Predicting the onset of net carbon uptake by deciduous forests with soil temperature and climate data: A synthesis of FLUXNET data

Abstract

We tested the hypothesis that the date of the onset of net carbon uptake by temperate deciduous forest canopies corresponds with the time when the mean daily soil temperature equals the mean annual air temperature. The hypothesis was tested using over 30 site-years of data from 12 field sites where CO(2) exchange is being measured continuously with the eddy covariance method. The sites spanned the geographic range of Europe, North America and Asia and spanned a climate space of 16 degrees C in mean annual temperature. The tested phenology rule was robust and worked well over a 75 day range of the initiation of carbon uptake, starting as early as day 88 near Ione, California to as late as day 147 near Takayama, Japan. Overall, we observed that 64% of variance in the timing when net carbon uptake started was explained by the date when soil temperature matched the mean annual air temperature. We also observed a strong correlation between mean annual air temperature and the day that a deciduous forest starts to be a carbon sink. Consequently we are able to provide a simple phenological rule that can be implemented in regional carbon balance models and be assessed with soil and temperature outputs produced by climate and weather models.

Full text

Int J Biometeorol (2005) 49: 377–387
DOI 10.1007/s00484-005-0256-4
ORIGINAL ARTICLE
Dennis D. Baldocchi · T. A. Black · P. S. Curtis ·
E. Falge · J. D. Fuentes · A. Granier · L. Gu ·
A. Knohl · K. Pilegaard · H. P. Schmid · R. Valentini ·
K. Wilson · S. Wofsy · L. Xu · S. Yamamoto
Predicting the onset of net carbon uptake by deciduous forests
with soil temperature and climate data: a synthesis of FLUXNET data
Received: 23 June 2004 / Revised: 21 October 2004 / Accepted: 30 December 2004 / Published online: 2 February 2005
C

ISB 2005
Abstract We tested the hypothesis that the date of the
onset of net carbon uptake by temperate deciduous for-
est canopies corresponds with the time when the mean
daily soil temperature equals the mean annual air temper-
ature. The hypothesis was tested using over 30 site-years
of data from 12 field sites where CO
2
exchange is being
measured continuously with the eddy covariance method.
The sites spanned the geographic range of Europe, North
America and Asia and spanned a climate space of 16
◦
Cin
mean annual temperature. The tested phenology rule was
D. D. Baldocchi (

) · L. Xu
Ecosystem Science Division, Department of Environmental
Science, Policy and Management, University of California,
151 Hilgard Hall,
Berkeley, CA 94720, USA
e-mail: baldocchi@nature.berkeley.edu
T. A. Black
Faculty of Agricultural Sciences, University of British
Columbia,
Vancouver, BC, Canada
P. S. Curtis
Department of Evolution, Ecology & Organismal Biology, Ohio
State University,
Columbus, OH, USA
E. Falge
Bayreuth University,
Bayreuth, Germany
J. D. Fuentes
Department of Environmental Sciences, University of Virginia,
Charlottesville, VA, USA
A. Granier
INRA,
Champenoux, France
L. Gu
Environmental Science Division, Oak Ridge National
Laboratory,
Oak Ridge, TN, USA
A. Knohl
Max Planck Institute for Biogeochemistry, Jena Germany,
K. Pilegaard
RISOE,
Roskilde, Denmark
H. P. Schmid
Department of Geography, Indiana University,
Bloomington, IN, USA
R. Valentini
Department of Forest Science and Environment, Universita’ di
Tuscia,
Viterbo, Italy
K. Wilson
Atmospheric Turbulence and Diffusion Division, NOAA,
Oak Ridge, TN, USA
S. Wofsy
Department of Earth and Planetary Science, Harvard University,
Cambridge, MA, USA
S. Yamamoto
National Institute of Advanced Industrial Science and
Technology,
Tsukuba, Ibaraki 305-8569, Japan
Present address:
L. Xu
LICOR,
Lincoln, NE, USA
A. Knohl
ESPM, University of California,
Berkeley, CA, USA
robust and worked well over a 75 day range of the initia-
tion of carbon uptake, starting as early as day 88 near Ione,
California to as late as day 147 near Takayama, Japan. Over-
all, we observed that 64% of variance in the timing when net
carbon uptake started was explained by the date when soil
temperature matched the mean annual air temperature. We
also observed a strong correlation between mean annual air
temperature and the day that a deciduous forest starts to be
a carbon sink. Consequently we are able to provide a sim-
ple phenological rule that can be implemented in regional

