebands selected from the near-infrared, red edge, and visible regions for Chl estimation for leaves (Zhao et al., 2005) and even at the canopy scale (Coops et al., 2003; Ferri et al., 2004). Overall, this indicates that, in an attempt to identify universal Chl indices, it may be important to incorporate more of the sources of spectral variability in plant reflectance data by using several narrow wavebands, while just two wavebands can result in a more accurate predictive model when calibrated for a specific plant type.
Most research has focused on Chls and less attention has been paid to quantifying Cars from reflectance spectra. Significant overlap in the absorption features of Chl and Cars (Fig. 1) and the low concentrations of Cars with respect to Chl in most leaves can present difficulties in defining suitable spectral indices for Cars. Spectral indices based on three narrow bands in the near-infrared, red edge, and visible regions of the spectrum, have shown promise for estimating the ratio of Cars:Chl (Peñuelas et al., 1995; Merzlyak et al., 1999). Robust measures of leaf Cars:Chl ratios over a range of species have been obtained using the photochemical reflectance index (PRI) (Sims and Gamon, 2002). PRI is a normalized difference ratio of reflectance in narrow bands either side of the peak in green reflectance (550 nm) and was originally developed to estimate short-term fluctuations in different xanthophyll pigments (Gamon et al., 1992). A number of spectral indices have been developed to estimate total Cars from reflectance (Chappelle et al., 1992; Blackburn, 1998a; Datt, 1998), but these have tended to have somewhat limited transferability across species or scales (Blackburn, 1998b; Sims and Gamon, 2002). Nevertheless, an index employing three narrow bands in the visible and red edge (i.e. 510 nm, 550 or 700 nm) and near-infrared regions appears to offer a robust solution at the leaf scale (Gitelson et al., 2002; Gitelson and Merzlyak, 2004).
Only a small number of studies have examined the opportunities for quantifying Anths concentration from reflectance spectra. Gamon and Surfus (1999) observed increased reflectance in the red region associated with increased Anths in immature leaves and proposed a broadband red:green ratio as an index of Anths. However, a subsequent test of this index using a number of different species found no relationship with Anths (Sims and Gamon, 2002). This concurs with the findings of Neill and Gould (1999) who found broadband red reflectance to be independent of leaf Anths concentration. The presence of Chl is the main factor which obscures relationships between reflectance and Anths. In an attempt to overcome this problem, an index has been proposed that uses narrow band green (550 nm) and early red edge (700 nm) reflectance (Gitelson et al., 2001). This approach was refined to account for the effects of varying leaf cellular structure by also incorporating a near-infrared band and the index was found to be an accurate predictor of Anths over a range of species (Gitelson and Merzlyak, 2004).
Derivative spectra and continuum removal
A first derivative spectrum (Fig. 3) displays the variation with wavelength in the slope of the original reflectance spectrum (Fig. 2). Likewise, a second derivative calculates the slope of the first derivative, and so on. It has been suggested that spectral derivatives have important advantages over spectral reflectance, such as their ability to reduce variability due to changes in illumination or soil/litter reflectance (Curran et al., 1991; Elvidge and Chen, 1995). The majority of workers have used derivatives to define the wavelength position of the red edge (
RE) and have illustrated relationships between RE and total chlorophyll (Chl tot) concentration for both leaves and canopies (Lichtenthaler et al., 1996; Mariotti et al., 1996; Jago et al., 1999). A common method for defining RE has been to identify the wavelength position of the peak (maximum amplitude) of the first derivative in the region of the red edge. This peak can be clearly seen in Fig. 2 at approximately 710 nm. A number of alternative methods for calculating RE based on first and second derivatives have also been proposed (Miller et al., 1990; Dawson and Curran, 1998).
|
However, several studies have identified the existence of two or more peaks in first derivative spectra in the region of the red edge (Boochs et al., 1990; Clevers et al., 2004; Smith et al., 2004) and this can lead to a bimodal distribution of
RE values corresponding to low and high Chl concentrations. Thus, RE tends to jump at a certain threshold of Chl and this prevents predictive relationships being established between RE and Chl. The existence of double or multiple peaks in first derivative spectra has been attributed to the characteristic absorption spectra of Chl (le Maire et al., 2004) with a possible contribution from Chl fluorescence (Zarco-Tejada et al., 2003). In an attempt to overcome the problems of the RE jump, le Maire et al. (2004) introduced a new index that is based on the difference between the amplitude of two bands (approximately 30 nm wide) within the lower and upper wavelength sides of the red edge feature in the first derivative. This ‘Double Difference Index’ showed good predictive accuracy for Chl at the leaf scale. An alternative solution is provided by Cho and Skidmore (2006) who developed a method for calculating RE that does not depend on characterizing a peak in the first derivative. The technique uses linear extrapolation of straight lines fitted to the flanks of the red edge feature in the first derivative. This avoids the RE jump and ensures that RE varies continuously and is closely correlated with Chl concentration (measured as total nitrogen in their study). Furthermore, the authors demonstrate how this approach can be applied to imaging spectrometer data that may have a lower spectral resolution than field or laboratory-based spectroradiometers. The amplitude of first and second derivatives of reflectance at individual and multiple combinations of wavelengths has been correlated with Chl concentrations at leaf and canopy scales (Boochs et al., 1990; Adams et al., 1999; Endo et al., 2000) as has the amplitude of the first derivative of pseudo absorbance (the logarithm of the reciprocal of reflectance) (Yoder and Pettigrew-Crosby, 1995; Blackburn, 1999). A small number of studies have also demonstrated relationships between the amplitude of spectral derivatives and Cars at leaf and canopy scales (Blackburn, 1998b; Blackburn and Steele, 1999; Guan et al., 2005). A ‘Brown Pigment Index’ has been developed using second derivative spectra to measure the distance between the position of the reflectance red edge and the wavelength with the minimum reflectance slope in the near-infrared (Peñuelas et al., 2004). This index was correlated with the concentrations of oxidized compounds in stressed leaves.
