Monitoring pigment‐driven vegetation changes in a low‐Arctic tundra ecosystem using digital cameras


Monitoring pigment-driven vegetation changes in a low-Arctic
tundra ecosystem using digital cameras

ALISON L. BEAMISH,1,� NICHOLAS C. COOPS,2 TXOMIN HERMOSILLA,2 SABINE CHABRILLAT,3 AND BIRGIT HEIM1

1Alfred Wegener Institute, Periglacial Research, Telegrafenberg A45, 14473 Potsdam, Germany
2Integrated Remote Sensing Studio (IRSS), Faculty of Forestry, University of British Columbia, 2424 Main Mall,

Vancouver, British Columbia V6T1Z4 Canada
3Helmholtz Centre Potsdam (GFZ), German Research Centre for Geosciences, Telegrafenberg A17, 14473 Potsdam, Germany

Citation: Beamish, A. L., N. C. Coops, T. Hermosilla, S. Chabrillat, and B. Heim. 2018. Monitoring pigment-driven
vegetation changes in a low-Arctic tundra ecosystem using digital cameras. Ecosphere 9(2):e02123. 10.1002/ecs2.2123

Abstract. Arctic vegetation phenology is a sensitive indicator of a changing climate, and rapid assess-
ment of vegetation status is necessary to more comprehensively understand the impacts on foliar condition
and photosynthetic activity. Airborne and space-borne optical remote sensing has been successfully used
to monitor vegetation phenology in Arctic ecosystems by exploiting the biophysical and biochemical
changes associated with vegetation growth and senescence. However, persistent cloud cover and low sun
angles in the region make the acquisition of high-quality temporal optical data within one growing season
challenging. In the following study, we examine the capability of “near-field” remote sensing technologies,
in this case digital, true-color cameras to produce surrogate in situ spectral data to characterize changes in
vegetation driven by seasonal pigment dynamics. Simple linear regression was used to investigate relation-
ships between common pigment-driven spectral indices calculated from field-based spectrometry and red,
green, and blue (RGB) indices from corresponding digital photographs in three dominant vegetation com-
munities across three major seasons at Toolik Lake, North Slope, Alaska. We chose the strongest and most
consistent RGB index across all communities to represent each spectral index. Next, linear regressions were
used to relate RGB indices and extracted leaf-level pigment content with a simple additive error propaga-
tion of the root mean square error. Results indicate that the green-based RGB indices had the strongest rela-
tionship with chlorophyll a and total chlorophyll, while a red-based RGB index showed moderate
relationships with the chlorophyll to carotenoid ratio. The results suggest that vegetation color contributes
strongly to the response of pigment-driven spectral indices and RGB data can act as a surrogate to track
seasonal vegetation change associated with pigment development and degradation. Overall, we find that
low-cost, easy-to-use digital cameras can monitor vegetation status and changes related to seasonal foliar
condition and photosynthetic activity in three dominant, low-Arctic vegetation communities.

Key words: hyperspectral; low-Arctic; red, green, and blue indices; true-color digital photography; vegetation
pigments.

Received 20 January 2018; accepted 24 January 2018. Corresponding Editor: Debra P. C. Peters.
Copyright: © 2018 Beamish et al. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
� E-mail: abeamish@awi.de

INTRODUCTION

Changes to the functioning of Arctic ecosys-
tems, such as shifts in photosynthetic activity, net
primary productivity, and species composition
influence global climate change and the resulting

feedbacks (Zhang et al. 2007, Bhatt et al. 2010,
Parmentier and Christensen 2013). Climatic
changes have been accompanied by broad-scale
shifts in Arctic vegetation community composi-
tion and species distribution, as well as fine-scale
shifts in individual plant reproduction and

 ❖ www.esajournals.org 1 February 2018 ❖ Volume 9(2) ❖ Article e02123

info:doi/10.1002/ecs2.2123
http://creativecommons.org/licenses/by/3.0/


phenology (Walker et al. 2006, Bhatt et al. 2010,
Elmendorf et al. 2012a, Bjorkman et al. 2015, Pre-
v�ey et al. 2017). Changes in vegetation phenol-
ogy can impact overall ecosystem functioning as
a result of mismatched species–climate interac-
tions. This in turn can impact species photosyn-
thetic activity and growth through increased
vulnerability to events such as frost, soil satura-
tion, or disruption to species chilling require-
ments (Inouye and McGuire 1991, Yu et al. 2010,
Cook and Wolkovich 2012, Høye et al. 2013,
Wheeler et al. 2015).

In Arctic ecosystems, snow pack conditions
and timing of snowmelt, rather than temperature,
are primary drivers of the onset of vegetation phe-
nology (Billings and Bliss 1959, Bjorkman et al.
2015). Changing snowmelt dynamics due to
changes in winter precipitation and spring tem-
peratures have been shown to directly influence
tundra vegetation phenology throughout the
growing season (Bjorkman et al. 2015). Thus, the
onset of the growing season, or the leaf-out stage,
acts as a benchmark of phenology, and identifica-
tion and characterization of this stage are key to
understanding the overall and subsequent phe-
nology and fitness of tundra vegetation (Iler et al.
2013, Wheeler et al. 2015). The timing of maxi-
mum expansion and elongation of leaves and
stems and when vegetation is at, or near, peak
photosynthetic activity, or peak greenness, are
also important benchmarks of phenological
phases. Peak greenness marks the climax of the
growing season, the timing and magnitude of
which can indicate ecosystem functioning, and
acts as a benchmark for long-term monitoring of
tundra productivity (Bhatt et al. 2010, 2013). A
final benchmark of phenological phase is the end
of the growing season, or senescence. In combina-
tion with leaf-out, senescence dictates growing
season length and has implications for overall sea-
sonal productivity and carbon assimilation (Park
et al. 2016). Characterizing key biophysical prop-
erties of Arctic vegetation associated with these
three benchmark phenological phases is impor-
tant for accurate monitoring and quantification of
local and regional changes in Arctic vegetation
and in turn Arctic and global changes to energy
and carbon cycling, as well as associated feed-
backs. The presence and seasonal changes in veg-
etation pigment content can be used to infer
biophysical properties including photosynthetic

activity and foliar condition at key phenological
phases of Arctic vegetation.
The major photosynthetic pigment groups of

chlorophyll and carotenoids absorb strongly in
the visible spectrum creating unique spectral sig-
natures (Curran 1989, Gitelson and Merzlyak
1998, Gitelson et al. 2002, Coops et al. 2003).
Chlorophyll pigments are the dominant factor
controlling the amount of light absorbed by a
plant and therefore photosynthetic potential,
which ultimately dictates primary productivity.
Carotenoids (carotenes and xanthophylls) are
responsible for absorbing incident radiation and
providing energy to the photosynthetic process
and are often used to provide information on the
physiological status of vegetation (Young and
Britton 1990, Bartley and Scolnik 1995). The third
major pigment group of anthocyanins (water-
soluble flavonoids) has a less concise function in
vegetation providing photoprotection (Steyn
et al. 2002, Close and Beadle 2003), drought and
freezing protection (Chalker-Scott 1999), and
they have been shown to play a role in recovery
from foliar damage (Gould et al. 2002). Photo-
synthetic pigments influence a wide range of
plant functioning from photosynthetic efficiency
to protective functions (Demmig-Adams and
Adams 1996) and can be measured destructively
through laboratory analysis or non-destructively
using high spectral resolution remote sensing
(Gitelson et al. 1996, 2001, 2002, 2006).
The development and degradation of chloro-

phyll pigments as a result of vegetation emer-
gence, stress, or senescence causes a distinct shift
in spectral signatures in the visible spectrum to
shorter wavelengths due to a narrowing of the
major chlorophyll absorption feature (550–
750 nm) and an overall reduction in spectral
absorption (Ustin and Curtiss 1990). Carotenoids
(yellow to orange) and anthocyanins (blue to red)
have less straightforward seasonal, and therefore
spectral, shifts. The spectral absorption by carote-
noid pigments can result in mixed or masked
spectral absorption with chlorophyll signals due
to overlapping absorption features in the visible
spectrum (400–700 nm). Carotenoids and antho-
cyanins often have relatively higher concentra-
tions in senesced and young leaves when
chlorophyll concentrations are relatively low
(Tieszen 1972, Sims and Gamon 2002, Stylinski
et al. 2002). However, in general, the predictable

