Short-term responses of ecosystem carbon fluxes
to experimental soil warming at the Swiss alpine treeline
Frank Hagedorn Æ Melissa Martin Æ Christian Rixen Æ
Silvan Rusch Æ Peter Bebi Æ Alois Zu
¨
rcher Æ Rolf T. W. Siegwolf Æ
Sonja Wipf Æ Christophe Escape Æ Jacques Roy Æ Stephan Ha
¨
ttenschwiler
Received: 17 November 2008 / Accepted: 12 February 2009 / Published online: 1 April 2009
Ó Springer Science+Business Media B.V. 2009
Abstract Climatic warming will probably have
particularly large impacts on carbon fluxes in high
altitude and latitude ecosystems due to their great
stocks of labile soil C and high temperature sensitiv-
ity. At the alpine treeline, we experimentally warmed
undisturbed soils by 4 K for one growing season with
heating cables at the soil surface and measured the
response of net C uptake by plants, of soil respiration,
and of leaching of dissolved organic carbon (DOC).
Soil warming increased soil CO
2
effluxes instanta-
neously and throughout the whole vegetation period
(?45%; ?120gCmy
-1
). In contrast, DOC leach-
ing showed a negligible response of a 5% increase
(NS). Annual C uptake of new shoots was not
significantly affected by elevated soil temperatures,
with a 17, 12, and 14% increase for larch, pine,
and dwarf shrubs, respectively, resulting in an
overall increase in net C uptake by plants of
20–40 g C m
-2
y
-1
. The Q
10
of 3.0 measured for
soil respiration did not change compared to a 3-year
period before the warming treatment started, sug-
gesting little impact of warming-induced lower soil
moisture (-15% relative decrease) or increased soil
C losses. The fraction of recent plant-derived C in
soil respired CO
2
from warmed soils was smaller than
that from control soils (25 vs. 40% of total C
respired), which implies that the warming-induced
increase in soil CO
2
efflux resulted mainly from
mineralization of older SOM rather than from
stimulated root respiration. In summary, one season
of 4 K soil warming, representative of hot years, led
to C losses from the studied alpine treeline ecosystem
by increasing SOM decomposition more than C gains
through plant growth.
Keywords Carbon Climate change
Dissolved organic carbon Soil organic matter
Soil respiration Stable isotopes
Temperature Treeline
Introduction
Feedbacks between terrestrial ecosystems and the
ongoing climatic changes are one of the key uncer-
tainties in predicting global warming (Davidson and
Janssens 2006; Heimann and Reichstein 2008).
Although the effects of rising temperatures on single
F. Hagedorn (&) S. Rusch A. Zu
¨
rcher S. Wipf
Swiss Federal Institute for Forest, Snow and Landscape
Research (WSL), 8093 Birmensdorf, Switzerland
e-mail: hagedorn@wsl.ch
M. Martin C. Rixen P. Bebi
WSL Institute for Snow and Avalanche Research SLF,
Flu
¨
elastrasse 11, 7260 Davos, Switzerland
R. T. W. Siegwolf
Paul-Scherrer Institut, 5232 Villigen-PSI, Switzerland
C. Escape J. Roy S. Ha
¨
ttenschwiler
Centre of Functional Ecology and Evolution, CEFE-
CNRS, 1919 route de Mende, 34293 Montpellier, France
123
Biogeochemistry (2010) 97:7–19
DOI 10.1007/s10533-009-9297-9
processes have been studied intensively, a compre-
hensive understanding of whole ecosystem responses
to global warming still remains elusive (see meta-
analysis by Rustad et al. 2001). Arctic and alpine
regions are likely to be particularly affected by
climate warming because observed temperature
increases in these areas are higher than anywhere
else (IPCC 2007; Rebetez and Reinhard 2008).
Moreover, high latitude/altitude ecosystems might
be more sensitive than other types of ecosystems
because plant growth is often limited by low
temperature (Ko
¨
rner 1998), and soil respiration is
more sensitive to warming at lower temperatures
(Kirschbaum 1995). In addition, high latitude/altitude
ecosystems store the greatest fraction of their carbon
stocks in soils (IPCC 2007), and compared to
temperate ecosystems, cold soils were found to
comprise more labile soil organic matter (SOM)
because decomposition and humification proceses are
slow (Jenny 1926; Sjo
¨
gersten et al. 2003).
