Content uploaded by Matthias Peichl
Author content
All content in this area was uploaded by Matthias Peichl on Jan 07, 2017
Content may be subject to copyright.
Bringing Color into the Picture:
Using Digital Repeat Photography
to Investigate Phenology Controls
of the Carbon Dioxide Exchange
in a Boreal Mire
Matthias Peichl,
1
* Oliver Sonnentag,
2
and Mats B. Nilsson
1
1
Department of Forest Ecology and Management, Swedish University of Agricultural Sciences Umea
˚, 90183 Umea
˚, Sweden;
2
De
´partement de ge
´ographie, Universite
´de Montre
´al, Montreal, Quebec, Canada
ABSTRACT
Mire vegetation phenology is closely linked to the
ecosystem carbon cycle but rarely monitored and
quantified with high temporal resolution. In this
study, we use digital repeat photography to explore
phenology as a control of the carbon dioxide (CO
2
)
exchange measured by eddy covariance (EC) in a
minerogenic boreal mire in northern Sweden over
2 years (2011–2012). Strong correlations and sea-
sonal hysteresis effects were observed between the
green chromatic coordinate (g
cc
) derived from the
digital image archive and leaf area index, day
length, and growing degree-day sum (GDDS). Dif-
ferences in GDDS between the 2 years were the
main control on the between-year variations in the
spring patterns of g
cc
. Periods with lower water
table level coincided with an increase of the red
chromatic coordinate. The onset and magnitudes of
EC-derived photosynthetic CO
2
uptake (that is,
gross ecosystem production, GEP) and net ecosys-
tem CO
2
exchange (NEE) during the spring green-
up of vascular plants were more closely related to
those of g
cc
than to those of air temperature and
photosynthetically active radiation. In contrast,
abiotic variables controlled GEP during the summer
period when vascular plant canopy cover was fully
developed. Stepwise regression analysis suggested
that g
cc
contributed substantially in explaining
variations in GEP during spring and autumn. Over
both growing seasons, g
cc
was well correlated with
GEP (r
2
= 0.68), NEE (r
2
= 0.58), and ecosystem
respiration (r
2
= 0.50). Overall, we show that dig-
ital repeat photography provides an inexpensive
and effective method for the continuous quantifi-
cation of the phenological patterns of the vascular
plant community in mire ecosystems. Our results
suggest that vegetation phenology is an important
control of the mire CO
2
exchange and should be
considered in both experimental and modeling
studies to better account for the separate effects of
phenology and abiotic drivers on mire carbon
dynamics.
Key words: boreal landscape; climate; canopy
greenness; carbon cycle; eddy covariance; digital
repeat photography; peatland; phenology; RGB
chromatic coordinates.
Received 14 December 2013; accepted 2 September 2014;
published online 13 November 2014
Author contributions MP designed the study, analyzed the data,
wrote the paper, OS designed the study, contributed to data analysis,
contributed to the paper writing, MBN designed the study, contributed to
the paper writing, provided funding for the study.
*Corresponding author; e-mail: Matthias.Peichl@slu.se
Ecosystems (2015) 18: 115–131
DOI: 10.1007/s10021-014-9815-z
2014 Springer Science+Business Media New York
115
INTRODUCTION
Boreal mires cover 3% of the land area but con-
stitute about a fourth to a third of the global pool of
soil organic carbon (Gorham 1991; Turunen and
others 2002) providing a small but persistent con-
temporary net carbon uptake of 15–30 gCm
-2
y
-1
(Gorham 1991; Turunen and others 2002; Roulet
and others 2007; Nilsson and others 2008). How-
ever, the mire carbon balance and specifically the
biosphere–atmosphere exchange of one of the most
important greenhouse gases, carbon dioxide (CO
2
),
is highly sensitive to both biotic and abiotic controls
(Gorham 1991; Dise 2009; Lund and others 2010;
Bu and others 2011). Abiotic variables such as
precipitation (Robroek and others 2009), temper-
ature (Lindroth and others 2007; Lund and others
2010), and water table level (Yurova and others
2007; Robroek and others 2009; Sulman and others
2010) have been commonly suggested as the main
drivers of mire carbon dynamics. The biotic con-
trols related to temporal vegetation dynamics may,
however, also substantially affect the mire CO
2
exchange (Lund and others 2010; Bu and others
2011; Laine and others 2012). Consequently, the
relationships between CO
2
fluxes and abiotic con-
trols vary seasonally as a result of vegetation
development (Savage and others 2013). Although
abiotic effects on the mire CO
2
exchange have been
investigated by numerous previous studies, the role
of phenology, defined as the study of the timing of
recurring biological life cycle events and the causes
of their temporal change due to abiotic and biotic
forces (Lieth 1974), has been less examined.
Understanding the feedbacks between the CO
2
exchange and its two main controls, that is, abiotic
drivers and vegetation phenology (Figure 1), is
essential to estimate the development of mire
ecosystem structure and carbon balance given the
predicted changes in future climate (IPCC 2013;
Richardson and others 2013).
Vegetation phenology describes the temporal
development of vegetation properties including
biomass, species composition, leaf area index (LAI),
canopy height, and leaf physiology (for example,
nitrogen, chlorophyll, and carotenoid contents)
(Migliavacca and others 2011; Sonnentag and
others 2011,2012; Bauerle and others 2012). These
temporal changes in vegetation structure and
physiology drive CO
2
exchange processes via their
control on photosynthesis, autotrophic respiration,
and plant litter production. In addition, vegetation
structure affects the efficiency of how plants absorb
and reflect incoming light, especially over green
wavelengths (Vogelmann and Moss 1993; Bubier
and others 1997; Sonnentag and others 2007a).
Consequently, a strong correlation may be ex-
pected between canopy greenness related to phe-
nological events or stages and the ecosystem-scale
CO
2
exchanges (Graham and others 2006; Ahrends
and others 2009; Migliavacca and others 2011). On
the other hand, vegetation phenology itself is dri-
ven by abiotic controls such as temperature, water
availability, and photoperiod (Bryant and Baird
2003;Ko
¨rner and Basler 2010). Specifically, the
concept of the growing degree-day sum (GDDS)
has been traditionally used within the wide field of
phenological research as main predictor of phe-
nology in other natural and managed terrestrial
ecosystems (Russelle and others 1984; Richardson
and others 2013).
Tracking the temporal evolution of vegetation
status through repeated measurements of biomass,
LAI, and leaf chemical properties usually requires
labor-intensive and/or destructive methods
including the manual recording of key phenologi-
cal events through visual observation and/or the
sampling of vegetation through clipping (Wilson
and others 2007; Crimmins and Crimmins 2008;
Richardson and O’Keefe 2009; Richardson and
others 2013). Alternatively, a variety of costly
ground-based radiometers provide spectral vegeta-
tion properties that were shown to be linked to
plant physiology and ecosystem CO
2
exchange
(Eklundh and others 2011; Migliavacca and others
2011; Chojnicki 2013). Remote sensing phenology
products from satellite and airborne imagery are
valuable but limited to a coarse temporal and spa-
tial resolution (Schubert and others 2010; Harris
Figure 1. Conceptual diagram of the three-way interac-
tions among phenology, climate, and ecosystem carbon
dioxide (CO
2
) exchange components.
116 M. Peichl and others
and Dash 2011). Thus, there is a need for an
inexpensive, simple, and automated method for the
continuous quantification of phenological infor-
mation on daily time and ecosystem spatial scales.