 ❖ www.esajournals.org 2 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



and distinct spectral features created by changes
in vegetation pigment content allow inferences of
ecological parameters such as photosynthetic
activity, nutrient concentration, and biomass
(Mutanga and Prins 2004, Mutanga and Skidmore
2004, Asner and Martin 2008, Ustin et al. 2009).

Of particular interest for in situ and laboratory
spectroscopy of vegetation and phenology are the
spectral characterization of the xanthophyll cycle
(carotenoids) and the ratio between photosynthetic
pigments of carotenoids and chlorophyll as they
relate to radiation use efficiency and in turn radia-
tive transfer models (Blackburn 2007, Garbulsky
et al. 2011, Gamon et al. 2016). Radiative transfer
models are used to simulate leaf and canopy spec-
tral reflectance and transmittance, and have been
the basis for the inverse determination of biophysi-
cal and biochemical parameters of vegetation
based on spectral reflectance (F�eret et al. 2011).
Even at lower spectral resolution, the importance
of pigments and pigment dynamics related to the
xanthophyll cycle can be seen in available products
for recent satellite missions; for example, the Euro-
pean Space Agency’s (ESA) Sentinel-2 provides
many pigment-driven vegetation indices (see
https://www.sentinel-hub.com/develop/documenta
tion/eo_products/Sentinel2EOproducts).

While highly valuable, in situ reflectance spec-
troscopy, like optical remote sensing, is challeng-
ing for remote Arctic field sites. Remote locations,
challenges associated with field-based monitor-
ing, and atmospheric and illumination conditions
of optical satellite imagery of these areas limit
available optical data, hindering phenological
studies. To complement the use of airborne and
satellite imagery to measure vegetation pigments
and classify biophysical properties in Arctic
ecosystems, the use of true-color digital photogra-
phy and red, green, and blue (RGB) indices to
infer photosynthetic activity and foliar condition
is a promising area of study and should be exam-
ined (Anderson et al. 2016, Beamish et al. 2016).
Research has shown that Arctic vegetation
changes are complex with strong site and species-
specific responses through space and time
(Walker et al. 2006, Bhatt et al. 2010, Elmendorf
et al. 2012b, Bjorkman et al. 2015, Prev�ey et al.
2017). True-color digital photography represents
both a simple and cost-effective way to increase
data volume both spatially and temporally to cap-
ture site and species-specific responses. Data from

consumer-grade digital cameras have extremely
high spatial resolution and high data collection
capabilities and are less weather dependent than
airborne and optical satellite remote sensing as
varying illumination can be easily corrected
(Richardson et al. 2007, Nijland et al. 2014).
Large-scale, true-color, repeatable digital photog-
raphy networks in temperate ecosystems such as
the PhenoCam network (http://phenocam.sr.
unh.edu/webcam/) highlight the existing ecologi-
cal applications of this method. Additional work
by Richardson et al. (2007) and Coops et al. (2012)
demonstrates the link between ground-based
true-color camera systems and satellite-scale
(Landsat and MODIS) observations. In addition
to phenological parameters, well-established RGB
indices derived from consumer-grade digital
cameras have been shown to accurately identify
vegetation cover in Arctic tundra and temperate
forests (Richardson et al. 2007, Ide and Oguma
2010, Nijland et al. 2014, Beamish et al. 2016) and
can be closely related to gross primary production
in grasslands, temperate forests, Arctic tundra,
and wetlands (Ahrends et al. 2009, Migliavacca
et al. 2011, Westergaard-Nielsen et al. 2013,
Anderson et al. 2016), as well as containing a sen-
sitivity to changes in plant photosynthetic pig-
ments (Ide and Oguma 2013).
In this study, we examine the capability of

in situ, true-color digital photography to act as a
surrogate for in situ spectral data in assessing pig-
ment-driven vegetation changes associated with
three key seasons representing early, peak, and
late season, in three dominant, low-Arctic tundra
vegetation communities. To do this, we asked the
following research questions: (1) What are the
relationships between RGB indices and in situ,
pigment-driven spectral indices? (2) How do
these relationships change with community type
and season? (3) To what extent do the indices rep-
resent actual chlorophyll and carotenoid content?
We conclude with proposing the RGB indices best
suited as proxies for in situ spectral data for moni-
toring seasonal vegetation change and the related
pigment dynamics in a low-Arctic tundra.

METHODS

Study site
Data were collected at the Toolik Field Station

(68°62.570 N, 149°61.430 W) on the North Slope of

 ❖ www.esajournals.org 3 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.

https://www.sentinel-hub.com/develop/documentation/eo_products/Sentinel2EOproducts
https://www.sentinel-hub.com/develop/documentation/eo_products/Sentinel2EOproducts
http://phenocam.sr.unh.edu/webcam/
http://phenocam.sr.unh.edu/webcam/


the Brooks Range, in north central Alaska in the
growing season of 2016. The Toolik Area is repre-
sentative of the southern Arctic Foothills, a phys-
iographic province of the North Slope (Walker
et al. 1989). Vegetation is a combination of moist
tussock tundra, wet sedge meadows, and dry
upland heaths. Data were acquired in the Toolik
Vegetation Grid, a 1 9 1 km long-term monitor-
ing site established by the National Science Foun-
dation (NSF) as part of the Department of
Energy’s R4D (Response, Resistance, Resilience,
and Recovery to Disturbance in Arctic Ecosys-
tems) project (Fig. 1). Vegetation monitoring
plots (1 9 1 m) are positioned in close proximity
to equally spaced site markers delineating the
intersection of the Universal Transverse Mercator
(UTM) coordinates and each point within the
grid. A subset of the grid was sampled for the
purpose of this study representing three distinct
and dominant vegetation communities as
defined by Bratsch et al. (2016) (Fig. 1). The com-
munities include moist acidic tundra (MAT),
moist non-acidic tundra (MNT), and moss tun-
dra (MT). A detailed description and representa-
tiveness (% cover of the Toolik Vegetation Grid)
of the three vegetation communities sampled

within the Long-term Toolik Vegetation Grid are
shown in Table 1.

Digital photographs
In order to capture the three major seasons of

early, peak, and late, we acquired true-color digi-
tal photographs on three days in the 2016 grow-
ing season (June 16, day of year [DOY] 165; July
11, DOY 192; August 17, DOY 229). Images were
taken at nadir approximately 1 m off the ground
with a white 1 9 1 m frame for registration of
off nadir images (Fig. 1). According to the Toolik
Lake Environmental Data Centre (https://toolik.
alaska.edu/edc/, 2017) phenology record that
monitors 15 dominant species at the Toolik Field
Station, the first image acquisition date occurred
two weeks after the average date of first leaf-out
(May 29, DOY 149) accurately capturing the peak
of the early leaf-out phase. The second acquisi-
tion took place one week after average date of
last flower petal drop (July 3, DOY 185) and
within one week of recorded full green-up (end
of June). The final acquisition took place three
weeks after average first day of fall color change
(July 27, DOY 209) and within the range of
recorded peak fall colors (mid-August).