Plant growth responses to increased temperature
have been studied in situ in both arctic (see review by
Dormann and Woodin 2002) and alpine (Kudo and
Suzuki 2003) ecosystems. Results indicate delayed
growth stimulation by 2 or more years after initiation
of warming in these ecosystems dominated by slow-
growing plants of determinate growth (Shaver et al.
1986; Parsons et al. 1994). The magnitude of growth
responses and thus of net C uptake rates, however,
have been found to be highly ecosystem and plant
species specific (e.g., Hartley et al. 1999). As the
alpine treeline has received little attention in exper-
imental warming studies (but see Danby and Hik
2007), growth responses to warming in these ecosys-
tems remain largely unknown.
Since temperature drives soil respiration rates
(Davidson and Janssens 2006), we might expect
increasing C losses from ecosystems under climatic
warming. The stimulation of soil respiration by
increased temperatures, however, could be counter-
balanced by changing carbon inputs from plants
(Oberbauer et al. 2007), by declining soil moisture
(Saleska et al. 1999), by an ‘acclimatization’ through
physiologically adapting microbial communities, or
by declining resource availability (Luo et al. 2001;
Melillo et al. 2002). Moreover, soil respiration is
essentially driven by recent photosynthates (Ho
¨
gberg
et al. 2001), and thus, the response of soil CO
2
effluxes to rising temperatures also depends on how
plants and their C allocation to below-ground sinks
respond to warming (Schindlbacher et al. 2009).
In contrast to soil respiration, little is known about
the temperature dependency of DOC leaching as the
second major pathway of C loss from ecosystems
(Harrison et al. 2008). Although DOC production is
at least partly microbial driven, concentrations of
DOC in the field are generally only weakly related to
temperature since (1) water fluxes are important co-
drivers (2) inputs of fresh litter vary across seasons
(3) DOC is a net product of DOC generation and
consumption, and (4) DOC is also released by
physico–chemical processes (Michalzik et al. 2001).
Laboratory and soil transplant studies using ‘undis-
turbed’ soil columns suggest that the Q
10
of DOC
production is much lower than that of soil respiration
(Christ and David 1996; Harrison et al. 2008).
However, interpretation of these short-term experi-
ments is limited by the physical disturbance of soils
during sampling.
Although C fluxes between ecosystems and the
atmosphere have been studied intensively, the reasons
for intra-annual variability and the responses to
extreme meteorological conditions, such as the hot
and dry year 2003, are still not well understood (Ciais
et al. 2005; Heimann and Reichstein 2008). A
promising approach to improve our knowledge of the
complex impacts of climatic changes and of climate
variability is the in situ manipulation of climatic
conditions. So far, experimental warming in colder
climates has mainly occurred at high latitudes (Oechel
et al. 1993; Oberbauer et al. 2007) and in boreal forests
(Niinisto
¨
et al. 2004; Bronson et al. 2008). Results
indicate ecosystem-dependent responses in C fluxes
with initial C losses in dry tundra and boreal forests,
but dampened effects under anoxic conditions. The
only study at high altitude was done in a dry alpine
meadow in Colorado, and showed that soil heating had
stronger indirect than direct effects on soil C cycling
by changing plant species composition and inducing
moisture limitations for soil respiration (Saleska et al.
1999).
Our study aimed at estimating how carbon fluxes
in alpine treeline ecosystems with undisturbed soils
and thick organic layers respond to in situ soil
warming. In our experiment, we warmed soils by a
constant 4 K throughout the snow-free period with
heating cables laid out on the soil surface. The soil
warming was conducted within a long-term CO
2
8 Biogeochemistry (2010) 97:7–19
123
enrichment study, which provides a unique
13
C label
for recent plant-derived C in soils. Our objectives
were (1) to estimate the in situ temperature depen-
dency of soil respiration and DOC leaching (2) to
determine if soil warming induces a moisture limi-
tation for soil C fluxes and plant growth; and (3) to
quantify how the 1-year warming treatment, as a
proxy for hot years, affects the C balance of alpine
treelines.