378
carbon balance models and be assessed with soil and tem-
perature outputs produced by climate and weather models.
Keywords Phenology
.
Eddy covariance
.
CO
2
exchange
.
Canopy photosynthesis
Introduction
The timing of leaf out, across the temperate deciduous for-
est biome, has major implications on the seasonal variation
of numerous ecosystem/atmosphere interactions. For ex-
ample, the transition between dormant and vegetated states
of deciduous forests causes an abrupt change in surface
albedo, aerodynamic roughness and the surface conduc-
tance to water, carbon dioxide and pollutant transfer. To-
gether, these variables alter the availability of energy and its
partitioning into sensible and latent heat exchange(Blanken
et al. 2001; Moore et al. 1996; Wilson and Baldocchi 2000).
A perturbation in the partitioning of energy is consequen-
tial because it alters the growth rate and ultimate depth
of the planetary boundary layer (McNaughton and Spriggs
1986). The modifications in boundary layer growth, in turn,
affect the diurnal course and amplitude of temperature
and humidity within the surface and planetary boundary
layer (Schwartz and Karl 1990). Furthermore, the absence
or presence of leaves can act as a switch for the forma-
tion of convective clouds through links with the properties
of the planetary boundary layers (Fitzjarrald et al. 2001;
Freedman et al. 2001; Schwartz and Crawford 2001).
The timing of leaf-out marks the beginning of the pho-
tosynthetic season for a deciduous forest and is a major
determinant of its duration (White et al. 1999). With re-
gards to terrestrial carbon cycling, the length of carbon
uptake period has much predictive power about the spatial
variation of the net annual carbon exchange of ecosystems
(NEE) across a latitudinal and continental gradient of de-
ciduous forests (Baldocchi et al. 2001)—the length of the
carbon uptake period explained 80% of the spatial vari-
ance in annual NEE. It has also been reported—on both
experimental (Black et al. 2000; Goulden et al. 1996b;
Schmid et al. 2000; Schmid et al. 2003) and theoretical
bases (White et al. 1999)—that the timing of leaf out pro-
vides partial explanation for the year-to-year variability in
NEE at individual sites; additional controlling factors on
NEE include presence and absence of snow, drought, and
summer cloudiness.
Lately, phenology has received added attention because
of its role as a surrogate in detecting global climate change
(Jackson et al. 2001; Penuelas and Filella 2001; White
et al. 2003). Phenological indices measured across Europe
(Menzel and Fabian 1999) and North America (Cayan et al.
2001) and interannual measurements of biosphere green-
ness, as observed by satellites (Myneni et al. 1997; Tucker
et al. 2001), are detecting a trend towards earlier springtime
leaf-out, portending a potential signal of global warming. In
order to simulate the implications of changing phenology
on biosphere-atmosphere interactions, models that com-
pute the biogeochemical cycling of water, carbon and nu-
trients, atmospheric chemistry, weather and climate need
algorithms that can predict the timing of leaf expansion,
the initiation of photosynthesis and the onset of net carbon
uptake by deciduous forests (White et al. 1997).
Dates of bud break, leaf unfolding, and commencement
of photosynthesis have been used to characterize aspects of
forest phenology (Brugger et al. 2003;Guetal.2003).
Physiological mechanisms for the timing of springtime
phenological events involve a need for dormancy and are
triggered by interactions between growth-promoting and
inhibiting hormones (Schaber and Badeck 2003). The re-
lease of these hormones seems to be triggered by an
accumulation of winter chilling, a critical photoperiod
and springtime warming. Historically, phenological mod-
els have used photoperiod and cumulative heat and chilling
units as independent variables (Chuine et al. 2003; Kramer
et al. 2000; Nizinski and Saugier 1988; Raulier and Bernier
2000; Spano et al. 1999). While this class of models has
many practical applications, it is highly empirical. Conse-
quently, its generality is limited because the threshold sum
of heat and chill units, that determines the date of a partic-
ular phenological event, needs to be calibrated at each site
and for each mix of species (Kramer et al. 2000; Raulier and
Bernier 2000; Taylor 1974). We also note that there can be
considerable imprecision with identifying the specific date
of a phenological event due to natural temporal and spatial
variability and sampling errors. Phenological metrics like
bud break, leaf unfolding and the onset of photosynthesis
are not synchronous and can occur for an extended period
(Brugger et al. 2003; Morecroft et al. 2003). For example, it
takes 11 days for European oak leaves to transcend between
budbreak and a physiological state that maintains a positive
carbon balance (Morecroft et al. 2003). Considerable spa-
tial variation in the timing of phenology will occur within a
woodland, too, due to the presence of multiple species and
because of microclimate variations. Sampling error is yet
another source of variation. Detecting the precise date of a
phenological event depends upon the sampling frequency
and sampling area associated with manual observations or
with the frequency of passage and pixel size associated
with remote sensing instruments mounted on a satellite or
tower.
Our goal is to assess a simple phenological rule that has
a detection criterion that is based on biophysical principles
and can be assessed with information that is commonly
available at weather stations and from weather and clima-
tological forecast models (e.g. air and soil temperature).
We propose and test the hypothesis that: “the date of the
onset of net carbon uptake by a temperate deciduous forest
corresponds with the time when the mean soil tempera-
ture equals its mean annual air temperature”. This working
hypothesis was generated from observing the timing of
leaf-out near Oak Ridge, Tennessee. There, trees tended to
leaf out when the soil temperature was near 13
◦
C (Taylor
1974), a temperature that corresponds with the mean annual
air temperature of the region. While this anecdotal obser-
vation requires further scrutiny with a larger database, we
contend that there may be mechanistic justifications for this
hypothesis. For example, soil temperature acts as a proxy

