Spatial and Seasonal Variability of the Stable Carbon Isotope

PUBLICATIONS
Journal of Geophysical Research: Biogeosciences
RESEARCH ARTICLE
10.1002/2015JG003193
Key Points:
• We show systematic spatial variability
13
in the δ C of soil CO2 and soil CO2 flux
in complex terrain
• At the landscape level, we found
13
different seasonal trends in the δ C of
soil CO2 and soil CO2 flux
• The spatial variability of soil moisture
13
is a strong predictor of the δ C of soil
CO2 flux at the landscape scale
Supporting Information:
• Supporting Information S1
Correspondence to:
D. A. Riveros-Iregui,
[email protected]
Citation:
Liang, L. L., D. A. Riveros-Iregui, and
D. A. Risk (2016), Spatial and seasonal
variabilities of the stable carbon isotope
composition of soil CO2 concentration
and flux in complex terrain, J. Geophys.
Res. Biogeosci., 121, 2328–2339,
doi:10.1002/2015JG003193.
Received 23 AUG 2015
Accepted 21 AUG 2016
Accepted article online 25 AUG 2016
Published online 14 SEP 2016
Spatial and seasonal variabilities of the stable carbon
isotope composition of soil CO2 concentration
and flux in complex terrain
Liyin L. Liang1,2, Diego A. Riveros-Iregui3, and David A. Risk4
1
School of Natural Resources, University of Nebraska–Lincoln, Lincoln, Nebraska, USA, 2School of Science and
Environmental Research Institute, University of Waikato, Hamilton, New Zealand, 3Department of Geography, University of
North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA, 4Department of Earth Sciences, St. Francis Xavier University,
Antagonish, Nova Scotia, Canada
Abstract Biogeochemical processes driving the spatial variability of soil CO2 production and flux are well
studied, but little is known about the variability in the spatial distribution of the stable carbon isotopes that
make up soil CO2, particularly in complex terrain. Spatial differences in stable isotopes of soil CO2 could
indicate fundamental differences in isotopic fractionation at the landscape level and may be useful to inform
modeling of carbon cycling over large areas. We measured the spatial and seasonal variabilities of the δ13C of
soil CO2 (δS) and the δ13C of soil CO2 flux (δP) in a subalpine forest ecosystem located in the Rocky Mountains
of Montana. We found consistently more isotopically depleted values of δS and δP in low and wet areas of the
landscape relative to steep and dry areas. Our results suggest that the spatial patterns of δS and δP are
strongly mediated by soil water and soil respiration rate. More interestingly, our analysis revealed different
temporal trends in δP across the landscape; in high landscape positions δP became more positive, whereas in
low landscape positions δP became more negative with time. These trends might be the result of differential
dynamics in the seasonality of soil moisture and its effects on soil CO2 production and flux. Our results
suggest concomitant yet independent effects of water on physical (soil gas diffusivity) and biological
(photosynthetic discrimination) processes that mediate δS and δP and are important when evaluating the
δ13C of CO2 exchanged between soils and the atmosphere in complex terrain.
1. Introduction
In complex terrain, the morphology of the landscape can influence the spatial distribution of resources such as
surface available energy [Korkalainen and Laurén, 2006], soil moisture [Emanuel et al., 2010], soil temperature
[Lundquist et al., 2008; Liang et al., 2014], or soil carbon and nitrogen [Webster et al., 2008]. Spatial variability in
biophysical resources has been found to lead to spatially organized heterogeneity in fluxes of soil CO2 in mountainous, forested ecosystems [Riveros-Iregui and McGlynn, 2009]. Investigating the way in which topography
mediates the transformation of soil carbon is required for improved understanding of the physical and biological
processes responsible for carbon exchange between terrestrial ecosystems and the atmosphere [Riveros-Iregui
et al., 2012; Adams et al., 2014]. The ratio of the stable isotopes of carbon (13C/12C) is commonly used to
examine carbon transformation along the soil-plant-atmosphere continuum [Bowling et al., 2009; Hu et al.,
2010; Riveros-Iregui et al., 2011b]. While recent studies have identified strong feedback between topography
and soil CO2 concentration and flux [Riveros-Iregui et al., 2007, 2012], little is known about the effects that
topography can pose to the spatial and temporal variabilities of the stable carbon isotopes (δ13C) that make
up that soil CO2, particularly in complex terrain.
©2016. American Geophysical Union.
All Rights Reserved.
LIANG ET AL.
Much of our understanding of environmental controls on δ13C of terrestrial ecosystems is derived from the
seminal work of Farquhar and colleagues more than three decades ago [Farquhar et al., 1982], which elucidated
fundamental relations that control the exchange of CO2 at the scale of single leaves. Discrimination against
13
CO2 can occur along the photosynthetic pathway when CO2 enters the plant through the stomata, and
variables such as air temperature, soil moisture, vapor pressure deficit, or photosynthetic photon flux density
influence the degree of isotopic discrimination [Farquhar et al., 1982, 1989]. At the ecosystem level, a large
fraction of the ecosystem-respired CO2 may come from recently produced carbohydrates [e.g., Flanagan
et al., 1996; Högberg et al., 2001; Knohl et al., 2005], and thus, the estimated δ13C value of these carbohydrates
is usually extrapolated across multiple spatial scales [e.g., Lloyd and Farquhar, 1994]. General ecosystem C
VARIABILITY OF THE δ13C IN SOIL CO2
2328
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
models typically include empirical relationships between environmental variables, the δ13C of recent
carbohydrates, and the δ13C of ecosystem-respired CO2. However, these models generally do not account for
the spatial heterogeneity in C fluxes derived from spatially distributed soil moisture and soil temperature
[Riveros-Iregui et al., 2011a; Liang et al., 2014] or for the landscape variability in ecosystem type, stage of development, or disturbance [Gifford, 2003].
Studies that examine stable isotopes in forest ecosystems typically focus on the vertical variability of the δ13C
of different forest compartments, including soil CO2 [e.g., Ehleringer et al., 2000; Fessenden and Ehleringer,
2003], leaves at various heights in the canopy [Vogel, 1978; Schleser and Jayasekera, 1985; Medina et al.,
1991], or CO2 in canopy air [Buchmann et al., 1996; Schaeffer et al., 2008a; Bowling et al., 2009; Riveros-Iregui
et al., 2011b]. However, little attention has been given to the spatial variability of the δ13C of forest carbon
across heterogeneous landscapes (although see Tucker et al. [2014] for comparisons between the δ13C of soil
respiration of a forest patch and a meadow patch during winter). Previous studies have recognized that the
assumption of spatially homogeneous composition of δ13C of forest carbon is often violated [Pataki et al.,
2003]; however, this assumption has not been rigorously tested in heterogeneous landscapes.