Fig. 1. Toolik Vegetation Grid located in the Toolik Research Area on the North Slope of the Brooks Range in
northern Alaska and a late season example plots of the three vegetation communities monitored in the study.
MAT, moist acidic tussock tundra; MNT, moist non-acidic tundra; MT, moss tundra.

 ❖ www.esajournals.org 4 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.

https://toolik.alaska.edu/edc/
https://toolik.alaska.edu/edc/


Images were acquired with a consumer-grade
digital camera (Panasonic DM3 LMX, Osaka,
Japan) in raw format, between 10:00 h and
14:00 h, under uniform cloud cover to reduce the
influence of shadow and illumination differ-
ences. The digital camera collects color values by
means of the Bayer matrix (Bayer 1976) with
individual pixels coded with a red, green, and
blue value between 0 and 256. RGB values of
each pixel in each image were extracted using
ENVI+IDL (Version 4.8; Harris Geospatial, Boul-
der, Colorado, USA). To reduce the impact of
non-nadir acquisitions, photo registration was
undertaken using the 1 9 1 m apart. Table 2
shows the normalized RGB channels and RGB
indices calculated based on the normalized red,
green, and blue values of the digital pho-
tographs, hereafter referred to as RGB indices.

Digital camera indices 2G-RB and nG have
been used to monitor vegetation phenology and
biomass extensively (Richardson et al. 2007, Ide
and Oguma 2013, Beamish et al. 2016). Addition-
ally, we defined new RGB indices to examine
their representativeness to selected pigment-
driven spectral indices (Table 3). We chose these
RGB indices as they represented as closely as
possible the mathematical formulas of selected
pigment-driven indices presented in the Field-
based spectral data section.

Field-based spectral data
Within one week of the digital photographs,

field-based spectral measurements of each vege-
tation plot were acquired using a GER 1500 field
spectroradiometer (Spectra Vista Corporation,
New York, New York, USA). Spectral data were

Table 1. Description of the three vegetation communities monitored in this study.

Community Description Detailed description % cover

Moist acidic
tussock
tundra

Occurs on soils with pH < 5.0–5.5
and is dominated by dwarf erect
shrubs such as Betula nana and
Salix pulchra, graminoids species
(Eriophorum vaginatum), and
acidophilous mosses

Betula nana–Eriophorum vaginatum. Dwarf-shrub, sedge,
moss tundra (shrubby tussock tundra dominated by
dwarf birch, Betula nana). Mesic to subhygric, acidic,
moderate snow. Lower slopes and water-track margins.
Mostly on Itkillik I glacial surfaces

6.1

Salix pulchra–Carex bigelowii. Dwarf-shrub, sedge, moss
tundra (shrubby tussock tundra dominated by diamond-
leaf willow, Salix pulchra). Subhygric, moderate snow,
lower slopes with solifluction

Moist
non-acidic
tundra

Dominated by mosses,
graminoids (Carex bigelowii), and
prostrate dwarf shrubs (Dryas
integrifolia)

Carex bigelowii–Dryas integrifolia, typical subtype;
Tomentypnum nitens–Carex bigelowii, Salix glauca subtype:
nontussock sedge, dwarf-shrub, moss tundra (moist
non-acidic tundra). Mesic to subhygric, non-acidic
(pH > 5.5), shallow to moderate snow. Solifluction areas
and somewhat unstable slopes. Some south-facing slopes
have scattered glaucous willow (Salix glauca)

5.8

Mossy
tussock
tundra

A moist acidic tussock tundra-
type community dominated by
sedges (E. vaginatum) and
abundant Sphagnum spp.

Eriophorum vaginatum–Sphagnum; Carex bigelowii–
Sphagnum: tussock sedge, dwarf-shrub, moss tundra
(tussock tundra, moist acidic tundra). Mesic to
subhygric, acidic, shallow to moderate snow, stable. This
unit is the zonal vegetation on fine-grained substrates
with ice-rich permafrost. Some areas on steeper slopes
with solifluction are dominated by Bigelow sedge
(Carex bigelowii)

54.2

Table 2. Calculated red, green, and blue (RGB) indices
from RGB channels of digital photographs.

Index Formula Source

nG G/(R+B+G)
nR R/(R+B+G)
nB B/(R+B+G)
GR nG/nR
GB nG/nB
2G-RB 2 9 nG � (nR + nB) Richardson et al. (2007)

Table 3. Pseudo pigment-driven red, green, and blue
(RGB) indices calculated from RGB channels of digi-
tal photographs.

Indices Formula

nBG (nB�nG)/(nB+nG)
nRG (nR�nG)/(nR+nG)
RGr ((nR�nG)/nR)
nG�1 (1/nG)
2R-GB 2 9 nR � (nG + nB)

 ❖ www.esajournals.org 5 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



acquired on June 14 (DOY 165), July 8 (DOY
189), and August 23 (DOY 235) in 2016 with a
spectral range of 350–1050 nm, 512 bands, a
spectral resolution of 3 nm, a spectral sampling
of 1.5 nm, and an 8° field of view. Data were
acquired under clear weather conditions between
10:00 and 14:00 local time corresponding to the
highest solar zenith angle. In each plot, radiance
data were acquired at nadir approximately 1 m
off the ground resulting in an approximately
15 cm diameter ground instantaneous field of
view. To reduce noise and characterize the spec-
tral variability in each plot, an average of nine
point measurements of upwelling radiance (Lup)
in each of the 1 9 1 m plots was used. Down-
welling radiance (Ldown) was measured as the
reflection from a white Spectralon© plate. Sur-
face reflectance (R) was processed as:

R ¼ Lup
Ldown

� 100 (1)

Reflectance spectra (0–100%) were prepro-
cessed with a Savitzky-Golay smoothing filter
(n = 11), and to remove sensor noise at the spec-
tral limits of the radiometer, data were subset to
400–985 nm.

Using the spectral reflectance data, we calcu-
lated common pigment-driven vegetation indices,
focusing on vegetation color, and those produced
as remote sensing products by the ESA and the
North American Space Agency (NASA); a brief
description of each is provided below (Table 4).
The indices that were calculated from the field-
based GER data are hereafter referred to as pig-
ment-driven spectral indices.

The Photochemical Reflectance Index (PRI) is an
indicator of photosynthetic radiation use efficiency
and was first proposed by Gamon et al. (1992).

The Plant Senescence Reflectance Index (PSRI) is
sensitive to senescence-induced reflectance changes
from changes in chlorophyll and carotenoid con-
tent and was proposed by Merzlyak et al. (1999).
The Pigment Specific Simple Ratios (PSSRa and b)
aim to model chlorophyll a and b, respectively
(Blackburn 1998, 1999, Sims and Gamon 2002).
Chlorophyll Carotenoid Index (CCI) was devel-
oped to track the phenology of photosynthetic
activity of evergreen species by Gamon et al.
(2016). The carotenoid reflectance indices (CRI1
and CRI2) use the reciprocal reflectance at 508 and
548 nm and 508 and 698 nm, respectively, in order
to remove the effect of chlorophyll (Gitelson et al.
2002). The anthocyanin reflectance indices (ARI1
and ARI2) developed by Gitelson et al. (2001,
2006) are designed to reduce the influence of
chlorophyll absorption to isolate the anthocyanin
absorption by taking the difference between the
reciprocal of green (550 nm) and the red-edge
(700 nm). ARI2 includes a multiplication by the
near infrared (NIR) to reduce the influence of leaf
thickness and density.