379
Table 1 List of the field sites used in this analysis, their location, mean annual air temperature and citations describing site characteristics
and meteorological measurements
Site Genera Lat Long Mean annual
temperature
citation
Prince Albert, Saskatchewan Populus 53 N 106 W 0.6 Black et al. (2000)
Douglas Lake, Michigan Populus/Quercus 45 N 84 W 6.2 Schmid et al. (2003)
Borden, Ontario Acer/Populus 44 N 79 W 6.4 Lee et al. (1999)
Collelongo Italy Fagus 41 N 13 E 6.5 Valentini et al. (1996)
Takayama, Japan Betula 36 N 137 E 7.3 Yamamoto et al. (1999)
Harvard Forest, Massachusetts Quercus/Acer 42 N 72 W 7.4 Goulden et al. (1996b)
Hainich, Germany Fagus 51N 10 E 7.5 Knohl et al. (2003)
Soroe, Denmark Fagus 55 N 11 E 7.6 Pilegaard et al. (2001)
Hesse, France Fagus 48 N 7 E 9.2 Granier et al. (2002)
Morgan Monroe, Indiana Quercus 39 N 86 W 11.8 Schmid et al. (2000)
Oak Ridge, Tennessee Quercus/Acer 36 N 84 W 14.9 Wilson and Baldocchi (2001)
Ione, California Quercus 38 N 120 W 16.5 Baldocchi et al. (2004)
for accumulated chill and heat units as it is an integrator
of these measures due to the soil’s thermal inertia and en-
ergy balance. Secondly, deciduous trees have evolved to
be in synchrony with their climate to minimize the expo-
sure of their young leaves to freezing and possibly lethal
temperatures during the spring; the probability of frost be-
comes quite low once mean daily air temperature exceeds
its annual mean temperature.
We test this hypothesis using continuous CO
2
flux mea-
surements, an approach that has been adopted in other
recent phenological studies (Gu et al. 2003; Suni et al.
2003). The analysis is based on over 30 site-years of me-
teorological and carbon flux data from 12 temperate de-
ciduous forest sites that are associated with the FLUXNET
project (Baldocchi et al. 2001). The datasets examined here
were acquired from sites that are distributed across the ge-
ographic domains of Europe, North America and Asia and
they span 16
◦
C in mean annual temperature. An advantage
of quantifying phenology with eddy covariance measure-
ments includes its capacity to make nearly continuous mea-
surements and to sample a large area, as characterized by
its flux footprint (Schmid 2002).
Materials and methods
We restricted this phenological study to FLUXNET study
sites whose trees have broad leaves and deciduous habits.
The key genera at the sites used in this analysis include Pop-
ulus (aspen), Acer (maple), Quercus (oak), Betula (birch)
and Fagus (beech). Forests at the majority of sites inspected
formed closed canopies; their leaf area indices ranged be-
tween 3 and 6 and their tree heights ranged between 8
and 30 m. With regards to age structure, most of the sites
were second-growth forests and were less than 120 years
old. Characteristics of the sites used in this analysis, and
primary references describing additional site details, are
summarized in Table 1.
A common set of meteorological and eddy flux measure-
ments were acquired from each of the sites involved in
the analysis. The datasets scrutinized here included long-
term and simultaneous measurements of air and soil tem-
perature and net ecosystem CO
2
exchange between the
forest and the atmosphere. Air temperature was measured
above the forest stands with aspirated and shielded sensors.
Soil temperatures were measured with either thermistor
or thermocouple sensors; most sites had soil temperature
measurements at 2, 5, 8 or 10 cm depths. Daily means
were computed from the original 30 min datasets using
the mean diurnal course gap-filling method (Falge et al.
2001).
CO
2
flux densities were measured across the forest-
atmosphere interface with the eddy covariance method
(Baldocchi et al. 1988). In Europe, flux measurements
systems were based on closed path CO
2
sensors (Aubinet
et al. 2000), while in North America, both open and closed
path CO
2
sensors were used; independent studies show
that there is no bias by using either an open or closed path
sensor system to measure CO
2
exchange (Billesbach et al.
2004; Suyker and Verma 1993).
Efforts have been made by the AmeriFlux and FLUXNET
communities to inter-compare CO
2
flux measurements and
meteorological measurements with a roving, calibration
system (Baldocchi et al. 2001; Billesbach et al. 2004). In
general, the absolute errors in eddy flux measurements of
CO
2
exchange are less than 15%, with the application of
proper corrections (Goulden et al. 1996a; Hollinger et al.
2004; Massman and Lee 2002).
The date of onset of net carbon uptake was identified as
the day when daily integrated net CO
2
exchange (NEE)
experienced a transition from its winter respiration phase
to its spring/summer assimilation phase (Fig. 1). To deduce
this date with objective and statistical rigor, we regressed
measurements of daily NEE on day of year, using a
subsets of data from this springtime source-sink transition
period. The method has a clean and distinct signal with
a quantifiable error. For the case shown in Fig. 1, the
statistical variation in detecting the precise date of the
zero crossing is ±2 days, based on the 95% confidence
interval.