Here we address the need to understand the spatial variability in the δ13C of forest carbon at spatial scales that
are greater than individual forest patches (e.g., at scales from 1000 m2 to 10 km2). We examined the spatial
variability of the δ13C of bulk soil CO2 (δS) and the δ13C of soil CO2 flux estimated by using two different
methods: a two end-member mixing relationship known as the Keeling plot (δP) [Keeling, 1958, 1961] and a
CO2 gas diffusion method (δJ) [Davidson, 1995; Bowling et al., 2009]. We collected these measurements by using
34 soil respiration plots distributed across a 3.9 km2 forested watershed of the Rocky Mountains of Montana.
We hypothesized that the morphology of the landscape can mediate the spatial patterns of the isotopic
composition (δ13C) of soil CO2 and flux in the same way that it mediates the magnitude of soil CO2
concentrations and flux [Riveros-Iregui et al., 2012]. We further hypothesized that soil moisture plays an
important role in the spatial and temporal variabilities of the δ13C of soil CO2 and flux at the landscape scale.
The information presented here is necessary to enhance our understanding of soil carbon transformation over
large areas and should contribute to our continuous refinement of the mechanistic understanding of
ecosystem processes and models.
2. Methods
2.1. Site Description
This study was conducted in the Tenderfoot Creek Experimental Forest (TCEF) located in the Little Belt
Mountains of central Montana, USA. The elevation of TCEF ranges from 1840 to 2421 m. Mean annual
precipitation is 880 mm with 70% falling as snow, and peak snowpack accumulations occur between late
March and mid-April [Farnes et al., 1995]. Mean annual temperature is 0°C, and the growing season typically
extends from early or mid-June to the end of August. The two dominant landscapes at this site are upland
forests and riparian meadows, both of which are vegetated by C3 plants only. This forest is a characteristic
of other subalpine forests of the northern Rocky Mountains and exhibits a broad set of microtopographic
features, including slope, aspect, convergent, and divergent areas, all of which mediate the variability of
the soil environment. Upland forests are mostly covered by Pinus contorta (lodgepole pine) in the overstory
and Vaccinium scoparium (Whortleberry) in the understory, whereas riparian meadows are covered by
Calamagrostis canadensis (blue joint reed grass) [Mincemoyer and Birdsall, 2006].
Thirty-four (34) soil CO2 sampling plots across a 3.9 km2 watershed were used in this study (Figure 1); these were
part of 62 plots installed between 2004 and 2005 (see Pacific et al. [2008] and Riveros-Iregui and McGlynn [2009]
for details). The spatial location of the soil CO2 plots was intended to capture the complete range of the relative
soil wetness potential of the forest (i.e., the full range from wet to dry locations). A quantitative assessment of the
microtopography of this site in relation to soil CO2 plot location was conducted by Riveros-Iregui and McGlynn
[2009], concluding that the location of all soil CO2 sampling plots was a characteristic of the topographic
heterogeneity (i.e., the full spectrum of soil moisture microenvironments) of the landscape. Twenty-four (24)
of those soil plots were located in forested uplands, whereas 10 soil plots were located in riparian areas.
At each soil plot, three PVC gas wells of similar volume were installed at 5 cm, 20 cm, and 50 cm, respectively.
The gas wells were made of 15 cm sections of 5.25 cm (inside diameter) PVC inserted into a hole augered to
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2329
Journal of Geophysical Research: Biogeosciences
Figure 1. Spatial distribution of the sample locations (34 sites) across the
Tenderfoot Creek Experimental Forest, located in Central Montana.
Sampling locations included 10 riparian meadow sites (circles) and 24
upland forest sites (triangles). Color shading represents elevation (m).
10.1002/2015JG003193
17, 20, or 50 cm, as described by Pacific
et al. [2008]. In all cases, the top of the
PVC was capped with a rubber stopper
(size 11) through which passed two sections of PVC tubing (4.8 mm inside diameter Nalgene 180 clear PVC, Nalge
Nunc International, Rochester, N.Y., USA)
that extended above the ground surface.
In the case of the deeper wells, the PVC
was left open at the bottom (20 and
50 cm, respectively). In the case of the
shallowest wells, the bottom of the PVC
was capped with a rubber stopper (size
11) and the PVC was drilled into from its
sides, leaving openings at a 5 cm depth
to facilitate gas equilibrium between the
gas well and the soil. Above the ground
surface, the tubing was joined with connectors to prevent gas leaks between
measurements. A total of 123 gas wells
were used in this study.
2.2. Data Collection
Measurements were taken at each of the 34 plots about twice a week over the period between 7 July
and 8 August 2012 (23 sampling days in total). At each plot, we collected three measurements of
volumetric water content (VWC) by using a portable meter (Hydrosense, Campbell Scientific, Logan,
UT, USA) that was inserted through the top 12 cm of soil. VWC measurements are reported as percent
values (m3 m3 × 100). At the same time, we measured VWC; we collected three measurements of soil
temperature by using a 12 cm soil thermometer (Reotemp Instruments, San Diego CA) for each site,
integrating from 0 to 12 cm depth.
Soil gas samples were collected from the gas wells at three depths (5 cm, 20 cm, and 50 cm) at each of the 34
plots. Soil gas samples were taken on different days, usually 3–4 days apart, allowing enough time for the soil
gas atmosphere to equilibrate. On each sampling day, 120 ml of soil gas were extracted from each gas well
with a syringe over a period of 2 min. An infrared gas analyzer (GM-70, Vaisala, Woburn, MA) was connected
in line with the sampling setup to circulate the air in the gas well and ensure that no major changes in CO2
concentration occurred inside the gas well and Nalgene tubing while sampling (such changes in CO2 concentrations could be indicative of leaks in the sampling setup). The samples were immediately transferred into a
180 ml air sample Tedlar bag (SKC Inc., Eighty Four, PA, USA). Within 4 h of collection, Tedlar bags were connected to a Cavity Ring-Down Spectroscopy (CRDS) analyzer (model 2101-i Picarro Inc., Sunnyvale, California)
to measure the δ13C composition of soil gas samples in the Tedlar bag.
According to manufacturer specifications, the CRDS analyzer offers a precision of δ13C that is better than 0.3‰.
Our routine checks using two standard tanks yielded a repeatability of the instrument better than 0.2‰.
However, owing to the reported drift in δ13C values of the CRDS analyzer over longer periods [Vogel et al.,
2012], we performed a quality check that consisted of running one standard sample at the beginning of each
day and one standard sample every ~90 min between soil gas samples. We also ran one standard sample after
all soil gas samples were analyzed on each day. The measured precision of the standard gas is 0.2‰ from 31
repeated measurements. We used all measurements of standard gas to calculate the daily drift of the CRDS
analyzer and used this information to correct the δ13C measurements of soil gas on a daily basis.