Vegetation pigment concentration
To estimate how accurately plot-level indices

represent actual leaf-level pigment content,
leaves and stems (n = 213) of the dominant vas-
cular species in a subset of the sampled plots
were collected at early, peak, and late season for
chlorophyll and carotenoid analysis. Samples
were placed in porous tea bags and preserved in
a silica gel desiccant in an opaque container for
up to 3 months until pigment extraction (Esteban
et al. 2009). Each sample was homogenized by
grinding with a mortar and pestle. Approxi-
mately 1.00 mg (�0.05 mg) of homogenized
sample was placed into a vial with 2 mL of

Table 4. Pigment-driven spectral indices.

Indices Short Formula Source

Photochemical Reflectance Index PRI (q533 � q569)/(q533 + q569) Gamon et al. (1992)
Plant Senescence Reflectance Index PSRI (q678 � q498)/q748 Merzlyak et al. (1999)
Pigment Specific Simple Ratio PSSR PSSRa = q800/q671 Blackburn (1998, 1999);

PSSRb = q800/q652 Sims and Gamon (2002)
Chlorophyll Carotenoid Index CCI (q531 � q645)/(q531 + q645) Gamon et al. (2016)
Carotenoid Reflectance Index 1 CRI1 (1/q508) � (1/q548) Gitelson et al. (2002)
Carotenoid Reflectance Index 2 CRI2 (1/q508) � (1/q698) Gitelson et al. (2002)
Anthocyanin Reflectance Index 1 ARI1 (1/q550) � (1/q700) Gitelson et al. (2001)
Anthocyanin Reflectance Index 2 ARI2 q800 9 [(1/q550) � (1/q700)] Gitelson et al. (2001)

 ❖ www.esajournals.org 6 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



dimethylformamide (DMF). Vials were then
wrapped in aluminum foil to eliminate any
degradation of pigments due to UV light and
stored in a fridge (4°C) for 24 h. Samples were
measured into a cuvette prior to spectrophoto-
metric analysis. Bulk pigment concentrations
were then estimated using a spectrophotometer
measuring absorption at 646.8, 663.8, and
480 nm (Porra et al. 1989). Absorbance (A) val-
ues at specific wavelengths were transformed
into lg/mg concentrations of chlorophyll a, Chla;
chlorophyll b, Chlb; total chlorophyll, Chla+b; car-
otenoids, Car, using the following equations:

Chla ¼ 12:00 � A663:8 � 3:11 � A646:8 (2)
Chlb ¼ 20:78 � A646:8 � 4:88 � A663:8 (3)
Chlaþb ¼ 17:67 � A646:8 þ 7:12 � A663:8 (4)

Car ¼ A480 � Chla
� �

=245 (5)

The chlorophyll a to chlorophyll b (Chla:b) and
the chlorophyll to carotenoid ratios (Chl:Car)
were calculated by dividing Eq. 2 by Eq. 3 and
Eq. 4 by Eq. 5, respectively. Pigment concentra-
tion was calculated as the average concentration
of the dominant species in each plot.

Data analysis
We examined the potential of digital camera

RGB data as a proxy for identified pigment-
driven vegetation indices using existing and new
RGB indices. Simple linear regression was per-
formed between RGB indices defined in Tables 2
and 3, and hyperspectral PRI, PSRI, PSSR, CCI,
CRI, and ARI indices defined in Table 4 (App-
endix S1: Table S1). The RGB and spectral indices
from significant RGB/spectral linear regressions
were then chosen for linear regression with Chla,
Chlb, Chla:b, and Car:Chl for each vegetation com-
munity to explore how well the chosen indices
represent actual leaf-level pigment content.

As we assume RGB indices can be used in
place of pigment-driven spectral indices, we
wanted to include the error associated with spec-
tral indices as a proxy for leaf-level pigments to
more accurately estimate uncertainties. To do
this, we used a simple additive error propagation
of root mean square error of the selected RGB/
pigment and spectral/pigment regressions using
the following equations:

RMSE ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPn

t¼1 ðXobs;t � Xmodel;tÞ2
n

s
(6)

RMSEprop ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðRMSEz � yÞ2 þ RMSEx � y

� �2q
(7)

where RMSE is the root mean square error, Xobs;t
and Xmodel;t are the actual and predicted values
of the linear regression t, respectively, and where
RMSEprop is the propagated RMSE of RGB~
pigment regressions using the RMSE of the
RGB~pigment regressions (RMSEz � y) and the
RMSE of the spectral~pigment regressions
(RMSEx � y; Appendix S1: Table S2). All analyses
were performed in R (R Development Core Team
2007), and an alpha of 0.05 was used.

RESULTS

RGB indices as a surrogate for pigment-driven
spectral indices
The strength of the relationships between the

RGB indices and the pigment-driven spectral
indices was variable between vegetation commu-
nities (Appendix S1). The most consistent and
strongest regressions across the three vegetation
communities were selected and are presented in
Table 5 and Fig. 2. The strongest three regres-
sions were observed between RGr and ARI2 in
MT (R2 = 0.77, P < 0.01, RMSE = 0.04), followed
by RG and PSRI in MAT (R2 = 0.75, P < 0.01,
RMSE = 0.05) and MNT (R2 = 0.75, P < 0.01,
RMSE = 0.04; Fig. 2). PSRI had the strongest
relationships across the three communities, while
the carotenoid reflectance indices of CRI1 and
CRI2 and the anthocyanin index of ARI1 had the
weakest relationships. In general, MT had the
strongest and most consistent regressions across
all indices followed by MAT and finally MNT.
We chose the following six RGB indices, 2G-RB,
2R-GB, nG, nG�1, RG, and RGr, representing the
pigment-driven indices of PSSRb, PRI, PSSRa,
CCI, PSRI, and ARI2, respectively, for linear
regression with leaf-level pigment content.

RGB indices as a surrogate for leaf-level
pigment content
The relationships between selected RGB indices

and pigment content suggest high variability in
both fit and uncertainty across the three vegeta-
tion communities (Table 6). In general, the indices

 ❖ www.esajournals.org 7 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



performed best across all communities when pre-
dicting Chla with the strongest observed in MNT.
nG, representing PSSRa, had the single strongest
regression with Chla in MNT (R

2 = 0.61, P < 0.01,
RMSE-P = 0.89), and this relationship was also
moderate in MAT and MT. Chla+b and Chla:b also
showed weak to moderate correlations, but per-
formance across communities was more variable
than Chla. Both Chla+b and Chla:b showed moder-
ate correlations with all indices except 2R-GB in
MNT and MAT, respectively. 2R-GB, representing
PRI, also had a moderate relationship with Chl:
Car in MT (R2 = 0.53, P < 0.01, RMSE = 0.21) but
poor performance in the other two communities.
The indices performed worst when predicting
Chlb with R

2 ≤ 0.27 in all cases.