380
Fig. 1 An example of daily net
ecosystem CO
2
exchange
measurements (NEE)asa
function of time. The solid line
represents the linear regression
and the dashed lines are the
95% confidence interval.
During this transition period a
linear regression was fit through
the data and the ‘leaf-out’ date
was computed by inverting that
regression and solving for when
NEE was zero
On a physiological and fundamental level, it may be
preferable to detect the onset of canopy photosynthesis
rather than rely on transitional observations of NEE. How-
ever, the use of canopy photosynthesis can introduce an
additional source of error and imprecision because canopy
photosynthesis is a derived quantity that is assessed by sub-
tracting an indirect estimate of ecosystem respiration from
direct measurements of NEE (Falge et al. 2002). Ecosystem
respiration (R
eco
), for example, is deduced from night mea-
surements (which are uncertain due to turbulent mixing)
and is calculated during the daytime with a temperature-
dependent function (Falge et al. 2001). During the dynamic
spring growth period, temperature response functions, that
are commonly used to assess ecosystem respiration are apt
to fail because growth respiration is accelerating during
this transitional period and it enhances ecosystem respira-
tion (Xu et al. 2004).
Our working hypothesis presumes that there is a corre-
spondence between the date of the initiation of net carbon
uptake and the day when mean daily-averaged soil temper-
ature crosses a line corresponding with the mean annual
air temperature. Since a tree is unable to sense the mean
annual air temperature a priori, we decided to approximate
mean annual air temperature (
T ) using a low-pass, digital
recursive filter (Hamming 1989). The low-pass character
of this recursive filter provides us with a method that ap-
proximates the temporal mean by weighting the current air
temperature with its history:
T
t
= (1 −α)T
t−1
+αT
t
(1)
The mean air temperature, at time t, is updated based on
its previous mean value, at t−1, and the most recent daily-
averaged air temperature, T
t
. The weighting factor, α,is
defined as:
α = exp

−
t
τ

(2)
We computed mean air temperature using a 730 day (2
year) time constant, τ , and a 1 day sampling interval, t.
Results and discussion
Before we attempt to test our working hypothesis across
a network of field sites, it is crucial that we demonstrate
whether or not this concept works across a representative
sampling of study sites (Figs. 2, 3). Figure 2 shows a test
of the hypothesis for a beech forest in Denmark; this site
is near the most northerly and coolest end of the deciduous
forest biome range in the FLUXNET network. We observe
that there is very close correspondence between the date
when net CO
2
exchange crossed the zero line and when
mean soil temperature (computed with a 3 day running
mean) crossed the isotherm representing the mean annual
air temperature at that site. A second example is shown for
a contrasting case, a site near Oak Ridge, Tenn. This site
is near the most southern and warmest end of the decidu-
ous forest biome (Fig. 3). Like the Danish forest, there is a
close correspondence between the date of the zero crossing
of canopy CO
2
exchange and when the soil temperature
crosses the line representing the mean annual air temper-
ature at that site. Despite the fact that the commencement
of net CO
2
uptake near Oak Ridge, Tenn. occurs about 30
days earlier and the mean annual air temperature is about

381
Fig. 2 Seasonal course in daily
integrated net ecosystem CO
2
exchange (NEE), mean air
temperature computed with a
digital recursive filter, mean air
temperature computed with a
digital recursive filter and the
mean daily soil temperature at
2 cm (based on a 3 day running
mean). These data were
acquired over a beech forest in
Denmark during 1999. The
double-headed arrow identifies
when soil temperature matched
mean annual air temperature
and when NEE crossedzero(the
lower horizontal line)
6
◦
C warmer than in Denmark, the general relationship ‘that
a critical soil temperature identifies the onset of net carbon
uptake by the ecosystem’ holds.
How well the match between daily mean soil tempera-
ture and mean annual air temperature provides a precise
gauge for predicting the onset of net CO
2
uptake for the en-
tire database is quantified in Fig. 4. We report that a linear
regression between the dependent (the day T
soil
equaled
mean annual air temperature) and independent (the day
NEE equaled zero) variables accounts for 64% of the vari-
ance. Furthermore, the slope of the regression was close to,
but significantly different from one (0.929±0.21) and the
intercept was 17.1±25.3 days. Because the dependent and
independent variables have sampling and measurement er-
rors, we also computed the geometric mean regression. In
this case, the slope was 1.15±0.344 and the intercept was
−8.57±41.5.
Another question we can pose and address is: ‘how well
can climatological data describe when a deciduous forest
transcends from being a carbon source to a sink?’ Figure 5
shows that the start-date of net carbon uptake becomes
earlier, in a linear fashion, as the local climate (mean annual
air temperature) becomes warmer. Overall, perturbations in
mean annual air temperature explain 69% of the variance
in the start-date of net carbon uptake.
In sum, both phenology rules seem robust by working
well over a 75 day range of the initiation of carbon uptake,
starting as early as day 88 in near Ione, California to as
late as day 147 near Takayama, Japan. However, the results
shown here may not be universal for all functional plant
types and climate zones. For example, Suni et al. (2003)
reported that soil temperature was not a good indicator
for signaling the onset of photosynthesis across the boreal
forest biome. On the other hand, they found that air temper-
ature was a good indicator of the onset of photosynthesis
for conifers at high latitudes, but they found no unifying re-
lationship that held across the boreal forest biome. Hence,
we advise the reader to apply the functional relationship
between soil temperature and the onset of carbon uptake
only to deciduous broadleaved forests. Close inspection of
Fig. 4 shows that the significant outliers were associated
with measurements at Harvard Forest, in Massachusetts,
where net carbon uptake starts later than one would expect
based on soil temperature measurements. This site is near
the eastern edge of the North American continent and is a
locale subject to much climatic variability during the spring
due to the passing of warm and cold air masses; examin-
ing over 40 years of mean air temperature we found that
the daily mean temperature ranges between 0 and 20
◦
C
around the expected date of leaf out, approximately day
120. So soil temperature may not queue the phenology of
net carbon uptake as well at this site as others. We also
add that our phenology rule does not work well for a de-
ciduous forest in the boreal zone, where the mean annual
temperature is close to zero centigrade and snow cover
keeps soil temperature close to zero during the winter and
early spring (Griffis et al. 2003). Consequently, soil temper-
ature in this region crosses the isotherm representing mean
air temperature much sooner than when leaves emerge
(Fig. 6).
Measurements of net CO
2
exchange have the potential
for assessing the timing of leaf-out if we know the time