2.3. Isotopic Mixing Relationships
Under steady state conditions, the δ13C of soil CO2 flux is equivalent to the δ13C of soil CO2 production
[Cerling, 1984]. We used two different methods, both commonly used, to estimate this value. The first method
was the Keeling plot method [Keeling, 1958, 1961] whose product is hereafter called δP, whereas the second
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2330
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
method is a diffusion-based relationship
[Davidson, 1995; Bowling et al., 2009]
whose product is hereafter called δJ.
The Keeling plot method is based on the
relationship between the isotopic signature of soil CO2 and atmospheric CO2 as
expressed in
δS ¼ C a ðδa δR Þ
1
þ δR ;
CS
(1)
where C and δ represent the CO2 concentration ([CO2], ppm) and its stable
carbon isotopic composition (δ13C [‰]),
respectively. The subscripts S, a, and R
represent values for the soil, background
air, and the apparent respiratory source,
respectively. In our experimental setup,
a Keeling plot consisted of a regression
Figure 2. Estimation of the stable carbon isotopic composition of soil between the observed carbon isotopic
respiration via the Keeling plot approach, noted as δP [Keeling, 1958,
composition of soil CO2 at three depths
1961]. The open circles represent the CO2 concentrations (ppm) and stable
(i.e., δS for 5, 20, and 50 cm) and above
13
carbon isotopic composition (δ C, ‰) of soil CO2 collected at 5, 20, and
air (δa), and the inverse CO2 concentra50 cm depths. The closed circle represents the CO2 concentration (Ca) and
tions for the same four points (CS for 5,
isotopic composition (δa) of background ambient air. Using these four
points, we derived the apparent isotopic signature of soil respiratory source 20, and 50 cm and Ca). The above-air
(δR; closed square). δP (closed triangle) was obtained by subtracting 4.4‰
values for Ca and δa were collected at a
from δR.
single location near the headwaters of
the watershed at an elevation of 70 cm
above the soil surface and throughout the study period. The means of all measured Ca and δa were 384
± 7.6 ppm and 11.1 ± 1.2‰, respectively. In particular, the seemingly more negative values of δa are due to
the proximity of these measurements to the soil, as it is well established that this value varies with height within
the canopy (Bowling et al., 2009). We used time-specific measurements of Ca and δa from each day to calculate
each Keeling plot. We did not collect location-specific Ca or δa, and thus, we assumed that both values were homogenous across the study site. Previous studies have demonstrated that the variability of Ca or δa is negligible compared to the much larger variability of these quantities in the soil [Riveros-Iregui et al., 2008; Bowling et al., 2015;
Mauer et al., 2016]. We then calculated the intercept of the regression (δR) as shown in Figure 2. We employed
the ordinary least squares regression and quantified the uncertainty by using the standard error of the intercept,
according to Zobitz et al. [2006]. Any regression with a standard error of the intercept above 1‰ was discarded
(i.e., 59 out of 285 records), ensuring that only the most robust regressions were used in further analysis
[Riveros-Iregui et al., 2011b]. The δP was calculated by subtracting 4.4‰ (diffusive depletion constant) from the estimated δR [Bowling et al., 2009]. Being a depth-averaged technique (where the atmosphere is the uppermost level),
the δP solution is not overly sensitive to errors at any particular location. Keeling plots are known to show curvature
under non-steady-state conditions [Nickerson and Risk, 2009; Bowling et al., 2015]. Nickerson and Risk [2009] used
numerical simulations combined with measurements of the δ13C of soil CO2 flux to show that Keeling plots are
nonlinear when diffusivities are high. Their work also demonstrates that nonsteady state conditions must be very
low (or nonexistent) to allow for the development of linear Keeling plots. In our case, each Keeling plot showed an
R2 greater than 0.95, suggesting that our measurements were sampled at or near steady state and that the theoretical diffusive depletion constant (4.4‰) between 12C and 13C could be applied to estimate δP.
The second method is a diffusion-based relationship derived by Davidson [1995] that uses δS measured at a
single depth and δa to estimate δJ. Bowling et al. [2009] further proposed a solution for the Davidson [1995]
model as follows:
δJ ¼
LIANG ET AL.
C S ðδS 4:4Þ C a ðδa 4:4Þ
;
1:0044ðC S C a Þ
VARIABILITY OF THE δ13C IN SOIL CO2
(2)
2331
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
where the CS/Ca are the soil/air CO2 concentrations, respectively. In principle, equation (2) is useful because δJ
can be obtained for any soil depth. Note that this model allows for the calculation of δJ independently of the
diffusivity coefficient and all related parameters (including air porosity, tortuosity, temperature, and pressure)
as they are the same for 12CO2 and 13CO2 at each point (see original derivation by Davidson [1995]). We
looked for systematic differences between the δP and δJ techniques by comparing both values as calculated
from equations (1) and (2) and using our field measurements. We further evaluated this comparison by
simulating the concentrations of 12CO2 and 13CO2 in an ideal soil profile, following the idealized diffusive
steady state model described in detail by Cerling [1984],
ϕi
z2
C iS ¼ i Lz (3)
þ C ia
2
DS
where i denotes the 12CO2 or 13CO2, ϕ is the soil respiration rate, Ds is CO2 diffusivity in soil, L is the impermeable boundary at which CO2 production rate is assumed to be zero, z is the measured depth, and Ca is the
ambient air CO2 concentration. The concentrations of 12CO2 and 13CO2 can be further used to calculate
the depth profile of δS (see supporting information). Based on the simulated δS, we tested the sensitivity of
δP and δJ to the change of CO2, δS, or both by introducing artificial variability in both terms; that is, we allowed
soil CO2 and δS to fluctuate within reasonable boundaries (5% for soil CO2, 0.3‰ for δS) and assessed the
impact that this fluctuation—independently or combined—had on the estimated δP and δJ. In our case,
the δ13C of CO2 production was set at 27‰, the diffusivity of CO2 was 3 × 105 m2 s1, L = 1 m, and soil
temperature and soil moisture were assumed constant. Atmospheric pressure was set at 1 standard
atmosphere (1 atm) at 25°C, and the δ13C of atmospheric CO2 (8.3‰) and its concentration (394 ppm) were
based on Mauna Loa's records for 2012 (http://scrippsco2.ucsd.edu/). See supporting material for additional
information.
2.4. Data Analysis
We performed a three-way analysis of variance (ANOVA) for the measured δS to quantify differences in this
variable across soil depths (5, 20, and 50 cm), landscape locations (riparian meadows and upland forests),
and time (sampling date, 21 days). Prior to conducting the ANOVA, we tested the normality of the data by
using the Shapiro-Wilk test [Shapiro and Wilk, 1965; Royston, 1982] and examined the homogeneity of variances by using the Bartlett's test [Bartlett, 1937]. The Tukey's honestly significant difference (HSD) test was
used to process post-hoc ANOVA comparisons. We also tested the effects of landscape location and sampling
date on the calculated δP by using two-way ANOVA and incorporated the soil moisture and soil respiration
rate as the covariates to conduct ANCOVA analysis. We further examined the relationship between the calculated δP and environmental variables such as soil temperature (Ta), soil volumetric water content (VWC), and
soil respiration rate via linear and multiple linear regressions. We applied a bootstrap analysis to the multiple
linear regression among the estimated δP, the measured VWC, and the measured soil respiration rate.
