Area between Peaks Feature in the Derivative
Reflectance Curve as a Sensitive Indicator of
Change in Chlorophyll Concentration
Eric Ariel L. Salas1 and Geoffrey M. Henebry
Geographic Information Science Center of Excellence (GIScCE), South
Dakota State University, 1021 Medary Ave., Wecota Hall 506B, Brookings,
South Dakota 57007-3510
Abstract: Vegetation spectral features can detect chlorophyll concentrations. Two
key spectral features evident in the first derivative (FD) of reflectance constitute the
two main peaks: one located around 685–705 nm and the other near 710–725 nm.
We propose that the area between peaks (ABP) can be used as a sensitive indicator
of changes in the photosynthetic pigments at leaf level and demonstrate it using a
high-spectral-resolution dataset of maize leaves collected by Gitelson and coworkers
(2005). We find significant high positive correlations (r2 > 0.90) between chloro-
phyll concentrations and both the ABP and its continuum length feature.
INTRODUCTION
The most promising results for detecting the presence of plant stress are obtained
by studying the sharp increase in reflectance of green vegetation normally between
the range 670 and 780 nm, known as the vegetation red-edge, as mentioned by Horler
et al. (1983). To resolve the red-edge feature, relatively narrow spectral bands are
required. Filella and Peñuelas (1994) pointed out that high-spectral-resolution spec-
trometry offers the potential to estimate the relationship between the red-edge and the
chlorophyll concentration for vegetation. Spectroscopy can detect many features in
the reflected spectrum, such as the large increase in reflectance at the red-edge, the
broad absorption features caused by liquid water in the leaf, and perhaps much finer
absorption features caused by biochemicals within the leaf. As Clark (1999)
observed, it gives finer details over a broader wavelength range and with greater
precision.
In spectroscopy, derivative methods are a familiar term for spectral discrimina-
tion. It is a qualitative and quantitative technique to accentuate and assess small
structural differences between spectra as defined by O’Haver et al. (1982). Research-
ers thus far have addressed applications of spectral derivatives in remote sensing
(Demetriades-Shah et al., 1990; Philpot, 1991; Peñuelas et al., 1994). In the case of
vegetation spectra, the first derivative curve reveals the red-edge inflection points
1Corresponding author: [Link]@[Link]
315
GIScience & Remote Sensing, 2009, 46, No. 3, p. 315–328. DOI: 10.2747/1548-1603.46.3.315
Copyright © 2009 by Bellwether Publishing, Ltd. All rights reserved.
316 SALAS AND HENEBRY
(REIP) through the occurrence of peaks and troughs that may explain physiological
changes of the vegetation. It provides important information about the spectral fea-
tures (Bruce and Li, 2001, p. 1542).
Horler et al. (1983) and Curran et al. (1990) previously demonstrated an explicit
relation between the wavelength of the REIP and the chlorophyll content of leaf sam-
ples. Laboratory studies showed a strong correlation between red-edge position and
leaf chlorophyll content, such as the result from Gitelson et al. (1996). Moreover, the
position of the REIP, as shown in modeling by Guyot et al. (1992), is controlled by
factors such as canopy Leaf Area Index (LAI), leaf chlorophyll content, and leaf incli-
nation angle.
A more in-depth analysis was conducted by Guyot et al. (1992) to compare spec-
tral shifts characterized by the wavelength position of the REIP with varying chloro-
phyll concentration of maple and horse chestnut leaves. Results showed the presence
of four inflection points, shown in four distinguishable peaks at 685–705, near 710,
720, and 740 nm. The same peaks define a vegetation feature termed the “double-
peak” around the red-edge boundaries as found by Zarco-Tejada et al. (2002, p. 1438;
2003, p. 291). Gitelson et al. (1996) related the positions of the each individual peak
to variations of chlorophyll concentrations.