Time series of simple color indices such as excess
green (ExG; Woebbecke and others 1995)orthe
green chromatic coordinate (g
cc
; Gillespie and
others 1987) derived from the red–green–blue
(RGB) color channel information of the digital
image archive have been used to relate changes of
canopy greenness to those of canopy structure (for
example, LAI) and CO
2
fluxes (primarily GEP) in
croplands (Meyer and Neto 2008; Sakamoto and
others 2012), deciduous and coniferous forests
(Richardson and others 2007,2009; Ahrends and
others 2009), grasslands and pastures (Migliavacca
and others 2011; Sonnentag and others 2011), and
arctic tundra (Westergaard-Nielsen and others
2013). Compared to the ExG, chromatic indices
have the advantage of suppressing effects from
changes in illumination scenes (Sonnentag and
others 2012). Moreover, color and reflectance
properties of Sphagnum mosses are strongly influ-
enced by their moisture content over the red and
near infrared wavelengths (Bryant and Baird 2003;
Gaalen and others 2007). Thus, the red chromatic
coordinate (r
cc
) might be useful in identifying
periods of vegetation moisture status related to
changes in the water table level (WTL) and its ef-
fects on vegetation phenology and CO
2
exchange
in peatland ecosystems. To our knowledge, how-
ever, so far no study has examined the link be-
tween peatland canopy color changes, water
availability, and CO
2
fluxes. Possible complications
due to complex and differential contributions and
interactions of color information from each of the
vascular plant canopy and non-vascular moss
ground cover (Bubier and others 1997; Graham
and others 2006; Sonnentag and others 2007a)
may have hampered the interest in applying this
technique in peatland ecosystems in the past.
In this study, we examined the link between
vegetation phenology and mire CO
2
exchange
using digital repeat photography and eddy covari-
ance (EC) measurements over 2 years made over a
boreal minerogenic mire in northern Sweden. The
main objectives were (i) to investigate the useful-
ness of chromatic coordinates (specifically g
cc
) de-
rived from a digital image archive to describe
changes in mire phenology, (ii) to determine con-
trols from ecosystem vegetation structure (for
example, LAI, canopy roughness length) and abi-
otic variables (for example, temperature, radiation)
on g
cc
and mire phenology, and (iii) to investigate
the relationship between mire phenology and
ecosystem CO
2
exchange. Our hypotheses were
that (i) g
cc
derived from digital repeat photography
provides a useful method for continuous and
automatic quantification of the vegetation phe-
nology in mire ecosystems, (ii) phenology as de-
scribed by g
cc
is closely related to ecosystem
vegetation structure, (iii) changes in the WTL and
associated plant moisture status are reflected in the
patterns of the r
cc
, and (iv) phenology is a major
control on the mire CO
2
fluxes, especially on GEP
during the spring green-up of vascular plants.
MATERIALS AND METHODS
Site Description
The study was conducted at Degero
¨Stormyr
(6411¢N, 1933¢E), an oligotrophic, minerogenic,
mixed mire system covering 6.5 km
2
and located in
the Kulba
¨cksliden Experimental Forest of the
Svartberget Long-Term Experimental Research
(LTER) facility near the town of Vindeln, county of
Va
¨sterbotten, Sweden. The mire area is situated on
a highland 270 m.a.s.l. between two major rivers,
Umea
¨lven and Vindela
¨lven, approximately 70 km
from the Gulf of Bothnia. The depth of the peat is
generally between 3 and 4 m, but depths up to 8 m
have been measured. The deepest organic deposits
correspond to an age of approximately 8000 years
and peat accumulation was initiated about 5500
BP. The mire catchment is predominantly drained
by the small creek Vargstugba
¨cken toward north-
west (Nilsson and others 2008).
The micro-topography is dominated by an irreg-
ular mosaic of carpets and lawns, with only sparse
occurrences of hummocks. The vascular plant
community of the mire consists mainly of cotton-
grass (Eriophorum vaginatum L.), tufted bulrush
(Trichophorum cespitosum L. Hartm.), cranberry
(Vaccinium oxycoccos L.), bog rosemary (Andromeda
polifolia L.), and cloudberry (Rubus chamaemorus L.)
with both mud sedge (Carex limosa L.) and Rannon
rush (Scheuzeria palustris L.) occurring more spar-
sely. In addition, Sphagnum majus Russ. C. Jens is
found on the bottom of the carpets whereas
Sphagnum lindbergii Schimp. and Sphagnum balticum
Russ. C. Jens are common for the lawns. The
hummocks are dominated by Sphagnum fuscum
Schimp. Klinggr. and Sphagnum rubellum Wils.
(Nilsson and others 2008; Laine and others 2012).
Total (moss capitula and vascular plants) above-
ground biomass is 141 ±45 g m
-2
(Laine and
others 2012).
The 30-year (1961–1990) mean annual precipi-
tation and air temperature are 523 mm and +1.2C,
Phenology Controls on Peatland Carbon Dioxide Exchange 117
respectively, whereas the mean air temperatures in
July and January are +14.7 and -12.4C, respec-
tively (Alexandersson and others 1991). The length
of the vegetation period (defined as the period
during which the mean daily temperature remains
above +5C) is 156 ±15 days. The snow cover
normally reaches a depth of up to 0.6 m and lasts
for approximately 6 months.
Eddy Covariance Measurements
The EC technique was used to measure the net
exchange of CO
2
at a sampling frequency of 20 Hz.
The system consisted of a three-dimensional sonic
anemometer (model 1012R3 Solent, Gill Instru-
ments, Lymington, UK) and a closed path infrared
gas analyzer (IRGA; model 6262, LI-COR, Lincoln,
NE, USA). The sonic anemometer was mounted on
the tower at a height of 1.8 m on a 1.0-m-long
boom and heated during the winter months. Fluxes
were calculated in real-time by the EcoFlux soft-
ware (In Situ Flux AB, Ockelbo, Sweden) according
to the EUROFLUX methodology (Aubinet and
others 1999) and stored as 30 min averages. Raw
data processing included a twofold coordinate
rotation by planar fit and a frequency correction for
the signal attenuation in the sampling path. The
CO
2
measurements from the IRGA were calibrated
against two reference gases approximately every
2 weeks. The energy balance closure was previ-
ously estimated to be 96% at this site (Sagerfors
and others 2008; Peichl and others 2013). More
details regarding instrumental set up and data
processing were previously outlined by Sagerfors
and others (2008) and Nilsson and others (2008).
Based on different estimates from three different
footprint models (Schuepp and others 1990; Hsieh
and others 2000; Kljun and others 2004), the
distances of the mean growing season peak source
and 90% fetch of the measured fluxes to the tower
during daytime (nighttime) ranged between
15–21 m (22–26 m) and 82–215 m (117–375 m),
respectively (Peichl and others 2014).
Missing half-hourly NEE data (due to instrument
and power failure, and so on) of up to 2 h were
filled with linear interpolation. Larger gaps during
the winter season were filled with the mean diurnal
variation (MDV) approach using a 14-day window
(described by Sagerfors and others 2008). During
the summer, NEE was filled by using standard
empirical regression relationships with air temper-
ature and radiation (Reichstein and others 2005).
Filled data represented 22 and 9% of the annual
time series for the years 2011 and 2012, respec-
tively. Separate estimates of GEP and ecosystem
respiration (ER) were derived with the procedure
described by Barr and others (2004). Negative NEE
values indicate net CO
2
uptake whereas positive
NEE values indicate emission of CO
2
. To facilitate
the direct comparison between the two component
fluxes, both GEP (by definition a negative flux) and
ER are presented as positive fluxes.
