Biological and physical influences on soil 14CO2 seasonal dynamics in a temperate hardwood forest

While radiocarbon ( 14C) abundances in standing stocks of soil carbon have been used to evaluate rates of soil carbon turnover on timescales of several years to centuries, soil-respiredCO2 measurements are an important tool for identifying more immediate responses to disturbance and climate change. Soil 1CO2 data, however, are often temporally sparse and could be interpreted better with more context for typical seasonal ranges and trends. We report on a semi-high-frequency sampling campaign to distinguish physical and biological drivers of soil 1CO2 at a temperate forest site in northern Wisconsin, USA. We sampled CO2 profiles every three weeks during snow-free months through 2012 in three intact plots and one trenched plot that excluded roots. Respired 1CO2 declined through the summer in intact plots, shifting from an older C composition that contained more bomb 14C to a younger composition more closely resembling present 14C levels in the atmosphere. In the trenched plot, respired 1CO2 was variable but remained comparatively higher than in intact plots, reflecting older bomb-enriched 14C sources. Although respired 1CO2 from intact plots correlated with soil moisture, related analyses did not support a clear cause-and-effect relationship with moisture. The initial decrease in 1CO2 from spring to midsummer could be explained by increases in 14Cdeplete root respiration; however, 1CO2 continued to decline in late summer after root activity decreased. We also investigated whether soil moisture impacted vertical partitioning of CO2 production, but found this had little effect on respired1CO2 because CO2 contained modern bomb C at depth, even in the trenched plot. This surprising result contrasted with decades to centuries-old pre-bomb CO 2 produced in lab incubations of the same soils. Our results suggest that root-derived C and other recent C sources had dominant impacts on respired 1CO2 in situ, even at depth. We propose that 1CO2 may have declined through late summer in intact plots because of continued microbial turnover of root-derived C, following declines in root respiration. Our results agree with other studies showing declines in the 14C content of soil respiration over the growing season, and suggest inputs of new photosynthates through roots are an important driver.