DISCUSSION

The relationships between selected RGB
indices and pigment-driven spectral indices sup-
port the capability of digital cameras to act as a
surrogate for in situ spectral data and for the
study of seasonal vegetation changes associated
with pigment dynamics in dominant low-Arctic
vegetation communities. Evidence for the capa-
bility of digital cameras as an ecological monitor-
ing tool of vegetation phenology, biomass, and
productivity and the relationships of these
parameters to spectrally derived vegetation
indices exists in a variety of ecosystems (Ahrends
et al. 2009, Coops et al. 2010, Ide and Oguma
2010, Migliavacca et al. 2011, Westergaard-Nielsen
et al. 2013, Nijland et al. 2014, Anderson et al.
2016, Beamish et al. 2016). Our study adds to this
knowledge base by expanding to pigment-driven

vegetation indices chosen for their indication of
photosynthetic efficiency, a parameter of high
interest in vegetation remote sensing. In addi-
tion, we showcase the utility of a ground-based
camera system in an ecosystem characterized by
challenging acquisition conditions for spectral
data.
There are some technical considerations when

using these types of digital cameras that need to
be considered such as consumer-grade digital
cameras typically have limited range charge cou-
ple device/complementary metal oxide semicon-
ductor sensors and employ automatic exposure
adjustments with changing environmental
brightness and local illumination. This in turn
affects the comparability of images collected
under different conditions; however, this can be
overcome through the use of band ratios which
are insensitive to brightness. Local illumination
differences caused by direct sunlight can still
introduce error; however, the Arctic is subject to
unique illumination conditions because of the
low solar zenith angle, long periods of daylight,
and frequent cloud cover leading to frequent dif-
fuse illumination conditions. Previous research
has suggested collection of data under uniform
cloud cover (diffuse conditions) is ideal (Ide and
Oguma 2010). All images in this study were col-
lected under uniform cloud cover by choice;
however, when considering a framework of
high-frequency (daily or hourly) data collection,
these phenomena should be taken into consider-
ation. Additionally, this study does not take into
account the impact of atmospheric scattering
common in shorter wavelengths (blue) of air-
and space-borne platforms. This phenomenon

Table 5. The most consistent, significant linear regressions between red, green, and blue (RGB) indices and
pigment-driven spectral indices in the three communities.

Spectral RGB

MAT MNT MT

R2 P-value RMSE R2 P-value RMSE R2 P-value RMSE

PRI 2R-GB 0.46 <0.01 0.04 0.29 <0.01 0.04 0.53 <0.01 0.03
PSRI RG 0.78 <0.01 0.05 0.75 <0.01 0.04 0.70 <0.01 0.05
PSSRa nG 0.72 <0.01 0.01 0.60 <0.01 0.01 0.71 <0.01 0.01
PSSRb 2G-RB 0.64 <0.01 0.04 0.53 <0.01 0.03 0.51 <0.01 0.03
CCI nG�1 0.56 <0.01 0.04 0.58 <0.01 0.02 0.73 <0.01 0.02
CRI1 GB 0.40 <0.01 0.18 0.16 0.02 0.14 0.34 <0.01 0.14
CRI2 RB 0.09 <0.01 0.05 �0.01 0.43 0.04 0.17 0.03 0.02
ARI1 RG 0.30 <0.01 0.10 0.17 0.01 0.07 0.42 <0.01 0.07
ARI2 RGr 0.63 <0.01 0.06 0.43 <0.01 0.04 0.77 <0.01 0.04

Note: Bold values represent moderate or stronger (R2 > 0.40) significant linear regressions.

 ❖ www.esajournals.org 8 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



would presumably reduce the correlation in the
blue channel, and thus, its use should be mini-
mized when inferring relationships between air-
and space-borne platforms. However, the use of
the blue channel in ground-based remote sensing

as in this study is less of a concern as atmo-
spheric scattering is minimal at this near-sensing
scale. Another consideration is that, digital cam-
eras contain what is known as a Bayer color filter
array that combines one blue, one red, and two

Fig. 2. Simple linear regression between the six best RGB indices and pigment-driven spectral indices across
three seasons in each community type. MAT, moist acidic tussock tundra; MNT, moist non-acidic tundra; MT,
moss tundra; RMSE, root mean square error.

 ❖ www.esajournals.org 9 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



green sensors into an image pixel making green
the most sensitive channel in the camera to mir-
ror the sensitivity of the human eye (Bayer 1976).

From the standpoint of the human eye and
digital cameras, all of the best performing RGB
indices except one provide a measure of vegeta-
tion greenness and the weaker relationships seen
with the carotenoid and anthocyanin (non-green
pigments) indices are indicative of the cameras
overall green bias or green sensitivity. The best
performing RGB green band-based indices of

2G-RB, nG, nG�1, RG, and rGR had the strongest
relationships (R2 > 0.43) with the pigment-
driven spectral indices of PSSRb, PSSRa, CCI,
PSRI, and ARI2, respectively, across all three veg-
etation communities. The moderate to strong
relationships observed in all three communities
suggest these RGB indices can be used to moni-
tor seasonal vegetation changes associated with
pigment-driven color changes, mostly related to
the amount of green or chlorophyll pigments, in
dominant low-Arctic vegetation communities

Table 6. Simple linear regression between the selected red, green, and blue (RGB) indices and the spectral index
they are representing in parentheses, and pigment content with the propagated mean absolute percentage
error (MAPE-P).

Pigment lg/mg Veg RGB (spectral)

(a)

2G-RB (PSSRb) 2R-GB (PRI) nG (PSSRa)

R2 P-value RMSE RMSE-P R2 P-value RMSE RMSE-P R2 P-value RMSE RMSE-P

Chla MAT 0.38 0.00 0.31 0.48 0.17 0.02 0.35 0.52 0.39 0.00 0.30 0.42
MNT 0.59 0.01 0.12 0.22 0.06 0.26 0.18 0.25 0.61 0.01 0.12 0.22
MT 0.41 0.01 0.30 0.48 0.23 0.06 0.34 0.50 0.40 0.01 0.30 0.47

Chlb MAT �0.02 0.45 0.12 0.18 �0.01 0.44 0.12 0.19 �0.01 0.43 0.12 0.18
MNT 0.27 0.09 0.07 0.11 �0.14 0.84 0.09 0.11 0.26 0.09 0.07 0.11
MT 0.08 0.18 0.20 0.29 0.06 0.20 0.20 0.28 0.08 0.18 0.20 0.28

Chla+b MAT 0.30 0.00 0.20 0.31 0.15 0.03 0.22 0.33 0.31 0.00 0.20 0.28
MNT 0.53 0.02 0.08 0.15 �0.02 0.39 0.12 0.16 0.54 0.01 0.08 0.15
MT 0.24 0.05 0.28 0.44 0.16 0.10 0.30 0.43 0.24 0.05 0.28 0.42

Chla:b MAT 0.50 0.00 0.32 0.53 0.15 0.03 0.41 0.61 0.50 0.00 0.32 0.51
MNT �0.02 0.38 0.28 0.41 0.33 0.06 0.23 0.37 0.01 0.34 0.28 0.39
MT 0.12 0.13 0.32 0.36 0.04 0.25 0.33 0.41 0.12 0.13 0.32 0.39

Chl:Car MAT 0.19 0.01 0.18 0.25 0.12 0.04 0.19 0.26 0.20 0.01 0.18 0.26
MNT 0.01 0.34 0.21 0.30 �0.14 0.93 0.22 0.31 0.01 0.34 0.21 0.31
MT 0.23 0.06 0.19 0.27 0.53 0.00 0.15 0.21 0.24 0.05 0.19 0.29

(b)

nG�1 (CCI) RG (PSRI) RGr (ARI2)