Measurements of absorption feature depths have been obtained by fitting a continuum to vegetation reflectance spectra (Kokaly and Clark, 1999), which is a similar approach to the use of spectral derivatives, in the sense that it can enhance spectral features of interest while suppressing unwanted effects. This approach has been extended by normalizing to the band depth at the centre and the area of the absorption feature (Curran et al., 2001). Stepwise regression was then used to identify optimal combinations of band depths for providing accurate estimates of Chl tot, a and b in dried and ground pine needles. Furthermore, continuum removal analysis was found to be superior to the use of standard derivative analysis for quantifying tree canopy biochemical properties from airborne imaging spectrometry data, though total N was examined rather than pigment concentrations specifically (Huang et al., 2004).
Radiative transfer (RT) models
A number of physically based models have been developed which account for the interactions of incident radiation with leaves and canopies and are able to simulate reflectance spectra. In the context of the remote sensing of pigments, such models have been used in both forward and inverse modes. In forward mode, RT models have often been used to generate large databases of reflectance spectra using wide ranges and combinations of the input variables that describe the biochemical and biophysical properties. Such databases have then been used to identify spectral variables or indices that are sensitive and robust predictors of pigment concentrations. At the leaf scale, such studies have been undertaken by using the PROSPECT (le Maire et al., 2004) and LIBERTY models (Coops and Stone, 2005). Other studies have been carried out using coupled leaf and canopy RT models to understand the way in which leaf reflectance properties are influenced by the larger number of controlling factors at canopy scale (Demarez and Gastellu-Etchegorry, 2000). Coupled models have enabled the development and refinement of spectral indices, for estimating foliar Chl concentrations from canopy reflectance data, that are insensitive to factors such as canopy structure, illumination geometry, and soil/litter reflectance (Broge and Leblanc, 2000; Daughtry et al., 2000). Such approaches have also been used in defining predictive relationships that have been applied to hyperspectral imagery to generate maps of Chl (Haboudane et al., 2002; Zarco-Tejada et al., 2004b, 2005a).
The numerical inversion of RT models based on measured reflectance spectra has also been used to quantify pigments. At the leaf scale, Chl has been predicted accurately using the PROSPECT (Jacquemoud et al., 1996; Demarez et al., 1999; Renzullo et al., 2006) and LEAFMOD (Ganapol et al., 1998) models in inverse mode. The inversion of coupled leaf and canopy RT models has been demonstrated for quantifying Chl from canopy hyperspectral data (Jacquemoud et al., 1995, 2000). A useful hybrid approach has been proposed which uses spectral indices that are correlated with foliar pigment concentrations as the merit (error minimization) function in the numerical inversion of RT models (Zarco-Tejada et al., 2001). It has been demonstrated that an index based on the ratio of reflectance in narrow bands in the red edge (e.g. at 750 and 710 nm), provides a suitable merit function for the inversion of coupled leaf and canopy models, with which to predict Chl from airborne hyperspectral data (Zarco-Tejada et al., 2001, 2004a; Sampson et al., 2003).
RT model inversion techniques offer the potential of a more generic approach to quantifying pigments from hyperspectral data than spectral pigment indices and other approaches that are based on empirical calibrations. However, intercomparisons have demonstrated that the predictive accuracy of different canopy models varies according to the canopy variable being estimated (Jacquemoud et al., 2000) and the method used for the inversion (Weiss et al., 2000). It has also been recognized that measurement and model uncertainities and model equifinality make inversion an ill-posed problem (Combal et al., 2002). Solutions can be found by using a priori information concerning the canopy under investigation, but this requirement could restrict the operational use of the inverse modelling approach.