Materials and methods
Study site description
The soil warming study was carried out at Stillberg
(2,180 m a.s.l.) in the Central Alps near Davos,
Switzerland, where a long-term research site was
established in the late 1950s to study climate–growth
relationships (Senn and Scho
¨
nenberger 2001). Long-
term average annual precipitation is 1,050 mm, mean
maximum snow depth is 1.50 m, mean annual
temperature is 1.4°C, and average January and July
temperatures are -5.8°C and 9.4°C, respectively.
The terrain is rather steep, with slopes of 25–30°
facing north–east. Parent material is Paragneiss. Soil
types are sandy Ranker and Podzols (Lithic Cry-
umbrepts and Typic Cryorthods). The organic layers
are Humimors dominated by 5–20 cm thick Oa
horizons (Bednorz et al. 2000; Hagedorn et al.
2008). Soil characteristics are given in Table 1.
Experimental set-up
The combined CO
2
enrichment—soil warming exper-
iment was conducted at the upper end of an
afforestation experiment established in 1975 slightly
above the natural treeline. Three treeline species, Larix
decidua L., Pinus cembra L., and Pinus uncinata
Ramond, were planted across an area of 5 ha spanning
an altitudinal range of 2,080–2,230 m a.s.l. Thirty-two
years later, the trees at the upper end of the plantation
are currently approximately 2 m tall and they form a
sparse open canopy with dense understory vegetation
composed predominantly of ericaceous dwarf shrubs,
such as Vaccinium myrtillus, Vaccinium uliginosum,
and Empetrum hermaphroditum, and of common
herbaceous species, such as Gentiana punctata,
Homogyne alpina, and Melampyrum pratense.
CO
2
enrichment
In 2001, we established experimental CO
2
enrich-
ment within a relatively homogeneous 2,500 m
2
area
(Ha
¨
ttenschwiler et al. 2002). Forty plots, each with an
individual Larix or Pinus tree in the centre and at
least 1 m apart from each other, were organized into
10 groups of four neighbouring plots to facilitate
logistics of CO
2
distribution and regulation. Five of
these ten groups were randomly assigned to an
elevated CO
2
treatment, while the remaining groups
served as controls, resulting in a split-plot design. The
20 elevated CO
2
plots were enriched with CO
2
using
a FACE-set-up, in which pure CO
2
is released
through 24 vertically hanging laser-punched drip
irrigation tubes fixed on steel frames enclosing an
area of 1.1 m
2
(Ha
¨
ttenschwiler et al. 2002). Concen-
trations of CO
2
were recorded every 10 min in all
cardinal points in one plot per group. These concen-
trations were used to adjust the CO
2
addition. The
multiple-year growing season average of CO
2
con-
centration was 566 ± 75 ppm
v
under elevated
CO
2
and 370 ± 3 ppm
v
under ambient CO
2
(see
Ha
¨
ttenschwiler et al. 2002; Handa et al. 2006). In
Table 1 Soil properties of
the treeline ecosystem at
2,200 m a.s.l., Stillberg,
Switzerland
a
Fine earth per volume soil
b
Particle sizes [63 lm
Horizon Depth
(cm)
Soil density
a
(g cm
-3
)
Fraction sand
b
(%)
pH
(CaCl
2
)
Soil organic
C (%)
C/N mass
ratio
Oi 7–6 0.07 ND ND 46.3 46.2
Oe 6–5 0.13 ND ND 45.1 28.1
Oa 5–0 0.16 ND 4.2 40.8 27.2
AE 0–20 0.85 60 3.8 4.7 21.3
Bh 20–45 0.91 61 4.0 3.9 25.1
Bs 45–80 1.10 65 4.2 4.2 29.1
Biogeochemistry (2010) 97:7–19 9
123
2006, we modified the CO
2
enrichment system by
putting the steel frames and the vertical tubes down to
a height of 50 cm thereby adding CO
2
to the dwarf
shrubs and lower parts of trees, In addition, laser
perforated tubes were woven into the tree crowns and
adjusted to match 570 ppm
v
CO
2
using a portable
IRGA. The mean CO
2
concentrations under elevated
CO
2
were 580 ± 60 ppm
v
within the rings and
555 ± 70 ppm
v
in the tree crowns.