In a study by Kruse et al. (1988), a technique was introduced for extracting spec-
tral features using the continuum. A continuum in a spectrum is defined by finding
the high points and fitting a straight line segment between these points. Applying this
technique to the first derivative spectral peaks can reveal another important vegeta-
tion feature such as the area between two peaks. Salas (2004) used this technique,
dealing with the first two pronounced peaks of the derivative spectra, and saw that the
changing (deepening or narrowing) of the “double-peak” had caused the shape of the
curve within the red-edge range to vary.
By integrating the previous findings of Gitelson and Merzlyak (1994), which
made it possible to create sensitive indices directly related to Chl, with the behaviors
of the area between and under the red-edge double-peak curve, a different view of the
vegetation pigment responses can be obtained. This view may address index limita-
tions (Stark et al., 2000, p. 242) and explain the movement of spectral peaks at longer
wavelengths, which Gitelson et al. (1996) mentioned were not often completely
resolved.
In this paper we aim to describe the change of the shape of the double-peak
through the use of the area between peaks (ABP). Moreover, we evaluate the hypoth-
esis that the area between peaks and the continuum line (Fig. 1) are sensitive to
changes in chlorophyll concentrations, and assess how the movement of peak loca-
tions relates to vegetation status.
MATERIALS AND METHODS
Photosynthetic Pigments
The method in the determination of the leaf level chlorophyll content of maize
follows that in Gitelson et al. (2005), as we used the same dataset. Reflectance mea-
surements were collected in the range 400 to 900 nm using a black plastic polyvinyl
chloride leaf clip, with a 2.3 mm diameter bifurcated fiber-optic attached to both an
CHANGE IN CHLOROPHYLL CONCENTRATION 317
Fig. 1. The first derivative (FD) curve of reflectance with the double-peak, area between peaks
(ABP) (shaded), and the continuum connecting the peaks.
Ocean Optics USB2000 spectroradiometer and to an Ocean Optics LS-1 tungsten
halogen light source. The system has a sampling interval of 0.3 nm and a spectral
resolution of around 1.5 nm (details by Viña et al., 2004).
With the leaf clip, individual leaves are held with a 60° angle relative to the bifur-
cated fiber optic. A Spectralon reflectance standard (99% reflectance) was scanned
for each leaf sample. The reflectance factor at each wavelength was calculated as the
ratio of the upwelling leaf radiance to the upwelling radiance of the standard, and
averaged across 10 separate scans made for each leaf. All scans were corrected for the
instrument’s dark current. After spectral readings, the measured areas of leaves were
punched and total Chl (a and b) was determined analytically.
Area between Peaks and the Continuum
The ABP computation started by filtering the noise using a linear (mixture) com-
bination of means of the field spectra measured from seven points, as used by Datt et
al. (2003). The shapes and features that we expected to find in the hyperspectral data
were thus preserved in the fitted model.
The Pythagorean Theorem proved as an efficient way to compute the length of
the continuum. We applied equation 1, where λ1 is the wavelength of the first peak, λ2
is the wavelength of the second peak, R1 is the FD reflectance of the first peak, and
R2 is the FD reflectance of the second peak.
2 2
Continuum = ( λ1 – λ2 ) + ( R1 – R2 ) . (1)
Location of the peaks and the FD values were the data needed for this step. For
ease of calculation, the SAMS software (Spectral Analysis and Management System,
developed by and downloaded from the Center for Spatial Technologies and Remote
Sensing at the University of California, Davis; [Link] was used to
analyze the absorption feature. Basically, SAMS employs the same principle as the
composite-area analysis discussed by Meriam (1966) to calculate areas. The compos-
ite-area method integrates smaller areas, usually rectangular, into a bigger area. Inte-
gration starts at the first peak, then moves along the wavelength range, and ends at the
location of the second peak. Once areas are computed, a ratio between the area under
318 SALAS AND HENEBRY
the function (in a specified spectrum interval) and the area under the straight line con-
necting the maxima (the continuum line) is performed. Equation 2 shows the absorp-
tion feature equation:
Au
a = 1 – ------- , (2)
Ac
where a is the absorption feature, Au is the area under the curve, and Ac is the area
under the continuum line.
