Soil Biology & Biochemistry 35 (2003) 1467–1483

www.elsevier.com/locate/soilbio

Explaining temporal variation in soil CO2 efflux in a mature

spruce forest in Southern Germany
Jens-Arne Subke*, Markus Reichstein, John D. Tenhunen
Department of Plant Ecology, University of Bayreuth, 95440 Bayreuth, Germany
Received 22 June 2002; received in revised form 14 July 2003; accepted 21 July 2003

Abstract
An open dynamic chamber system was used to measure the soil CO2 efflux intensively and continuously throughout a growing season in a
mature spruce forest (Picea abies) in Southern Germany. The resulting data set contained a large amount of temporally highly resolved
information on the variation in soil CO2 efflux together with environmental variables. Based on this background, the dependencies of the soil
CO2 efflux rate on the controlling environmental factors were analysed in-depth. Of the abiotic factors, soil temperature alone explained 72%
of the variation in the efflux rate, and including soil water content (SWC) as an additional variable increased the explained variance to about
83%. Between April and December, average rates ranged from 0.43 to 5.15 mmol CO2 m22 s21 (in November and July, respectively) with
diurnal variations of up to 50% throughout the experiment. The variability in wind speed above the forest floor influenced the CO2 efflux rates
for measuring locations with a litter layer of relatively low bulk density (and hence relatively high proportions of pore spaces). For the
temporal integration of flux rates for time scales of hours to days, however, wind velocities were of no effect, reflecting the fact that wind
forcing acts on the transport, but not the production of CO2 in the soil. The variation in both the magnitude of the basal respiration rate and the
temperature sensitivity throughout the growing season was only moderate (coefficient of variation of 15 and 25%, respectively). Soil water
limitation of the CO2 production in the soil could be best explained by a reduction in the temperature-insensitive basal respiration rate, with
no discernible effect on the temperature sensitivity. Using a soil CO2 efflux model with soil temperature and SWC as driving variables, it was
possible to calculate the annual soil CO2 efflux for four consecutive years for which meteorological data were available. These simulations
indicate an average efflux sum of 560 g C m22 yr21 (SE ¼ 22 g C m22 yr21). An alternative model derived from the same data but using
temperature alone as a driver over-estimated the annual flux sum by about 7% and showed less inter-annual variability. Given a likely shift in
precipitation patterns alongside temperature changes under projected global change scenarios, these results demonstrate the necessity to
include soil moisture in models that calculate the evolution of CO2 from temperate forest soils.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: Carbon cycle; Open dynamic chamber; Picea abies; CO2 efflux; Soil temperature; Soil water content

1. Introduction average temperatures (Parton et al., 1987; Houghton, 1995).

It is the reaction of this largest of all pools to changes in
Recent publications indicate that the terrestrial biosphere climate, which will determine whether ecosystems will
is acting as a C sink (Valentini et al., 2000; Schimel et al., continue to absorb CO2 from the atmosphere, or whether
2001; IPCC, 2001), thus mitigating a potential global increased decomposition of soil organic matter (SOM) will
eventually turn present C sinks into C sources.
warming due to radiative forcing by anthropogenic emis-
The decomposition of SOM is a function of environ-
sions of so-called greenhouse gases (mostly CO2 and CH4,
mental variables (both physical and chemical) and the
but also N2O and halocarbons). Most of the C bound in the
composition of the SOM. The changes in chemical
terrestrial biosphere is found in the soil (IPCC, 2000), with a
composition of organic material with age affect the rate at
general trend of increasing C storage with decreasing annual
which certain fractions of the total SOM pool can be
decomposed (Berg et al., 1996; Coûteaux et al., 1998; Berg,
* Corresponding author. Present address: Dipartimento di Scienze
Ambientali, Seconda Università di Napoli, Via Vivaldi, 43, 81100
2000). While most of freshly added organic material
Caserta, Italy. Tel.: þ39-0823-274656; fax: þ39-0823-274605. decomposes readily after a few years (Bohn et al., 1985),
E-mail address: jens.subke@unina2.it (J.-A. Subke). the remainder becomes part of a more inert, or stable C pool
0038-0717/$ - see front matter q 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0038-0717(03)00241-4
1468 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

within the soil. In the absence of disturbance, SOM

accumulation can continue over centuries or even millennia
(Jenny, 1980; Bohn et al., 1985; Liski, 1997). The
temperature sensitivity of this old fraction of SOM is
critical, since a global trend towards higher annual mean
temperatures would create a positive feedback for global
change. If this persistent fraction of SOM is tolerant of
rising temperatures, as is indicated by some results (Liski
et al., 1999), the present C sink may prevail or even
strengthen (Grace and Rayment, 2000; Thornley and
Cannell, 2001). However, this temperature insensitivity of
old SOM has been disputed by Ågren (1999), who cites
other results indicating that old SOM is indeed sensitive to
temperature. Trumbore (2000), however, cautions that
extrapolations of research findings (on which the models
mentioned here are founded) have to account for the
heterogeneity of the C stock since storage calculations will
otherwise under-estimate short-term storage and over-
estimate long-term storage of C. Changes in the allocation
of assimilated C to roots and root turnover following
climatic change may provide yet another process that affects
C storage below ground (Norby and Jackson, 2000) which is
only poorly understood, and hence not reflected in
ecosystem models.
The picture emerging from these ongoing trends in
ecosystem research shows that reliable predictions about the Fig. 1. The ground vegetation of 50 £ 50-m2 grid-map in Weidenbrunnen 2
behaviour of ecosystems with respect to their C storage with the dominant species for each of the 2.5 £ 2.5 m2 indicated. Numbers
potential are only possible if the below-ground processes are indicate the approximate positions of the soil chamber locations, E is the
better understood. There is a considerable amount of location of the eddy correlation sensor, and dark dots are mature trees.
literature covering the efflux dependencies of CO2 from
a closed cover dominated by the grasses Deschampsia
forest soils, covering a wide range of sampling strategies.
flexuosa and Calamagrostis villosa, in large monospecific
However, most studies are based on sporadic sampling of
patches. Small patches of Urtica dioiica and ‘nurseries’ of
the soil CO2 efflux, or misrepresent seasonal effects by too
dense 2 to 4-yr-old P. abies patches also occur (Fig. 1).
short sampling campaigns. We have made an in-depth
analysis of the factors influencing the CO2 efflux from soil
based on data collected throughout an entire growing season 2.2. Sampling system
in a mature spruce forest located in a small mountain range
in Southern Germany. To capture the temporal variation in CO2 efflux on short
(diurnal) and extended time scales (one growing season), a
sampling system was constructed which was capable of
2. Methods continuously recording the instantaneous efflux rate from
multiple sampling positions. Great care was taken in the
2.1. Site description construction of the sampling chambers to avoid measuring
artefacts, in particular from pressure reductions in the
The ‘Weidenbrunnen 2’ site is a 112-yr-old Norway chamber space which inherently cause problems with open
spruce stand (Picea abies (L.) Karst.) at about 760 m chamber designs. Fig. 2 shows one open dynamic chamber
elevation in the Fichtelgebirge, a mountain range in installed at the site, showing the chamber design with air
northern Bavaria (SE Germany; 508080 N, 118520 E). The intakes of ambient and chamber air. A schematic diagram of
local soil type was classified as cambic podzol over granitic the gas path between each of the five chambers and the gas
bedrock characterised by low pH values (3.3 –3.9; Heindl analyser is shown in Fig. 3, while a more detailed
and Bott, 1995). Soil litter and the organic horizon had an description of the design and instrumental set-up can be
average thickness of 1.6 and 15 cm, respectively, with found in Subke (2002).
roughly equal thickness in the Of and Oh horizons. Average The instantaneous soil CO2 efflux from each of five
tree height was 27 m with a tree density of about 312 chambers was measured sequentially once every hour.
trees ha21 and a leaf area index of 7.2 (E. Falge, personnel Chamber tests have shown that the presence of the chamber
communication). The understorey was characterised by only slightly alters the temperature of the topsoil, and had no
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1469

each of the five lids between the three collars of each

measuring location (generally once every 1 to 3 d), allowed
sufficient time for the environmental conditions within each
collar to be unaffected by the previous presence of a
chamber lid.