R2 P-value RMSE RMSE-P R2 P-value RMSE RMSE-P R2 P-value RMSE RMSE-P

Chla MAT 0.37 0.00 0.31 0.49 0.35 0.00 0.31 0.46 0.38 0.00 0.31 0.45
MNT 0.60 0.01 0.12 0.22 0.56 0.01 0.13 0.21 0.55 0.01 0.13 0.15
MT 0.39 0.01 0.31 0.47 0.35 0.02 0.32 0.47 0.36 0.02 0.31 0.47

Chlb MAT �0.02 0.49 0.12 0.19 �0.02 0.44 0.12 0.19 0.00 0.36 0.12 0.19
MNT 0.26 0.09 0.07 0.11 0.04 0.29 0.08 0.11 0.03 0.30 0.08 0.11
MT 0.08 0.19 0.20 0.27 0.08 0.18 0.20 0.28 0.08 0.18 0.20 0.28

Chla+b MAT 0.30 0.00 0.20 0.32 0.28 0.00 0.21 0.31 0.31 0.00 0.20 0.30
MNT 0.53 0.02 0.08 0.15 0.43 0.03 0.09 0.14 0.43 0.03 0.09 0.11
MT 0.24 0.05 0.28 0.41 0.23 0.06 0.28 0.42 0.23 0.06 0.28 0.41

Chla:b MAT 0.51 0.00 0.31 0.54 0.43 0.00 0.34 0.52 0.43 0.00 0.34 0.51
MNT 0.00 0.35 0.28 0.39 0.22 0.12 0.25 0.37 0.21 0.12 0.25 0.35
MT 0.12 0.13 0.32 0.39 0.10 0.16 0.32 0.39 0.11 0.15 0.32 0.40

Chl:Car MAT 0.19 0.01 0.18 0.27 0.20 0.01 0.18 0.26 0.23 0.01 0.18 0.24
MNT 0.01 0.34 0.21 0.30 �0.05 0.46 0.21 0.31 �0.06 0.47 0.22 0.28
MT 0.24 0.05 0.19 0.29 0.36 0.02 0.17 0.27 0.35 0.02 0.17 0.23

Notes: Veg, vegetation community; Pig, pigment content. Bold values represent moderate or stronger (R2 > 0.40) significant lin-
ear regressions. MAT, moist acidic tussock tundra; MNT, moist non-acidic tundra; MT, moss tundra; RMSE, root mean square
error. See Tables 1–3 for definitions of both spectral and RGB indices.

 ❖ www.esajournals.org 10 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



(Fig. 2). Though we found an overall weakness
in RGB indices to accurately predict leaf-level
pigment content, a number of significant weak to
moderate relationships between Chla and Chla+b
with the green RGB indices suggest they do cap-
ture seasonal changes in chlorophyll content.
Moderate to weak relationships between pig-
ment content and pigment-driven spectral
indices have also been reported, even with data
collected concurrently at the leaf level, due to
variations in species-specific plant structure and
developmental stage (Sims and Gamon 2002).
The pigment content presented in this study is
an approximation using the mean of dominant
vascular species and does not take into account
the influence of moss or standing litter, both of
which influence greenness, especially at early
and late season when vascular vegetation is not
fully expanded. This combination of roughly esti-
mated pigment content, a greenness signal that is
not only composed of vascular species, especially
at early and late season, and species-specific
characteristics could explain the weak correla-
tions and the differences between communities.
However, it should be noted that all indices
demonstrated a significant moderate correlation
in at least one community with pigment content.

The five pigment-driven spectral indices used
in this study target different pigment groups;
however, all are relevant for the monitoring of
foliar condition and photosynthetic activity sug-
gesting the cameras can also indirectly infer
changes to non-green pigment groups through
an absence or changes in the greenness or chloro-
phyll pigments. Though the green RGB indices
were generally the most consistent and had the
strongest relationships, we also found that PRI is
well represented by 2R-GB and nR in MT and
MAT (Table 5; Appendix S1). Since the pigment-
driven spectral index of PRI is a prominently
used index for estimating photosynthetic light
use efficiency (Gamon et al. 1992, Pe~nuelas et al.
1995, Gamon and Oecologia 1997), the use of
digital camera nR to monitor this parameter is a
particularly interesting result.

Current operational satellite missions provide
an excellent opportunity for global monitoring of
foliar condition with relatively high spatial reso-
lution. Here we focused on exploring the spectral
information in the visible wavelength region
related to tundra vegetation color, driven by

pigment dynamics. The broadband spectral set-
tings of major operational satellite missions are
at first consideration not optimal for capturing
the detailed spectral reflectance of vegetation as
represented by narrowband spectral vegetation
indices. However, we show that RGB color val-
ues from consumer-grade digital cameras mea-
suring even broader-band spectral information
show correlations with reputed pigment-driven
spectral indices such as CCI, PSSRa, and ARI2
indices. Our results suggest vegetation color con-
tributes strongly to the response of these hyper-
spectral indices. The use of narrowband spectral
indices related to vegetation color and pigment
dynamics in order to monitor vegetation status
and condition is already occurring with the Earth
Observation System products from the Sentinel-2
multispectral satellite and will become more
common in the future with upcoming hyperspec-
tral satellite missions such as the Environmental
Mapping and Analysis Program (EnMAP)
planned for 2020.

CONCLUSIONS

Results of this study support the utility of digi-
tal cameras to act as a surrogate for in situ spec-
tral data to monitor pigment dynamics as a
result of seasonal changes in a low-Arctic ecosys-
tem. The RGB indices using the green band per-
form best as proxies for pigment-driven spectral
indices. We highlight nG as a proxy for PSSRa in
particular because of moderate to strong relation-
ships with both spectral and pigment data sug-
gesting this RGB index can track changes in
chlorophyll a content. We also suggest 2G-RB as
a proxy for PSSRb, nG�1 for CCI, RG for PSRI,
and RGr as a proxy for ARI1. Though the accu-
racy of pigment prediction for these indices is
not as strong, there is evidence that RGB indices
do track seasonal changes. This method repre-
sents a promising gap-filling tool and compli-
mentary data source for optical remote sensing
of vegetation in logistically and climatically
challenging Arctic ecosystems. The implementa-
tion of low-cost time-lapse systems or nadir
point measurements by an observer with con-
sumer-grade digital cameras is highly feasible
and proven in Arctic tundra ecosystems, and this
study increases the possible applications of this
method.

 ❖ www.esajournals.org 11 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



ACKNOWLEDGMENTS

This research was supported by EnMAP science
preparatory program funded under the DLR Space
Administration with resources from the German Fed-
eral Ministry of Economic Affairs and Energy (support
code: DLR/BMWi 50 EE 1348) in partnership with the
Alfred Wegener Institute in Potsdam. Funding has also
been provided through an NSERC Doctoral post-
graduate scholarship awarded to AB. SC acknowl-
edges support from the European Union’s Horizon
2020 Research and Innovation Programme (Grant No:
689443) via the iCUPE project (Integrative and Com-
prehensive Understanding on Polar Environments).
The authors would like to thank the logistical support
provided by Toolik Research Station and Skip Walker
of the Alaska Geobotany Center at the University of
Alaska, Fairbanks. We would also like to thank Marcel
Buchhorn and the HySpex Lab at the University of
Alaska, Fairbanks for calibration of the spectrometer
and MB and SW for providing GIS data of the Toolik
Area. Finally, we would like to thank Robert Guy from
the Faculty of Forestry at the University of British
Columbia for laboratory support.