|
Prospects and future directions |
|---|
|
TopAbstractIntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
Intercomparison of hyperspectral approaches
When attempting to identify a robust, generic solution, there is currently only limited evidence with which to rank the performance of the range of existing hyperspectral approaches for quantifying plant pigments. Many studies have tested a number of approaches using data from only one or a small number of different species and there has been a paucity of comprehensive testing of the whole range of approaches outlined above across many species. Useful intercomparisons have been performed across a large number of species using a limited number of different hyperspectral approaches (Sims and Gamon, 2002) and across a limited number of species using a large number of different approaches (le Maire et al., 2004). Interestingly, there is some agreement between these studies in terms of the optimal approaches that are identified. Both studies found that the best predictors of leaf Chl concentrations were simple ratio and normalized difference ratios that were modified to incorporate a waveband in the blue region as a correction for leaf surface (specular) reflectance. However, the wavelengths of the optimum three narrow bands for use in these indices differed between the two studies. Intercomparison studies such as these could usefully be extended by incorporating hyperspectral approaches to pigment quantification that are based on the numerical inversion of RT models, by testing methods at the canopy scale as well as the leaf scale and evaluating approaches for estimating independently Chl a and b, Cars, and Anths.
Scaling
Most research in this field has concentrated upon deriving relationships between the reflectance properties of leaves and their pigment concentrations. Indeed, such work has indicated that the best reflectance based indices performed better that the current commercial instrument (SPAD) for non-destructive Chl estimates (Richardson et al., 2002; le Maire et al., 2004). These leaf scale techniques may be valuable in some applications, but, in order to exploit the opportunities offered by imaging spectrometry for synoptic, simultaneous, and spatially continuous information, it is important that suitable methods are available for deriving estimates of foliar pigment concentrations from canopy scale reflectance spectra. Several strategies are available for the analysis of canopy spectra (Zarco-Tejada et al., 2001), these involve: applying relationships between reflectance and pigments derived at the leaf scale directly to canopy reflectance spectra; deriving relationships directly between canopy reflectance and ground-based measurements of pigment concentrations; using canopy RT models to scale up relationships between reflectance and pigments derived at the leaf scale; inverting canopy reflectance through coupled canopy and leaf RT models in order to estimate pigment concentrations; using hybrid approaches that employ a combination of the previous four strategies. While the advantages and disadvantages of these different strategies are reasonably well characterized in general terms and there are some useful studies that have explicitly tested one or two of the techniques, there is scope for much more extensive and systematic intercomparison. Research could usefully be directed towards producing guidelines to potential users of hyperspectral remote sensing of pigments. It may be beneficial to assess the relative merits of the five strategies according to: (i) different plant functional types or crop types; (ii) user requirements for absolute or relative accuracy of pigment estimates; (iii) the need to develop a local solution or a more general one that will operate across a range of vegetation types or environments; (iv) the requirement for data concerning pigments per unit leaf mass (often referred to as concentration) or per unit ground area (density); (v) the specific pigment or combination of pigments that are being quantified; (vi) the sensing and analytical capabilities of intended users; and (vii) the availability of prior information and ancillary data concerning the vegetation under investigation. There are also questions concerning the most effective strategies at particular scales of investigation and whether different approaches, within the limits of currently available relationships or models, become optimal as the investigations move from individual plants through canopies, stands, fields and communities up to landscapes, regions and biomes.
New methods of hyperspectral data analysis
A number of analytical approaches have emerged mainly from laboratory spectroscopy studies which hold promise for quantifying pigments from hyperspectral data. The capabilities of techniques such as modified Gaussian modelling (Sunshine and Pieters, 1993) and independent component analysis (Chen and Wang, 2001) for decomposing spectra and modelling component absorption features have been demonstrated and are worthy of further investigation in the present context. However, a technique that is already receiving some scrutiny and is showing considerable promise is that of wavelet decomposition (Graps, 1995). In laboratory spectroscopy, the ability of wavelet decomposition to remove the effects of background spectral variation when quantifying concentrations of components with fine absorption from mixtures has been demonstrated (Mittermayr et al., 2001). In the present context, this offers the potential for both removing the effects of broader absorption features from the narrower features of specific pigments and for dealing with factors which affect broader regions of vegetation reflectance spectra such as the leaf or canopy structure and soil/litter response. It has also been shown in the context of quantitative spectroscopy that wavelet decomposition has a capacity for noise suppression and it can deal with difficult situations where background varies between calibration and prediction data sets. These are potentially valuable characteristics for dealing with hyperspectral remotely-sensed imagery of vegetation. Wavelet decomposition has the potential to capture more of the information contained within high resolution spectra than narrow-band index approaches and offers the prospect of developing robust and extendible methods for pigment determinations.