2.3. Installation of the soil respiration system in the field

Five sample locations within the stand were chosen for

soil respiration measurements, and three collars were
installed at each of these locations. Collars were inserted
to a depth of between 1 and 2 cm into the soil, with about
50 cm spacing between collars at each location, and
Fig. 2. Open dynamic soil chamber (design adapted from Rayment and remained in place for the duration of the experiment. To
Jarvis, 1997). Air is drawn from the chamber through the lateral canal, and capture potential differences in the soil CO2 efflux due to the
ambient air passively follows the pressure gradient through the centrally
ground cover, two sampling locations were selected for each
mounted inlet tube. The intake of ambient air for the differential
concentration measurements can be seen next to the inlet tube. of the most dominant ground vegetation types, and a third
location was located in a ‘nursery’ of 2-yr-old P. abies
discernible effect on the soil moisture (Subke, 2002). To plants (Fig. 1). The chambers were usually positioned as far
avoid any artefact due to prolonged presence of the as possible from mature trees (between 3 and 4 m) with the
chamber, however, only readings of the instantaneous exception of chamber 3, where the collars were within 0.5
CO2 efflux rate obtained within 24 h of positioning a and 1 m of a mature tree. Within each collar, all above-
chamber lid on a collar were considered valid data. Moving ground parts of the vegetation had been removed before

Fig. 3. Schematic diagram of the gas path between the five chambers and the infrared gas analyser (IRGA). A multiplexer controlled the switching of solenoid
valves within the pumping unit (shaded area), thus directing the flow from the five chambers and a calibration line (top) sequentially to the IRGA. A data logger
(not shown) recorded all relevant readings at 1 min intervals.
1470 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

measurements were made and any new growth removed can be simulated. Using meteorological data that was
during the season. Collars were installed 48 h before available for the entire year and measurements of some soil
measurements commenced to avoid artefacts due to the qualities at the site, it was possible to simulate the organic
installation. Throughout this text, each of the 15 soil collars SWC over 7 weeks for which good measurements (from five
will be referred to according to the measuring location and probes) were available. Since the data from the measuring
the number of the collar within this location. For example, probes was both discontinuous and inconsistent, only SWC
the collar description ‘5_2’ refers to the second collar of values calculated using this model are used in the analysis.
sampling location 5 (Fig. 1).
2.6. Data analysis
2.4. Soil respiration measurements
2.6.1. CO2 efflux
Readings of the differential CO2 concentration and the The soil CO2 efflux was analysed with respect to its
flow rate were obtained at 1 min intervals. The soil dependence on temperature ðTÞ; SWC, and wind forcing ðuÞ :
respiration rate could be calculated from the respective
variables according to FðsoilÞ ¼ fðTÞ £ fðSWCÞ þ fðuÞ : ð2Þ
Cdif fk While T and SWC both act on the production of CO2 by
Fsoil ¼ ; ð1Þ
A autotrophic or heterotrophic respiration, u affects the
where Cdif is the differential CO2 concentration between physical transport of CO2 from the soil to the atmosphere.
chamber air and ambient (in mmol mol21), f is the flow rate The amount emitted due to pressure induced pumping relates
(in l min21), A is the chamber base area (315 cm2), and k is a to a quantity of CO2 stored in the soil pores, which can be
constant factor combining the conversions of CO2 concen- visualised as a buffer between the soil and the atmosphere
trations from (mmol mol21) to (mmol m23) and for f from that is depleted under turbulent and replenished under calm
(l min21) to (m3 s21). conditions, and the long-term average of this flux value is
The CO2 differential signal was checked for stability to therefore zero. Accordingly, the dependence on wind
ensure steady-state conditions within the chamber, and the induced pressure fluctuations is only used for instantaneous
average value for the soil CO2 efflux of the last 3 min of a CO2 efflux measurements, while efflux averages are tested
10 min measuring interval were recorded. for T and SWC dependence only. Different possible
relationships between the environmental variables and the
2.5. Correlating measurements CO2 efflux were tested for each of these functions (Sections
2.6.2– 2.6.4). These are developed from existing equations,
The production of CO2 within the soil is basically a and the model parameters are estimated to fit the function to
biochemical process and thus responds strongly to vari- the measurement data using multivariate, non-linear
ations in temperature. This dependence may change with the regression. All regression fits were performed using the
age of the organic matter (roughly corresponding to software PV-Wave version 6.21.
increasing depth within the soil), and also with the
availability of water for the relevant biochemical reactions. 2.6.2. Temperature functions
Accordingly, temperature probes and soil moisture sensors The Arrhenius type function (Eq. (3)) as described by
were installed near the soil collars. At each of the five Lloyd and Taylor (1994) is widely accepted as a realistic
locations, a temperature profile was sampled at 5, 10 and description of the fundamental temperature dependence of
30 cm depths once every 30 min. The soil water content soil respiration. The Q10 function (Eq. (4)) was used as an
(SWC) was recorded for the upper 10 cm of the organic alternative exponential temperature relationship since it is
layer (Theta Probes, Delta-T devices Ltd, Cambridge, UK). also widely used. It is noted, however, that the concept of a
Since the variation in wind speed has been hypothesised to strict Q10 relationship for soil respiration processes has been
affect the transport of CO2 from the soil (Kimball and criticised on the basis of the variation of this factor itself
Lemon, 1971), wind velocity data, which was available with temperature and SWC (Howard and Howard, 1993;
from an eddy correlation sensor operated at the same site Lloyd and Taylor, 1994; Kutsch, 1996). The basic
(Fig. 1), was also considered for analysis. difference between these two functions is that in Eq. (3),
Measurements of SWC were not consistent throughout the temperature sensitivity decreases with increasing
the year owing to the varying number of soil moisture temperature, while in Eq. (4), the relationship is constant
probes used. To adjust an apparent bias due to the throughout the temperature range. In order to assess the
misrepresentation of the stand SWC by a too small number deviations of these relatively complex relationships from a
of probes, an existing stand process model was employed to simple linear one, a linear dependence of soil CO2 efflux on
simulate the water content of the organic layer. This process the soil temperature was also included (Eq. (5))
based model (PROXEL, Reichstein, 2001) includes a multi-
layer soil compartment, in which the movement of water fðTÞ ¼ Rref eE0 ðð1=56:02Þ2ð1=Tþ46:02ÞÞ ; ð3Þ
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1471

ðT2Tref =10Þ; where su is the standard deviation of the horizontal wind

fðTÞ ¼ Rref Q10 ð4Þ
velocity (recorded at 20 Hz and aggregated into 1 s
fðTÞ ¼ Rref þ mðT 2 Tref Þ; ð5Þ averages; standard deviations were formed for 10 min
intervals for these 1 s averages) and c is a linear parameter
where T is the soil temperature (in 8C). All other model
fitted during the regression routine.
parameters were fitted by the multivariate non-linear
regression. Rref is the soil respiration rate at the reference
2.7. Calculating the annual soil C efflux
temperature Tref (here set to 10 8C, the median temperature
for Weidenbrunnen 2
of the data set), E0 is an exponential parameter affecting the
temperature sensitivity (which is related to the activation
The results of the data analysis described so far allow the
energy in the Arrhenius equation, see Lloyd and Taylor,
calculation of soil CO2 efflux rates for given environmental
1994), and m is the slope of the linear regression.
conditions. This allows the simulation of the soil CO2 efflux
for periods over which relevant input data are available but
2.6.3. Soil water content functions
for which no measurements took place. It is thus possible to
Soil CO2 efflux data were not tested against the SWC
calculate efflux sums that can be compared either to the total
alone, since this environmental variable shows less diurnal
stand flux of C, or for different periods to assess the
or seasonal variation than temperature, so that effects due to
temporal variability due to climatic variations.
SWC would be masked by the influence of temperature.
The input for such a soil model created from the regression
Instead, SWC sensitivity of soil CO2 efflux was tested
results is a continuous data set with all relevant environ-
simultaneously to the temperature dependence by multiply-
mental variables. The interest in this context is in long-term
ing each of the T functions with one of the SWC functions
flux sums, so that the function influencing the transport of
described in this section.
CO2 ðfðuÞ Þ is of no relevance, since its average contribution to
Eq. (6) is a modified version of the model proposed by
the efflux is zero (see Section 4 for more detail). For the
Bunnell et al. (1977), while Eq. (7) is an alternative
remaining variables (T and SWC), a long-term continuous
formulation derived from a Gompertz function after
data set could be constructed from a number of sources.
Janssens et al. (2002):
Soil temperature measurements form part of a measuring
SWC routine at an intensive research site immediately adjacent to
fðSWCÞ ¼ ; ð6Þ
SWC1=2 þ SWC the Weidenbrunnen 2 plot (data supplied by the Bayreuth
ða2b£SWCÞ Institute for Terrestrial Ecosystem Research, BITÖK), and
fðSWCÞ ¼ e2e ; ð7Þ data were available from 1 April, 1997 to the 1 April, 2001.
where SWC is the volumetric soil water content (m3 Following the results of the regression analysis, only the
water m23 soil), SWC1=2 is the soil water content, at which temperature at 5 cm depth was required, which was
half the maximum respiration (i.e. under conditions without aggregated into hourly averages (from 10-min interval
water stress at a given temperature) occurs, and a and b are data). Short gaps in the data set (, 2 h) were filled by linear
both data set specific constants. Both the Bunnell and interpolation. For longer gaps (up to 36 h), the average was
Gompertz model for soil water limitation included in their formed from the temperature readings taken at the same
original form a constraint for limitation due to high SWCs. time of day on the preceding and following days and a
However, no meaningful parameters could be found for correction of a linear trend of the averaged values performed
equations including soil CO2 efflux limitation due to high to maintain continuity. Following the described steps, most
SWCs, presumably owing to sufficient drainage of the upper data gaps could be filled, but one longer period of missing
soil layers at the Weidenbrunnen 2 site. Accordingly, Eqs. data remained (8 August –12 September, 2000). Preliminary
(6) and (7) only contain the functions describing limitations calculations showed that annual sums calculated from
due to dry soil conditions. temperature data, aggregated into hourly averages, differed
from sums calculated from data aggregated into daily
2.6.4. Wind forcing function averages by less than 0.2%. In order to simplify the gap
Vertical movement of air may be induced by pressure filling of the remaining gap, data were aggregated into daily
differences that occur at the soil surface. This ‘pumping’ averages. It was possible to create a simple model based on
motion may represent a significant means of physical (1) air temperature (also measured at an adjacent site and
transport of CO2 from the soil, and the open chamber design data supplied by BITÖK), (2) the temperature at 5 cm depth
(in contrast to closed chamber models) allows this natural averaged over the preceding 10 d, and (3) an approximation
process to occur within the chamber space. The variation in of the lower soil temperature. For the same periods in 1997–
wind speed ðsuÞ had been found to be an appropriate 1999, this model produced calculated temperatures in good
surrogate for pressure fluctuations (Subke, 2002). The agreement with those measured (linear fit for measured vs.
function for wind forcing took the simple linear form of modelled temperatures: r 2 ¼ 0:91; s:d: ¼ 0:222; n ¼ 36;
P , 0:0001), so that a realistic modelling of the temperature
fðuÞ ¼ csu; ð8Þ at 5 cm depth for the period in 2000 can be assumed.
1472 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