The method by SAMS may be confirmed manually by applying composite-area
analysis, with few integral calculus concepts, integration of the sub-parabolic areas,
and utilization of the equation of the sub-parabolic curve.
Bootstrapping Regression
In bootstrapping statistics, multiple samples are developed by random sampling
with replacement from the sample domain as mentioned by Efron and Tibshirani
(1993). Applying the bootstrap regression is similar to running a regression model for
each bootstrap data set and computing the beta coefficients (β) that are used to predict
the value of the dependent variable, in this case the Chl. In a nutshell, the bootstrap
analogy is applied to test the bootstrapped difference in regression coefficients (β0
and β1 in a regression equation). We also used the method to check other properties of
the general linear regression’s estimators, such as the bias and standard error.
Red-Edge Chlorophyll Index
The red-edge chlorophyll index (CIred-edge) was suggested by Gitelson et al.
(2006) for Chl determination in leaves. The relationship between analytical Chl
and the CIred-edge measured with a linear best-fit function was r2 > 0.94. We chose
CIred-edge to contrast results we derived from ABP. The detail of the index is shown in
equation 3:
CI red-edge = (NIR/Red-edge) – 1, (3)
where NIR is the average reflectance from 770 nm to 800 nm and Red-edge is the
average reflectance in the range 720–730 nm.
RESULTS
The basic statistics of the samples are presented in Table 1. Laboratory analytical
extraction of pigments varied widely, with a broad range of total Chl values from
200.69 to 804.04 mg m–2.
The first derivative of the reflectance of maize is shown in Figure 2, where two
groups are observed. The first group corresponds to the six samples with maximum
peaks occurring at a shorter wavelength, around 685–705 nm, while the second group
has the remaining nine samples with their maximum values at a longer wavelength,
near 710–725 nm. Maximum peaks on a shorter wavelength have range values for
CHANGE IN CHLOROPHYLL CONCENTRATION 319
Table 1. Chlorophyll Content Statistics Used in This Studya
Chlorophyll (mg m–2)
Pigment Min Max Mean Median
Chl a 160.93 651.13 363.95 346.13
Chl b 39.77 152.92 86.08 86.51
Total Chl 200.69 804.04 450.03 432.64
aTotal number of samples used is 15.
Fig. 2. First derivative of reflectance of maize showing the double-peak, one around 685–705
nm and the other near 710–725 nm. No third or fourth peaks were evident in the vegetation
considered.
Chl from 270 to 350 mg m–2. Higher ranges for Chl are observed for those maximum
peaks falling at longer wavelengths, 360–800 mg m–2 compared to the first group.
Area between Peaks
No distinct differences are seen in the range of values of the ABP for the two
groups. The first has a range of 0.015 to 0.022, while the second spans 0.005 to 0.025.
Nonetheless, the first group has a narrower range (0.007) than the second group
(0.02). This disparity can be explained (details below) when the movement between
peaks and the continuum line are taken into consideration.
ABP Relationship to Chlorophyll Concentrations
The ABP exhibits sensitivity to chlorophyll (Chl) concentrations, including Chl
a, Chl b, and total Chl, in samples with a maximum peak at shorter wavelengths (Fig.
3). High negative correlations are observed due to the rise of the first peak caused by
320 SALAS AND HENEBRY
Fig. 3. Linear relationships between Chl versus the ABP for maize samples with a maximum
peak on the shorter wavelength. Best-fit functions for each pair are represented by the
regression lines, significant at p = 0.05.
Fig. 4. Linear relationships between Chl versus the ABP for maize samples with a maximum
peak on the longer wavelength. Best-fit functions for each pair are represented by the
regression lines, significant at p = 0.05.
vegetation maturity as Chl increases, which in turn decreases the ABP. The rise of the
peak ceases when Chl concentration exceeds 45µg cm–2 (François et al., 2004, p. 13).