Bootstrap is a random resampling with replacement method that allows for the quantification of uncertainty
(including measures of bias, standard error, and confidence intervals) of sample estimates [Efron, 1979]. The
10,000-iteration bootstrap analysis and resulting confidence surfaces were intended to provide information
regarding the interaction of the two measured variables (VWC and soil respiration rate) and the prediction
error in the estimated δP, across their full range of variability in VWC and soil respiration. All statistical analyses
and data processing were performed by using MATLAB R2011b (The MathWorks Inc., Natick, MA, USA) and R
Studio (Version 0.99.903, R Studio Inc., Boston, MA).
3. Results
3.1. Measurements and Spatial Variability of δS
Across all 34 plots, the δ13C of soil CO2 (δS) varied from 29.7 to 10.0‰, with consistently less negative
values near the surface in both riparian meadows and upland forests (Figure 3a). The δS was significantly different among depths (ANOVA, p < 0.0001), landscape locations (riparian meadows and upland forests,
p < 0.0001), sampling date (p < 0.0001), and the interaction between landscape locations and depths
(p = 0.0002). No significant effects were found between sampling date and depth (p = 0.46), sampling date
and landscape locations (p = 0.32), or all three (p = 0.45). Upon closer look, the HSD test indicated that δS
was systematically more negative in riparian sites than upland forests (Figure 3a) with average differences
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2332
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
Figure 3. (a) Measured isotopic composition of soil CO2 (δS) at all riparian meadows and upland forest sites at 5, 20, and 50 cm. The dashed gray lines are the theoretical δS profiles according to an idealized diffusive steady state model [Cerling, 1984, 1991] under different soil respiration rates, as indicated. (b) Variability of δS
from different depth in relation to changing soil volumetric water content (VWC). Probability distributions of measured δS at 5 cm (c), 20 cm (d), and 50 cm (e) at
different VWC. The dashed lines represent the mean δS for each probability distribution (18.3, 21.1, and 21.6‰, respectively).
of 2.4‰ (95% confidence interval (CI): (1.9 2.8‰)), and more negative in deeper layers than shallow soils
(Figures 3a and 3b) with average difference of 2.8‰ between 5 and 20 cm (CI: (2.4 3.3‰)) and 0.5‰ between
20 and 50 cm (CI: (0.1 1.0‰)). Furthermore, our analysis showed that landscape positions with consistently
higher VWC also exhibited systematically more negative δS than those landscape positions with low VWC
(Figures 3a and 3b) and that this trend was consistent at all three depths (Figures 3c–3e), leading to a
prominent spatial organization of δS along the range of observed soil moisture.
3.2. Estimates of δP and δJ
When comparing estimates of the δ13C of soil CO2 flux derived from the Keeling plot method (δP) and diffusionbased relationship (δJ), we found significant differences between the two at 5 cm (p < 0.0001) but not at 20 cm
(p = 0.98) or 50 cm (p = 0.96) (Figure 4). These results were further confirmed by the idealized diffusive steady
state model (equation (3)). An analysis of the sensitivity of δP and δJ to the variability in CS and δS at different
depths showed that estimates of δJ are much more sensitive to small variability in CS and δS in shallow horizons
than in deeper soil horizons, whereas estimates of δP showed similar sensitivity to changes in CS and δS across all
depths (see Figure S1 in the supporting information). Thus, based on the comparison of the measured δP and δJ
with depth (Figure 4) and the sensitivity of each to idealized diffusive steady state conditions, we chose to use
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2333
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
the δP values derived from the Keeling plot
approach in all further analysis given their
greater confidence, particularly for the
surface layers.
The estimated δP ranged from 32.4 to
23.8‰ for all 34 soil plots. Similar to the
spatial variability exhibited by δS, estimates
of δP were significantly more negative in
riparian meadows (30.7 ± 0.5‰) than
upland forests (28.1 ± 0.1‰) (Figure 5a),
with an average difference of 1.7‰ (CI:
(1.3 2.2‰)) across all sites. Two-way
ANOVA results showed that landscape
position (p < 0.0001), sampling date
(p < 0.0001), and their interaction
(p = 0.017) had significant effects on δP,
indicating a strong spatial and temporal
Figure 4. Comparison of the isotopic composition of soil CO2 flux esti- variability of δP. Furthermore, incorporating
the VWC and respiration rate with landmated by the Keeling plot approach (δP) and the diffusion approach
(δJ) across different depths. The error bars represent the standard error
scape position and time, the ANCOVA
of the intercept of all Keeling plots. These results show higher agreeresults showed that both VWC
ment by both techniques at deeper soil horizons (20 and 50 cm) than at
(p < 0.0001) and respiration rate (RS)
shallow horizons (5 cm).
(p = 0.039) have significant effects on δP,
although the effect of the interaction
between landscape position and sampling date on δP was not significant (p = 0.11). Furthermore, we found
no significant effects from the interaction between RS and VWC on δP (p = 0.47). No statistical relation was found
between δP and soil temperature (Figure 5c). At the landscape scale the spatial variability of δP appeared to be
systematically mediated by VWC and soil respiration rate (RS) (Figures 5b, 5d, and 5e). A multiple linear regression among δP, VWC, and RS showed that VWC (p < 0.0001) and RS (p = 0.016) had significant effects on δP, with
more negative δP at higher VWC and RS (R2 = 0.335, p < 0.0001) (Figure 5b). Furthermore, a 10,000-iteration
bootstrap analysis of this multiple linear regression revealed much lower uncertainty at low RS and low VWC
and greater uncertainty at high RS and high VWC (Figure 5b). While VWC and RS were correlated (R2 = 0.25,
p < 0.01), these predictors did not exhibit multicollinearity on the multiple linear regression (variable inflation
factor = 1.503) and their interaction did not have a significant effect on δP (p = 0.0832).
3.3. Spatial and Temporal Variabilities of δP
To further illustrate the spatial and temporal variabilities of δP at the landscape level, we combined the emergent multiple linear regression model (Figure 5b) with results from a recently developed ecohydrological
model developed for this same site, which includes VWC and RS modules [Riveros-Iregui et al., 2011a]. This
analysis sought to examine the potential variability of δP under highly variable soil water content and soil
respiration rates, characteristic for this site. Our results revealed strong spatial variability of δP across the
landscape, which closely followed differences in VWC imposed by the topography and drainage patterns
(Figure 6). More importantly, these effects are noticeable across the growing season (i.e., seasonal changes
from June to August) although the magnitude of the differences between high and low areas of the
landscape seems to change across the season.