382
Fig. 3 Seasonal course in daily
integrated net CO
2
exchange
(NEE) and the mean daily soil
temperature at 8 cm. These data
were acquired over an
oak/maple forest near Oak
Ridge, Tenn. during 1996. The
double-headed arrow identifies
when soil temperature matched
mean annual air temperature
and when NEE crossedzero(the
lower horizontal line)
Fig. 4 The empirical
relationship between the date
when mean daily soil
temperature equals mean annual
air temperature and when daily
net ecosystem carbon exchange
(NEE) crosses zero. The solid
line represents the linear
regression through the data and
the dashed line represents the
95% confidence interval
delay between when leaves unfold and when canopy pho-
tosynthesis matches soil respiration. We detected the date
of leaf-out at a few selected sites using light transmission
measurements through the canopy; leaf-out was identified
when the fraction of beam penetration through the forest
decreased, relative to its winter deciduous state. Our detec-
tion of the date of leaf out, with CO
2
flux measurements,
accounted for over 94% of the variance in the timing of
leaf-out observed with light measurements. The statistics
associated with the linear regression possessed a non-zero
intercept of 22.1±29.1 days and the regression slope that
was less than one (0.744±0.243) (Fig. 7). We also observed

383
Fig. 5 The relationship
between the mean annual air
temperature and the day when
net carbon of deciduous forests
uptake commences. The solid
line represents the linear
regression through the data and
the dashed line represents the
95% confidence interval
Fig. 6 Seasonal course in daily
integrated net ecosystem CO
2
exchange (NEE), the mean daily
soil temperature at 8 cm
(computed with a 3 day running
mean) and mean air
temperature, computed with a
digital recursive filter. These
data were acquired over an
aspen forest in the southern
portion of the boreal zone of
Canada during 2000
that the onset of net carbon uptake, relative to the date of
leaf-out, becomes more delayed as the start of growing
season becomes later.
To compensate for the bias between the date of observed
leaf-out and the onset of net carbon uptake, we produced
a transformed metric using the regression between the ob-
served and inferred dates of leaf-out, discussed in Fig. 7.We
next compared this transformed metric for identifying the
date of leaf out against the date when mean daily soil tem-
perature crosses mean annual air temperature (Fig. 8). For
the dataset in hand, we found that the mean date of leaf out,
detected using soil and air temperature, was ±114.9±14.9

384
Fig. 7 The relationship
between the day net ecosystem
CO
2
exchange crossed zero and
the date that leaf out was
observed. Data were from sites
in Michigan, Tennessee,
Saskatchewan and California
Fig. 8 A test of the timing of
leaf out as detected by the date
when soil temperature crosses
the mean air temperature, as
computed with a recursive
digital filter. Data on the
dependent axis were assessed
by applying the empirical
relation between the observed
date of leaf out and that detected
with CO
2
flux measurements.
Statistical analysis (via a paired
t-test) indicates there is no
significant difference between
the data on the dependent and
independent axes
days and the mean date detected with the transformed CO
2
flux measurements was 116.6±14.3 days. Further analysis
of the data, using Student’s paired t statistic, indicates that
there was no significant difference between the two means
on the 5% probability level (t=0.413; P=0.60; 29 df). A
linear regression between the independent and dependent
variables explained 61% of the variance, had a slope of
0.744 and an intercept of 28.7.
There are several sources of variation associated with the
results in Figs. 5 and 8 that merit further discussion. Soil
temperature was not measured at uniform depths across
the network, so this source of variation may contribute to
some ‘noise’ introduced into the cross-site comparison. In
general, we attempted to minimize this source of variation
by: (1) using soil temperatures measured in the area of
the main root activity, 5 to 16 cm; (2) by relying on daily