The ABP is more sensitive to Chl concentrations in samples with a maximum
peak at longer wavelengths (Fig. 4). In contrast to the samples with a maximum peak
at a shorter wavelength (first group), the samples show much higher coefficients of
determination (r2 > 0.90) for total Chl when fitted against the area between peaks.
Gitelson et al. (1996) found that the second peak is related to the increased Chl
concentration; hence, it becomes the main peak of the first derivative curve (Fig. 5A).
The graph also confirms the already established concepts of the blue-shift and red-
shift of the location of the maximum peak as previously investigated by Rock et al.
(1988) and Clevers and Buker (1991).
Since ABP is sensitive to the movement of the second peak (Fig. 5B), the
decrease of the ABP is instigated by Chl increases (Fig. 4). As Chl increases, the ABP
decreases in response as the location of the maximum peak moves further to a longer
CHANGE IN CHLOROPHYLL CONCENTRATION 321
Fig. 5. Linear relationships between Chl (A) and ABP (B) versus the location of the maximum
peak on the longer wavelength (second group) for maize. Best-fit functions for each pair are
represented by the regression lines, significant at p = 0.05.
wavelength. François et al. (2004) also pointed to the emergence of the second peak
as the concentration of chlorophyll increases.
Changes of the Continuum Length in Relation to Chl
The minimum-distance straight line that connects the double-peak is referred to
as “the continuum length” (see Fig. 1). It provides a metric to track the variation of
the ABP and the movement of the maximum peaks. The decrease of the ABP is a
result of the shifting of the peaks of the second group to a longer wavelength (due to
Chl increase) and not because of the changes of the peak positions of the first group.
Figure 6A illustrates how the length of the continuum varies directly with the increase
of Chl content for the second group, in contrast to Figure 6B, which shows weaker
correspondence. The continuum is seen to be sensitive to the shifting of the maximum
peak located on the longer wavelength.
François et al. (2004) found a reduced shift (715–720 nm) for the second group
compared to the shift of the first group (705–715 nm). The same results are observed
in this study, with the first group having a shift from 685 to 705 nm and a reduced
shift for the second group, 710 to 725 nm. However, even for a diminished shift, the
322 SALAS AND HENEBRY
Fig. 6. Variation of the continuum length for samples with the maximum peak on the longer
wavelength, second group, in relation to the increasing Chl content.
shifting to a longer wavelength of the second group still captures the increase in chlo-
rophyll concentrations.
ABP in Relation to Both Peaks
Figure 7 further illustrates the movement of all peaks, wavelength-wise, in rela-
tion to the ABP. Results from the second group (Fig. 7A) support that the maximum
peak (A.1) is most affected by shifts to longer wavelengths because of the decrease of
the ABP. The other peaks, regardless of where they are located (first or second
group), do not appear to be affected by the increase of the ABP, with a pattern main-
taining the same wavelength peak locations (A.1, B.1, and B.2).
Red-Edge Chlorophyll Index
The resulting linear best-fit function between CI and Chl is shown in Figure 8.
The CI explains 95% of the total Chl variability. This high r2 is manifested when all
samples, without regard of the location of the peaks, are taken all together in the
analysis. However, the coefficient of determination decreased to 0.87 when samples
that have peaks on shorter wavelengths are taken out of the equation. We note that
samples with peaks on shorter wavelength show a stronger linearity for CI vs. Chl,
with a higher r2 = 0.96.
DISCUSSION
The spectral derivatives in the neighborhood of the red-edge have an informative
structure. The red-edge inflection point, easily seen as a peak, is a wavelength posi-
tion that is affected by the changing of chlorophyll content as concluded by Clevers
and Buker (1991) and Gitelson et al. (1996). A multiple-peak feature may be
revealed: four to six peaks according to François et al. (2004) for simulated spectra,
four peaks according to Gitelson et al. (1996), and more than one peak from other
authors: Zarco-Tejada et al. (2003), Clevers et al. (2004), and Kooistra et al. (2004).