4. Discussion
It has been established that topography and landscape morphology can play an essential role in mediating
soil and ecosystem carbon transformation at the landscape scale [Riveros-Iregui and McGlynn, 2009; Emanuel
et al., 2011; Riveros-Iregui et al., 2012; Adams et al., 2014] as a result of the spatial distribution of biophysical
variables such as soil water, soil temperature, and soil nutrients. Our working hypothesis was that the
morphology of the landscape, and in particular soil moisture, can mediate the spatial patterns of the isotopic
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2334
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
Figure 5. (a) A comparison of Keeling plot between the aggregated δP at riparian meadows and upland forest sites. (b) A multiple linear regression between δP and soil
water content (VWC), and soil respiration rate (RS), including an estimated confidence surface interval proposed by Liu and Lin [2009] and derived from a 10,000-iteration
bootstrap. Correlations between (c) δP and soil temperature, (d) VWC, and (e) RS revealed that VWC and RS are stronger predictors of the variability in δP than soil
temperature. The error bars represent the standard error of the intercept of all Keeling plots.
Figure 6. Modeled spatial and seasonal variabilities of δP across the entire watershed based on the relationship between δP and soil moisture (VWC) and soil respiration
(RS), shown in Figure 5b. The likely range and spatial patterns of δP assessed for three different time periods: (a) June 28, (b) August 8, and (c) August 23.
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2335
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
composition (δ13C) of soil CO2 and flux in the same way that it mediates the magnitude of soil CO2 concentrations and efflux. We found clear spatial and temporal patterns in δS and δP across the entire watershed during the growing season, which appeared to be mediated by VWC, soil respiration rate, and their interactions.
Furthermore, our empirical model suggests that there are confounding effects from physical and biological
controls on the δ13C composition of soil CO2 flux in complex terrain.
4.1. Physical Effects on δS and δP
Plot-level comparisons between δS measured various depths (5 cm versus 20 cm or 20 cm versus 50 cm)
(Figure 3b) and VWC, revealed that the δS was systematically less negative in the shallow soil relative to the
deeper soil (Figure 3a). Previous studies have reported similar depth-dependence of δS [Cerling, 1984;
Cerling et al., 1991; Davidson, 1995] under steady state conditions, which is likely caused by (1) faster molecular
diffusion at shallow depths and enhanced gas exchange between soil CO2 and the overlying atmospheric CO2
and (2) reduced soil gas diffusivity at deeper depths [Suwa et al., 2004]. Our findings further suggest that in
complex terrain, the depth-dependence of δS is also strongly influenced by soil moisture (Figure 3).
Regression analysis of the relationship between environmental factors, i.e., soil temperature and soil moisture,
and δP, showed that δP is strongly related to VWC. Thus, accurate representation of δS and δP at the landscape
level should include information regarding the spatial variability of soil moisture and its differential effects on
the rate of soil respiration. However, we found no soil temperature effects on δP in our study (Figure 5a). This
might be because of a hypothesized strong autotrophic contribution to soil respiration during the growing
season at this site, according to previous studies [Riveros-Iregui et al., 2007, 2011a]. The autotrophic component
of soil respiration is more sensitive to photosynthesis and soil moisture rather than soil temperature, which
could result in the lack of relationship between δP and soil temperature found in this study.
Similarly, wind-induced variability could potentially affect soil CO2 diffusion and the δS value of shallow soil, as
it has been observed for CO2 profiles in snow [Bowling and Massman, 2011]. Under windy conditions, the diffusivity ratio between 13CO2 and 12CO2 deviates from the 4.4‰ diffusion constant [Bowling and Massman,
2011], since wind can cause a pressure gradient between the soil and the atmosphere and induce a small
but persistent airflow within the soil [Bowling and Massman, 2011]. Advective effects could be strong in riparian areas due to low canopy cover and associated wind. However, previous studies have shown that wind
advection does not significantly affect the diffusivity ratio of CO2 for buried gas wells [Kayler et al., 2010;
Moyes et al., 2010], as those deployed in our study. While we did not collect wind speed in riparian areas that
would allow us to further evaluate this effect, our data quality control, namely, the standard error of the intercept of each Keeling plot lower than 1‰ and the R2 of each soil Keeling plot greater than 0.95 (section 2.3),
suggest that in our study steady state conditions were met during data collection and the wind-induced
effect is negligible. Nonetheless, the confounding effects of covarying factors can be problematic and remain
to be resolved experimentally to further our mechanistic understanding and uncertainty estimation [e.g.,
Ogle and Pendall, 2015].
4.2. Plant Physiological Effects on δS and δP
Several studies have investigated the variability in the δ13C composition of ecosystem respiration or soil CO2
flux as a response to variable soil water content, either at the stand level or across broad geographic climate
gradients [Ekblad and Högberg, 2001; Bowling et al., 2002; Pataki et al., 2003; Ekblad et al., 2005; Knohl et al.,
2005; Lai et al., 2005; Alstad et al., 2007; Schaeffer et al., 2008b] in ways that are consistent with our understanding of carbon isotopic exchange between the leaf and the atmosphere [Farquhar et al., 1989]. These studies hypothesized that the variability in the δ13C of respired CO2 (from soil or canopy) is partially dependent
on leaf-level photosynthetic discrimination and derived respiration by plants and root symbionts. Our results
confirmed the relationships reported by those studies: δP increases (less negative) with decreasing soil moisture (Figures 5b and 5d). Our results also suggest that the effects of soil moisture on photosynthetic processes
and resulting soil CO2 are evident across gradients of soil moisture (Figures 3b–3e) in heterogeneous landscapes. Thus, in complex terrain, the spatial variability of δS and δP may be described, at least partially, as a
function of landscape position. This spatial variability is likely indicative of plant physiological responses
along topographic gradients and the interdependent autotrophic and heterotrophic contributions to total
forest respired 13CO2 across wet and dry portions of the landscape. Recent studies at another subalpine forest
in Colorado have shown that fine-scale spatial variability in topography can have significant species-specific
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2336
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
effects on tree growth patterns as well as water-limiting conditions, mediating the physiological response of
trees to moisture in variable landscape positions [Adams et al., 2014]. Nonetheless, further examination into
the physiological mechanisms driving this spatial variability across topographic gradients is required.
4.3. Model Implementation and Isotopic Mixing Relationship Considerations
The potential for seasonal variability in δP at the landscape scale was observed after coupling the emergent multiple linear regression model (Figure 5b) with an ecohydrological model, previously developed
for this site [Riveros-Iregui et al., 2011a]. Although the multiple linear regression model only explains 33%
of the variation in δP, our results suggest systematic spatial and temporal differences in δP, mainly
mediated by changes in soil water content throughout the growing season (Figure 6). To the best of
our knowledge our study is the first to show the extent of this isotopic variability. By examining the seasonal
dynamics of modeled δP on a 5 m grid cell from June 28 to August 23, we can see that a substantial portion of the
landscape exhibits a decrease in δP across the landscape. At the landscape level, the modeled δP started out
more uniform but it became more spatially complex, as ridges dried out and riparian areas stayed productive.