The capabilities of a wavelet-based technique have been examined using reflectance spectra and pigment data collected for a range of plant species at leaf and canopy scales (Blackburn, 2006). For the combined data set and all of the individual vegetation types, methods based on wavelet decomposition appreciably outperformed the narrow-band spectral indices tested and stepwise selection of narrowband reflectance. However, the relative performance of the three different techniques employed for selecting the wavelet coefficients for use in predictive models varied between vegetation types. There was considerable variability in the performance of predictive models according to the wavelet function used for spectral decomposition and the optimum wavelet functions differed between vegetation types and between individual pigments within the same vegetation type. The research indicates that wavelet analysis holds promise for the accurate determination of Chl a and b and Cars, but further work is needed to refine the approach. Such findings are supported by other preliminary work that has tested the abilities of wavelet decomposition for deriving predictive models for the remote sensing of total leaf nitrogen (Ferwerda and Jones, 2005) and Chl (Reum and Zhang, 2005).
Pigments under investigation
There is considerable scope for more extensive investigations of the hyperspectral remote sensing of pigments other than total Chl. The small number of studies that have investigated Cars and Anths, and even Chl a and b individually, belie the potential that information on their concentrations could provide. While a number of studies have demonstrated useful spectral indices for estimating these pigments at the leaf scale, the transferability across species and up to the canopy scale needs further investigation. Moreover, the range of strategies (as discussed in the section entitled ‘Scaling’) for achieving estimates of pigments other than Chl from canopy measurements have yet to be tested. Of particular importance in achieving this will be the extension of leaf and coupled leaf and canopy RT models to incorporate explicitly and individually the absorption coefficients of the range of pigments other than total Chl.
Some pigments, namely phytochromes, cryptochrome, and phototropin, present within leaves act as photoreceptors and allow plants to sense and respond to variations in the intensity, quality, and direction of solar irradiation through photomorphogenesis. While photoreceptor concentrations are very small compared to Chls, Cars, and Anths, they do have distinct absorption spectra and some impact upon the reflectance spectra of leaves and this offers the opportunity of characterizing photoreceptors in situ from reflectance spectra. While spectral indices designed for the major groups of pigments may ignore or discard residual information contained within the spectrum concerning photoreceptors, it may be possible that spectral decomposition techniques such as wavelet analysis could be employed, which have been shown to be sensitive to Cars in the presence of much higher concentrations of Chl (Blackburn, 2006). Indeed, spectroscopic methods have already been successfully developed to measure phytochromes in the presence of chlorophyll in leaf extracts (Lamparter et al., 1994) and spectral indices for phytochromes have been applied to airborne hyperspectral imagery (Almeida and De Souza Filho, 2004), although their predictive accuracy was not tested. The composition, genesis, and functional roles of photoreceptors is under intense investigation and the development of non-invasive techniques for characterizing photoreceptors at leaf and canopy scales would facilitate substantial advances in this area of plant physiology.
Fusion of hyperspectral and LIDAR remote sensing
At the canopy scale, structural variables such as LAI, plant architecture, and the three-dimensional distribution of different plants can all have important influences on measured reflectance spectra and generally act to confound relationships between reflectance and leaf pigment concentrations. A possibility for measuring these structural properties is offered by LIDAR (Light Detection and Ranging) remote sensing. Large-footprint LIDARs that record the entire return pattern of each laser pulse enable detailed assessment of vegetation vertical structure, but with limited spatial sampling (a transect) and resolution. The more widely available small-footprint discrete return imaging LIDARs offer full spatial coverage and high spatial resolution. It has been demonstrated that forest canopy surface elevation models derived from such instruments can be used to create spatial filters that can be applied to (h-resolution) imaging spectrometer data from forests to extract spectral information from tree crowns, while removing extraneous spectral information from canopy gaps (Blackburn, 2002). This approach improved the accuracy of pigment estimates at the stand scale. Furthermore, recent investigations have demonstrated the possibility of measuring canopy LAI from imaging LIDAR (Riano et al., 2004) and using this to scale-up estimates of foliar Chl concentration, derived from hyperspectral data, to quantify total canopy Chl content (Solberg et al., 2005). In addition, in the context of the hyperspectral remote sensing of pigments, LIDAR data could be used to derive site-specific model parameter ranges in generating look up tables for RT model inversion (Kotz et al., 2004). Therefore, particularly for heterogenous canopies, there is considerable scope for further research to investigate the combined use of LIDAR and hyperspectral data.
Emerging applications
While the analytical approaches for hyperspectral remote sensing of plant pigments continue to be developed and refined, a wide range of applications of these techniques are emerging across various spatial scales. At the leaf scale, pigment indices applied to high resolution hyperspectral images of leaves has revealed the spatial patterns of damage to tissue and photosynthetic functioning in response to freezing and interacting light and temperature stresses (Nicotra et al., 2003). Pigment indices applied to leaf spectra have provided efficient and sensitive measurements of large numbers of leaves that have been used to reveal variations in physiological stress along an elevation gradient and differences in tolerance between two coniferous species (Richardson et al., 2001) and to assess the magnitude of physiological response to moderate changes in light and soil nutrient availability in temperate deciduous trees (Baltzer and Thomas, 2005). The non-destructive in situ monitoring of apple development has been achieved by using hyperspectral measurements of Chl and Anths and this has provided a reliable method for determining optimum harvest date (Herold et al., 2005). In the context of precision agriculture, pigment indices applied to data from field-based hyperspectral instruments have been used to determine crop nitrogen fertilizer requirements (Zillman et al., 2006), to identify high grain-yielding genotypes (Gutierrez-Rodriguez et al., 2004), to predict wheat grain protein content (Wang et al., 2004), and to quantify gross primary production (Gitelson et al., 2006). Pigment indices applied to airborne imaging spectrometry data have been used to produce accurate maps of within-field yield variability in cotton crops (Zarco-Tejada et al., 2005b), the physiological condition of forests (Sampson et al., 2000) and swamp vegetation (Gan et al., 2002), and the distribution of different vegetation types (Fuentes et al., 2001; Almeida and De Souza Filho, 2004). This array of examples reflects the increasing value of hyperspectral remote sensing of plant pigments across a range of pure and applied scientific disciplines.