Abiotic Measurements
Instruments for measurements of abiotic variables
were mounted on the same tower as the EC sen-
sors. Photosynthetically active radiation (PAR) was
measured as photosynthetic photon flux density by
a quantum sensor (SKP 215, Skye Instruments Ltd,
Powys, UK) mounted at the top of the tower (that is,
4 m above the ground). In addition, air temperature
(T
a
) and relative humidity (RH) were measured by a
MP100 temperature and moisture sensor (Rotronic
AG, Bassersdorf, Switzerland) inside a self-venti-
lated radiation shield mounted 1.8 m above the
ground. The T
a
and RH data were used to derive the
atmospheric vapor pressure deficit (VPD). Rainfall
was measured using a tipping-bucket (ARG 100,
Campbell Scientific, Logan, Utah, USA) 4 m away
from the tower during April–October and corrected
for a (10%) underestimation of the precipitation (P)
(Sagerfors and others 2008). For the remaining part
of the year, P data were obtained from a nearby
(1 km) standard meteorological station which is
part of a reference climate monitoring program at
the Vindeln experimental forests. Gaps in the time
series of abiotic variables (due to instrument failure)
were filled either directly (that is, for P) or with
seasonal regression relationships (that is, for PAR,
T
a
, and RH) with respective data from the nearby
(13 km away) meteorological station at the
Svartberget Research Station. Soil temperatures (T
s
)
at 2, 10, 18, 26, 34, and 42 cm depths were mea-
sured by thermistors mounted in sealed, water-
proof, stainless steel tubes (TOJO Skogsteknik,
Dja
¨kneboda, Sweden) in a lawn community 100 m
northeast of the flux tower. The WTL below the
peat surface was measured in a lawn community
100 m northeast of the flux tower using a float and
counterweight system attached to a potentiometer
(Roulet and others 1991). All abiotic data were
collected at 10-s intervals and averaged to 30 min
mean values. Day length was determined from day
of year, latitude, and obliquity of the ecliptic
(Forsythe and others 1995). Growing degree days
(GDD) were calculated for each day (i) by taking the
average of the daily maximum and minimum T
a
(T
max
and T
min
, respectively) and subtracting a base
temperature (T
base
)of10C, with negative GDD
118 M. Peichl and others
values set to zero (that is, GDD = 0 for all days with
mean daily T
a
£10C) (equation 1a). For each
year, the GDDS was subsequently derived from the
cumulative sum of the GDD for all days between the
first day (that is, January 1, i= 1) and the last day
(j) of the summation period (equation 1b):
GDDi¼max 0;Tmaxi þTmini
2Tbase
ð1aÞ
GDDSj¼X
j
i¼1
GDDi:ð1bÞ
Estimation of Vegetation Canopy
Structure
To investigate the relationship between canopy
greenness and vegetation structure, we determined
the surface roughness length (z
0
) as proxy for
canopy height and the vascular plant LAI. Surface
roughness length was estimated using the hori-
zontal wind speed at measurement height and
iteration (n= 50) of the zero-displacement height
and z
0
as described by Sonnentag and others
(2011). LAI was determined in a nearby experi-
mental plot bordering the main footprint area by
clipping and scanning the biomass of seven micro-
plots every 2–3 weeks from May to October.
Digital Repeat Photography
In April 2011, we installed a digital point-and-
shoot camera (A480, Canon, Tokyo, Japan) 50 m
away from the flux tower roughly overseeing the
EC footprint area from a south-eastern viewing
direction (Figure 2). The digital camera was con-
tained in an outdoor camera housing (VT-EH10;
Vitek Industrial Video Products, Valencia, CA,
USA) mounted on a vertical pole at 3 m height
above the ground surface with a downward looking
viewing angle of 15. Images from the digital
camera were taken at hourly intervals and stored as
uncompressed 24-bit JPEG files on 4 GB memory
flash cards. To perform repeat photography with
the Canon A480, we wrote an intervalometer script
for the Canon Hack Development Kit (http://chdk.
wikia.com). The camera set up was in place from
about mid-April (just before snowmelt) to the mid/
end of November (until just after the first snowfall
event) in both years.
Chromatic Coordinates
To describe changes in peatland canopy color while
suppressing the effects of changes in scene illumi-
nation, we followed the procedure outlined in
Sonnentag and others (2012). In brief, for each
image of the digital image archive we calculated the
Figure 2. Seasonal color
changes at the Degero
¨
Stormyr mire; red frame
indicates the region of
interest used for the
image analysis (Color
figure online).
Phenology Controls on Peatland Carbon Dioxide Exchange 119
red, green, and blue chromatic coordinates (r
cc,
g
cc
,
and b
cc
)as
rcc ¼R
RþGþBð2aÞ
gcc ¼G
RþGþBð2bÞ
bcc ¼B
RþGþB;ð2cÞ
where R,G, and Bare the digital numbers (0–255)
of the red, green, and blue image channels. Next,
we created 3-day r
cc
,g
cc
,andb
cc
time series for
2011 and 2012 by assigning the 90th percentile of
all available hourly r
cc
,g
cc
, and b
cc
values within a
3-day window to the center day of a discrete (non-
overlapping) moving window (Sonnentag and
others 2012).
Regression Analysis of Biotic and Abiotic
Controls on the CO
2
Exchange
We used a stepwise approach of nonlinear regres-
sion analysis to test the explanatory power of biotic
and abiotic variables for estimating GEP and ER
following Shaver and others (2007). As a first step,
the best function type (that is, linear, exponential,
or hyperbolic) between predicted variable and each
explanatory variable was selected based on the
coefficient of determination (r
2
) and root mean
square error (RMSE). In the second step, multiple
nonlinear abiotic models were developed by adding
one explanatory variable at a time to a multiplica-
tive model until no further increase in r
2
was ob-
served. The final abiotic models selected for GEP
(equation 3) and ER (equation 4) were
GEPa¼c1c2PAR
c1PAR þc2
ðc3exp c4Ta
ðÞ
Þ
ðc5þc6WTLÞð3Þ
ERa¼c7exp c8Ta
ðÞ
;ð4Þ
where c
1–8
represent the model coefficients and
subscript aindicates the abiotic model estimate. As
a final step, the functional relationship between
CO
2
fluxes and g
cc
was added as biotic control to
create biotic–abiotic models for each GEP and ER
(equations 5,6)
GEPba ¼c1c2PAR
c1PAR þc2
ðc3exp c4Ta
ðÞ
Þ
ðc5þc6WTLÞðc7þc8gccÞð5Þ
ERba ¼ðc9exp c10 Ta
ðÞ
Þðc11 þc12 gccÞ;ð6Þ
where the subscript ba indicates the biotic–abiotic
model estimate. A similar analysis was conducted
to identify the relative importance of the main
abiotic controls for explaining variations in g
cc
. All
data processing was performed using the NonLin-
earModel.fit function in the statistical toolbox of
the software package MATLAB 2012b (The Math-
Works Inc., Natick, Massachusetts, USA).
RESULTS
Ecosystem Vegetation Structure and
Abiotic Controls on Mire Canopy Color
The temporal development of the destructively
sampled vascular plant LAI was synchronous with
g
cc
(Figure 3A). Overall, g
cc
explained 82% of the
Figure 3. ATemporal patterns of 3-day averages of the green chromatic coordinate (g
cc
) and destructively sampled leaf
area index (LAI) in 2012 and Bregression relationship between g
cc
and LAI for the entire growing season (black line) and
spring (green line and symbols) and autumn (yellow line and symbols). Regression equation and r
2
are given for the entire
growing season regression (Color figure online).
120 M. Peichl and others
variation in LAI in 2012 (Figure 3B). Moreover, the
exponential relationship between LAI and g
cc
also
showed a hysteresis pattern with lower LAI at gi-
ven g
cc
during the early growing season compared
to the late growing season.
The surface roughness length (z
0
) started to in-
crease by about 2–3 weeks earlier than the onset of
g
cc
(Figure 4). From May onward, z
0
and g
cc
were
coupled showing a similar slope of increase and
simultaneous peak in July. However, z
0
and g
cc
started to diverge again in August and remained
decoupled during the rest of the autumn period
when vascular plant vegetation turned from green
to brown while maintaining its vertical structure
and thus its surface drag effect.
The g
cc
was closely related to day length with
similar hysteresis loops in both years (Figure 5).
Deviations from this trajectory were related to be-
tween-year variations in abiotic conditions causing
differences in the onset of the spring green-up,
maximum g
cc
, and the termination date for the
vegetation period. However, the onset of senes-
cence (that is, the decline of g
cc
) occurred at the
same time in both years, that is, when day length
decreased to below 18 h.