Soil warming
The warming experiment was installed in half of the
plots (10 at ambient CO
2
and 10 at elevated CO
2
)
during summer 2006. On the ground surface of each
plot, woven under the aboveground shoots of the
dwarf shrubs, 26 m of 420 W-heating cables (SPS-
S004026, Chromalox-ETIREX, Soissons, France)
were laid out in spirals with a 5 cm distance between
neighbouring cables. Preliminary tests were con-
ducted during 2005 and 2006 to determine the cable
length, density and power required to achieve the
target temperature, while at the same time assuring
that maximum cable surface temperatures would not
exceed 45°C. The 4 K soil warming treatment was
achieved by switching the power supply on and off in
1 min intervals and was applied continuously (day
and night). In 2007, the warming treatment began
directly after snowmelt (23rd May) and was switched
off just before the site was covered in snow for winter
(17th October) to avoid an interaction between soil
temperature and snowmelt or snow cover.
Sampling and field measurements
Air and soil temperatures were measured at -3, -5
and -10 cm in the soil plus 5, 10, 20, 100 and
200 cm in 4–10 plots per warming treatment using
temperature sensors (Hobo Pro v2, Onset Computer
Corporation and ibuttons, Maxim Integrated Products
DS1922L). Air humidity was measured with ibutton
DS1923 sensors. Volumetric soil moisture was
repeatedly measured by Frequency Domain Reflec-
tometry with a Theta sonde ML2x probe (Delta-T,
UK) at fixed locations in all 40 plots. Readings were
converted using a soil-specific equation derived from
a lab calibration with soil from all groups. Gravi-
metric soil moisture was determined in mid July 2007
by taking five soil samples (0–5 cm depth) per plot
with a soil corer (2 cm) and measuring the decrease
in weight after drying soil samples at 105°C.
Soil solution was collected in all plots by installing
two ceramic suction cups (SoilMoisture Equipment
Corp., Santa Barbara, USA) per plot at 3–7 cm depth.
All suction cups were located within the Oa horizon
that dominated the organic layer. At each sampling
event, we collected soil water by evacuating suction
cups with a constant 400 hPa for about 16 h (over-
night). In addition to the suction cups, we installed
zero-tension lysimeters (8 9 8 cm plexiglass plates
with a PE-net) at 5 cm soil depth. The sampling
devices were installed at a fixed depth because space
within the 1.1 m
2
plots was too limited to open pits
for identifying diagnostic horizons. All lysimeters
were connected to 1-l glass bottles buried in the
ground. After collection, soil water samples were
stored in cooling boxes for transport to the institute
for analysis.
CO
2
-efflux from soils and its d
13
C
Soil respiration was measured in the field with
permanently installed PVC collars (10-cm ID and a
height of 5 cm) and a LI-COR 6400-09 soil chamber
connected to a LI-COR-820 portable system for data
collection. One PVC collar per plot (total n = 40)
was pressed to a depth of 2 cm into the organic layer
in between dwarf shrubs in the middle of the plots.
Soil respiration rates were estimated from increases
in CO
2
concentrations with time after scrubbing the
chamber air to ambient CO
2
levels. The pump rate
through the system was kept small with
0.2 ml min
-1
. For
13
CO
2
measurements, we closed
the chambers with PVC-lids at least 4 h after
stopping the CO
2
enrichment to avoid contamination
with
13
C-depleted CO
2
. Twenty to thirty minutes
after closing the collars, we took gas samples by
retrieving 15 ml of air with 20 ml syringe through a
septum and by injecting the air in 12 ml pre-
evacuated glass vials closed with an airtight rubber
septum (volume of 12 ml, Exetainer gas testing vials,
Labco Limited, High Wycombe, UK). The glass vials
were evacuated with a vacuum of 800 hPa immedi-
ately before the sampling.