|
Conclusion |
|---|
|
TopAbstractIntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
The dynamics of plant pigments relate strongly to the physiological status of plants, therefore information concerning the temporal and spatial variations of pigments can be a valuable indicator of a range of key properties and processes in vegetation and the wider ecosystem. The forthcoming generation of spaceborne hyperspectral sensors in combination with airborne and field instruments will open up the possibility of acquiring high resolution reflectance spectra from leaf to biome scales. Research over the last few decades has focused on developing analytical methods for extracting pigment concentrations from hyperspectral data and some convincing approaches have been developed, although no universal solutions have emerged. Progress can be made through extended intercomparison of techniques, the parallel refinement of empirically-derived approaches and RT modelling, and emphasis on defining the optimum strategies, given different circumstances and user requirements, for scaling methods up to the canopy level. A series of technological and analytical opportunities exist for enhancing the robustness, extendability and scope of hyperspectral methods for quantifying plant pigments. A range of studies have already benefited from the current capabilities and it can be anticipated that improvements in techniques will promote a wider diversity of applications.
|
References |
|---|
|
TopAbstractIntroductionThe development of hyperspectral...Prospects and future directionsConclusionReferences |
|---|
Adams ML, Philpot WD, Norvell WA. (1999) Yellowness index: an application of spectral 2nd derivatives to estimate chlorosis of leaves in stressed vegetation. International Journal of Remote Sensing 20 3663–3675.[CrossRef]
Almeida TIR and De Souza Filho CR. (2004) Principal component analysis applied to feature-oriented band ratios of hyperspectral data: a tool for vegetation studies. International Journal of Remote Sensing 25 5005–50023.[CrossRef]
Archetti M. (2000) The origin of autumn colours by coevolution. Journal of Theoretical Biology 205 625–630.[CrossRef][ISI][Medline]
Archetti M and Brown SP. (2004) The coevolution theory of autumn colours. Proceedings of the Royal Society of London B 271 1219–1223.[Medline]
Baltzer JL and Thomas SC. (2005) Leaf optical responses to light and soil nutrient availability in temperate deciduous trees. American Journal of Botany 92 214–223.
Bartley GE and Scolnik PA. (1995) Plant carotenoids: pigments for photoprotection, visual attraction, and human health. The Plant Cell 7 1027–1038.[CrossRef][ISI][Medline]
Barker DH, Seaton GGR, Robinson SA. (1997) Internal and external photoprotection in developing leaves of the CAM plant Cotyledon orbiculata. Plant, Cell and Environment 20 617–624.[CrossRef]
Blackburn GA. (1998a) Spectral indices for estimating photosynthetic pigment concentrations: a test using senescent tree leaves. International Journal of Remote Sensing 19 657–675.[CrossRef]
Blackburn GA. (1998b) Quantifying chlorophylls and carotenoids from leaf to canopy scales: an evaluation of some hyperspectral approaches. Remote Sensing of Environment 66 273–285.[CrossRef]
Blackburn GA. (1999) Relationships between spectral reflectance and pigment concentrations in stacks of broadleaves. Remote Sensing of Environment 70 224–237.[CrossRef][ISI]
Blackburn GA. (2002) Remote sensing of forest pigments using airborne imaging spectrometer and LIDAR imagery. Remote Sensing of Environment 82 311–321.[CrossRef]
Blackburn GA. (2006) Wavelet decomposition of hyperspectral data: a novel approach to quantifying pigment concentrations in vegetation. International Journal of Remote Sensing (in press).
Blackburn GA and Steele CM. (1999) Towards the remote sensing of matorral vegetation physiology: relationships between spectral reflectance, pigment and biophysical characteristics of semi-arid bushland canopies. Remote Sensing of Environment 70 278–292.[CrossRef]
Boochs F, Dockter K, Kupfer G, Kuhbauch W. (1990) Shape of the red edge as a vitality indicator for plants. International Journal of Remote Sensing 11 1741–1754.[CrossRef]
Broge NH and Leblanc E. (2000) Comparing the prediction power and stability of broadband and hyperspectral vegetation indices for estimation of green leaf area index and canopy chlorophyll density. Remote Sensing of Environment 76 156–172.[CrossRef]
Broge NH and Mortensen JV. (2002) Deriving green crop area index and canopy chlorophyll density of winter wheat from spectral reflectance data. Remote Sensing of Environment 81 45–57.