Introduction
The presence of large 14 C gradients in soil makes 14 C a potentially sensitive tool for detecting changes in respiration sources.The dynamic range of 14 C in putative respiratory substrates is often many times larger than for δ 13 C: deep soils generally contain an abundance of organic matter that is deplete in 14 C due to radioactive decay and the older age of deep carbon, while near-surface soils reflect litter additions containing "bomb C", a legacy of aboveground thermonuclear weapons testing in the early 1960s (Gaudinski et al., 2000;Trumbore, 2000).Root and microbial respiration also often have different 14 C abundance, with root-derived CO 2 more closely resembling the recent atmosphere.This distinction has been employed to partition total soil respiration into heterotrophic (R h ) and autotrophic (R a ) components (Czimczik et al., 2006;Hahn et al., 2006;Hicks Pries et al., 2013; C. L. Phillips et al.: Biological and physical influences on soil 14 CO 2 seasonal dynamics Schuur and Trumbore, 2006).While the distinctions between deep and shallow, and between R h and R a end-members are useful for partitioning, the large 14 C range in potential CO 2 sources may also accentuate seasonal and synoptic variability in soil 14 CO 2 .Although 14 C measurements have proven useful for identifying changes in respiratory sources following disturbance and climatic change (Czimczik et al., 2006;Hicks Pries et al., 2013;Hirsch et al., 2003;Schuur and Trumbore, 2006), our understanding of these effects could be improved with more information on 14 CO 2 seasonal trends.
Several temporal studies have suggested that seasonal variation in soil-respired 14 CO 2 may be large, and may therefore encode information about seasonal dynamics of respiratory sources.Gaudinski et al. (2000) found soil-respired 14 CO 2 decreased by approximately 40 ‰ between May and December at Harvard Forest, a temperate deciduous system.Similarly, ecosystem-respired 14 CO 2 at a tundra site in Alaska decreased over the summer by as much as 20‰ (Hicks Pries et al., 2013).Schuur and Trumbore (2006), however, found a large increase of 84 ‰ between June and August at a boreal forest site in Alaska.Unfortunately, temporal density in data sets with repeated sampling is generally very sparse, providing little information from which to fully describe seasonal variability or identify environmental drivers.
To help address this gap, in 2011-2012 we conducted a study of respired 14 CO 2 dynamics at Willow Creek eddy covariance site, a temperate semi-deciduous forest in northern Wisconsin, USA.Our goal was to examine soil 14 CO 2 dynamics through the growing season, and evaluate whether soil emissions also influenced atmospheric 14 CO 2 dynamics.In this paper, we present our soil 14 CO 2 observations and evaluate potential physical and biological processes underlying seasonal variation.Specifically, we evaluated impacts on soil 14 CO 2 from the following processes: 1. Seasonal shifts in relative contributions of R h and R a .
2. Seasonal changes in relative contributions of deep and shallow CO 2 production.
3. Seasonal changes in 14 C of R h , reflecting shifts in microbial substrates.
Although not an exhaustive list, by focusing on these processes we hoped to tease apart the relative influences of plant activity, microbial activity, and soil physical properties on respired 14 CO 2 variability.
Investigating influences from these sources may help illuminate the utility and limitations of 14 CO 2 for understanding soil metabolism.To our knowledge there has been no previous investigation of whether 14 CO 2 of R h varies seasonally, and R h has been assumed to be isotopically static at seasonal to interannual timescales for partitioning heterotrophic and autotrophic respiration (Hicks Pries et al., 2013;Schuur and Trumbore, 2006) and for modeling rates of soil organic matter turnover (Torn et al., 2002).If heterotrophic 14 C varies seasonally, this would indicate that the quality of soil C destabilized through time has greater environmental sensitivity than is presently represented by most soil biogeochemistry models.The effects of soil moisture and gas diffusion on respired 14 CO 2 are also largely unexplored.Although soil moisture and gas diffusion can play roles in regulating deep versus shallow CO 2 production (Davidson et al., 2006;Phillips et al., 2012), gas diffusion is often neglected in favor of biological explanations for why sources of soil respiration vary through time.A simultaneous assessment of the relative influences on 14 CO 2 by soil physical factors in addition to plant and microbial activity provides a check on existing assumptions and tendencies.

Methods
To evaluate influences of plant and microbial activity and soil physical factors, we measured surface CO 2 flux rates and subsurface profiles of CO 2 , 14 CO 2 , and δ 13 CO 2 in three intact soil plots and one plot that was trenched to exclude roots to 1 m depth.The trenched plot did not have spatial replication; therefore, a limitation of this study is that the treatments could not be statistically compared.Observations from the trenched plot, however, allowed us to examine in situ dynamics of microbially respired 14 CO 2 through time, in the absence of live roots, which we compared with more common in vitro microbial respiration measurements from laboratory soil incubations.We used comparisons of the intact and trenched plots to estimate the relative contributions of R h and R a to total soil respiration.Subsurface profile measurements were used to estimate CO 2 and 14 C contributions from each soil horizon.
In addition, we employed a one-dimensional (1-D) soil CO 2 diffusive transport model to simulate how variations in the rate and isotopic composition of CO 2 production would be expected to impact 14 CO 2 of soil air and surface flux.We used simulations as a second, independent approach for estimating 14 CO 2 of microbial production from observations of soil air.