Plant shoot growth was measured in order to
estimate carbon uptake and how the 1-year warming
treatment affected C balance at the site (see calcu-
lations below). The length of all new shoots of both
10 Biogeochemistry (2010) 97:7–19
123
tree species was measured during early autumn of
2006 and 2007. For the dominant dwarf shrubs
Vaccinium myrtillus, V. uliginosum and Empetrum
hermaphroditum, three measurements of each species
per plot were used to estimate mean 2006 and 2007
shoot growth.
Chemical analysis
All solution samples from the field were passed
through 0.45-lm cellulose-acetate filters (Schleicher
& Schuell, ME25) within 2 days of collection and
then stored at 4°C until analysis. Microbial biomass
was determined with the chloroform-fumigation-
extraction method (Vance et al. 1987) using a
soil:solution ratio of 1:5 for the extraction with
0.5 M K
2
SO
4
. Concentrations of dissolved organic C
in soil extracts and waters were determined with a
TOC/TN analyser (TOC-V Shimadzu Corp. Tokyo,
Japan).
Stable isotopes
The d
13
C values of soil CO
2
were measured with a
gasbench II linked to a mass spectrometer (Delta Plus
XL, Thermo Finnigan, Bremen interfaced with a
Delta-S Finnigan MAT, Bremen, Germany) after
depressurising the vials with a needle (Joos et al.
2008). CO
2
concentrations of gas samples were
calculated from the calibration line with standard
gas samples of known CO
2
concentrations (340 and
5,015 ppm). Results of the C isotope analysis were
expressed in d units (%). The d
13
C values were
referenced to the Pee Dee Belemnite (PDB) standard.
Calculations and statistics
Net plant C uptake
The warming effect on net C uptake by plants was
roughly estimated by multiplying measured growth
effects by estimates of plant biomass production.
Biomass accumulation by trees was obtained by first
estimating growth functions from measured tree
height changes 0, 4, 6, 10, 15, 20, and 30 years after
planting seedlings at the Stillberg site (Senn and
Scho
¨
nenberger 2001; P. Bebi and C. Rixen,
unpublished data). The fitted functions were then
applied to the trees in the soil warming experiment.
Finally, species-specific allometric regressions
between height and total biomass (aboveground
and coarse roots) based on 25 trees excavated at the
study site (Bernoulli and Ko
¨
rner 1999) were used to
calculate the net increase in tree biomass from 2006
to 2007. Biomass productivity of dominant dwarf
shrubs in 2007 was similarly calculated from shoot
length measurements, applying allometric relation-
ships established from understory plots at the
Stillberg site (n = 21) (M. Martin and S. Wipf,
unpublished data). Fine root productivity for all
plants combined, not included in the above calcu-
lations, was determined from ingrowth core data and
13
C measurements of fine roots (Handa et al. 2008).
For both trees and dwarf shrubs, net C uptake by
plants was calculated by multiplying biomass by
0.5.
Temperature dependencies of soil respiration and
DOC concentrations were estimated by the Q
10
function:
R ¼ R
10
Q
ððT 10Þ=10Þ
10
: ð1Þ
In which R is the measured soil respiration or DOC
concentration, R
10
is the simulated soil respiration or
DOC concentration at 10°C, Q
10
is the temperature
sensitivity (over a range of 10°C), and T is the soil
temperature. The R
10
and Q
10
were estimated by
fitting measured data using the Levenberg–Marquard
algorithm (Origin 7.0, OriginLab).
Annual soil CO
2
effluxes were calculated from
applying the fitted Q
10
functions to measured mean
daily soil temperatures for each of the plots.
Stable isotopes
The d
13
C of soil-respired was calculated by a mixing
model with the sampled CO
2
from the soil chambers
being composed of ambient CO
2
and soil respired
CO
2
(Subke et al. 2004). The fraction of soil-respired
‘new’ C (f
new
) derived from the CO
2
addition (being
respired from the rhizosphere or litter) was calculated
by relating the difference in
13
C of respired CO
2
between ambient and elevated CO
2
to the difference
in
13
C in plant leaves (mean of tree needles and dwarf
shrub leaves).
Biogeochemistry (2010) 97:7–19 11
123