Buscaglia HJ and Varco JJ. (2002) Early detection of cotton leaf nitrogen status using leaf reflectance. Journal of Plant Nutrition 25 2067–2080.[CrossRef]
Carter GA and Knapp AK. (2001) Leaf optical properties in higher plants: linking spectral characteristics to stress and chlorophyll concentration. American Journal of Botany 88 677–684.
Carter GA and Spiering BA. (2002) Optical properties of intact leaves for estimating chlorophyll concentration. Journal of Environmental Quality 31 1424–1432.
Chalker-Scott L. (1999) Environmental significance of anthocyanins in plant stress responses. Photochemistry and Photobiology 70 1–9.
Chappelle EW, Kim MS, McMurtrey JE. (1992) Ratio analysis of reflectance spectra (RARS): an algorithm for the remote estimation of the concentrations of chlorophyll A, chlorophyll B and the carotenoids in soybean leaves. Remote Sensing of Environment 39 239–247.[CrossRef][ISI]
Chen J and Wang XZ. (2001) A new approach to NIR spectral data analysis using independent component analysis. Journal of Chemical Information and Computer Sciences 41 992–1001.[CrossRef][ISI][Medline]
Chen L, Huang JF, Wang FM, Tang YL. (2006) Comparison between backpropagation neural network and regression models for estimation of pigment content in rice leaves and panicles using hyperspectral data. International Journal of Remote Sensing (in press).
Cho MA and Skidmore AK. (2006) A new technique for extracting the red edge position for hyperspectral data: the linear extrapolation method. Remote Sensing of Environment 101 181–193.[CrossRef]
Clevers JGPW, Kooistra L, Salas EAL. (2004) Study of heavy metal contamination in floodplains using the red edge position in spectroscopic data. International Journal of Remote Sensing 25 3883–3895.
Close DC and Beadle CL. (2003) The ecophysiology of foliar anthocyanin. The Botanical Review 69 149–161.[CrossRef]
Combal B, Baret F, Weiss M, Trubuil A, Mace D, Pragnere A, Myneni R, Knyazikhin Y, Wang L. (2002) Retrieval of canopy biophysical variables from bidirectional reflectance: using prior information to solve the ill-posed inverse problem. Remote Sensing of Environment 84 1–15.[CrossRef]
Coops N, Drury S, Smith ML, Martin M, Ollinger S. (2002) Comparison of green leaf eucalypt spectra using spectral decomposition. Australian Journal of Botany 50 567–576.[CrossRef]
Coops NC and Stone C. (2005) A comparison of field-based and modelled reflectance spectra from damaged Pinus radiata foliage. Australian Journal of Botany 53 417–429.[CrossRef]
Coops NC, Stone C, Culvenor DS, Chisholm LA, Merton RN. (2003) Chlorophyll content in eucalypt vegetation at the leaf and canopy scales as derived from high spectral resolution data. Tree Physiology 23 23–31.[ISI][Medline]
Curran PJ, Dungan JL, Macler BA, Plummer SE. (1991) The effect of a red leaf pigment on the relationship between red-edge and chlorophyll concentration. Remote Sensing of Environment 35 69–75.[CrossRef][ISI]
Curran PJ, Dungan JL, Peterson DL. (2001) Estimating the foliar biochemical concentration of leaves with reflectance spectrometry. Remote Sensing of Environment 76 349–359.[CrossRef]
Dash J and Curran PJ. (2004) The MERIS terrestrial chlorophyll index. International Journal of Remote Sensing 25 5403–5413.[CrossRef]
Datt B. (1998) Remote sensing of chlorophyll a, chlorophyll b, chlorophyll a+b and total carotenoid content in eucalyptus leaves. Remote Sensing of Environment 66 111–121.[CrossRef][ISI]
Daughtry CST, Walthall CL, Kim MS, Brown de Colstoun E, McMurtrey JE. (2000) Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance. Remote Sensing of Environment 74 229–239.[CrossRef]
In Davies KM (Ed.). Plant pigments and their manipulation. Annual plant reviews (2004) Oxford, UK Blackwell Publishing Vol. 14.
Dawson TP and Curran PJ. (1998) A new technique for interpolating the reflectance red edge position. International Journal of Remote Sensing 19 2133–2139.[CrossRef]
Demarez V and Gastellu-Etchegorry P. (2000) A modelling approach for studying forest chlorophyll content. Remote Sensing of Environment 71 226–238.[CrossRef]
Demarez V, Gastellu-Etchegorry P, Mougin E, Marty G, Proisy C, Dufrene E, Le Dantec V. (1999) Seasonal variation of leaf chlorophyll content of a temperate forest. Inversion of the PROSPECT model. International Journal of Remote Sensing 20 879–894.[CrossRef]
Demming-Adams B and Adams WW. (1996) The role of xanthophyll cycle carotenoids in the protection of photosynthesis. Trends in Plant Science 1 21–27.