Comparison of the temperature measured in Weiden- the derivation of SWC for 1999 (Reichstein, 2001), the
brunnen 2 at 5 cm depth during this study, and the SWC could be modelled from 1 April, 1997 to 31 March,
contemporary data from the Weidenbrunnen flux-tower 2001. Data gaps (due to missing precipitation data) never
site showed a good correlation (r 2 of 0.99 for daily averages exceeded more than 7 d, and were filled using the seasonal
for 145 d; data range: , 1 to . 16 8C). However, there was a average of previous and following years.
consistent and significant ðP , 0:001Þ deviation from the
1:1 line for the two temperature averages, possibly owing to
a difference in stand structure or a slight difference in the 3. Results
burial depth of the respective temperature probes. For the
modelling of soil CO2 efflux from the Weidenbrunnen 2 site, 3.1. Daily and seasonal patterns of soil CO2 efflux
the temperature readings from Weidenbrunnen were
corrected according to y ¼ 0:878x þ 1:31: 3.1.1. Seasonal and daily flux pattern
Long-term data for the SWC of the organic layer were Soil respiration was measured continuously from 28
not available. However, using the model described for April to 3 December, 1999 with two long gaps in July and

Fig. 4. Seasonal course of measured soil temperatureðn ¼ 3432Þ; modelled SWC of the organic layer, and measured soil CO2 efflux ðn ¼ 2429Þ in 1999. All
data are hourly averages.
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1473

September/October (2 and 4 weeks, respectively) due to grouped averages ¼ 0.050, X 2 of collar-averages ¼ 0.072,
instrument failure. The daily average soil CO2 efflux rate 0.042 and 0.038, with n ¼ 65; 23, 21 and 21, respectively),
(i.e. flux rates averaged for all collars measured in 1 d) so that all flux averages could be treated as a true spatial
ranged from 0.58 mmol m22 s21 on 16 November to average of the stand.
3.72 mmol m22 s21 on 26 July. The instantaneous CO2
efflux rate could be as low as 0.43 mmol m22 s21 (17 3.1.2. Temperature and soil water content dependence
November, collar 4_2) and as high as 5.15 mmol m22 s21 Parameter fits were performed for the temperature
(on 19 July, collar 3_1). The range of the daily average CO2 dependence functions alone, as well as for all combinations
efflux rate varied between 0.24 and 2.59 mmol m22 s21 (on of temperature and SWC functions
15 November and 31 May, respectively), with greater
FðsoilÞ ¼ fðTÞ fðSWCÞ ; ð9Þ
variations occurring in summer when the efflux rate is
greatest (Fig. 4). SWC limitation only occurred during short where fðTÞ is one of Eqs. (3) –(5), and fðSWCÞ takes the value 1
periods in summer, when the SWC dropped below or is one of Eq. (6) or (7). Best fits of hourly flux-average
0.2 m3 m23 in the organic layer, and the soil CO2 efflux data to Eqs. (3) –(5) were achieved for the soil temperature
rate was reduced despite high soil temperatures. at 5 cm depth. All three temperature response functions
Typical daily courses of the CO2 efflux rates are plotted fitted the data well, and multiplying each of the temperature
in Fig. 5, showing a marked increase from about 2 h after functions by one of the SWC limitation functions improved
sunrise to about mid-afternoon and a slow decline through- the fit to the data (Table 1).
out the night until the following morning. Soil CO2 efflux Out of all nine regression models indicated in Table 1,
rates usually peaked well after midday but before the the combination of the Lloyd and Taylor (1994) temperature
maximum temperature at 5 cm depth was recorded. Daily model and the SWC model after Bunnell et al. (1977) (Eqs.
time courses of CO2 efflux were generally continuous but (3) and (6)) were chosen as for further data analysis, due to
could show considerable variation between hourly readings the slightly better value for the adjusted coefficient of
(as, for example, on the afternoon of 20 August in Fig. 5). correlation (adj. r 2).
Flux measurements from all five collars collected within
1 h were aggregated, thus yielding a temporally and 3.1.3. Interactions between the temperature and moisture
spatially averaged soil CO2 efflux estimate. Since all dependence
chambers were moved between collars of a location The results presented in Table 1 clearly show the
simultaneously, the resulting averages represented three dependence of soil CO2 efflux on the SWC. In order to
different spatial averages (for collars 1, 2, and 3 of all test whether the temperature dependence of the CO2 efflux
locations, respectively). Temperature regressions (using in turn depends on the SWC, the hourly efflux averages were
daily averages of CO2 flux and temperature and Eq. (3)) for divided into SWC classes (between 0.20 and 0.32 m3 m23,
data obtained under conditions of no soil water limitation SWC classes had a width of 0.01 m3 m23, above and below
(see below) showed that the three spatial averages did not this range, classes contained a wider range of values to
differ significantly from a regression using all data (X 2 of allow sufficient numbers of data points for regression

Fig. 5. Soil CO2 efflux rate (symbols and left axis) and soil temperature profile (lines and right axis) measured over 4 d in August 1999. Data are instantaneous
flux readings from all three collars at location 4 (i.e. collars 4_1, 4_2, and 4_3).
1474 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