The differences in the onset of spring green-up
(that is, rise in g
cc
) were also reflected by patterns of
the GDDS, which increased earlier in 2011 than in
2012 (Figure 6). We noted that a threshold of
about 50 GDDS was required as a trigger for a rise
Figure 4. Temporal
patterns of 3-day averages
for green chromatic
coordinate (g
cc
) and
surface roughness length
(z
0
) in 2011. Line indicates
polynomial fit through z
0
data points for visual
purposes.
Figure 5. Hysteresis type
regression relationship
between day length and
Agreen chromatic
coordinate (g
cc
) and B
gross ecosystem
production (GEP) in 2011
and 2012. Red dotted lines
visualize the mean
trajectory line; red arrows
indicate trajectory
direction from spring to
autumn; upward and
downward arrows indicate
the onset of spring
vegetation green-up and
senescence, respectively
(Color figure online).
Phenology Controls on Peatland Carbon Dioxide Exchange 121
in g
cc
and the associated start of the vegetation
green-up period.
Regression analysis determined T
a
as the main
control on g
cc
explaining 77% of its variance alone
in form of an exponential relationship, compared to
day length and PAR which explained 59 and 34%,
respectively (Table 1). Combined models of T
a
, day
length and PAR did not further increase the
explanatory power of the T
a
model.
The between-year differences (2011 relative to
2012) of the 3-day mean red chromatic index (r
cc
)
and WTL were negatively correlated (Figure 7).
Other ecosystem vegetation structure or abiotic
controls on r
cc
were not observed.
Relationship Between Mire Phenology,
Abiotic Variables, and Ecosystem CO
2
Exchange
The net uptake of CO
2
in 2011 started about
2 weeks earlier in spring, reached a greater maxi-
mum uptake of CO
2
in the summer and decreased
by about 2 weeks earlier at the end of the growing
season compared to 2012 (Figure 8A). An earlier
spring onset and autumn decrease, higher spring/
early summer values and summer maxima were
also observed for GEP and ER in 2011 than in 2012
(Figure 8B, C). Similarly to the CO
2
flux patterns,
g
cc
showed an earlier onset of the vegetation green-
up, higher spring/early summer values and earlier
senescence in 2011 than in 2012 (Figure 8D). In
contrast to the CO
2
fluxes, however, the magni-
tudes of the summer maxima of g
cc
were similar in
2011 and 2012. In both years, r
cc
steadily increased
throughout the growing season and started to de-
cline only in the late autumn when soil tempera-
ture reached the freezing point (Figure 8E, I).
Meanwhile, b
cc
decreased throughout the growing
season to a minimum in July and subsequently
increased during autumn (Figure 8F). Between-
Figure 6. ASeasonal patterns of 3-day averages for
green chromatic coordinate (g
cc
) and growing degree-day
sum (GDDS) and Bregression relationship between g
cc
and GDDS in 2011 and 2012. Dotted black line visualizes
the approximate GDDS threshold for spring vegetation
green-up.
Figure 7. Regression relationship of the between-year
(2011 relative to 2012) differences of the growing season
3-day mean red chromatic coordinate (r
cc diff
) and water
table level (WTL
diff
).
Table 1. Regression Analysis of Variables (That
is, Air Temperature, T
a
; Photosynthetically Active
Radiation, PAR; and Day Length, DL) Explaining
the Green Chromatic Coordinate (g
cc
)
Model Adj. R
2
RMSE
g
cc
=f
1
(PAR) 0.34 0.023
g
cc
=f
2
(DL) 0.59 0.018
g
cc
=f
2
(T
a
) 0.77 0.014
g
cc
=f
2
(T
a
)9f
2
(DL) 9f
1
(PAR) 0.78 0.013
f
1
=a+b9x.
f
2
=a9exp(b 9x).
RMSE = root mean square error; Adj. R
2
= adjusted coefficient of determination.
All regressions are significant at P <0.05.
122 M. Peichl and others
year differences in the r
cc
and b
cc
time series oc-
curred during July, October, and November, cor-
responding to periods with pronounced differences
in WTL (that is, lower WTL in 2011 was associated
with higher r
cc
and lower b
cc
than in 2012;
Figure 8E, F, J). Photosynthetically active radiation
Figure 8. Between-year variability in the 3-day averages of Anet ecosystem exchange (NEE), Bgross ecosystem pro-
duction (GEP), Cecosystem respiration (ER), Dgreen chromatic coordinate (g
cc
), Ered chromatic coordinate (r
cc
), Fblue
chromatic coordinate (b
cc
), Gair temperature (T
a
), Hsoil temperature (T
s
) at 2 cm depth, Iphotosynthetically active
radiation (PAR), and Jwater table level (WTL) in 2011 and 2012. Dotted horizontal line in A,Hand Jindicates zero line;
dashed vertical line in A–Cindicates separate contributions from mosses and vascular vegetation to NEE, GEP, ER, and g
cc
.
The color-shaded boxes highlight different periods with notable between-year differences in the respective variable (Color
figure online).
Phenology Controls on Peatland Carbon Dioxide Exchange 123
Figure 9. Temporal patterns for normalized 3-day averages of gross ecosystem production (GEP), green chromatic
coordinate (g
cc
), air and soil temperature (T
a
,T
s
), and photosynthetically active radiation (PAR) for 2011 and 2012. Arrows
indicate the timing of the spring onset (defined as the timing when the normalized time series curve rises above the zero
line) of PAR, T
a
,T
s
, GEP, and g
cc
, respectively.
Figure 8. continued
124 M. Peichl and others
was higher in late April and July of 2011 than for
the same period in 2012 (Figure 8G). Air temper-
ature increased earlier and remained higher during
April in 2011 than in 2012 (Figure 8H). For the
remaining growing season, T
a
was similar for both
years except for a few warmer events in 2011.
Based on T
s
at 2 cm depth, shallow soil thaw and
freezing occurred 1–2 weeks earlier and later,
respectively, in 2011 compared to 2012 (Figure 8I).
The spring onset (defined as the timing when the
normalized time series curve rises above the zero
line) of GEP lagged that of T
a
and PAR but coin-
cided with the shallow soil thawing and the onset
of the green-up (that is, increase in g
cc
) in both
years (Figure 9). Noteworthy, GEP started even
before shallow soil thawing and the onset of the
green-up in 2011, which indicates a contribution to
GEP from the uppermost moss capitula and/or al-
gae exposed to warmer T
a
and standing snowmelt
water on top of the frozen soil layer. The rate of
increase in GEP during the subsequent green-up
generally followed that of g
cc
, whereas day-to-day
variations in GEP were further modified by fluc-
tuations in T
a
and PAR as indicated by the simul-
taneous timing of their peaks and drops.
Furthermore, the timing of peak GEP and g
cc
was
synchronous with that of T
a
. Relative to 2012, a
faster decline of GEP occurred in autumn 2011
which corresponded to earlier decreases in PAR and
g
cc
, whereas autumn T
a
was similar in both years.
The pooled data from both years suggested a
strong correlation (r
2
= 0.68, P<0.001) between
g
cc
and GEP (Figure 10A). This link was especially
strong in spring and autumn but less pronounced
during the summer period when g
cc
reached its
maximum values. Moreover, there was a hysteresis
in the correlation between g
cc
and GEP with higher
GEP in spring than in autumn at the same g
cc
.
Furthermore, g
cc
was also correlated with ER
(r
2
= 0.50, P<0.01) and NEE (r
2
= 0.58,
P<0.001) although not as strongly as compared
with GEP (Figure 10B, C).