This trend is likely because in riparian areas plants achieve optimal photosynthetic conditions in response to
higher summer temperatures, whereas toward the ridges plants respond to lower soil moisture conditions.
Further studies are needed to fully examine the linkage between plant photosynthesis and δP in complex terrain.
An important outcome of this study is that appropriate methodologies need to be used for interpreting
isotopic differences with confidence. In shallow soils, the δJ estimate is quite sensitive to possible errors
in isotopic signature or concentration, and this would be expected as a direct consequence of the steep
concentration and isotopic gradients at the surface-atmosphere interface that are being interpreted
when solving for δJ. Though the Keeling approach has been criticized for its lack of accommodation
for nonsteady-state diffusion processes [Nickerson and Risk, 2009], it should be noted that δJ and δP
are both steady state solutions requiring the assumption of equilibrium. From this study, however, the
Keeling approach represented a more certain technique for interpreting near-surface differences
because concentration-isotopic data across the depth profile will buffer against slope (and intercept)
change resulting from error at any one depth (Figure 4). The δJ is advantageous in that it requires measurements at only one soil depth combined with atmospheric background air, meaning that it also may
be useful to estimate different signatures of production or as a diagnostic depth-specific tool to evaluate
changes in the signature of production or the degree of non-steady-state activity through depth.
However, we suggest that δJ estimates should be interpreted with caution near the surface-atmosphere
interface. The inherent sensitivity to error does not systematically bias δJ estimates or make them wrong,
but simply more prone to mislead, especially where errors in isotopic measurements or concentration
could exist. Because we found that the results of the δP and δJ approaches were quite similar at 20
and especially at 50 cm depth, we abandoned the use of the δJ for interpreting patterns in the shallow
profile for extrapolation across the landscape. This gives us more confidence that the observed landscape patterns are real and not an artifact of methodology.
It is also important to note that the Davidson [1995] model assumes a homogeneous, uniformly diffusive soil
column whereas the Keeling approach is depth-averaged. Combining both methods could serve as a proxy
for near-surface advection effects and aid in the characterization of the environmental conditions under
which measurements were taken. When estimates of δJ do not exhibit isotopic depletion near the soil surface,
it signals that the soil transport regime is indeed diffusive. Future modeling studies should further investigate
observed differences between δJ and δP near the surface and the impact on landscape estimates of isotopic
exchange, particularly in areas with variable soil moisture.
5. Conclusions
Our results suggest that the spatiotemporal variability of δS and δP is influenced by soil water content
and soil respiration in complex terrain, and as such its spatiotemporal variability is the combined result
of the seasonal change in CO2 production (including autotrophic and heterotrophic sources) and transport (including moisture-driven effects on soil gas diffusivity). Areas with low soil respiration rates had
less negative δP, which is consistent with our understanding of the physical effects of diffusive mixing
of soil gas with overlain atmospheric CO2 [Cerling, 1984]. These conclusions should cause some caution
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2337
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
to studies directed to estimate the dynamics of δS or δP at the landscape level, especially in areas of variable soil moisture. Furthermore, our results suggest that the parameterization of soil moisture variability
in space and time is needed to accurately predict and model the δ13C of soil CO2 flux in heterogeneous
landscapes and areas of pronounced topography.
Acknowledgments
The soil CO2 isotope data and computer
code used in this manuscript are available for collaborative use in the supporting information. Please contact the
corresponding author, Diego RiverosIregui ([email protected]), prior to use.
This work was supported by the USDA
award 2012-67019-19360 and the NSFNebraska EPSCoR First Award Program
to D. R.-I. D. R.-I. acknowledges support
from a Visiting Faculty Fellowship at
National Center for Atmospheric
Research while writing this manuscript.
We appreciate logistic collaboration
from the Tenderfoot Creek
Experimental Forest and the USDA,
Forest Service, Rocky Mountain
Research Station, including Robert
Keane and Elaine Sutherland, and field
assistance from William A. Avery, Kendra
Kaiser, Erin C. Seybold, and Timothy P.
Covino. Two anonymous reviewers and
the associate editor provided valuable
suggestions for the improvement of this
manuscript.
LIANG ET AL.
References
Adams, H. R., H. R. Barnard, and A. K. Loomis (2014), Topography alters tree growth–climate relationships in a semi-arid forested catchment,
Ecosphere, 5(11), art148, doi:10.1890/ES14-00296.1.
Alstad, K. P., C.-T. Lai, L. B. Flanagan, and J. R. Ehleringer (2007), Environmental controls on the carbon isotope composition of ecosystemrespired CO2 in contrasting forest ecosystems in Canada and the USA, Tree Physiol., 27(10), 1361–1374.
Bartlett, M. S. (1937), Properties of sufficiency and statistical tests, Proc. R. Soc. Lond. A Math. Phys. Sci., 160(901), 268–282, doi:10.2307/96803.
Bowling, D. R., and W. J. Massman (2011), Persistent wind-induced enhancement of diffusive CO2 transport in a mountain forest snowpack,
J. Geophys. Res., 116, G04006, doi:10.1029/2011JG001722.
Bowling, D. R., W. J. Massman, S. M. Schaeffer, S. P. Burns, R. K. Monson, and M. W. Williams (2009), Biological and physical influences on the
carbon isotope content of CO2 in a subalpine forest snowpack, Niwot Ridge, Colorado, Biogeochemistry, 95, 37–59.
Bowling, D. R., J. E. Egan, S. J. Hall, and D. A. Risk (2015), Environmental forcing does not induce diel or synoptic variation in carbon isotope
content of forest soil respiration, Biogeosciences, 12, 5143–5160, doi:10.5194/bg-12-5143-2015.
Bowling, D., N. McDowell, B. Bond, B. Law, and J. Ehleringer (2002), 13C content of ecosystem respiration is linked to precipitation and vapor
pressure deficit, Oecologia, 131(1), 113–124.
Buchmann, N., W.-Y. Kao, and J. R. Ehleringer (1996), Carbon dioxide concentrations within forest canopies-variation with time, stand
structure, and vegetation type, Glob. Chang. Biol., 2(5), 421–432, doi:10.1111/j.1365-2486.1996.tb00092.x.
Cerling, T. E. (1984), The stable isotopic composition of modern soil carbonate and its relationship to climate, Earth Planet. Sci. Lett., 71(2),
229–240, doi:10.1016/0012-821X(84)90089-X.
Cerling, T. E. (1991), Carbon dioxide in the atmosphere; evidence from Cenozoic and Mesozoic Paleosols, Am. J. Sci., 291(4), 377–400,
doi:10.2475/ajs.291.4.377.
Cerling, T. E., D. K. Solomon, J. Quade, and J. R. Bowman (1991), On the isotopic composition of carbon in soil carbon dioxide, Geochim.