The presence of multiple peaks creates a dilemma as to the decision of selecting the
CHANGE IN CHLOROPHYLL CONCENTRATION 323
Fig. 7. Locations of the peaks for samples with the maximum peak on the longer wavelength,
second group (A), and for samples with the maximum peak on shorter wavelength, first group
(B) in relation to increasing ABP.
“real” maximum peak on the first derivative of the reflectance to serve as an REIP
index for vegetation studies. Despite the possibility that several local peaks observed,
two main and prominent peaks, or double-peak, are widely recognized. This study
addressed the vegetation feature as first peak on a shorter wavelength and second
peak on a longer wavelength.
It is seen that the maximum peak on a longer wavelength (second peak) is con-
sidered the main peak of the first derivative curve; its movement is found to be highly
correlated with the ABP. Changes of the pigments are also seen to have caused the
shifting, causing evidently high negative correlations between ABP vs. Chl.
In the situation when the first peak is dominant, an increase in magnitude of the
ABP does not necessarily follow an increase of the photosynthetic pigment. Although
the dominance of the first peak is related to the yellowing of leaves (Gitelson et al.,
1996, p. 503) or lower chlorophyll content (François et al., 2004, p. 14), it is also
related to the vegetation maturity. When the first peak goes up to a maximum, up to
45 µg Chl/cm2 (ibid., p. 13), the ABP diminishes in response.
324 SALAS AND HENEBRY
Fig. 8. Linear best-fit function for Chl (total) versus red-edge chlorophyll index (CI red-edge)
combining all samples.
Table 2. Bootstrap Regression Statistics Using Observation Resampling
(Correlation Model) to Obtain Properties of β0 and β1a
A. Peak on longer wavelength B. Peak on shorter wavelength
Least square Bootstrap Least square Bootstrap
β0 β1 β0 β1 β0 β1 β0 β1
Bias 0 0 –3.10 –62.88 0 0 –8.91 349.88
Standard error 47.01 2756 47.43 3,214 92.44 4,969 216.02 12,277
95% confidence interval
Lower bound 773.73 –27,758 791.54 –26,969 321.33 –3,1291 343.17 –27,461
Upper bound 1,003.80 –14,269 948.53 –15,363 834.63 –3,697 737.52 –5,381
aUsing
dataset for ABP and Chl from maize samples with a maximum peak on the longer
wavelength (A) and maximum peak on the shorter wavelength (B). The statistics are based on
500 replications for each sample from the original dataset.
From the assessment of the ABP against the photosynthetic pigment, it is clear
that the new vegetation feature plays a role in detecting small changes of the pigment
concentrations, e.g. total Chl (all r2 > 0.75). Validation using bootstrap regression
produces the results shown in Table 2, with comparable values of β0 and β1 when we
resample the observations (correlation model). Bootstrap does a good job in estimat-
ing the properties of the beta coefficients, because the results from the method are not
very different from the classical method. The resulting biases for the bootstrap are
small (e.g., –3.10 and –62.88 for β0 and β1, respectively) and the standard errors are
comparable, especially for data samples from peaks on longer wavelengths. Results
indicate that the ABP approach could supplement other methods to quantify Chl con-
centrations in plants.
The distance between peaks is a component of the ABP. Variations of the magni-
tude of the continuum length is related to the movement of the peak on the longer
CHANGE IN CHLOROPHYLL CONCENTRATION 325
wavelength. François et al. (2004) stated that the maximum of the peaks slowly shifts
toward higher wavelengths, whether the maximum peak is at the shorter wavelength
or the longer wavelength. However, we find that the shifting of the peak located on
the shorter wavelength is not associated with the changing of the continuum length.
An advantage of the ABP method employed in this study is that it is robust, as
relatively little error due to the utilization of SAMS is induced during the whole dou-
ble-peak analysis. The authors also conducted manual computations of the ABP that
have resulted in slightly higher r2 (all r2 > 0.80) when assessed against Chl. However,
the latter is only for substantiation purposes and not recommended as an approach for
ABP calculation as it takes up much time and is computationally expensive.