The stepwise nonlinear regression analysis sug-
gested that including g
cc
as an explanatory variable
into a biotic–abiotic model for estimating GEP im-
proved the explanatory power for the spring
(r
2
= 0.88) and autumn periods (r
2
= 0.97) com-
pared to that of the abiotic model (r
2
= 0.77 and
0.74 in spring and autumn, respectively) (Table 2).
Relative to the abiotic model, no further improve-
ment was obtained by including g
cc
during the
summer (r
2
= 0.85 and 0.85, respectively) and for
the entire growing season (r
2
= 0.93 and 0.94,
respectively). In contrast to GEP, ER was well ex-
plained by T
a
alone during spring, summer, and the
entire growing season (r
2
= 0.79, 0.83, and 0.87,
respectively). However, including g
cc
improved the
ER model during the autumn period (from
r
2
= 0.71 to 0.89).
DISCUSSION
Ecosystem Vegetation Structure and
Abiotic Controls on Mire Phenology
We observed abiotic controls on all three chromatic
coordinates, that is, g
cc
,r
cc
,andb
cc
. For instance,
between-year differences in r
cc
and b
cc
occurred
during periods with contrasting WTL. It is likely
that differences in the moisture content of the
moss and physiological changes in the vascular
plants during moisture stress resulted in differences
in the absorption and reflectance of the red and
blue fraction of incoming light by both vascular
and non-vascular vegetation (Vogelmann and
Figure 10. Growing season (all data points) and seasonal
(refer to colors) regression relationships between the 3-
day means of the green chromatic coordinate (g
cc
) and A
net ecosystem exchange (NEE), Bgross ecosystem pro-
duction (GEP), and Cecosystem respiration (ER) in 2011
(circles) and 2012 (triangles) (Color figure online).
Phenology Controls on Peatland Carbon Dioxide Exchange 125
Moss 1993; Bryant and Baird 2003; Graham and
others 2006; Gaalen and others 2007). For in-
stance, Graham and others (2006) observed larger
reflectance at red wavelengths (with the maximum
difference occurring at 675 nm) in dry moss (Tor-
tula princeps) compared to moist moss, which agrees
well with the response of higher r
cc
during periods
with lower WTL observed in our study. Our find-
ings also suggest a link between the low WTL and
high r
cc
during July and the subsequent earlier
decline of GEP in August of 2011, compared to
2012. This finding indicates some potential in using
the r
cc
as indicator for plant moisture stress.
Moisture-related changes in the reflectance over
the red and near infrared wavelength have been
previously identified for peatland vegetation (Bry-
ant and Baird 2003; Gaalen and others 2007). Less
information exists on moisture effects on plant
reflectance over blue wavelengths. However, given
that the g
cc
was similar during the dry periods in
both years in our study, the moisture response of
the b
cc
observed in our study might have also
simply resulted from a residual effect due to
changes in r
cc
at constant g
cc
. Thus, more work is
needed to better understand the separate effects
from vascular and non-vascular plant communities
on chromatic coordinates and phenology in mire
ecosystems. Overall, out of the three chromatic
coordinates investigated in this study, g
cc
repre-
senting the overall peatland canopy greenness
proved the most useful proxy for mire phenology
and CO
2
exchange. Given the strong control of g
cc
on the ecosystem CO
2
fluxes, it is thus also nec-
essary to further understand the drivers of mire
phenology itself.
Among the main ecosystem properties, LAI was
the one most closely related to g
cc
. Other ongoing
work at the site recently indicated that the corre-
lation between LAI and g
cc
appears to be consistent
also across years (unpublished data). These findings
further suggest that mire canopy greenness is dri-
ven primarily by changes in the vascular plant
canopy development. The observed seasonal hys-
teresis in this correlation might result from seasonal
changes in leaf nitrogen and/or chlorophyll con-
tents causing additional variations in peatland
canopy color (Keenan and others 2014; Yang and
others 2014). Thus, further improvements in the
understating of temporal patterns in g
cc
could pos-
sibly be achieved by including additional informa-
tion on leaf nitrogen and/or chlorophyll contents as
previously suggested for forest ecosystems (Keenan
and others 2014). Nevertheless, our results suggest
that g
cc
may, after initial calibration against manual
measurements, serve as a useful proxy for LAI and
thus mire vegetation development.
Table 2. Stepwise Nonlinear Regression Analysis of Variables (That is, Green Chromatic Coordinate, g
cc
; Air
Temperature, T
a
; Photosynthetically Active Radiation, PAR; and Water Table Level, WTL) Explaining Gross
Ecosystem Production (GEP) and Ecosystem Respiration (ER)
Model Spring Summer Autumn Growing season
Adj. R
2
RMSE Adj. R
2
RMSE Adj. R
2
RMSE Adj. R
2
RMSE
GEP
Single variable model
GEP = f
1
(g
cc
) 0.70 0.47 n.s. n.s. 0.91 0.14 0.68 0.83
GEP = f
2
(T
a
) 0.77 0.47 0.69 0.82 0.59 0.31 0.82 0.61
GEP = f
3
(PAR) 0.31 0.80 0.59 0.95 0.74 0.25 0.52 1.00
GEP = f
1
(WTL) 0.22 0.43 0.47 1.09 0.16 0.45 0.54 1.07
Abiotic model
GEP = f
2
(T
a
)9f
3
(PAR) 9f
1
(WTL) 0.77 0.24 0.85 0.59 0.74 0.27 0.93 0.44
Biotic–abiotic model
GEP = f
1
(g
cc
)9f
2
(T
a
)9f
3
(PAR) 9f
1
(WTL) 0.88 0.17 0.85 0.59 0.97 0.10 0.94 0.39
ER
Single variable model
ER = f
1
(g
cc
) 0.55 0.29 n.s. n.s. 0.89 0.11 0.60 0.63
ER = f
2
(T
a
) 0.79 0.25 0.83 0.48 0.71 0.19 0.87 0.36
Biotic–abiotic model
ER = f
1
(g
cc
)9f
2
(T
a
) 0.64 0.26 n.s. n.s. 0.89 0.12 0.89 0.32
Model function f
1
=a+b9x; f
2
=a9exp(b 9x); f3¼abx
axþb:
RMSE = root mean square error; Adj. R
2
= adjusted coefficient of determination; n.s. = not significant at P <0.05.
126 M. Peichl and others
Estimating leaf area in peatland ecosystems using
spectral or color indices of vegetation remains
challenging because of the short and multi-layer
canopy structure overlaying a non-vascular plant
community background (Sonnentag and others
2007a,b). Moreover, conventional optical leaf area
meters (for example, LAI-2200 Plant Canopy Ana-
lyzer by LI-COR) are limited to estimating the plant
area index including non-photosynthesizing plant
components and require comprehensive correction
algorithms (Ryu and others 2010). Our comparison
of g
cc
with the surface roughness length (z
0
) showed
a divergence in autumn caused by the increasing
presence of dead stand biomass. Furthermore, WTL
draw-down following spring flood, exposing non-
photosynthesizing vegetation material, resulted in
an immediate increase of the z
0
that might be
interpreted mistakenly as an earlier onset of vege-
tation development than actually observed in g
cc
and GEP. Thus, the use of g
cc
may serve as a better
proxy for green leaf area and thus photosyntheti-
cally active tissue compared to conventional leaf
area meters or canopy roughness estimates.
Day length and the circadian clock have been
previously identified as important controls on the
seasonal patterns of plant photosynthesis (Romero
and Valverde 2009; Bauerle and others 2012). Re-
cently, Bauerle and others (2012) reported that the
autumn decline in photosynthesis of tree saplings
was caused by photoperiodic down-regulation
overriding the effects from vegetation greenness
and phenological stage. In our study, we found that
the seasonal development of g
cc
also generally fol-
lowed day length in both years indicating an
internal control of vegetation development, which
is further modified by between-year differences in
abiotic drivers, that is, T
a
and cloudiness effects on
PAR. However, although a similar hysteresis was
observed between day length and GEP, the devia-
tions of GEP from this trajectory were closely
associated with those of g
cc
. For instance, the
slower decline of GEP in autumn 2012 (with day
length decreasing from 19 to 17 h) matched the
slower decline of g
cc
during the same period. Fur-
thermore, the marginal contribution of day length
in explaining variations in GEP in the stepwise
regression analysis suggests that photoperiodic
down-regulation of plant photosynthesis in au-
tumn may be less important than vegetation phe-
nology in this northern peatland ecosystem.