Table 1 plotted, with SWC1/2 ¼ 0.172 m3 m23 multiplied by a value

Parameters and coefficients of determination for all combinations of of Rref ¼ 3:57 (both values are taken from the temperature
temperature and soil water content dependencies
and SWC regression in Table 1).
Moisture limitation Temperature function
function
Lloyd and Taylor Q10 Linear 3.1.4. CO2 soil efflux due to pressure pumping
The hypothesised influence of vertical air pumping on the
None Rref 2.05 ^ 0.01 2.03 ^ 0.01 2.10 ^ 0.01 CO2 efflux from the soil acts only on the gas transport from
T-par 304 ^ 8 2.61 ^ 0.07 0.199 ^ 0.004
the uppermost soil layer and not on the production of CO2.
adj. r 2 0.72 0.70 0.74
Since the pressure fluctuations therefore only affect the
Gompertz Rref 2.65 ^ 0.13 2.58 ^ 0.11 2.60 ^ 0.12 instantaneous efflux situation, it was not deemed useful to
T-par 403 ^ 8 3.64 ^ 0.10 0.286 ^ 0.015
a 0.364 ^ 0.079 0.452 ^ 0.086 0.167 ^ 0.107
average fluxes and environmental variables over any length
b 8.38 ^ 1.13 8.09 ^ 1.14 7.90 ^ 1.45 of time. Best-fit regressions were applied to the instan-
adj. r 2 0.83 0.82 0.82 taneous soil CO2 efflux data for each collar separately. Only
Bunnell Rref 3.57 ^ 0.13 3.66 ^ 0.15 3.22 ^ 0.10
valid flux data for periods when wind-data was available
T-par 403 ^ 8 3.65 ^ 0.10 0.355 ^ 0.014 were included. The regression model took the form of Eq. (2),
SWC1/2 0.172 ^ 0.015 0.188 ^ 0.017 0.116 ^ 0.010 with Eq. (3) for fðTÞ ; Eq. (6) for fðSWCÞ and Eq. (8) for fðuÞ : In
adj. r 2 0.83 0.82 0.82 order to compare the effect of wind forcing on the efflux data,
‘T-par’ refers to the respective parameters of the temperature sensitive a second set of regressions were performed with fðuÞ set to
parts of Eqs. (3)-(5), all other parameters are the same as for Eqs. (3) –(7). zero. Variables used in the regression are the soil temperature
The coefficient of determination has been adjusted for the respective at 5 cm depth and SWC of the organic layer for T and SWC,
numbers of parameters; n ¼ 822 for all regressions
respectively, and the standard deviation of the horizontal
wind-speed 10 min previous to an efflux reading for su:
analysis; Fig. 6) and temperature regressions (Eq. (3)) were As can be seen from Table 2, only about half of the
performed for each of these classes. The model parameter chambers show a slight improvement in the value of the
Rref represents the soil CO2 efflux rate at 10 8C, i.e. it gives a adjusted r 2 if fðuÞ is included (‘adj. r 2 ðuÞ’ compared to
measure of the magnitude of the efflux and is temperature ‘adj: r 2 ðu ¼ 0Þ’). Strikingly, all collars located in D.
insensitive, whereas the parameter E0 indicates the flexuosa patches (locations 3 and 4) show a better fit for
sensitivity of the efflux to temperature changes. By plotting regressions including the wind function, while most of the
both parameters against the different SWC classes, a clear remaining collars show little or no improvement in the
trend in the magnitude of Rref is revealed, while the coefficient of determination. The value of c was signifi-
temperature sensitive parameter shows no trend with cantly different from zero for collars 3 and 4, suggesting that
changing SWC (Fig. 6). The curve according to the SWC here an increased variability in wind speed is indeed
limitation model by Bunnell et al. (1977) (Eq. (6)) is also positively correlated with the soil CO2 efflux rate.

Fig. 6. Variations in the regression parameters Rref (triangles and left axis) for T05 ¼ 10 8C and E0 (circles and right axis) for different SWC classes. The line is
the effect of SWC limitation found for simultaneous regression of T and SWC (hatched grey line: Rref ¼ 3:57 £ ðSWC=SWC þ 0:172Þ;; compare Table 1).
Error bars are standard errors of the parameter estimation.
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1475

3.2. Spatial variation

0.102
0.028

0.012
0.097
2.82

0.10

0.82
0.82
5_3

353

226
14
Direct comparisons between soil respiration rates at
20.332 different collars were not possible, since only five collars
0.068

0.016
0.105
were measured simultaneously. In order to test for
2.32

0.10

0.77
0.76
5_2

348

251
17 characteristic differences between the sampling locations,
the instantaneous flux rate of CO2 from each collar at

In the nomenclature of collars, the first number indicates the chamber location (1_5 as in Fig. 1), and the second number indicates the collar within each chamber location (1_3).
between 9.5 and 10.5 8C from the entire growing season
0.253
0.064

0.035
0.091
3.17

0.20

0.85
0.85
5_1

413

250
15

(excluding periods when SWC became limiting) were

compared. A total of 523 measurements had been
conducted within this temperature range (between 80
0.081
0.536

0.014
0.122

0.645
2.19

0.09

0.67
4_3

349

272

and 121 for each of the locations), and no effect of

22

hysteresis (i.e. whether temperature previous to the

measurements had been either above or below the
0.141
0.616

0.024
0.092

temperature range) was detectable.

2.23

0.13

0.81
0.78
4_2

404

295
19

CO2 efflux rates of locations 1 and 3 were found to differ

significantly from all other locations, and from each other
(1-way ANOVA and Tukey test, F4500 ¼ 92:1; P , 0:05),
0.192
0.869

0.039
0.140
2.63

0.21

0.72
0.67
4_1

380

272
22

while locations 2, 4 and 5 show virtually identical rates. No

significant difference due to the ground vegetation type (also
indicated in Fig. 7) was apparent for data pooled for either
The last row shows values of the adjusted r 2 for regressions with f ðuÞ set to zero (parameters of these regressions not shown).
0.039
0.296

0.008
0.133
3.18

0.09

0.41
0.39
3_3

198

241

collars or locations.
17

3.3. Annual soil CO2 efflux for Weidenbrunnen 2

0.105
0.416

0.018
0.136
2.77

0.12

0.58
0.56
3_2

305

234
19

Fig. 8 shows the course of the soil temperature, SWC

and the soil CO2 efflux rate calculated using Eqs. (3) and
0.228
0.671

0.038
0.181

0.525
4.83

0.33

0.55

(6) with the parameters stated in Table 1. The regression

3_1

271

204
23

of all soil CO2 efflux data obtained for SWC . 0.2

against Eq. (3) yielded the parameters Rref ¼ 2:09 ^ 0:01
and E0 ¼ 354 ^ 8 (n ¼ 702; adj. r 2 ¼ 0.82). A compari-
20.056
0.034

0.008
0.101
2.21

0.07

0.82
0.82
2_3

391

210

son of the efflux modelled from the two regressions

19

illustrates the effect of SWC limitation in the summer

months (Fig. 9).
0.059
0.054

0.013
0.119

Based on the temperature and SWC regression results,

2.45

0.10

0.62
0.62
2_2

383

223
23

the total soil CO2 efflux over the 4-yr-period could be

estimated. Given the extent of the available soil
temperature data, annual totals were calculated from 1
0.063
0.014

0.014
0.117
2.23

0.10

0.74
0.74
2_1

362

228
19
Regression parameters of each of the soil collars for FðT;SWC;uÞ

April of each year to 31 March of the following year,

and annual sums are given in Table 3.
According to the rainfall data supplied by BITÖK,
20.047
0.103

0.009
0.066

1997 was an extremely dry year. Comparison with

2.86

0.07

0.89
0.89
1_3

383

300
10

rainfall sums recorded over 14-d intervals showed that

while the annual precipitation sum was lower than in all
previous and following years, technical problems with
0.041
0.266

0.011
0.107
2.45

0.09

0.78
0.77
1_2

362

216

the rainfall instruments contributed to considerable under-

15

estimations of the annual total. The discrepancies

between the hourly data (on which the SWC model is
Collara

0.094
0.331

0.016
0.116
2.60

0.11

0.58
0.56

based) and the alternative fortnightly measurements

1_1

310

222
20

occurred in winter and between late June and late July.

During the 4 weeks between 26 June and 22 July, the
adj. r 2 ðu ¼ 0Þ

SWC model under-estimated actual SWC by an equiv-

SE SWC1=2

alent of about 38 mm rainfall. The prolonged water stress

adj. r 2 ðuÞ
SWC1/2
Table 2

SE Rref
SE E0

indicated for the second half of 1997 (Fig. 8), however,

SE c
Rref

a
E0

is consistent with the 14-d interval rainfall data.

n
c
1476 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

Fig. 7. Average soil respiration rates at around 10 8C of the five sampling locations. Error bars indicate 95% confidence intervals, means with the same letters
are not significantly different ðP ¼ 0:05Þ:

Fig. 8. Four year time course of (a) soil temperature (black line and left axis) and SWC (grey line and right axis), and (b) the modelled soil CO2 efflux rate.
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1477

Fig. 9. Modelled soil CO2 efflux for 1999. Grey line: using temperature and SWC data with regression parameters as stated in Table 1 (for Lloyd and Taylor-
type temperature and the Bunnell type SWC dependence); black line: using temperature data only and regression results as stated in the text.