Cosmochim. Acta, 55(11), 3403–3405.
Davidson, G. R. (1995), The stable isotopic composition and measurement of carbon in soil CO2, Geochim. Cosmochim. Acta, 59(12),
2485–2489.
Efron, B. (1979), Bootstrap methods: Another look at the Jackknife, Ann. Stat., 7, 1–26.
Ehleringer, J. R., N. Buchmann, and L. B. Flanagan (2000), Carbon isotope ratios in belowground carbon cycle processes, Ecol. Appl., 10(2),
412–422.
13
Ekblad, A., and P. Högberg (2001), Natural abundance of C in CO2 respired from forest soils reveals speed of link between tree
photosynthesis and root respiration, Oecologia, 127, 305–308.
13
Ekblad, A., B. Boström, A. Holm, and D. Comstedt (2005), Forest soil respiration rate and δ C is regulated by recent above ground weather
conditions, Oecologia, 143(1), 136–42, doi:10.1007/s00442-004-1776-z.
Emanuel, R. E., D. A. Riveros-Iregui, B. L. McGlynn, and H. E. Epstein (2011), On the Spatial Heterogeneity of Net Ecosystem Production in
Complex Landscapes, Ecosphere, 2(7), art86, doi:10.1890/ES11-00074.1.
Emanuel, R. R. E., H. E. H. Epstein, B. L. B. McGlynn, D. L. Welsch, D. J. Muth, and P. D'Odorico (2010), Spatial and temporal controls on
watershed ecohydrology in the northern Rocky Mountains, Water Resour. Res., 46, W11553, doi:10.1029/2009WR008890.
Farnes, P. E., R. C. Shearer, W. W. McCaughey, and K. J. Hansen (1995), Comparisons of hydrology, geology, and physical characteristics
between Tenderfoot Creek Experimental Forest (East Side) Montana, and Coram Experimental Forest (West Side) Montana. Final Report
RJVA-INT-92734, Bozeman, Mont.
Farquhar, G. D., M. H. O'Leary, and J. Berry (1982), On the relationship between carbon isotope discrimination and the intercellular carbon
dioxide concentration in leaves, Aust. J. Plant Physiol., 9, 121–137.
Farquhar, G. D., J. R. Ehleringer, and K. T. Hubick (1989), Carbon isotope discrimination and photosynthesis, Annu. Rev. Plant. Physiol. Plant.
Mol. Biol., 40(1), 503–537.
13
Fessenden, J. E., and J. R. Ehleringer (2003), Temporal variation in δ C of ecosystem respiration in the Pacific Northwest: Links to moisture
stress, Oecologia, 136(1), 129–36, doi:10.1007/s00442-003-1260-1.
Flanagan, L. B., J. R. Brooks, G. T. Varney, S. C. Berry, and J. R. Ehleringer (1996), Carbon isotope discrimination during photosynthesis and the
isotope ratio of respired CO2 in boreal forest ecosystems, Global Biogeochem. Cycles, 10(4), 629–640, doi:10.1029/96GB02345.
Gifford, R. M. (2003), Plant respiration in productivity models: Conceptualisation, representation and issues for global terrestrial carbon-cycle
research, Funct. Plant Biol., 30(2), 171–186, doi:10.1071/FP02083.
Högberg, P., A. Nordgren, N. Buchmann, A. F. Taylor, A. Ekblad, M. N. Högberg, G. Nyberg, M. Ottosson-Löfvenius, and D. J. Read (2001), Largescale forest girdling shows that current photosynthesis drives soil respiration, Nature, 411(6839), 789–92, doi:10.1038/35081058.
Hu, J., D. J. P. Moore, D. A. Riveros-Iregui, S. P. Burns, and R. K. Monson (2010), Modeling whole-tree carbon assimilation rate using observed
transpiration rates and needle sugar carbon isotope ratios, New Phytol., 185(4), 1000–1015, doi:10.1111/j.1469-8137.2009.03154.x.
Kayler, Z. E., E. W. Sulzman, W. D. Rugh, A. C. Mix, and B. J. Bond (2010), Characterizing the impact of diffusive and advective soil gas
transport on the measurement and interpretation of the isotopic signal of soil respiration, Soil Biol. Biochem., 42(3), 435–444,
doi:10.1016/j.soilbio.2009.11.022.
Keeling, C. D. (1958), The concentration and isotopic abundances of atmospheric carbon dioxide in rural areas, Geochim. Cosmochim. Acta,
13(4), 322–334, doi:10.1016/0016-7037(58)90033-4.
Keeling, C. D. (1961), The concentration and isotopic abundances of carbon dioxide in rural and marine air, Geochim. Cosmochim. Acta,
24(3-4), 277–298, doi:10.1016/0016-7037(61)90023-0.
13
Knohl, A., R. Werner, W. Brand, and N. Buchmann (2005), Short-term variations in δ C of ecosystem respiration reveals link between
assimilation and respiration in a deciduous forest, Oecologia, 142(1), 70–82.
Korkalainen, T., and A. Laurén (2006), Using phytogeomorphology, cartography and GIS to explain forest site productivity expressed as tree
height in southern and central Finland, Geomorphology, 74(1–4), 271–284, doi:10.1016/j.geomorph.2005.09.001.
VARIABILITY OF THE δ13C IN SOIL CO2
2338
Journal of Geophysical Research: Biogeosciences
10.1002/2015JG003193
13
Lai, C. T., J. R. Ehleringer, A. J. Schauer, P. P. Tans, D. Y. Hollinger, K. T. Paw U, J. W. Munger, and S. C. Wofsy (2005), Canopy-scale δ C of
photosynthetic and respiratory CO2 fluxes: Observations in forest biomes across the United States, Glob. Chang. Biol., 11(4), 633–643.
Liang, L., D. A. Riveros-Iregui, R. E. Emanuel, and B. L. McGlynn (2014), A simple framework to estimate distributed soil temperature from discrete
air temperature measurements in data-scarce regions, J. Geophys. Res. Atmos., 119, 407–417, doi:10.1002/2013JD020597.Received.
Liu, W., and S. Lin (2009), Construction of exact simultaneous confidence bands in multiple linear regression with predictor variables constrained in an ellipsoidal region, Statistica Sinica, 19, 213–232.
Lloyd, J., and G. D. Farquhar (1994), 13C discrimination during CO2 assimilation by the terrestrial biosphere, Oecologia, 99(3–4), 201–215,
doi:10.1007/BF00627732.
Lundquist, J. D., N. Pepin, and C. Rochford (2008), Automated algorithm for mapping regions of cold-air pooling in complex terrain,
J. Geophys. Res., 113, D22107, doi:10.1029/2008JD009879.
Maurer, G. E., A. M. Chan, N. A. Trahan, D. J. Moore, and D. R. Bowling (2016), Carbon isotopic composition of forest soil respiration in the
decade following bark beetle and stem girdling disturbances in the Rocky Mountains, Plant Cell Environ., 39(7), 1513–1523, doi:10.1111/
pce.12716.