Possibly, given that maximum day length occurs
before the growing season peak at high latitudes,
the onset of vegetation senescence might lag the
seasonal cycle of day length to extend and maxi-
mize the length of the plant CO
2
uptake period in
northern boreal mires.
Our findings suggest that the concept of the
GDDS is also useful in explaining and predicting
vegetation phenology in mire ecosystems. Overall,
we found that T
a
expressed as GDDS best explained
g
cc
and between-year differences in the onset of the
spring vegetation green-up. Similarly, Ide and
Oguma (2010) reported a strong correlation be-
tween inter-annual variations in green-up dates
and local spring temperature in forested dry and
wetland ecosystems in Japan. Furthermore, a cer-
tain GDDS threshold, which has long been known
to exist for various plant species (Russelle and
others 1984), may also exist for the mire vascular
plant community. Thus, temperature patterns prior
to the growing season appear to be important with
respect to determining the timing of the GDDS
threshold required for triggering the vegetation
green-up. This further implies that future climatic
changes in the pre-growing season temperature
patterns may potentially result in significant alter-
ations of mire phenology and CO
2
exchange.
Relationship Between Mire Phenology,
Abiotic Variables and Ecosystem CO
2
Exchange
We observed a strong correlation between g
cc
de-
rived from the digital image archive and the mire
ecosystem CO
2
exchange, especially with GEP. The
between-year differences in the onset and devel-
opment of GEP during the spring vegetation green-
up period were strongly correlated with g
cc
. Fur-
thermore, vegetation phenology, described by g
cc
,
was identified as an important control on GEP
during the early growing season. During this green-
up period, vegetation development may therefore
be considered as the limiting factor setting a po-
tential baseline for GEP, which then is further
modified by fluctuations in the abiotic drivers such
as PAR and T
a
. Thus, greater GEP, ER, and en-
hanced net CO
2
uptake in 2011 compared to 2012
were likely the results of both earlier and enhanced
vegetation development during spring and more
favorable abiotic conditions during summer
weather events with higher T
a
and PAR.
In contrast, vegetation greenness played a less
important role in controlling the temporal pattern
of GEP in the summer period during which LAI and
chlorophyll content remain relatively constant at
their maximum. At this stage, variations in abiotic
controls govern the CO
2
flux patterns. For instance,
reduced PAR during a cloudy period in July 2012
Phenology Controls on Peatland Carbon Dioxide Exchange 127
was directly reflected in simultaneous reductions in
NEE, GEP, and ER. Moreover, similar maxima of
the g
cc
suggest that between-year differences in
abiotic controls primarily determined the different
magnitudes of the growing season maxima of the
ecosystem CO
2
fluxes.
The observed correlations of g
cc
with NEE and ER
might have resulted from temporal changes in GEP
and plant biomass which affected NEE but also ER
via the autotrophic respiration component (Chapin
and others 2006). Because both the heterotrophic
respiration component of ER and plant growth are
temperature driven (commonly resulting in a
strong temporal autocorrelation between T
a
and
plant biomass during the summer), ER is usually
related to T
a
. This is further supported in this study
where T
a
was statistically the strongest predictor of
ER. However, it is important to note that a con-
siderable amount of ER is produced from auto-
trophic plant respiration even in peatland
ecosystems (Hardie and others 2009) and thus di-
rectly linked to biomass, rather than to T
a
. Fur-
thermore, the between-year pattern and timing of
the switch from continuously higher net CO
2
up-
take, GEP, and ER fluxes during the early summer
to lower net CO
2
uptake, GEP, and ER fluxes dur-
ing the later summer in 2011 (relative to 2012)
corresponded directly to those observed for g
cc
.In
comparison, this clear between-year pattern was
not apparent in the T
a
and PAR time series. Thus,
differences in the amount and development of
plant biomass, as described by g
cc
, were a major
driver of the between-year variations in NEE, GEP,
and ER in this study.
It is noteworthy that, although vegetation
development and hence canopy greenness ap-
peared as an important control for the spring onset
and temporal patterns of the ecosystem CO
2
fluxes
(specifically GEP) during certain periods (that is,
spring and autumn) in this study, vegetation phe-
nology itself was related to abiotic controls, which
thus indirectly exerts effects on the ecosystem CO
2
exchange. Disentangling interactions among biotic
and abiotic controls and their direct and indirect
effects on the mire ecosystem CO
2
exchange re-
mains therefore a future research challenge (Tekl-
emariam and others 2010).
CONCLUSIONS
We used chromatic coordinates derived from a
digital image archive to investigate mire phenology
and its control on the CO
2
exchange estimated with
the EC method over 2 years (2011–2012) in a
minerogenic mire in northern Sweden. Based on
our findings we conclude:
Seasonal patterns of the green and red chromatic
coordinates (g
cc
and r
cc
) were well related to those
of the destructively sampled vascular plant leaf
area and of the water table level, respectively.
Moreover, the g
cc
was able to distinguish green
plant biomass from non-photosynthesizing plant
material. In addition, the r
cc
might provide useful
additional information on the water table dynam-
ics and related plant moisture status. Thus, the use
of chromatic coordinates derived from digital
repeat photography may serve as an inexpensive,
simple but effective method for the automated
and continuous quantification of mire phenology.
The main control on g
cc
and on the onset of the
spring vegetation green-up was air temperature
expressed as the growing degree-day sum (GDDS).
The existence of a GDDS threshold as trigger for
mire vascular plant green-up suggests that pre-
growing season temperature patterns and its
modification by future climatic changes may have
significant implications for mire phenology and
CO
2
exchange.
Between-year differences in the timing of the
onset and flux magnitudes of net ecosystem CO
2
uptake (NEE), gross ecosystem production (GEP),
and ecosystem respiration (ER) in spring could be
well explained by the differences in mire phenol-
ogy, described by g
cc
. Moreover, g
cc
was also an
important control on GEP and ER during the
autumn senescence period. In contrast, between-
year differences in the growing season maxima
and short-term fluctuations in NEE, GEP, and ER
were predominantly governed by abiotic controls,
specifically by air temperature and incoming
photosynthetically active radiation. Thus, mire
vegetation development might control and set the
potential baseline, whereas abiotic drivers further
determine peak magnitudes and short-term vari-
ations in the net exchange of CO
2
and its compo-
nent fluxes, GEP and ER, in peatland ecosystems.
ACKNOWLEDGMENTS
This study was financed by the Swedish Research
Council for Environment, Agricultural Sciences
and Spatial Planning (Grant No. 2007-666). We
also acknowledge the Kempe Foundation for the
grants supporting the micrometeorological instru-
mentation. Support from the Integrated Carbon
Observation System (ICOS) Sweden research
infrastructure (Swedish Research Council) is also
acknowledged. We thank Pernilla Lo
¨fvenius for the
128 M. Peichl and others
maintenance of the digital camera set-up and
management of the digital image archive. We also
thank Jo
¨rgen Sagerfors, Paul Smith, and Mikaell
Ottosson Lo
¨fvenius for helpful discussions of this
research work and for maintenance of the eddy
covariance and meteorological instrumentation at
the field site.
REFERENCES
Ahrends HE, Etzold S, Kutsch WL, Stoeckli R, Bruegger R,
Jeanneret F, Wanner H, Buchmann N, Eugster W. 2009. Tree
phenology and carbon dioxide fluxes: use of digital photog-
raphy for process-based interpretation at the ecosystem scale.