The comparison between the capabilities of the ABP against the CI as an indica-
tor of change in chlorophyll concentration is vital. While the CI yields a higher coeffi-
cient of determination than the ABP (r 2 = 0.95 vs. r 2 = 0.60) when samples are
combined, it is a bit weaker (r2 = 0.87 vs. r2 = 0.91) when samples have a higher Chl
concentration. The possible explanation for the differences lies in the varying Chl that
may have stimulated the slight movement of the peaks, which the ABP is initially
designed for. However, the CI may have missed the movements at higher concentra-
tions. Although CI seems to be a robust method, an advantage of the ABP in revealing
peak shifting is something the CI fails to demonstrate.
It should be noted that this new feature provides an opportunity to look further
into the first derivative curve generated from hyperspectral data. Given that the ABP
is only a small component of the FD curve, smoothing procedures for signal denois-
ing must be approached with caution. How much filtering can be applied to the spec-
tra before a relevant feature, such as the ABP, disappears from the FD curve? The
answer may depend on the characteristics of the signature at hand. In this study, the
linear combination of means was found to be an efficient way to retain the two peaks.
CONCLUSIONS
Despite a limited number of samples, we are confident about the general implica-
tions made of our findings for two reasons: (1) each individual hyperspectral signa-
ture visibly defined the peaks in question; and (2) the Chl pigments were laboratory-
extracted from a leaf level with high accuracy.
The results of this study signify close relationships between the ABP feature,
including the continuum length, and chlorophyll concentrations in maize leaves. The
area between the two peaks found in the first derivative curve of a vegetation spec-
trum exhibited high sensitivity to the movement of the maximum peak located on a
longer wavelength. When vegetation is green or when Chl is high, reflectance
increases, followed by the decrease of the ABP, forcing the shift of the maximum
peak on the longer wavelength to even longer wavelength positions (red-shift). As
soon as the vegetation starts to yellow and chlorophyll decreases, the maximum peak
subsequently relocates to a shorter wavelength, as the results suggest. This condition
may be reversed during early stages of vegetation, with the maximum peak located at
a shorter wavelength before shifting to a longer wavelength when vegetation grows or
Chl builds up.
It is evident that continued yellowing and decrease of the photosynthetic
pigments could cause the ABP to drop in magnitude and almost certainly disappear.
326 SALAS AND HENEBRY
Further, these metrics (ABP and continuum length) do not indicate when the maxi-
mum peak on the longer wavelength (second peak) ceases to be the maximum and
when the maximum peak on the shorter wavelength (first peak) begins to prevail.
There is no means to tell what maximum level of Chl concentrations corresponds to
the dominance of the second peak or the shifting of the maximum peak from longer to
shorter wavelength, despite the fact that François et al. (2004) assigned a threshold
Chl value (45µg/cm 2 ) to indicate the shifting of the peaks. Additional study is
required to investigate this phenomenon and its significance for monitoring vegeta-
tion status.
The efficiency of the ABP is dependent on the presence of the double peak found
around 685–725 nm of the FD curve. Although our findings are specific to a hyper-
spectral dataset, we believe that that the methodology used in this study could also be
applied to other visible peaks, without regard to wavelength locations. Additional
research is essential to verify the ability of the ABP feature to gauge levels of Chl
concentrations for other crops, not only maize.
Finally, the investigation of the ABP and the method employed in this study real-
ize two goals that were set forth: (1) describe the characteristics of the double-peak,
with the continuum line alongside and its movement in relation to Chl concentrations;
(2) show through regression analysis that the ABP feature can become a good estima-
tor of chlorophyll concentrations for maize.
ACKNOWLEDGMENTS
The authors thank A. Gitelson for his insights and for use of his spectral data, S.
Wacker and M. Schickel for numerous discussions and critical comments on the man-
uscript, and J. G. P. W. Clevers for remarks on previous work by E. Salas. The
research was supported in part by a GIScCE graduate research fellowship.