4. Discussion increases by 10 8C. The wide use of this purely empirical

relationship in biological systems has been criticised by
4.1. Fundamental temperature dependence Lloyd and Taylor (1994) who point out that Q10 functions
of soil CO2 efflux systematically overestimate fluxes at high temperatures. In
the modified Arrhenius relationship for temperature and soil
A positive relationship between soil temperature and soil respiration proposed by Lloyd and Taylor, which was also
CO2 efflux rate is well established (Lundegårdh, 1926; used in this study, the activation energy decreases with
Singh and Gupta, 1977; Raich and Schlesinger, 1992) and is increasing soil temperature. This soil respiration model (Eq.
clearly reflected in the seasonal course of both variables in (3)) has since been supported by other researchers (Fang and
Fig. 4. The causal link of this positive correlation by the Moncrieff, 1999; Rayment and Jarvis, 2000), but the Q10
increased biological activity of both the autotrophic (roots, concept is still widely used (Boone et al., 1998; Davidson
e.g. Bouma et al., 1997) and the heterotrophic organisms et al., 1998; Morén and Lindroth, 2000), in some cases
(microbial communities and soil dwelling animals) in the alongside the Lloyd and Taylor-model (Fang et al., 1998;
soil and the increased diffusivity of CO2 under higher Buchmann, 2000). The regression results of both daily and
temperatures is also undisputed. The exact nature of this hourly averages of soil CO2 efflux obtained in this study
relationship, however, is less clear and receives consider- (Table 1) support the Lloyd and Taylor model as the more
able attention (see Janssens et al., 2002 for a review). Of all realistic concept for soil CO2 efflux measurements.
existing empirical relationships, only three were tested on The apparent good fit of a linear regression model
the data collected during this study These are (1) an (Fig. 10) presumably stems from larger scatter at high soil
Arrhenius type temperature response, (2) the Q10 concept, temperatures, where all functions describe the data
and (3) a simple linear regression. equally well, and the less represented measurements at
The Q10 relationship expresses the factor by which a low soil temperatures, where the linear model under-
biochemical reaction increases when the temperature estimates the measured efflux. Depending on soil types

Table 3
Total annual C loss through soil CO2 efflux

Annual mean T Precipitation T and SWC model T only model ‘T’-‘T&SWC’ Over-estimate
(8C) (mm) (g C m22 y21) (g C m22 y21) (g C m22 y21) (%)

1997 5.9 572a 497 588 91 15.5

1998 6.0 1300 566 581 15 2.6
1999 6.4 1170 592 602 10 1.5
2000 6.8 945 586 618 32 5.3
Mean ^ st. dev. 6.0b 1019b 560 ^ 43 597 ^ 16 37 ^ 37 6.2 ^ 6.3
CV 7.8 2.7

The difference between the two regression results is also expressed as the proportion of the efflux calculated with the temperature regression.
a
Precipitation data for 1997 likely to under-estimate actual precipitation sums; see text for detail.
b
Mean of the annual temperature and annual precipitation are calculated for the years 1993–2000 (data supplied by BITÖK).
1478 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

Fig. 10. Temperature dependence of soil CO2 efflux averaged over 24 h following the closing of the chamber lid. Open symbols are values obtained at average
SWCs of below 0.2 m3 m23. Regression curves are best fits for all data excluding conditions with SWC limitation; solid line: Lloyd and Taylor (1994), hatched
line: Q10, dotted line:linear function.

and different temperatures investigated, linear regressions increasing depth (Fig. 5). That the most shallow temperature
have been favoured to describe soil CO2 efflux (Anderson, sensor shows greatest diurnal variation is therefore obvious
1973; Koizumi et al., 1999). As with the Q10 model, and clearly documented in Fig. 5. Since the organic layer
however, it lacks a physiological basis and bears obvious with the greatest pool of easily decomposed C is hence
limitations for extrapolations within physiological scales, exposed to the environmental variable with greatest
which is certainly desirable for modelling purposes. fluctuation, the correlation between the CO2 efflux rate
According to the parameters in Table 1, for example, and the most shallow soil temperature can be expected.
soil respiration would become negative for temperatures For the same forest plot, Buchmann (2000) found that
below about 2 1 8C. removing the litter and organic layer of the forest floor (i.e.
The presented soil CO2 efflux data therefore show that the top 13 cm of soil) led to no significant reduction in the
the Lloyd and Taylor (1994) function, based on a soil CO2 efflux rate, suggesting that the main source of
temperature insensitive efflux rate at a set temperature CO2 is located below these strata. Yet, the correlation
ðRref Þ and a parameter representing the activation energy between temperature and soil CO2 efflux in the same study
ðE0 Þ that varies according to the soil temperature, is the was less at 15 cm depth than it was at 10 or 5 cm depth
fundamentally soundest description of the relationship (r 2 ¼ 0:70; 0.80 and 0.80, respectively). The only mean-
between soil temperature and soil CO2 efflux. Therefore, ingful interpretation of these apparently contradictory results
only the Lloyd and Taylor function was used for the further in the same study is that the deeper soil layers contribute the
analysis of the interaction between soil temperature and the bulk of CO2 without any temperature sensitivity, and the
soil CO2 efflux. more shallow layers contribute a small but changeable and
temperature dependent portion of the CO2 flux. However,
4.2. Temperature profile within the soil looking at short-term variability of soil CO2 efflux found in
our study, peak efflux rates regularly exceed twice the night-
Of the factors influencing the soil respiration rate, soil time values of within 24 h of measuring (data not shown, but
temperature shows the greatest short-term variation. The see Fig. 5). This pattern indicates that a large fraction of the
surface efflux of CO2 is the result of the heterotrophic and CO2 flux originates in those soil strata that are affected by the
autotrophic activity from the entire soil profile, and depends diurnal temperature cycle (i.e. the litter and organic layer).
on the substrate quality and the environmental conditions at The absence of a reduction in efflux after removal of these
all depths within this profile. The temperature within a given layers found by Buchmann (2000) is therefore likely to be
soil layer depends on the temperature in adjacent soil layers, attributed to measuring artefacts due to disturbance of the
air temperature and heating due to solar radiation. Greatest soil environment, and not a major contribution of the efflux
temperature fluctuations appear at the soil surface, due to from the mineral soil.
seasonal and diurnal fluctuations in both radiation and air Mariko et al. (2000), Davidson et al. (2000) and others
temperature. By means of thermal diffusion these fluctu- have pointed out the limitations of using a simple
ations propagate to deeper soil layers, resulting in a more temperature function from one soil depth only to describe
dampened and time-lagged temperature signal with the process operating throughout the profile and is
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1479

influenced by the heterogeneity of substrate and environ- exponential regression gives a good fit for data of the entire
mental factors. Splitting the temperature response function growing season, the actual correlation for measurements
into several components to represent the flux contribution taken within 24 h show a different dependence on soil
from different soil layers is one step towards a better temperature at 5 cm (Fig. 11a and b).
understanding of the origin of CO2 within the soil. However, A likely cause for this difference between short- and long-
this would increase the number of parameters in a regression term temperature dependence is the correlation of the efflux
model thus requiring more data points to yield a significant rate with the inappropriate temperature signal. That the
regression result (Draper and Smith, 1981, p. 298, temperature signal is dampened due to thermal inertia with
recommend that the number of observations exceeds the increasing depth has already been shown. If one presumes
number of parameters 5- to 10-fold). Soil respiration that the short-term variability in the efflux rate is due to the
studies, especially those conducted using manually operated temperature variation in the top-most layer alone, the
closed chambers, often do not provide sufficient numbers of correlating temperature signal from a deeper soil layer
observations to allow regressions of this kind. Despite would have to be corrected for the signal-dampening. If the
sufficient amounts of data obtained in our study, it was not efflux rates in Fig. 11b were plotted against a broader
possible to extract information about compound fluxes from temperature range, the daily temperature dependence curves
two different soil depths, and the temperature measured at a would resemble the seasonal temperature dependence more
depth of 5 cm alone was found to be adequate for a closely. The magnitude of the flux, however, is still affected
description of the surface CO2 flux. Closer analysis of the by temperatures from all depths, so that if one wanted to
interactions of variables showed a high degree of correlation decrease the scattering around the annual regression line,
between the temperature signals from 5 and 10 cm depth a more detailed knowledge of the soil temperature distri-
(r 2 ¼ 0:85; P , 0:001), resulting in these two temperatures bution, especially at the soil surface would be necessary.
to effectively act as one single variable. The correlation
between the more shallow temperatures and the temperature 4.3. Soil moisture limitations
at 30 cm depth was less, but including the deeper
temperature for the modelling of the flux did not result in Limitation of soil CO2 efflux due to either low or high
a significantly improved explanation of soil surface efflux SWCs have been described previously, and a multitude of
variation. regression models exist (Davidson et al., 2000; Janssens
However, there is evidence that the temperature closer to et al., 2002). Drought stress occurs as water becomes
the soil surface, where the most extreme short-term changes limiting for the normal metabolic activity of microbial
in temperature due to changing direct and diffuse radiation organisms or macroflora (Singh and Gupta, 1977). At the
take place, show even greater correlation with the efflux opposite end of the optimal range for respiratory activity, a
rate. The peak rates in Fig. 5 generally precede the peak of reduction due to high SWCs may occur as water limits the
the soil temperature curve at 5 cm. The most extreme peaks diffusion of gases in and out of soil pores. With no O2
are likely to result from times when the soil chamber was available for the aerobic decomposition process, CO2
exposed to relatively high radiation and the temperature at production is inhibited as well as its transport from the
the soil surface had increased considerably. Similar peaks soil pores to the atmosphere (Linn and Doran, 1984). No
were observed for all collars and at various times of the day. limitation due to high SWC was observed at our site,
If one looks at measurements taken at the same collar over probably owing to adequate drainage by the mineral soil (a
the course of several weeks, the limitations of using just one sandy loam with a clay content of less than 5%). The
temperature for the correlation becomes evident. While an limiting effect of low SWC however is pronounced, and