Clim Res 39:261–74.
Alexandersson H, Karlstro
¨m C, Larsson-Mccann S. 1991. Tem-
perature and precipitation in Sweden 1961–1990. Reference
normals. Meteorologi 81. Norrko
¨ping, Sweden: Swedish
Meteorological and Hydrological Institute (SMHI).
Aubinet M, Grelle A, Ibrom A, Rannik U, Moncrieff J, Foken T,
Kowalski A, Martin P, Berbigier P, Bernhofer C, Clement R,
Elbers J, Granier A, Grunwald T, Morgenstern K, Pilegaard K,
Rebmann C, Snijders W, Valentini R, Vesala T. 1999. Esti-
mates of the annual net carbon and water exchange of forests:
the EUROFLUX methodology. Adv Ecol Res 30:113–75.
Barr A, Black T, Hogg E, Kljun N, Morgenstern K, Nesic Z. 2004.
Inter-annual variability in the leaf area index of a boreal as-
pen-hazelnut forest in relation to net ecosystem production.
Agric For Meteorol 126:237–55.
Bauerle WL, Oren R, Way DA, Qian SS, Stoy PC, Thornton PE,
Bowden JD, Hoffman FM, Reynolds RF. 2012. Photoperiodic
regulation of the seasonal pattern of photosynthetic capacity
and the implications for carbon cycling. Proc Natl Acad Sci
109:8612–17.
Bryant RG, Baird AJ. 2003. The spectral behaviour of Sphagnum
canopies under varying hydrological conditions. Geophys Res
Lett 30:1134.
Bu Z, Hans J, Li H, Zhao G, Zheng X, Ma J, Zeng J. 2011. The
response of peatlands to climate warming: a review. Acta Ecol
Sin 31:157–62.
Bubier JL, Rock BN, Crill PM. 1997. Spectral reflectance mea-
surements of boreal wetland and forest mosses. J Geophys Res
Atmos 102:29483–94.
Chapin FSIII, Woodwell GM, Randerson JT, Rastetter EB, Lovett
GM, Baldocchi DD and others 2006. Reconciling carbon-cycle
concepts, terminology, and methods. Ecosystems 9:1041–50.
Chojnicki BH. 2013. Spectral estimation of wetland carbon
dioxide exchange. Int Agrophys 27:1–5.
Crimmins MA, Crimmins TM. 2008. Monitoring plant phenology
using digital repeat photography. Environ Manag 41:949–58.
Dise NB. 2009. Peatland response to global change. Science
326:810–11.
Eklundh L, Jin H, Schubert P, Guzinski R, Heliasz M. 2011. An
optical sensor network for vegetation phenology monitoring
and satellite data calibration. Sensors 11:7678–709.
Forsythe WC, Rykiel EJ Jr, Stahl RS, Wu H, Schoolfield RM.
1995. A model comparison for daylength as a function of
latitude and day of year. Ecol Model 80:87–95.
Gaalen KEV, Flanagan LB, Peddle DR. 2007. Photosynthesis,
chlorophyll fluorescence and spectral reflectance in Sphag-
num moss at varying water contents. Oecologia 153:19–28.
Gillespie AR, Kahle AB, Walker RE. 1987. Color enhancement of
highly correlated images. II. Channel ratio and ‘chromaticity’
transformation techniques. Remote Sens Environ 22:343–65.
Gorham E. 1991. Northern peatlands: role in the carbon cycle
and probable responses to climatic warming. Ecol Appl 1:182–
95.
Graham EA, Hamilton MP, Mishler BD, Rundel PW, Hansen
MH. 2006. Use of a networked digital camera to estimate net
CO
2
uptake of a desiccation-tolerant moss. Int J Plant Sci
167:751–8.
Hardie SML, Garnett MH, Fallick AE, Ostle NJ, Rowland AP.
2009. Bomb
14
C analysis of ecosystem respiration reveals that
peatland vegetation facilitates release of old carbon. Geoderma
153:393–401.
Harris A, Dash J. 2011. A new approach for estimating northern
peatland gross primary productivity using a satellite-sensor-
derived chlorophyll index. J Geophys Res Biogeosci
116:G04002.
Hsieh CI, Katul G, Chi T. 2000. An approximate analytical model
for footprint estimation of scaler fluxes in thermally stratified
atmospheric flows. Adv Water Resour 23:765–72.
Ide R, Oguma H. 2010. Use of digital cameras for phenological
observations. Ecol Inform 5:339–47.
IPCC. 2013: Climate change 2013: the physical science basis. In:
Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Bos-
chung J, Nauels A, Xia Y, Bex V, Midgley PM, Eds. Contri-
bution of working group I to the fifth assessment report of the
intergovernmental panel on climate change. Cambridge:
Cambridge University Press.
Keenan TF, Darby B, Felts E, Sonnentag O, Friedl M, Hufkens K,
O’ Keefe JF, Klosterman S, Munger JW, Toomey M, Rich-
ardson AD. 2014. Tracking forest phenology and seasonal
physiology using digital repeat photography: a critical assess-
ment. Ecol Appl . doi:10.1890/13-0652.1.
Kljun N, Calanca P, Rotach MW, Schmid HP. 2004. A simple
parameterisation for flux footprint predictions. Bound Layer
Meteorol 112:503–23.
Ko
¨rner C, Basler D. 2010. Phenology under global warming.
Science 327:1461–2.
Laine AM, Bubier J, Riutta T, Nilsson MB, Moore TR, Vasander
H, Tuittila ES. 2012. Abundance and composition of plant
biomass as potential controls for mire net ecosytem CO
2
ex-
change. Botany 90:63–74.
Lieth HH. 1974. Phenology and seasonality modeling. London:
Chapman and Hall. p 444.
Lindroth A, Lund M, Nilsson M, Aurela M, Christensen TR,
Laurila T, Rinne J, Riutta T, Sagerfors J, Stro
¨M L, Tuovinen J-
P, Vesala T. 2007. Environmental controls on the CO
2
ex-
change in north European mires. Tellus B 59:812–25.
Lund M, Lafleur PM, Roulet NT, Lindroth A, Christensen TR,
Aurela M, Chojnicki BH, Flanagan LB, Humphreys ER, Laurila
T, Oechel WC, Olejnik J, Rinne J, Schubert P, Nilsson MB.
2010. Variability in exchange of CO
2
across 12 northern
peatland and tundra sites. Glob Change Biol 16:2436–48.
Meyer GE, Neto JC. 2008. Verification of color vegetation indices
for automated crop imaging applications. Comput Electron
Agric 63:282–93.
Migliavacca M, Galvagno M, Cremonese E, Rossini M, Meroni
M, Sonnentag O, Cogliati S, Manca G, Diotri F, Busetto L,
Cescatti A, Colombo R, Fava F, Morra di Cella U, Pari E,
Siniscalco C, Richardson AD. 2011. Using digital repeat
photography and eddy covariance data to model grassland
Phenology Controls on Peatland Carbon Dioxide Exchange 129
phenology and photosynthetic CO
2
uptake. Agr Forest
Meteorol 151:1325–37.
Nilsson M, Sagerfors J, Buffam I, Laudon H, Eriksson T, Grelle A,
Klemedtsson L, Weslien P, Lindroth A. 2008. Contemporary
carbon accumulation in a boreal oligotrophic minerogenic
mire—a significant sink after accounting for all C-fluxes. Glob
Change Biol 14:2317–32.
Peichl M, Sagerfors J, Lindroth A, Buffam I, Grelle A, Kle-
medtsson L, Laudon H, Nilsson MB. 2013. Energy exchange
and water budget partitioning in a boreal minerogenic mire. J
Geophys Res Biogeosci 118:1–13.