385
Fig. 9 Seasonal variation in
mean daily soil at 2 and 32 cm
and mean air temperature
computed with a digital
recursive filter. The data were
collected at an oak woodland
field site in California during
2003. Both measures of soil
temperature crossed the mean
air temperature at day 90
Fig. 10 Seasonal course of
daily-integrated CO
2
flux and
canopy photosynthesis at Oak
Ridge, Tenn. during 1999.
Canopy photosynthesis was
computed by subtracting
understory eddy flux
measurements from the
overstory measurements
mean temperatures, a more conservative metric; and (3) by
applying a 3-day running mean to the soil temperatures.
Overall, site-to-site differences in the depth of soil temper-
ature measurements probably had a minor and secondary
effect on the results shown in Figs. 5 and 8. This claim is
supported with experimental data shown in Fig. 9. We ob-
serve that there was little difference when daily mean soil
temperature, at 2 or 32 cm depths, first crossed the isotherm
representing mean annual air temperature at an oak savanna
field site in California and at other sites where soil temper-
atures was measured at multiple depths as in Oak Ridge,
Tennessee and Soroe, Denmark (data not shown).

386
There may also be imprecision associated with using the
CO
2
flux cross-over date as a measure of leaf out, rather
than canopy photosynthesis. At most temperate deciduous
forest sites there will be some photosynthesis prior to this
date, which offsets soil respiration. But the temporal change
in both NEE and A
c
during spring is rapid and will only
cause a few days lag in the detected leaf out date, as shown
for a case near Oak Ridge, Tenn. (Fig. 10).
A strength of our approach, compared with traditional
phenological models based on heat degree units, is that our
method does not rely on an arbitrary heat unit threshold that
must be calibrated on a site-by-site basis. It is also worth
noting that the simple phenology scheme examined in this
report says nothing about photoperiod, which may also be
a source of variance and a weakness of the method we are
advocating here (Nizinski and Saugier 1988; Raulier and
Bernier 2000).
With regards to further work, we encourage a wider test-
ing of this scheme with remote sensing data at continental
scales. This exercise would involve predicting the seasonal
course of soil temperature at each pixel in the deciduous
forest biome and find the date when it matches the local
mean annual air temperature. Then one would compare
that product with remote sensing data of the green wave of
spring. The phenology algorithm could also use additional
validation against data from independent phenology net-
works (http://www.uwm.edu/∼mds/markph.html) and new
measurements being produced by the Moderate Resolu-
tion Imaging Spectroradiometer (MODIS) on the TERRA
satellite (Shabanov et al. 2003). Finally, we encourage col-
leagues to install video cameras at all FLUXNET sites and
record the state of the canopy each day.
Acknowledgements We thank the technicians, students and
postdoctoral students who helped collect data at all the field sites
and the funding agencies that supported the numerous team. The
senior author is supported by the NASA FLUXNET project and
DOE Terrestrial Carbon Program (DE-FG0203ER63638).
References
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 (2000) Estimates
of the annual net carbon and water exchange of European
forests: the EUROFLUX methodology. Adv Ecol Res 30:113–
175
Baldocchi DD, Hicks BB, Meyers TP (1988) Measuring biosphere-
atmosphere exchanges of biologically related gases with
micrometeorological methods. Ecology 69:1331–1340
Baldocchi D, Falge E, Gu LH, Olson R, Hollinger D, Running S,
Anthoni P, Bernhofer C, Davis K, Evans R, Fuentes J, Goldstein
A, Katul G, Law B, Lee XH, Malhi Y, Meyers T, Munger W,
Oechel W, Paw UKT, Pilegaard K, Schmid HP, Valentini R,
Verma S, Vesala T, Wilson K, Wofsy S (2001) FLUXNET:
A new tool to study the temporal and spatial variability of