LITERATURE CITED

Ahrends, H., S. Etzold, W. L. Kutsch, R. St€ockli, R.
Br€ugger, F. Jeanneret, H. Wanner, N. Buchmann,
and W. Eugster. 2009. Tree phenology and carbon
dioxide fluxes: use of digital photography for pro-
cess-based interpretation at the ecosystem scale.
Climate Research 39:261–274.

Anderson, H. B., L. Nilsen, H. Tømmervik, S. Karlsen,
S. Nagai, and E. J. Cooper. 2016. Using ordinary
digital cameras in place of near-infrared sensors to
derive vegetation indices for phenology studies of
high Arctic vegetation. Remote Sensing 8:847.

Asner, G. P., and R. E. Martin. 2008. Spectral and
chemical analysis of tropical forests: scaling from
leaf to canopy levels. Remote Sensing of Environ-
ment 112:3958–3970.

Bartley, G., and P. Scolnik. 1995. Plant carotenoids: pig-
ments for photoprotection, visual attraction, and
human health. Plant Cell 7:1027–1038.

Bayer, B. 1976. Color imaging array. U.S. Patent
3,971,065.

Beamish, A., W. Nijland, M. Edwards, N. Coops, and
G. Henry. 2016. Phenology and vegetation change
measurements from true colour digital photogra-
phy in high Arctic tundra. Arctic Science 2:33–49.

Bhatt, U., D. Walker, and M. Raynolds. 2010. Circum-
polar Arctic tundra vegetation change is linked to
sea ice decline. Earth Interactions 14:1–20.

Bhatt, U. S., D. A. Walker, M. K. Raynolds, P. A. Bie-
niek, H. E. Epstein, J. C. Comiso, J. E. Pinzon, C. J.
Tucker, and I. V. Polyakov. 2013. Recent declines in
warming and vegetation greening trends over pan-
Arctic tundra. Remote Sensing 5:4229–4254.

Billings, W., and L. Bliss. 1959. An alpine snowbank
environment and its effects on vegetation, plant
development, and productivity. Ecology 40:388–
397.

Bjorkman, A., S. Elmendorf, A. Beamish, M. Vellend,
and G. Henry. 2015. Contrasting effects of warming
and increased snowfall on Arctic tundra plant phe-
nology over the past two decades. Global Change
Biology 21:4651–4661.

Blackburn, G. 1998. Spectral indices for estimating
photosynthetic pigment concentrations: a test
using senescent tree leaves. International Journal of
Remote Sensing 19:657–675.

Blackburn, G. 1999. Relationships between spectral
reflectance and pigment concentrations in stacks of
deciduous broadleaves. Remote Sensing of Envi-
ronment 70:224–237.

Blackburn, G. A. 2007. Wavelet decomposition of
hyperspectral data: a novel approach to quantify-
ing pigment concentrations in vegetation. Interna-
tional Journal of Remote Sensing 28:2831–2855.

Bratsch, S. N., H. E. Epstein, M. Buchhorn, and D. A.
Walker. 2016. Differentiating among four Arctic
tundra plant communities at Ivotuk, Alaska using
field spectroscopy. Remote Sensing 8:51.

Chalker-Scott, L. 1999. Environmental significance of
anthocyanins in plant stress responses. Photochem-
istry and Photobiology 70:1–9.

Close, D., and C. Beadle. 2003. The ecophysiology of
foliar anthocyanin. Botanical Review 69:149–161.

Cook, B., and E. Wolkovich. 2012. Divergent responses
to spring and winter warming drive community
level flowering trends. Proceedings of the National
Academy of Sciences 109:9000–9005.

Coops, N., T. Hilker, C. Bater, M. Wulder, S. Nielsen,
G. McDermid, and G. Stenhouse. 2012. Linking
ground-based to satellite-derived phenological
metrics in support of habitat assessment. Remote
Sensing Letters 3:191–200.

Coops, N., T. Hilker, F. Hall, C. Nichol, and G. Drolet.
2010. Estimation of light-use efficiency of terrestrial
ecosystems from space: a status report. BioScience
63:788–797.

Coops, N., C. Stone, D. Culvenor, L. Chisholm, and R.
Merton. 2003. Chlorophyll content in eucalypt veg-
etation at the leaf and canopy scales as derived
from high resolution spectral data. Tree Physiology
23:23–31.

Curran, P. J. 1989. Remote sensing of foliar chemistry.
Remote Sensing of Environment 30:271–278.

 ❖ www.esajournals.org 12 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



Demmig-Adams, B., and W. W. Adams. 1996. The role
of xanthophyll cycle carotenoids in the protection
of photosynthesis. Trends in Plant Science 1:21–26.

Elmendorf, S. C., et al. 2012a. Plot-scale evidence of
tundra vegetation change and links to recent sum-
mer warming. Nature Climate Change 2:453–457.

Elmendorf, S., et al. 2012b. Global assessment of exper-
imental climate warming on tundra vegetation:
heterogeneity over space and time. Ecology Letters
15:164–175.

Esteban, R., et al. 2009. Alternative methods for sam-
pling and preservation of photosynthetic pigments
and tocopherols in plant material from remote
locations. Photosynthesis Research 101:77–88.

F�eret, J.-B., C. Franc�ois, A. Gitelson, G. Asner, K. Barry,
C. Panigada, A. Richardson, and S. Jacquemoud.
2011. Optimizing spectral indices and chemometric
analysis of leaf chemical properties using radiative
transfer modeling. Remote Sensing of Environment
115:2742–2750.

Gamon, J., F. Huemmrich, C. Wong, I. Ensminger, S.
Garrity, D. Hollinger, A. Noormets, and J. Pe~nue-
las. 2016. A remotely sensed pigment index reveals
photosynthetic phenology in evergreen conifers.
Proceedings of the National Academy of Sciences
113:13087–13092.

Gamon, S., and S. Oecologia. 1997. The photochemical
reflectance index: an optical indicator of photosyn-
thetic radiation use efficiency across species, func-
tional types, and nutrient levels. Oecologia 112:
492–501.

Gamon, J., J. Pe~nuelas, and C. Field. 1992. A narrow-
waveband spectral index that tracks diurnal
changes in photosynthetic efficiency. Remote Sens-
ing of Environment 41:35–44.

Garbulsky, M., J. Pe~nuelas, J. Gamon, Y. Inoue, and I.
Filella. 2011. The photochemical reflectance index
(PRI) and the remote sensing of leaf, canopy and
ecosystem radiation use efficiencies a review and
meta-analysis. Remote Sensing of Environment
115:281–297.

Gitelson, A. A., G. P. Keydan, and M. N. Merzlyak.
2006. Three-band model for noninvasive estimation
of chlorophyll, carotenoids, and anthocyanin con-
tents in higher plant leaves. Geophysical Research
Letters 33:LL1402.

Gitelson, A., and M. Merzlyak. 1998. Remote sensing
of chlorophyll concentration in higher plant leaves.
Advances in Space Research 22:689–692.

Gitelson, A. A., M. N. Merzlyak, and O. B. Chivku-
nova. 2001. Optical properties and nondestructive
estimation of anthocyanin content in plant leaves.
Photochemistry and Photobiology 74:38–45.

Gitelson, A. A., M. N. Merzlyak, and H. K. Lichten-
thaler. 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., Y. Zur, O. B. Chivkunova, and M. N.
Merzlyak. 2002. Assessing carotenoid content in
plant leaves with reflectance spectroscopy. Photo-
chemistry and Photobiology 75:272–281.