Peichl M, O
¨quist M, Ottosson-Lo
¨fvenius M, Ilstedt U, Sagerfors J,
Grelle A, Lindroth A, Nilsson MB. 2014. A 12-year record
reveals pre-growing season temperature and water table level
threshold effects on the net carbon dioxide uptake in a boreal
fen. Environ Res Lett 9:055006. doi:10.1088/1748-9326/9/5/
055006.
Reichstein M, Falge E, Baldocchi D, Papale D, Aubinet M, Ber-
bigier P, Bernhofer C, Buchmann N, Gilmanov T, Granier A,
Grunwald T, Havrankova K, Ilvesniemi H, Janous D, Knohl A,
Laurila T, Lohila A, Loustau D, Matteucci G, Meyers T, Mi-
glietta F, Ourcival J-M, Pumpanen J, Rambal S, Rotenberg E,
Sanz M, Tenhunen J, Seufert G, Vaccari F, Vesala T, Yakir D,
Valentini R. 2005. On the separation of net ecosystem ex-
change into assimilation and ecosystem respiration: review
and improved algorithm. Glob Change Biol 11:1424–39.
Richardson AD, O’Keefe J. 2009. Phenological differences be-
tween understory and overstory. In: Noormets A, Ed. Phe-
nology of ecosystem processes. New York: Springer. p 87–117.
Richardson AD, Jenkins JP, Braswell BH, Hollinger DY, Ollinger
SV, Smith M-L. 2007. Use of digital webcam images to track
spring green-up in a deciduous broadleaf forest. Oecologia
152:323–34.
Richardson AD, Braswell BH, Hollinger DY, Jenkins JP, Ollinger
SV. 2009. Near-surface remote sensing of spatial and temporal
variation in canopy phenology. Ecol Appl 19:1417–28.
Richardson AD, Keenan TF, Migliavacca M, Ryu Y, Sonnentag
O, Toomey M. 2013. Climate change, phenology, and phe-
nological control of vegetation feedbacks to the climate sys-
tem. Agric For Meteorol 169:156–73.
Robroek B, Schouten MGC, Limpens J, Berendse F, Poorter H.
2009. Interactive effects of water table and precipitation on
net CO
2
assimilation of three co-occurring Sphagnum mosses
differing in distribution above the water table. Glob Change
Biol 15:680–91.
Romero JM, Valverde F. 2009. Evolutionarily conserved pho-
toperiod mechanisms in plants. Plant Signal Behav 4:642–4.
Roulet N, Hardill S, Comer N. 1991. Continuous measurement of
the depth of water table, inundation, in wetlands with fluc-
tuating surfaces. Hydrol Process 5:399–403.
Roulet NT, Lafleur PM, Richard PJH, Moore TR, Humphreys ER,
Bubier J. 2007. Contemporary carbon balance and late
Holocene carbon accumulation in a northern peatland. Glob
Change Biol 13:397–411.
Russelle MP, Wilhelm W, Olson RA, Power JF. 1984. Growth
analysis based on degree days. Crop Sci. 24:28–32.
Ryu Y, Nilson T, Kobayashi H, Sonnentag O, Law BE, Baldocchi
DD. 2010. On the correct estimation of effective leaf area in-
dex: does it reveal information on clumping effects? Agric For
Meteorol 150:463–72.
Sagerfors J, Lindroth A, Grelle A, Klemedtsson L, Weslien P,
Nilsson M. 2008. Annual CO
2
exchange between a nutrient-
poor, minerotrophic, boreal mire and the atmosphere. J
Geophys Res 113:G01001.
Sakamoto T, Gitelson AA, Nguy-Robertson AL, Arkebauer TJ,
Wardlow BD, Suyker AE, Verma SB, Shibayama M. 2012. An
alternative method using digital cameras for continuous
monitoring of crop status. Agric For Meteorol 154–155:113–26.
Savage K, Davidson EA, Tang J. 2013. Diel patterns of auto-
trophic and heterotrophic respiration among phenological
stages. Glob Chang Biol 19:1151–9.
Schubert P, Eklundh L, Lund M, Nilsson M. 2010. Estimating
northern peatland CO
2
exchange from MODIS time series
data. Remote Sens Environ 114:1178–89.
Schuepp PH, Leclerc MY, MacPherson JI, Desjardins RL. 1990.
Footprint prediction of scalar fluxes from analytical solutions
of the diffusion equation. Bound Layer Meteorol 50:355–73.
Shaver GR, Street LE, Rastetter EB, Van Wijk MT, Williams M.
2007. Functional convergence in regulation of net CO
2
flux in
heterogeneous tundra landscapes in Alaska and Sweden. J
Ecol 95:802–17.
Sonnentag O, Chen JM, Roberts DA, Talbot J, Halligan KQ,
Govind A. 2007a. Mapping tree and shrub leaf area indices in
an ombrotrophic peatland through multiple endmember
spectral unmixing. Remote Sens Environ 109:342–60.
Sonnentag O, Talbot J, Chen JM, Roulet NT. 2007b. Using direct
and indirect measurements of leaf area index to characterize
the shrub canopy in an ombrotrophic peatland. Agric For
Meteorol 144:200–12.
Sonnentag O, Detto M, Vargas R, Ryu Y, Runkle BRK, Kelly M,
Baldocchi DD. 2011. Tracking the structural and functional
development of a perennial pepperweed, Lepidium latifolium
L., infestation using a multi-year archive of webcam imagery
and eddy covariance measurements. Agric For Meteorol
151:916–26.
Sonnentag O, Hufkens K, Teshera-Sterne C, Young AM, Friedl
M, Braswell BH, Milliman T, O’Keefe J, Richardson AD. 2012.
Digital repeat photography for phenological research in forest
ecosystems. Agric For Meteorol 152:159–77.
Sulman BN, Desai AR, Saliendra NZ, Lafleur PM, Flanagan LB,
Sonnentag O, Mackay DS, Barr AG, van der Kamp G. 2010.
CO
2
fluxes at northern fens and bogs have opposite responses
to inter-annual fluctuations in water table. Geophys Res Lett
37:L19702.
Teklemariam TA, Lafleur PM, Moore TR, Roulet NT, Humphreys
ER. 2010. The direct and indirect effects of inter-annual
meteorological variability on ecosystem carbon dioxide ex-
change at a temperate ombrotrophic bog. Agric For Meteorol
150:1402–11.
Turunen J, Tomppo E, Tolonen K, Reinikainen A. 2002. Esti-
mating carbon accumulation rates of undrained mires in
Finland–application to boreal and subarctic regions. Holocene
12:69–80.
Vogelmann JE, Moss DM. 1993. Spectral reflectance measure-
ments in the genus Sphagnum. Remote Sens Environ 45:
273–9.
Westergaard-Nielsen A, Lund M, Hansen BU, Tamstorf MP.
2013. Camera derived vegetation greenness index as proxy for
gross primary production in a low Arctic wetland area. ISPRS J
Photogramm Remote Sens 86:89–99.
Wilson D, Alm J, Riutta T, Laine J, Byrne KA, Farrell EP, Tuittila
E-S. 2007. A high resolution green area index for modelling
the seasonal dynamics of CO
2
exchange in peatland vascular
plant communities. Plant Ecol 190:37–51.
130 M. Peichl and others
Woebbecke DM, Meyer GE, Von Bargen K, Mortensen DA.
1995. Color indices for weed identification under various soil,
residue, and lighting conditions. Trans ASAE 38(1):259–269.
Yang X, Tang J, Mustard JF. 2014. Beyond leaf color: comparing
camera-based phenological metrics with leaf biochemical,
biophysical, and spectral properties throughout the growing
season of a temperate deciduous forest. J Geophys Res Bio-
geosci 119:181–91.
Yurova A, Wolf A, Sagerfors J, Nilsson M. 2007. Variations in net
ecosystem exchange of carbon dioxide in a boreal mire:
modeling mechanisms linked to water table position. J Geo-
phys Res 112:G02025.
Phenology Controls on Peatland Carbon Dioxide Exchange 131