Dunn JL, Turnbull JD, Robinson SA. (2004) Comparison of solvent regimes for the extraction of photosynthetic pigments from leaves of higher plants. Functional Plant Biology 31 195–202.[CrossRef]
Elvidge CD and Chen Z. (1995) Comparison of broad-band and narrow-band red and near-infrared vegetation indices. Remote Sensing of Environment 54 38–48.[CrossRef]
Endo T, Okuda T, Tamura M, Yasuoka Y. 2000. Estimation of photosynthetic rate of plants from hyperspectral remote sensing of biochemical content. 21st Asian conference on remote sensing, December 4–8, 2000, Taipei, Taiwan. Online proceedings: http://www.gisdevelopment.net/aars/acrs/2000/.
Fang Z, Bouwkamp J, Solomos T. (1998) Chlorophyllase activities and chlorophyll degradation during leaf senescence in non-yellowing mutant and wild type of Phaseolus vulgaris L. Journal of Experimental Botany 49 503–510.
Ferri CP, Formaggio AR, Schiavinato MA. (2004) Narrow band spectral indexes for chlorophyll determination in soybean canopies [Glycine max (L.) Merril]. Brazilian Journal of Plant Physiology 16 131–136.
Ferwerda JG and Jones S. (2005) Continuous wavelet transformations for hyperspectral feature detection. Proceedings of the 12th international symposium on spatial data handlingAustria University of Vienna pp. 12–14 July.
Filella I, Serrano L, Serra J, Peñuelas J. (1995) Evaluating wheat nitrogen status with canopy reflectance indices and discriminant analysis. Crop Science 35 1400–1405.
Fuentes DA, Gamon JA, Qui HL, Sims DA, Roberts DA. (2001) Mapping Canadian boreal forest vegetation using pigment and water absorption features derived from the AVIRIS sensor. Journal of Geophysical Research – Atmospheres 106 33565–33577.[CrossRef]
Gamon JA, Peñuelas J, Field CB. (1992) A narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency. Remote Sensing of Environment 41 35–44.[Medline]
Gamon JA and Surfus JS. (1999) Assessing leaf pigment content and activity with a reflectometer. New Phytologist 143 105–117.[CrossRef][ISI]
Gan FP, Wang RS, Ma AN, Yang SM. (2002) investigation on physiological status of regional vegetation using pushbroom hyperspectral imager data. Acta Botanica Sinica 44 983–989.
Ganapol BD, Johnson LF, Hammer PD, Hlavka CA, Peterson DL. (1998) LEAFMOD: a new within-leaf radiative transfer model. Remote Sensing of Environment 63 182–193.[CrossRef]
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.[CrossRef][ISI][Medline]
Gitelson AA and Merzlyak MN. (1997) Remote estimation of chlorophyll content in higher plant leaves. International Journal of Remote Sensing 18 2691–2697.[CrossRef][ISI]
Gitelson AA and Merzlyak MN. (2004) Non-destructive assessment of chlorophyll, carotenoids and anthocyanin content in higher plant leaves: principles and algorithms. In Stamatiadis S, Lynch JM, Shepers JS (Eds.). Remote sensing for agriculture and the environmentGreece Ella pp. 78–94.
Gitelson AA, Merzlyak MN, Chivkunova OB. (2001) Optical properties and nondestructive estimation of anthocyanin content in plant leaves. Photochemistry and Photobiology 74 38–45.[CrossRef][ISI][Medline]
Gitelson AA, Vina A, Ciganda V, Rundquist DC. (2005) Remote estimation of canopy chlorophyll content in crops. Geophysical Research Letters 32 Art. No. L08403.
Gitelson AA, Vina A, Verma SB, Rundquist DC, Arkebauer TJ, Keydan G, Leavitt B, Ciganda V, Burba GG, Suyker AE. (2006) Relationship between gross primary production and chlorophyll content in crops: implications for the synoptic monitoring of vegetation productivity. Journal of Geophysical Research – Atmospheres 111 Art. No. DO8S11.
Gitelson AA, Zur Y, Chivkunova OB, Merzlyak MN. (2002) Assessing carotenoid content in plant leaves with reflectance spectroscopy. Photochemistry and Photobiology 75 272–281.[CrossRef][ISI][Medline]
Goel PK, Prasher SO, Landry JA, Patel RM, Viau AA, Miller JR. (2003) Estimation of crop biophysical parameters through airborne and field hyperspectral remote sensing. Transactions of the ASAE 46 1235–1246.