REFERENCES
Bruce, L. M. and J. Li, 2001, “Wavelets for Computationally Efficient Hyperspectral
Derivative Analysis,” Transactions on Geoscience and Remote Sensing,
39:1540–1546.
Clark, R. N., 1999, “Spectroscopy of Rocks and Minerals, and Principles of Spectros-
copy,” in Manual of Remote Sensing, Volume 3, Remote Sensing for the Earth
Sciences, Rencz, A. N. (Ed.), New York, NY: John Wiley and Sons, 3–58.
Clevers, J. G. P. W. and C. Buker, 1991, “Feasibility of the Red-Edge Index for the
Detection of Nitrogen Deficiency,” in Proceedings of the 5th International
Symposium on Physical Measurements and Signatures in Remote Sensing,
Courchevel, France: ESA SP-319:165–168.
Clevers, J. G. P. W., Kooistra, L., and E. A. L. Salas, 2004, “Study of Heavy Metal
Contamination in River Floodplains Using the Red-Edge Position in Spectro-
scopic Data,” International Journal of Remote Sensing, 25(19):3883–3895.
Curran, P. J., Dungan, J. L., and H. L. Gholz, 1990, “Exploring the Relationship
Between Reflectance Red-edge and Chlorophyll Content in Slash Pine Leaves,”
Tree Physiology, 7:33–48.
CHANGE IN CHLOROPHYLL CONCENTRATION 327
Datt, B., McVicar, T. R., Van Niel, T. G., Jupp, D. L. B., and J. S. Pearlman, 2003,
“Preprocessing EO-1 Hyperion Hyperspectral Data to Support the Application of
Agricultural Indexes,” IEEE Transactions on Geoscience and Remote Sensing,
41(6):1246–1259.
Demetriades-Shah, T. H., Steven, M. D., and J. A. Clark, 1990, “High Resolution
Derivatives Spectra in Remote Sensing,” Remote Sensing of Environment,
33:55–64.
Efron, B. and R. J. Tibshirani, 1993, An Introduction to the Bootstrap, London, UK:
Chapman & Hall/CRC.
Filella, I. and J. Penuelas, 1994, “The Red-Edge Position and Shape as Indicators of
Plant Chlorophyll Content, Biomass, and Hydric Status,” International Journal
of Remote Sensing, 15:1459–1470.
François, C., Le Maire, G., and E. Dufrêne, 2004, “Towards Universal Broad Leaf
Chlorophyll Indices Using PROSPECT Simulated Database and Hyperspectral
Reflectance Measurements,” Remote Sensing of Environment, 89(1):1–28.
Gitelson, A. A., Keydan, G. P., and M. N. Merzlyak, 2006, “Three-Band Model for
Noninvasive Estimation of Chlorophyll, Carotenoids, and Anthocyanin Contents
in Higher Plant Leaves,” Geophysical Research Letters, 33, L11402 [doi:10.1029/
2006GL026457].
Gitelson, A. A. and M. Merzlyak, 1994, “Spectral Reflectance Changes Associated
With Autumn Senescence of Asculus Hippocastanum and Acer Platanoides
Leaves: Spectral Features and Relation to Chlorophyll Estimation,” Journal of
Plant Physiology, 143:286–292.
Gitelson, A. A., Merzlyak, M., and H. Lichtenthaler, 1996, “Detection of Red-Edge
Position and Chlorophyll Content by Reflectance Measurements Near 700 nm,”
Journal of Plant Physiology, 148:501–508.
Gitelson, A. A., Viña, A., Rundquist, D. C., Ciganda, V., and T. J. Arkebauer, 2005,
“Remote Estimation of Canopy Chlorophyll Content in Crops,” Geophysical
Research Letters, 32, L08403 [doi:10.1029/2005GL022688].