Fig. 11. Dependence of the instantaneous soil CO2 efflux on soil temperature at 5 cm depth for one collar only. Data in (a) are from the entire growing season,
data in (b) are taken within 24 h for the dates shown (all measurements made in 1999).
1480 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

the adequacy of the Bunnell model (Eq. (6)) for the shown experimentally by Kimball and Lemon (1971).
description is clearly documented in Fig. 6. Like most other Closed chamber systems generally exclude these natural
functions used to describe SWC limitations, this model is fluctuations to occur within the chamber space, so that no
empirically based. Papendick and Campbell (1981) show investigation into these effects has been reported from
that the mechanistically appropriate function scales the CO2 studies using this chamber type alone. Rayment and Jarvis
efflux with the cube of the SWC (Davidson et al., 2000). (2000) who also used an open chamber design found no
Given an appropriate parameterisation, this model produces improvement of the description of the efflux data by
virtually identical results to those of both the Bunnell and including the friction velocity in their model. The only
the Gompertz model, so that in this study, the Bunnell evidence of wind-induced pressure pumping stems from
model, with only one fitted parameter, was favoured. The micrometeorological measurements of soil respiration
exact relationship between SWC and the soil CO2 efflux rate (Baldocchi and Meyers, 1991; Arneth et al., 1998). Since
differs from one soil type to another (Howard and Howard, the causal mechanism of the mass flow in and out of the soil
1993), and is also likely to depend on adaptations by the soil pores is the variation in static air pressure, the direct use of
microbial communities to local climatic conditions. Severe this variable is most desirable. Wind is the movement of air
soil CO2 efflux limitations in more arid ecosystems, for due to gradients in air pressure, so that the variation in wind
example, do not occur until the SWC drops below about speed (expressed as su) is a reasonable surrogate for static
0.1 m3 m23 (Carlyle and Ba Than, 1988; Janssens et al., pressure fluctuations ðspÞ: Correlations between sp and the
2000, 2003). Since drought stress is not common at the friction velocity (up ; used as a surrogate by Rayment and
Weidenbrunnen 2 site, there is only minor environmental Jarvis (2000)) and between sp and u (as used by Arneth
pressure for microbial communities to develop appropriate et al., 1998) were both weaker than for su (data not shown).
adaptations. The value of about 0.2 m3 m23 for the The effect of pressure pumping on the soil CO2 efflux is
volumetric water content below which soil CO2 efflux likely to differ according to soil properties like soil bulk
occurs is similar to those reported in other studies from density and soil pore sizes. Using the Bowen Ratio/Energy
temperate and boreal regions of between 0.12 and Balance method and a closed dynamic soil chamber, Dugas
0.19 m3 m23 (temperate deciduous: Hanson et al., 1993; (1993) measured the CO2 flux from bare clay soil. The
Arneth et al., 1998; Davidson et al., 1998; boreal coniferous: reasonably good agreement between both methods with no
Gärdenäs, 2000). The logarithmic regression model describ- reported effect of wind speed indicates that the effect of
ing soil water limitation of soil CO2 efflux in an Asian pressure pumping is negligible for this particular soil type.
steppe ecosystem (Chen et al., 1999) indicates that similar For our data, an effect of pressure pumping could be
adaptation processes act in quite different ecosystem types. demonstrated for few of the 15 soil collars. However, there
Carlyle and Ba Than (1988), Kutsch and Kappen (1997), appears to be a correlation between the type of ground
and Reichstein et al. (2002) have suggested a dependence of vegetation a collar was located in and the influence of static
the temperature sensitivity on SWC, a result that is not pressure fluctuations: All six collars located in D. flexuosa
supported by our findings. Since dry conditions usually patches showed a higher coefficient of determination for the
coincide with high soil temperatures, the effect of either function including wind forcing, compared to only two of
variable becomes confounded with the other (Davidson the remaining nine collars (Table 2). No aboveground parts
et al., 1998). Given this dependence of variables, reduction of the ground vegetation were present within the chambers,
in temperature sensitivity may be due to an increase in but the litter and the Of layer were strongly influenced by the
temperature (a well-established relationship, see above) respective ground vegetation types. The thickness of the
rather than the supposed soil moisture effect. Regressions litter layer in the D. flexuosa patches varies between 2 and
using the Arrhenius type model (which already incorporates 5 cm, compared to 1 –1.4 cm for all other ground vegetation
a decrease in temperature sensitivity with increasing patches. The bulk density of the litter layer was found to
temperatures) showed for the data from Weidenbrunnen 2, range from 0.2 to 0.3 g cm23 for all sites, which is
that the decrease in soil CO2 efflux can be explained purely considerably less than that of the organic and mineral soil
by a reduction in the (temperature insensitive) basal (1.5 and 1.7 g cm23, respectively). It is therefore plausible
respiration rate (Fig. 6). Moisture effects, on the other that the effect of CO2 flushing from the soil pores due to
hand, have also been observed on the E0 -parameter, so that pressure fluctuations was only detected at those collars with
further experimental work is needed in order to clarify this a more substantial layer of low bulk density (and hence a
particular effect. greater pore volume).
The range of su recorded was 0.04 to 1.35 m s21
4.4. Soil CO2 efflux facilitation by wind forcing (mean ¼ 0.34, median ¼ 0.30, n ¼ 12 868), so that the
contribution of the pressure-induced efflux could be as large
Measurements of the soil CO2 efflux are very sensitive to as 1.17 mmol m22 s21 (by multiplying the maximum value
atmospheric pressure, and that static pressure fluctuations of su by 0.869, the largest value for parameter c in Table 2).
cause a mass flow in and out of the soil pores and hence However, it would be too simplistic to associate a given
increase the rate of diffusion of a gas from the soil has been value of su with a specific flux contribution, since
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1481

the meteorological conditions previous to the measurement systems used. An intercomparison of chamber types
(on a scale of hours to days) would have to be accounted for conducted within the framework of the EUROFLUX
first. The flux contribution due to pressure pumping is likely programme (Lankreijer et al., 2003) showed that open
to be considerable for gusts of wind following a period of dynamic chamber systems produce generally lower efflux
relative calm, while it should be smaller for similar given estimates than closed dynamic systems. Given the intensive
wind conditions if the soil pores have been ‘flushed’ by validation of the open dynamic chamber used in our study
pumping previously. (Subke, 2002), and the higher total sampling area
For the purpose of modelling the long-term soil CO2 (4712 cm2 ¼ 314 cm 2 £ 15 collars) in this study vs.
efflux, however, it is not sensible to include pressure 393 cm2 ( ¼ 79 cm2 £ 5 collars) in Buchmann (2000), the
fluctuations as an additional variable, since it only affects lower estimate of around 580 g C m22 yr21 is likely to be a
the gas transport mechanism and not the CO2 production more accurate representation of the actual annual efflux.
within the soil. Averaged over time, therefore, the net In most temperate ecosystems, SWC limitation only
contribution of the pressure related transport would be zero. occurs sporadically, and differs in frequency and duration
However, our results show that measurements done by between years (Fig. 8). Sampling strategies that rely on
chambers that do not measure under steady-state conditions periodic measurements rather than continuous flux readings
are likely to produce biased results owing to the present and therefore run the risk of under-estimating the effect of SWC
previous meteorological conditions. Similarly meteorologi- on soil CO2 efflux, if short periods of low SWC are not
cal techniques such as eddy covariance, which require sampled. The pronounced decrease in annual C loss from
minimum wind speeds and friction velocities to be the soil for 1997 illustrates the necessity to incorporate the
applicable, would over-estimate soil CO2 efflux, as only SWC sensitivity of soil CO2 efflux into global change
the ‘flushing’ of CO2 from the soil would be recorded and models. With global precipitation patterns changing along
not the relatively low efflux rate following times of high with regional annual mean temperatures, an estimate of
pressure fluctuations. future contributions to the total C budget by the soil is only
possible if the SWC sensitivity is known and can be
4.5. Annual soil CO2 efflux described mathematically.