ecosystem-scale carbon dioxide, water vapor, and energy flux
densities. B Am Meteorol Soc 82:2415–2434
Baldocchi DD, Xu L, Kiang N (2004) How plant functional-type,
weather, seasonal drought, and soil physical properties alter
water and energy fluxes of an oak-grass savanna and an annual
grassland. Agric For Meteorol 123:13–39
Billesbach D, Fischer M, Torn M, Berry J (2004) A portable eddy co-
variance system for the measurement of ecosystem-atmosphere
exchange of CO
2
, water vapor, and energy. J Atmos Ocean
Tech 21:639–650
Black T, Chen W, Barr A, Arain M, Chen Z, Nesic Z, Hogg E,
Neumann H, Yang P (2000) Increased carbon sequestration by
a boreal deciduous forest in years with a warm spring. Geophys
Res Lett 27:1271–1274
Blanken PD, Black TA, Neumann HH, den Hartog G, Yang PC,
Nesic Z, Lee X (2001) The seasonal water and energy exchange
above and within a boreal aspen forest. J Hydrol 245:118–
136
Brugger R, Dobbertin M, Krauchi N (2003) Phenological variation
of forest trees. In: Schwartz MD (ed) Phenology: an integrative
environmental science. Kluwer, Dordrecht, pp 255–268
Cayan D, Kammerdiener S, Dettinger M, Caprio J, Peterson D
(2001) Changes in the onset of spring in the western United
States. B Am Meteorol Soc 82:399–415
Chuine I, Kramer K, Hanninen H (2003) Plant development models.
In: Schwartz MD (ed) Phenology: an integrative environmental
science. Kluwer, Dordrecht, pp 217–235
Falge E, Baldocchi D, Olson R, Anthoni P, Aubinet M, Ch B, Burba
G, Ceulemans R, Clement R, Dolman H, Granier A, Gross P,
Grunwald T, Hollinger D, Jenson NO, Katul G, Keronen P,
Kowalski A, Lai CT, Law B, Meyers T, Moncrief J, Moors
EJ, Munger W, Pilegaard K, Rannik U, Rebmann C, Sukyer
A, Tenhunen J, Tu K, Verma S, Vesala T, Wilson K, Wofsy S
(2001) Gap filling strategies for defensible annual sums of net
ecosystem exchange. Agric For Meteorol 107:43–69
Falge E, Baldocchi D, Tenhunen J, Aubinet M, Bakwin P, Berbigier
P, Bernhofer C, Burba G, Clement R, Davis KJ (2002) Season-
ality of ecosystem respiration and gross primary production as
derived from FLUXNET measurements. Agric For Meteorol
113:53–74
Fitzjarrald D, Acevedo O, Moore K (2001) Climatic consequences
of leaf presence in the eastern United States. J Climate 14:598–
614
Freedman J, Fitzjarrald D, Moore K, RK S (2001) Boundary
layer clouds and vegetation-atmosphere feedbacks. J Climate
14:180–197
Goulden ML, Munger JW, Fan SM, Daube BC, Wofsy SC (1996a)
Measurements of carbon sequestration by long-term eddy
covariance: methods and a critical evaluation of accuracy.
Global Change Biol 2:169–182
Goulden ML, Munger JW, Fan S-M, Daube BC, Wofsy SC (1996b)
Exchange of carbon dioxide by a deciduous forest: response to
interannual climate variability. Science 271:1576–1578
Granier A, Pilegaard K, Jensen NO (2002) Similar net ecosystem
exchange of beech stands located in France and Denmark.
Agric For Meteorol 114:75–82
Griffis TJ, Black TA, Morgenstern K, Barr AG, Nesic Z, Drewitt GB,
Gaumont-Guay D, McCaughey JH (2003) Ecophysiological
controls on the carbon balances of three southern boreal forests.
Agric For Meteorol 117:53–71
Gu L, Post WM, Baldocchi DD, Black A, Verma S, Vesala T, Wofsy
S (2003) Phenology of Vegetation Photosynthesis. In: Schwartz
MD (ed) Phenology: an integrative science. Kluwer, Dordrecht,
pp 467–488
Hamming RW (1989) Digital filters, 3rd edn. Dover, Mineola, N.Y.
Hollinger DY, Aber J, Dail B, Davidson EA, Goltz SM, Hughes
H, Leclerc MY, Lee JT, Richardson AD, Rodrigues C, Scott
NA, Achuatavarier D, Walsh J (2004) Spatial and temporal
variability in forest-atmosphere CO
2
exchange. Global Change
Biol 10:1689–1706
Jackson RB, Lechowicz MJ, Li X, Mooney HA (2001) Phenology,
growth, and allocation in global terrestrial productivity. In: Roy
J, Saugier B, Mooney HA (eds) Terrestrial global productivity:
past, present, and future. Academic, San Diego, pp 61–
82
Knohl A, Schulze E-D, Kolle O, Buchmann N (2003) Large carbon
uptake by an unmanaged 250-year-old deciduous forest in
Central Germany. Agric For Meteorol 118:151–167