Gould KS, McKelvie J, Markham KR. (2002) Do anthocyanins function as antioxidants in leaves? Imaging of H2O2 in red and green leaves after mechanical injuryPlant. , Cell and Environment 25 1261–1269.[CrossRef]
Graps A. (1995) An introduction to wavelets. IEEE Computational Science and Engineering 2 50–61.[CrossRef]
Guan Y, Guo S, Liu J, Zhang X. (2005) Algorithms for the estimation of the concentrations of chlorophyll a and carotenoids in rice leaves from airborne hyperspectral data. Lecture Notes in Computer Science 3514 908–915.
Gutierrez-Rodriguez M, Reynolds MP, Escalante-Estrada JA, Rodriguez-Gonzalez MT. (2004) Association between canopy reflectance indices and yield and physiological traits in bread wheat under drought and well-irrigated conditions. Australian Journal of Agricultural Research 55 1139–1147.[CrossRef][ISI]
Haboudane D, Miller JR, Tremblay N, Zarco-Tejada PJ, Dextraze L. (2002) Integrated narrow band vegetation indices for prediction of crop chlorophyll content for application to precision agriculture. Remote Sensing of Environment 81 416–426.[CrossRef]
Hamilton WD and Brown SP. (2001) Autumn tree colours as a handicap signal. Proceedings of the Royal Society of London B 268 1489–1493.[Medline]
Hansen PM and Schjoerring JK. (2003) Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression. Remote Sensing of Environment 86 542–553.[CrossRef]
Härtel H and Grimm B. (1998) Consequences of chlorophyll deficiency for leaf carotenoids composition in tobacco synthesizing glutamate 1-semialdehyde aminotransferase antisense RNA: dependency on developmental age and growth light. Journal of Experimental Botany 49 535–546.
Herold B, Truppel I, Zude M, Geyer M. (2005) Spectral measurements on ‘Elstar’ apples during fruit development on the tree. Biosystems Engineering 91 173–182.[CrossRef]
Huang Z, Turner BJ, Dury SJ, Wallis IR, Foley WJ. (2004) Estimating foliage nitrogen concentration from HYMAP data using continuum removal analysis. Remote Sensing of Environment 93 18–29.[CrossRef]
Inoue Y and Peñuelas J. (2001) An AOTF-based hyperspectral imaging system for field use in ecophysiological and agricultural applications. International Journal of Remote Sensing 22 3883–3888.[CrossRef]
Jacquemoud S, Bacour C, Poilve H, Frangi JP. (2000) Comparison of four radiative transfer models to simulate plant canopies reflectance: direct and inverse mode. Remote Sensing of Environment 74 471–481.[CrossRef]
Jacquemoud S and Baret F. (1990) PROSPECT: a model of leaf optical properties spectra. Remote Sensing of Environment 34 75–91.[CrossRef]
Jacquemoud S, Baret F, Andrieu B, Danson FM, Jaggard K. (1995) Extraction of vegetation biophysical parameters by inversion of the PROSPECT+SAIL models on sugar-beet canopy reflectance data: application to TM and AVIRIS sensors. Remote Sensing of Environment 52 163–172.
Jacquemoud S, Ustin SL, Verdebout J, Schmuck G, Andreoli G, Hosgood B. (1996) Estimating leaf biochemistry using the PROSPECT leaf optical properties model. Remote Sensing of Environment 56 194–202.[CrossRef][ISI]
Jago RA, Cutler MEJ, Curran PJ. (1999) Estimating canopy chlorophyll concentration from field and airborne spectra. Remote Sensing of Environment 68 217–224.[CrossRef]
Jones HG, Archer N, Rotenburg E, Casa R. (2003) Radiation measurement for plant ecophysiology. Journal of Experimental Botany 54 879–889.
Kim MS, Daughtry CST, Chappelle EW, McMurtrey JE, Walthall CL. (1994) The use of high spectral resolution bands for estimating absorbed photosynthetically active radiation. ISPRS sixth international colloquium on physical measurements and signatures in remote sensing, Val d'Isère, France, 17–21 January 1994. European Space Agency, Paris 29 9–306.
Kokaly RF and Clark RN. (1999) Spectroscopic determination of leaf biochemistry using band-depth analysis of absorption features and stepwise linear regression. Remote Sensing of Environment 67 267–287.[CrossRef]
Kotz B, Schaepman M, Morsdorf F, Bowyer P, Itten K, Allgower B. (2004) Radiative transfer modeling within a heterogeneous canopy for estimation of forest fire fuel properties. Remote Sensing of Environment 92 332–344.[CrossRef]
Lamparter T, Hughes J, Hartmann E. (1994) A fully automated dual wavelength photometer for phytochrome measurements from chlorophyll containing extracts. Photochemistry and Photobiology 60 79–183.
Landsberg JJ, Prince SD, Jarvis PG, McMurtrie RE, Luxmoore RJ, Medlyn BE. (1996) Energy conversion and use in forests: an analysis of forest production in terms of radiation utilization efficiency (
). In Gholz HL, Nakane K, Shimoda H (Eds.). The use of remote sensi