The calculated soil CO2 efflux for the years 1997 to 2000
show considerable inter-annual variability (Table 3). During 4.6. Conclusion
years when annual precipitation sums are close to the long-
term average (as for the years 1998 to 2000), SWC The automated and continuous measurements with the
limitation leads to a reduction of between 9 and 33 g open dynamic chamber system have provided a powerful
C m22 yr21 (or about 1.5 –5.5%) of the respiration sum basis for a comprehensive analysis into the factorial
under conditions without SWC limitation. If longer periods dependence of soil CO2 efflux. The results provide detailed
of SWC limitation occur, as in 1997, C loss from the soil is information that can be used to parameterise ecosystem
reduced considerably. While the values stated in Table 3 are models. Owing to the continuous soil CO2 efflux data, periodic
likely to be an under-estimation of the annual rainfall, the events such as sporadic drying of the top-soil were detected
severity of the SWC limitation on the annual soil CO2 efflux and the SWC limitation could be included in the mathematical
is obvious. The annual results of about 570 g C m22 yr21 description of the soil CO2 efflux. The considerable inter-
for the years 1998 to 2000 compare well to the 560 ^ 17 g annual variability in the C flux sums underlines the necessity
C m22 yr21 reported by Raich and Schlesinger (1992) for to include the effect of the SWC in ecosystem models for
coniferous forests between 40 and 608 latitude. Buchmann humid as well as in arid and semi-arid ecosystems.
(2000) calculated an annual efflux of 710 g C m22 yr21 for a At the same time, the results also show in which areas
neighbouring stand ‘Weidenbrunnen 1’ (a 47-yr-old dense more research efforts have to be undertaken in order to
plantation of P. abies) based on measurements from 1998, understand the dynamics of the C balance of ecosystems. A
while the soil CO2 efflux rates in Weidenbrunnen 2 were simple extrapolation of the results would suggest that an
stated as even higher than in Weidenbrunnen 1, so that an increase in temperature would result in higher CO2
even greater annual sum would result. The theoretical efflux production from the soil (assuming unchanged SWC
rates reported by Buchmann with instantaneous flux rates of conditions). However, if the amount of organic C available
1 mmol m22 s21 at 0 8C and about 5 mmol m22 s21 at 15 8C for microbial decomposition remains unchanged, the total
for the Weidenbrunnen 2 site do not compare well with amount of CO2 efflux would remain constant. One study of
those of our study (compare Figs. 2 and 8). The reason for old SOM in boreal soils (Liski, 1997), for example, found
the higher estimate by Buchmann (2000) may be partly this particular fraction to be insensitive to temperature
explained by a misrepresentation of the actual stand soil changes. A more thorough analysis of specific rates of
CO2 efflux due to a small number of sampling locations (in decomposition for C pools of different stability within the
the mentioned study ‘four to five’ collars were used in the soil would be needed in order to predict the likely behaviour
stand), or a systematic error of either of the sampling of forest soils under a changed climate.
1482 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483

Acknowledgements Dugas, W.A., 1993. Micrometeorological and chamber measurements of

CO2 flux from bare soil. Agricultural and Forest Meteorology 67,
115 –128.
We would like to thank Jörg Gerchau and Gunnar Fang, C., Moncrieff, J.B., 1999. A model for soil CO2 production and
Lischeid of the Bayreuth Institute for Terrestrial Ecosystem transport. 1. Model development. Agricultural and Forest Meteorology
Research for supplying temperature and rainfall data. This 95, 225–236.
work was funded by the German Ministry for Education, Fang, C., Moncrieff, J.B., Gholz, H.L., Clark, K.L., 1998. Soil CO2 efflux
and its spatial variation in a Florida slash pine plantation. Plant and Soil
Science, Research and Technology (BMBF).
205, 135–146.
Gärdenäs, A., 2000. Soil respiration fluxes measured along a hydrological
gradient in a Norway spruce stand in south Sweden Skogaby. Plant and
References Soil 221, 273 –280.
Grace, J., Rayment, M., 2000. Respiration in the balance. Nature 404,
819 –820.
Ågren, G.I., 1999. Temperature dependence of old soil organic matter. Hanson, P.J., Wullschleger, S.D., Bohlman, S.A., Todd, D.E., 1993.
Ambio 29 (1), 55. Seasonal and topographic patterns of forest floor CO2 efflux from an
Anderson, J.M., 1973. Carbon dioxide evolution from two temperate, upland oak forest. Tree Physiology 13, 1– 15.
deciduous woodland soils. Journal of Applied Ecology 10, 361– 378. Heindl, B., Bott, H., 1995. Charakterisierung der Bodenvegetation im
Arneth, A., Kelliher, F.M., Gower, S.W., Scott, N.A., Byers, J.N., Einzugsgebiet Lehstenbach. In: Manderscheidt, B., Göttlein, A. (Eds.),
McSeveny, T.M., 1998. Environmental variables regulating soil carbon Wassereinzugsgebiet Lehstenbach-das BITÖK-Untersuchungsgebiet
dioxide efflux following clear-cutting of a Pinus radiata D. Don am Waldstein (Fichtelgebirge, NO-Bayern), Bayreuther Forum Ökolo-
plantation. Journal of Geophysical Research-Atmospheres 103, gie, 18. BITÖK, Bayreuth, Germany, pp. 49 –63.
5695–5705. Houghton, R.A., 1995. Changes in the storage of terrestrial carbon since
Baldocchi, D.D., Meyers, T.P., 1991. Trace gas exchange above the floor of 1850. In: Lal, R., Kimble, J., Levine, E., Stewart, B.A. (Eds.), Soils and
a deciduous forest. 1. Evaporation and CO2 efflux. Journal of Global Change, CRC Press, Boca Raton, pp. 45– 65.
Geophysical Research 96, 7271–7285. Howard, D.M., Howard, P.J.A., 1993. Relationships between CO2
Berg, B., 2000. Litter decomposition and organic matter turnover in evolution, moisture content and temperature for a range of soil types.
northern forest soils. Forest Ecology and Management 133, 13 –22. Soil Biology and Biochemistry 25, 1537–1546.
Berg, B., Ekbohm, G., Johansson, M.-B., McClaugherty, C., Rutigliano, F., IPCC, Intergovernmental Panel on Climate Change, 2000. In: Watson,
Virzo De Santo, A., 1996. Maximum decomposition limits of forest R.T., Noble, I.R., Bolin, B., Ravindranath, N.H., Verardo, D.J.,
litter types: a synthesis. Canadian Journal of Botany 74, 659 –672. Dokken, D.J. (Eds.), Special Report: Land Use, Land-Use Change
Bohn, H.L., McNeal, B.L., O’Conner, G.A., 1985. Soil Chemistry, Wiley, and Forestry: Summary for Policymakers, Cambridge University
New York, pp. 135–151. Press, Cambridge.
Boone, R.D., Nadelhoffer, K.J., Canary, J.D., Kaye, J.P., 1998. Roots exert IPCC, 2001. A Report of Working Group I of the Intergovernmental Panel
a strong influence on the temperature sensitivity of soil respiration. on Climate Change, Climate Change 2001: The Scientific Basis,
Nature 396, 570 –572. Cambridge University Press, Cambridge.
Bouma, T.J., Nielsen, K.L., Eissenstat, D.M., Lynch, J.P., 1997. Estimating Janssens, I.A., Meiresonne, L., Ceulemans, R., 2000. Mean soil CO2 efflux
respiration of roots in soil: interactions with soil CO2, soil temperature from a mixed forest: temporal and spatial integration. In: Ceulmans,
and soil water content. Plant and Soil 195, 221 –232. R.J.M., Veroustraete, F., Gond, V., van Rensbergen, J.B.H.F. (Eds.),
Buchmann, N., 2000. Biotic and abiotic factors controlling soil respiration Forest Ecosystem Modelling, Upscaling and Remote Sensing, Aca-
rates in Picea abies stands. Soil Biology and Biochemistry 32, demic Publishing, The Hague, pp. 19–31.
1625–1635. Janssens, I.A., Dore, S., Epron, D., Lankreijer, H., Buchmann, N., Longdoz,
Bunnell, F.L., Tait, D.E.N., Flanagan, P.W., van Cleve, K., 1977. Microbial B., Brossaud, J., Montagnani, L., 2003. Climatic influences on seasonal
respiration and substrate weight loss. I. A general model of the and spatial differences in soil CO2 efflux. In: Valentini, R., (Ed.), Fluxes
influences of abiotic variables. Soil Biology and Biochemistry 9, of Energy, Water and Carbon Dioxide of European Forests, Ecological
33–40. Studies, Springer, Berlin, pp. 235–256.
Carlyle, J.C., Ba Than, U., 1988. Abiotic controls of soil respiration beneath Jenny, H., 1980. The Soil Resource. Origin and Behaviour, Springer, New
an 18-year-old Pinus radiata stand in South-Eastern Australia. Journal York.
of Ecology 76, 654 –662. Kimball, B.A., Lemon, E.R., 1971. Air turbulence effects upon soil
Chen, S.Q., Cui, X.Y., Zhou, G.S., Li, L.H., 1999. Study on the CO2-release gas exchange. Soil Science Society of America Proceedings 35,
rate of soil respiration and litter decomposition in Stipa grandis Steppe 16 –21.
in Xilin River Basin, Inner Mongolia. Acta Botanica Sinica 41, Koizumi, H., Kontturi, M., Mariko, S., Nakadai, T., Bekku, Y., Mela, T.,
645– 650. 1999. Soil respiration in three soil types in agricultural ecosystems in
Coûteaux, M.M., McTiernan, K.B., Berg, B., Szuberla, D., Dardenne, P., Finland. Acta Agriculturæ Scandinavica, Section B, Soil and Plant
Bottner, P., 1998. Chemical composition and carbon mineralisation Science 49, 65–74.
potential of Scots Pine needles at different stages of decompostition. Kutsch, W., 1996. Untersuchung zur Bodenatmung zweier Ackerstandorte
Soil Biology and Biochemistry 30, 583–595. im Bereich der Bonhöveder Seenkette. EcoSys, Verein zur Förderung
Davidson, E.A., Belk, E., Boone, R.D., 1998. Soil water content and der Ökosystemforschung zu Kiel e.V., Kiel, Germany.
temperature as independent or compounded factors controlling soil Kutsch, W.L., Kappen, L., 1997. Aspects of carbon and nitrogen cycling in
respiration in a temperate mixed hardwood forest. Global Change soils in the Bornhöved Lake district. II. Modelling the influence of
Biology 4, 217 –227. temperature increase on soil respiration and organic carbon content in
Davidson, E.A., Verchot, L.V., Cattânio, J.H., Ackerman, I.L., arable soils under different managements. Biogeochemistry 39,
Carvalho, J.E.M., 2000. Effects of soil water content on soil 207 –224.
respiration in forests and cattle pastures of eastern Amazonia. Lankreijer, H., Janssens, I.A., Buchmann, N., 2003. Measurement of
Biogeochemistry 48, 53–69. soil respiration within the EUROFLUX project. In: Valentini, R.,
Draper, N.R., Smith, H., 1981. Applied Regression Analysis, second ed., (Ed.), Fluxes of Energy, Water and Carbon Dioxide of European
Wiley, New York. Forests, Ecological Studies, vol. 163. Springer, Berlin, pp. 37–54.
 J.-A. Subke et al. / Soil Biology & Biochemistry 35 (2003) 1467–1483 1483