Guyot, G., Baret, F., and S. Jacquemoud, 1992, “Imaging Spectroscopy for Vegetation
Studies,” in Imaging Spectroscopy: Fundamentals and Prospective Applications,
Toselli F. and J. Bodechtel (Eds), Dordrecht, The Netherlands: Kluwer, 145–165.
Horler, D. N. H., Dockray, M., and J. Barber, 1983, “The Red-Edge of Plant Leaf
Reflectance,” International Journal of Remote Sensing, 4:273–288.
Kooistra, L., Salas, E. A. L., Clevers, J. G. P. W., Wehrens, R., Leuven, R. S. E. W.,
Nienhuis, P. H., and L. M. C. Buydens, 2004, “Exploring Field Vegetation
Reflectance as an Indicator of Soil Contamination in River Floodplains,” Envi-
ronmental Pollution, 127(2):281–290.
Kruse, F. A., Calvin, W. M., and O. Seznec, 1988, “Automated Extraction of Absorp-
tion Features from Airborne Visible/Infrared Imaging Spectrometer (AVIRIS)
and Geophysical Environmental Research Imaging Spectrometer (GERIS)
Data,” in Proceedings of the Airborne Visible/Infrared Imaging Spectrometer
(AVIRIS) Performance Evaluation Workshop, Pasadena, CA: Jet Propulsion Lab-
oratory, JPL Publication 88-38, 62–75.
Meriam, J. L., 1966, Statics, 2nd ed. Toronto, Canada: John Wiley and Sons, Inc.,
chapter 5, art. 25, p. 169.
328 SALAS AND HENEBRY
O’Haver, T., Fell, A. F., Smith, G., Gans, P., Sneddon, J., Bezur, L., Michel, R. G.,
Ottaway, J. M., Miller, J. N., Ahmad, T. A., Chadburn, B. P., and Cottrell, C. T.,
1982, “Derivative Spectroscopy and Its Applications in Analysis,” Analytical
Proceedings, 19:22–46.
Peñuelas, J., Gamon, J. A., Fredeen, A. L., Merino, J., and C. B. Field, 1994, “Reflec-
tance Indices Associated with Physiological Changes in Nitrogen and Water-
Limited Sunflower Leaves,” Remote Sensing of Environment, 48:135–146.
Philpot, W. D., 1991, “The Derivative Ratio Algorithm: Avoiding Atmospheric
Effects in Remote Sensing,” IEEE Transactions on Geoscience Remote Sensing,
29(3):350–357.
Rock, B. N., Hoshizaki, T., and J. R. Miller, 1988, “Comparison of in Situ and Air-
borne Measurements of the Blue Shift Associated with Forest Decline,” Remote
Sensing of Environment, 24:109–127.
Salas, Eric A. L., 2004, “Vegetation REIP for Soil Contamination Indicator,” Univer-
sity of San Carlos Graduate Journal, 20: 210–217.
Stark, R., Gitelson, A., Grits, U., Rundquist, D., and Y. Kaufman, 2000, “New Tech-
nique for Remote Estimation of Vegetation Fraction: Principles, Algorithms and
Validation,” Aspects of Applied Biology, 60:241–246.
Viña, A., Gitelson, A. A., Rundquist, D. C., Keydan, G., Leavitt, B., and J. Schepers,
2004, “Monitoring Maize (Zea mays L.) Phenology with Remote Sensing,”
Agronomy Journal, 96:1139–1147.
Zarco-Tejada, P. J., Miller, J. R., Mohammed, G. H., Noland, T. L., and P. H. Samp-
son, 2002, “Vegetation Stress Detection Through Chlorophyll a + b Estimation
and Fluorescence Effects on Hyperspectral Imagery,” Journal of Environmental
Quality, 31:1433–1441.
Zarco-Tejada, P. J., Pushnik, J. C., Dobrowski, S., and S. L. Ustin, 2003, “Steady-
State Chlorophyll a Fluorescence Detection from Canopy Derivative Reflectance
and Double-Peak Red-Edge Effects,” Remote Sensing of Environment, 84:283–
294.