Medina, E., L. Sternberg, and E. Cuevas (1991), Vertical stratification of δ13C values in closed natural and plantation forests in the Luquillo
mountains, Puerto Rico, Oecologia, 87(3), 369–372, doi:10.1007/BF00634593.
Mincemoyer, S. A., and J. L. Birdsall (2006), Vascular flora of the tenderfoot creek experimental forest, little belt mountains, Montana,
Madroño, 53(3), 211–222.
Moyes, A. B., S. J. Gaines, R. T. W. Siegwolf, and D. R. Bowling (2010), Diffusive fractionation complicates isotopic partitioning of autotrophic
and heterotrophic sources of soil respiration, Plant Cell Environ., 33(11), 1804–19, doi:10.1111/j.1365-3040.2010.02185.x.
Nickerson, N., and D. Risk (2009), Keeling plots are non-linear in non-steady state diffusive environments, Geophys. Res. Lett., 36, L08401,
doi:10.1029/2008GL036945.
Ogle, K., and E. Pendall (2015), Isotope partitioning of soil respiration: A Bayesian solution to accommodate multiple sources of variability,
J. Geophys. Res. Biogeosci., 120, 221–236, doi:10.1002/2014JG002794.
Pacific, V., B. McGlynn, D. Riveros-Iregui, D. Welsch, and H. Epstein (2008), Variability in soil respiration across riparian-hillslope transitions,
Biogeochemistry, 91(1), 51–70, doi:10.1007/s10533-008-9258-8.
Pataki, D. E., J. R. Ehleringer, L. B. Flanagan, D. Yakir, D. R. Bowling, C. J. Still, N. Buchmann, J. O. Kaplan, and J. A. Berry (2003), The application
and interpretation of Keeling plots in terrestrial carbon cycle research, Global Biogeochem. Cycles, 17(1, 1022), doi:10.1029/2001GB001850.
Riveros-Iregui, D. A., and B. L. McGlynn (2009), Landscape structure control on soil CO2 efflux variability in complex terrain: Scaling from point
observations to watershed scale fluxes, J. Geophys. Res., 114, G02010, doi:10.1029/2008JG000885.
Riveros-Iregui, D. A., R. E. Emanuel, D. J. Muth, B. L. McGlynn, H. E. Epstein, D. L. Welsch, V. J. Pacific, and J. M. Wraith (2007), Diurnal hysteresis
between soil CO2 and soil temperature is controlled by soil water content, Geophys. Res. Lett., 34, L17404, doi:10.1029/2007GL030938.
Riveros-Iregui, D. A., B. L. McGlynn, H. E. Epstein, and D. L. Welsch (2008), Interpretation and Evaluation of Combined Measurement
Techniques for Soil CO2 Efflux: Discrete Surface Chambers and Continuous Soil CO2 Concentration Probes, J. Geophys. Res. Biogeosci., 113,
G04027, doi:10.1029/2008JG000811.
Riveros-Iregui, D. A., B. L. McGlynn, L. A. Marshall, D. L. Welsch, R. E. Emanuel, and H. E. Epstein (2011a), A watershed scale assessment of a
process soil CO2 production and efflux model, Water Resour. Res., 47, W00J04, doi:10.1029/2010WR009941.
Riveros-Iregui, D. A., J. Hu, S. P. Burns, D. R. Bowling, and R. K. Monson (2011b), An interannual assessment of the relationship between the
stable carbon isotopic composition of ecosystem respiration and climate in a high-elevation subalpine forest, J. Geophys. Res., 116,
G02005, doi:10.1029/2010JG001556.
Riveros-Iregui, D. A., B. L. McGlynn, R. E. Emanuel, and H. E. Epstein (2012), Complex terrain leads to bidirectional responses of soil respiration
to inter-annual water availability, Glob. Chang. Biol., 18(2), 749–756.
Royston, J. P. (1982), An extension of Shapiro and Wilk's W test for normality to large samples, J. R. Stat. Soc. Ser. C (Applied Stat., 31(2),
115–124, doi:10.2307/2347973.
Schaeffer, S. M., D. E. Anderson, S. P. Burns, R. K. Monson, J. Sun, and D. R. Bowling (2008a), Canopy structure and atmospheric flows in relation to
the δ13C of respired CO2 in a subalpine coniferous forest, Agric. For. Meteorol., 148(4), 592–605, doi:10.1016/j.agrformet.2007.11.003.
Schaeffer, S. M., J. B. Miller, B. H. Vaughn, J. W. C. White, and D. R. Bowling (2008b), Long-term field performance of a tunable diode
laser absorption spectrometer for analysis of carbon isotopes of CO2 in forest air, Atmos. Chem. Phys., 8(17), 5263–5277, doi:10.5194/
acp-8-5263-2008.
Schleser, G. H., and R. Jayasekera (1985), δ13C-variations of leaves in forests as an indication of reassimilated CO2 from the soil, Oecologia,
65(4), 536–542, doi:10.1007/BF00379669.
Shapiro, S. S., and M. B. Wilk (1965), An analysis of variance test for normality (complete samples), Biometrika, 52(3/4), 591–611, doi:10.2307/
2333709.
Suwa, M., G. G. Katul, R. Oren, J. Andrews, J. Pippen, A. Mace, and W. H. Schlesinger (2004), Impact of elevated atmospheric CO2 on forest floor
respiration in a temperate pine forest, Global Biogeochem. Cycles, 18, GB2013, doi:10.1029/2003GB002182.
Tucker, C. L., J. M. Young, D. G. Williams, and K. Ogle (2014), Process-based isotope partitioning of winter soil respiration in a subalpine
ecosystem reveals importance of rhizospheric respiration, Biogeochemistry, 121(2), 389–408, doi:10.1007/s10533-014-0008-9.
Vogel, F. R., L. Huang, D. Ernst, L. Giroux, S. Racki, and D. E. J. Worthy (2012), Evaluation of a cavity ring-down spectrometer for in-situ
13
observations of δ CO2, Atmos. Meas. Tech. Discuss., 5(4), 6037–6058.
Vogel, J. C. (1978), Recycling of carbon in a forest environment, Oecologia Plant., 13(1), 89–94.
Webster, K. L., I. F. Creed, F. D. Beall, and R. A. Bourbonnière (2008), Sensitivity of catchment-aggregated estimates of soil carbon dioxide
efflux to topography under different climatic conditions, J. Geophys. Res., 113, G03040, doi:10.1029/2008JG000707.
Zobitz, J. M., J. P. Keener, H. Schnyder, and D. R. Bowling (2006), Sensitivity analysis and quantification of uncertainty for isotopic mixing
relationships in carbon cycle research, Agric. For. Meteorol., 136(1–2), 56–75.
LIANG ET AL.
VARIABILITY OF THE δ13C IN SOIL CO2
2339