Linn, D.M., Doran, J.W., 1984. Effect of water-filled pore space on carbon Reichstein, M., 2001. Drought Effects on Carbon and Water Exchange in
dioxide and nitrous oxide production in tilled and nontilled soils. Soil Three Mediterranean Ecosystems, Bayreuther Forum Ökologie, vol. 89.
Science Society of America Journal 48, 1267– 1272. Bayreuther Institut für Terrestrische Ökosystemforschung, Bayreuth,
Liski, J., 1997. Carbon storage of forest soils in Finland, Department of Germany.
Forest Ecology Publications, University of Helsinki, Helsinki. Reichstein, M., Tenhunen, J.D., Roupsard, O., Ourcival, J.M., Rambal, S.,
Liski, J., Ilvesniemi, H., Mäkelä, A., Westman, C.J., 1999. CO2 emissions Dore, S., Valentini, R., 2002. Ecosystem respiration in two Mediterra-
from soil in response to climatic warming are overestimated—the nean evergreen Holm Oak forests: drought effects and decomposition
decomposition of old soil organic matter is tolerant of temperature. dynamics. Functional Ecology 16, 27– 39.
Ambio 28, 171 –174. Schimel, D.S., House, J.I., Hibbard, K.A., Bousquet, P., Ciais, P.,
Lloyd, J., Taylor, J.A., 1994. On the temperature dependence of soil Peylin, P., Braswell, B.H., Apps, M.J., Baker, D., Bondeau, A.,
respiration. Functional Ecology 8, 315–323. Canadell, J., Churkina, G., Cramer, W., Denning, A.S., Field,
Lundegårdh, H., 1926. Carbon dioxide evolution of soil and crop growth. C.B., Friedlingstein, P., Goodale, C., Heimann, M., Houghton,
Soil Science 23, 417–453. R.A., Melillo, J.M., Moore, B. III, Murdiyarso, D., Noble, I.,
Mariko, S., Nishimura, N., Mo, W., Matsui, Y., Kibe, T., Koizumi, H., Pacala, S.W., Prentice, I.C., Raupach, M.R., Rayner, P.J., Scholes,
2000. Winter CO2 flux from soil and snow surfaces in a cool-temperate R.J., Steffen, W.L., Wirth, C., 2001. Recent patterns and
deciduous forest, Japan. Ecological Research 15, 363–372. mechanisms of carbon exchange by terrestrial ecosystems. Nature
Morén, A.-S., Lindroth, A., 2000. CO2 exchange at the floor of a boreal 414, 169 – 172.
forest. Agricultural and Forest Meteorology 101, 1–14. Singh, J.S., Gupta, S.R., 1977. Plant decomposition and soil respiration in
Norby, R.J., Jackson, R.B., 2000. Root dynamics and global change: terrestrial ecosystems. Botanical Review 43, 449–528.
seeking an ecosystem perspective. New Phytologist 147, 3–12. Subke, J.-A., 2002. Forest Floor CO2 Fluxes in Temperate Forest
Papendick, R.I., Campbell, G.S., 1981. Theory and measurement of water Ecosystems. An Investigation of Spatial and Temporal Patterns and
potential. In: Parr, J.F., Gardner, W.R., Elliot, L.F. (Eds.), Water Abiotic Controls, Bayreuther Forum Ökologie, vol. 96. BITÖK,
Potential Relations in Soil Microbiology, Soil Science Society of Bayreuth, Germany.
America, Madison, pp. 1–22. Thornley, J.H.M., Cannell, M.G.R., 2001. Soil carbon storage response to
Parton, W.J., Schimel, D.S., Cole, C.V., Ojima, D.S., 1987. Analysis of temperature: an hypothesis. Annals of Botany 87, 591 –598.
factors controlling soil organic matter levels in Great Plains grasslands. Trumbore, S., 2000. Age of soil organic matter and soil respiration:
Soil Science Society of America Journal 51, 1173–1179. radiocarbon constraints on belowground C dynamics. Ecological
Raich, J.W., Schlesinger, W.H., 1992. The global carbon dioxide flux in soil Application 10, 399– 411.
respiration and its relationship to vegetation and climate. Tellus 44B, Valentini, R., Matteucci, G., Dolman, A.J., Schulze, E.-D., Rebmann, C.,
81–99. Moors, E.J., Granier, A., Gross, P., Jensen, N.O., Pilegaard, K., Lindroth,
Rayment, M.B., Jarvis, P.G., 1997. An improved open chamber system for A., Grelle, A., Bernhofer, C., Grünwald, T., Aubinet, M., Ceulemans, R.,
measuring soil CO2 effluxes in the field. Journal of Geophysical Kowalski, A.S., Vesala, T., Rannik, Ü., Berbiger, P., Loustau, D.,
Research 102, 28779–28784. Gudmundsson, J., Thorgeirsson, H., Ibrom, A., Morgenstern, K.,
Rayment, M.B., Jarvis, P.G., 2000. Temporal and spatial variation of soil Clement, R., Moncrieff, J., Montagnani, L., Minerbi, S., Jarvis, P.G.,
CO2 efflux in a Canadian boreal forest. Soil Biology and Biochemistry 2000. Respiration as the main determinant of carbon balance in European
32, 35 –45. forests. Nature 404, 861 –865.