Gould, K. S., J. McKelvie, and K. R. Markham. 2002.
Do anthocyanins function as antioxidants in
leaves? Imaging of H2O2 in red and green leaves
after mechanical injury. Plant, Cell and Environ-
ment 25:1261–1269.

Høye, T. T., E. Post, N. M. Schmidt, K. Trøjelsgaard,
and M. C. Forchhammer. 2013. Shorter flowering
seasons and declining abundance of flower visitors
in a warmer Arctic. Nature Climate Change 3:759–
763.

Ide, R., and H. Oguma. 2010. Use of digital cameras
for phenological observations. Ecological Informat-
ics 5:339–347.

Ide, R., and H. Oguma. 2013. A cost-effective monitor-
ing method using digital time-lapse cameras for
detecting temporal and spatial variations of snow-
melt and vegetation phenology in alpine ecosys-
tems. Ecological Informatics 16:23–34.

Iler, A. M., T. T. Høye, D. W. Inouye, and N. M. Sch-
midt. 2013. Nonlinear flowering responses to cli-
mate: Are species approaching their limits of
phenological change? Philosophical Transactions of
the Royal Society B 368:20120489.

Inouye, D. W., and D. A. McGuire. 1991. Effects of
snowpack on timing and abundance of flowering
in Delphinium nelsonii (Ranunculaceae): implica-
tions for climate change. American Journal of Bot-
any 78:997–1001.

Merzlyak, M., A. Gitelson, O. Chivkunova, and V. Rak-
itin. 1999. Non-destructive optical detection of pig-
ment changes during leaf senescence and fruit
ripening. Physiologia Plantarum 106:135–141.

Migliavacca, M., M. Galvagno, E. Cremonese, M. Ros-
sini, M. Meroni, O. Sonnentag, S. Cogliati, G.
Manca, F. Diotri, and L. Busetto. 2011. Using digital
repeat photography and eddy covariance data to
model grassland phenology and photosynthetic
CO2 uptake. Agricultural and Forest Meteorology
151:1325–1337.

Mutanga, A. Skidmore., and H. H. Prins. 2004. Predict-
ing in situ pasture quality in the Kruger National
Park, South Africa, using continuum-removed
absorption features. Remote Sensing of Environ-
ment 89:393–408.

Mutanga, O., and A. Skidmore. 2004. Hyperspectral
band depth analysis for a better estimation of grass
biomass (Cenchrus ciliaris) measured under con-
trolled laboratory conditions. International Journal

 ❖ www.esajournals.org 13 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.



of Applied Earth Observation and Geoinformation
5:87–96.

Nijland, W., D. R. Jong, D. Jong, and M. Wulder. 2014.
Monitoring plant condition and phenology using
infrared sensitive consumer grade digital cameras.
Agricultural and Forest Meteorology 184:98–106.

Park, T., S. Ganguly, H. Tømmervik, E. S. Euskirchen,
K.-A. Høgda, S. Karlsen, V. Brovkin, R. R. Nemani,
and R. B. Myneni. 2016. Changes in growing sea-
son duration and productivity of northern vegeta-
tion inferred from long-term remote sensing data.
Environmental Research Letters 11:084001.

Parmentier, F.-J., and T. Christensen. 2013. Arctic:
speed of methane release. Nature 500:529.

Pe~nuelas, J., I. Filella, and J. A. Gamon. 1995. Assess-
ment of photosynthetic radiation-use efficiency with
spectral reflectance. New Phytologist 131:291–296.

Porra, R. J., W. A. Thompson, and P. E. Kriedemann.
1989. Determination of accurate extinction coefficients
and simultaneous equations for assaying chlorophylls
a and b extracted with four different solvents: verifi-
cation of the concentration of chlorophyll standards
by atomic absorption spectroscopy. Biochimica et Bio-
physica Acta (BBA): Bioenergetics 975:384–394.

Prev�ey, J., et al. 2017. Greater temperature sensitivity
of plant phenology at colder sites: implications for
convergence across northern latitudes. Global
Change Biology 23:2660–2671.

R Development Core Team. 2007. R: a language and
environment for statistical computing. R Founda-
tion for Statistical Computing, Vienna, Austria.

Richardson, A., J. Jenkins, B. Braswell, D. Hollinger, S.
Ollinger, and M.-L. Smith. 2007. Use of digital web-
cam images to track spring green-up in a decidu-
ous broadleaf forest. Oecologia 152:323–334.

Sims, D. A., and J. A. Gamon. 2002. Relationships
between leaf pigment content and spectral reflec-
tance across a wide range of species, leaf structures
and developmental stages. Remote Sensing of
Environment 81:337–354.

Steyn, W. J., S. J. E. Wand, D. M. Holcroft, and G.
Jacobs. 2002. Anthocyanins in vegetative tissues: a
proposed unified function in photoprotection.
New Phytologist 155:349–361.

Stylinski, C., J. Gamon, and W. Oechel. 2002. Seasonal
patterns of reflectance indices, carotenoid pigments
and photosynthesis of evergreen chaparral species.
Oecologia 131:366–374.

Tieszen, L. L. 1972. The seasonal course of above-
ground production and chlorophyll distribution in
a wet arctic tundra at Barrow, Alaska. Arctic and
Alpine Research 4:307–324.

Ustin, S. L., and B. Curtiss. 1990. Spectral characteris-
tics of ozone-treated conifers. Environmental and
Experimental Botany 30:293–308.

Ustin, S., A. Gitelson, and S. Jacquemoud. 2009. Retrie-
val of foliar information about plant pigment sys-
tems from high resolution spectroscopy. Remote
Sensing of Environment 113:S67–S77.

Walker, M. D., D. A. Walker, K. R. Everett, and C.
Segelquist. 1989. Wetland soils and vegetation, arc-
tic foothills, Alaska. U.S. Fish and Wildlife Service
Biological Report 89, Washington, D.C., USA.

Walker, M., et al. 2006. Plant community responses to
experimental warming across the tundra biome.
Proceedings of the National Academy of Sciences
USA 103:1342–1346.

Westergaard-Nielsen, A., M. Lund, B. Hansen, and M.
Tamstorf. 2013. Camera derived vegetation green-
ness index as proxy for gross primary production
in a low Arctic wetland area. ISPRS Journal of Pho-
togrammetry and Remote Sensing 86:89–99.

Wheeler, H., T. Høye, N. Schmidt, J.-C. Svenning, and
M. Forchhammer. 2015. Phenological mismatch
with abiotic conditions—implications for flowering
in Arctic plants. Ecology 96:775–787.

Young, A., and G. Britton. 1990. Carotenoids and
stress. Pages 87–112. in R. G. Alscher and J. R.
Cummings, editors. Stress Responses in Plants:
Adaptation and Acclimation Mechanisms. Wiley-
Liss, New York, New York, USA.

Yu, H., E. Luedeling, and J. Xu. 2010. Winter and
spring warming result in delayed spring phenol-
ogy on the Tibetan Plateau. Proceedings of the
National Academy of Sciences 107:22151–22156.

Zhang, X., D. Tarpley, and J. T. Sullivan. 2007. Diverse
responses of vegetation phenology to a warming
climate. Geophysical Research Letters 34:L19405.

SUPPORTING INFORMATION

Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.
2123/full

 ❖ www.esajournals.org 14 February 2018 ❖ Volume 9(2) ❖ Article e02123

BEAMISH ET AL.

http://onlinelibrary.wiley.com/doi/10.1002/ecs2.2123/full
http://onlinelibrary.wiley.com/doi/10.1002/ecs2.2123/full