387
Kramer K, Leinonen I, Loustau D (2000) The importance of
phenology for the evaluation of impact of climate change on
growth of boreal, temperate and Mediterranean ecosystems, an
overview. Int J Biometeorol 44:67–75
Lee XH, Fuentes JD, Staebler RM, Neumann HH (1999) Long-term
observation of the atmospheric exchange of CO
2
with a
temperate deciduous forest in southern Ontario, Canada. J
Geophys Res 104:15975–15984
Massman WJ, Lee X (2002) Eddy covariance flux corrections
and uncertainties in long-term studies of carbon and energy
exchanges. Agric For Meteorol 113:121–144
McNaughton KG, Spriggs TW (1986) A mixed-layer model for
regional evaporation. Bound-Lay Meteorol 34:243–262
Menzel A, Fabian P (1999) Growing season extended in Europe.
Nature 397:659
Moore KE, Fitzjarrald DR, Sakai RK, Goulden ML, Munger JW,
Wofsy SC (1996) Seasonal variation in radiative and turbulent
exchange at a deciduous forest in central Massachusetts. J Appl
Meteorol 35:122–134
Morecroft MD, Stokes VJ, Morison JIL (2003) Seasonal changes
in the photosynthetic capacity of canopy oak (Quercus robur)
leaves: the impact of slow development on annual carbon
uptake. Int J Biometeorol 47:221–226
Myneni RB, Keeling CD, Tucker CJ, Asrar G, Nemani RR (1997)
Increased plant growth in the northern high latitudes from 1981
to 1991. Nature 386:698–702
Nizinski J, Saugier B (1988) A model of leaf budding and develop-
ment for a mature Quercus forest. J Appl Ecol 25:643–652
Penuelas J, Filella I (2001) Phenology: responses to a warming
world. Science 294:793–795
Pilegaard K, Hummelshoj P, Jensen NO, Chen Z (2001) Two years
of continuous CO
2
eddy-flux measurements over a Danish
beech forest. Agric For Meteorol 107:29–41
Raulier F, Bernier PY (2000) Predicting the date of leaf emergence for
sugar maple across its native range. Can J For Res 30:1429–1435
Schaber J, Badeck F-W (2003) Physiology-based phenology model
for tree species in Germany. Int J Biometeorol 47:197–201
Schmid HP (2002) Footprint modeling for vegetation atmosphere
exchange studies: a review and perspective. Agric For Meteorol
113:159–183
Schmid HP, Grimmond CSB, Cropley F, Offerle B, Su H-B (2000)
Measurements of CO
2
and energy fluxes over a mixed hardwood
forest in the mid-western United States. Agric For Meteorol
103:357–374
Schmid HP, Su H-B, Vogel CS, Curtis PS (2003) Ecosystem-
atmosphere exchange of carbon dioxide over a mixed hardwood
forest in northern lower Michigan. J Geophys Res DOI
10.1029/2002JD003011
Schwartz MD, Karl TR (1990) Spring phenology-natures exper-
iment to detect the effect of green-up on surface maximum
temperatures. Mon Weather Rev 118:883–890
Schwartz M, Crawford T (2001) Detecting energy balance
modifications at the onset of spring. Phys Geogr 22:394–409
Shabanov NV, Wang Y, Buermann W, Dong J, Hoffman S, Smith
GR, Tian Y, Knyazikhin Y, Myneni RB (2003) Effect of foliage
spatial heterogeneity in the MODIS LAI and FPAR algorithm
over broadleaf forests. Remote Sens Environ 85:410–423
Spano D, Cesaraccio C, Duce P, Snyder R (1999) Phenological
stages of natural species and their use as climate indicators. Int
J Biometeorol 42:124–133
Suni T, Berninger F, Vesala T, Markkanen T, Hari P, Makela A, Ilves-
niemi H, Hanninen H, Nikinmaa E, Huttula T, Laurila T, Aurela
M, Grelle A, Lindroth A, Arneth A, Shibistova O, Lloyd J (2003)
Air temperature triggers the recovery of evergreen boreal forest
photosynthesis in spring. Global Change Biol 9:1410–1426
Suyker AE, Verma S (1993) Eddy correlation measurements of CO
2
fluxusingaclosedpathsensor-theoryandfield-testsagainstan
open-path sensor. Bound-Lay Meteorol 64:391–407
Taylor F (1974) Phenodynamics of production in a mesic deciduous
forest. In: Leith H (ed) Phenology and seasonality modeling.
Springer, Berlin Heidelberg New York, pp 237–254
Tucker C, Slayback D, Pinzon J, Los S, Myneni R, Taylor M (2001)
Higher northern latitude normalized difference vegetation index
and growing season trends from 1982 to 1999. Int J Biometeorol
45:184-190
Valentini R, DeAngelis P, Matteucci G, Monaco R, Dore S,
Scarascia-Mugnozza GE (1996) Seasonal net carbon dioxide
exchange of a beech forest with the atmosphere. Global Change
Biol 2:199–207
White MA, Running SW, Thornton PE (1997) A continental phenol-
ogy model for monitoring vegetation responses to interannual
climate variability. Global Biogeochem Cycles 11:217–234
White M, Running S, Thornton P (1999) The impact of growing-
season length variability on carbon assimilation and evapotran-
spiration over 88 years in the eastern US deciduous forest. Int
J Biometeorol 42:139–145
White MA, Brunsell NA, Schwartz MD (2003) Vegetation phenol-
ogy in global change studies. In: Schwartz MD (ed) Phenology:
an integrative environmental science. Kluwer, Dordrect, pp
453–466
Wilson KB, Baldocchi DD (2000) Seasonal and interannual vari-
ability of energy fluxes over a broadleaved temperate deciduous
forest in North America. Agric For Meteorol 100:1–18
Wilson KB, Baldocchi DD (2001) Comparing independent estimates
of carbon dioxide exchange over five years at a deciduous forest
in the southern United States. J Geophys Res 106:34167–34178
Xu L, Baldocchi DD, Tang J (2004) How soil moisture, rain pulses
and growth alter the response of ecosystem respiration to temper-
ature. Global Biogeochem Cyles DOI 10.1029/2004GB002281
Yamamoto S, Murayama S, Saigusa N, Kondo H (1999) Seasonal
and inter-annual variation of CO
2
flux between a temperate
forest and the atmosphere in Japan. Tellus 51B:402–413