Site and soil description
The Willow Creek Ameriflux site is located in the Chequamegon National Forest of north central Wisconsin (45 • 48 W, 90 • 07 N), and is composed of mature, second growth hardwood trees approximately 80-100 yr old, dominated by sugar maple, basswood, and green ash (Acer saccharum Marshall, Tilia Americana L., Fraxinus pennsylvanica Marshall).Eddy covariance measurements have been made at the site since 1998, and plant and soil characteristics have been described in detail by others (Bolstad et al., 2004;Cook et al., 2004;Martin and Bolstad, 2005).In June 2011 we established a group of four soil plots centered about 30 m from the base of the eddy covariance tower (Fig. 1).In each plot we excavated a trench to 75 cm depth to characterize the profile and install instrumentation, removing soil in 10 cm increments to back-fill in the same order.Soils were deep and moderately permeable, formed from unsorted, coarse glacial till, and have evidence of mixing from windthrow, freeze-thaw, and earthworm activity.Texture in the four plots was classified as either sandy loams or loamy sands (mean texture in top 20 cm: 63 % sand, 31 % silt, 6 % clay, 5-12 % rock fragments).Soils lacked an O horizon, had an A horizon 8-12 cm in depth with a clear wavy boundary, followed by at least one B horizon, with variation among plots in iron depletions and accumulations, and finally a BC horizon starting at 50-60 cm with increased amounts of gravelly sand and gravel.We later found gas wells at and below 50 cm to be poorly drained until mid-summer.
We installed gas wells at 6 depths, at the interfaces between genetic horizons and several intermediate depths (nominal depths were 8, 15, 22, 30, 50, and 70 cm, with ≤ 3 cm variation across plots).We used a 2.5 cm diameter drill auger to create horizontal holes in the profile wall extending in 70-100 cm as permitted by stone content, and pounded gas wells into the holes.The wells were constructed of PVC pipe (70 to 100 cm long × 3 cm ID; inner volume 0.5 to 0.7 L), which were perforated along the bottom with a row of 1 cm diameter holes to exchange air with the surrounding soil, and wrapped in Tyvek ® polyethylene membrane to exclude water and soil macrofauna.Wells were staggered horizontally within a 15 cm range to reduce impacts on vertical CO 2 diffusion.Gas wells were capped at both ends, connected to the soil surface with two lengths of 1/8 polyethylene tubing, and the tubes were capped at the soil surface with plastic 2-way valves, which were housed in plastic enclosures.Thermistors were placed adjacent to each gas well to measure soil temperature (CS-107B, Campbell Scientific, Logan, Utah, USA), and TDR soil moisture probes were placed horizontally at 4 and 18 cm (CS-616, Campbell Scientific).Two sets of soil cores (5 cm diameter × 5 cm long) centered at 2. 5, 7.5, 12.5, 18, 30, 40, and 60 cm were also removed from each exposed profile for isotopic analysis (see below), and for analysis of texture, porosity, and moisture release at the Oregon State University Soil Science Physical Characterization Lab.
To create the trenched plot, we dug a trench 30 cm wide × 100 cm deep around all sides of a 2 m × 2 m plot, and lined the trench with 0.13 mm thick polyethylene vapor barrier to prevent in-growth of new roots before refilling the trench with soil.Trenching was completed in early September 2011.The plot did not contain any woody plants, and emerging herbaceous plants (mostly grass) were clipped to their root crowns throughout 2012.

Soil CO 2 flux and profile air
Soil surface CO 2 flux was measured using forced diffusion (FD) chambers and Vaisala GMP343 CO 2 sensors (Vaisala Corp, Helsinki, Finland), as described by Risk et al. (2011).
Each soil plot contained a FD soil chamber and atmospheric reference, and a co-located PVC soil collar for comparisons with the Licor-8100 soil flux system (Licor Environmental, Lincoln, NE, USA).FD CO 2 flux, temperature, and moisture were recorded hourly, and Licor CO 2 flux comparisons were made approximately every 3 weeks during the growing season.Soil profile CO 2 was measured with the Licor-8100 IRGA, by first circulating air through a soda-lime trap to remove CO 2 from the Licor internal volume and tubing, and then switching valves to shut off the CO 2 trap and circulate soil air between the gas well and Licor.Soil air was circulated in a closed loop for several minutes until concentrations stabilized.A 1 µm air filter and a 50 mL canister of Drierite ™ plumbed to the Licor inlet trapped particles and moisture from incoming soil air.The gas well tubing was also prepurged by removing and discarding 50 mL of air with a syringe before connecting the tubing to the Licor.
After measuring CO 2 , we sampled soil air for isotopic analysis using pre-evacuated 400 mL stainless steel canisters (Restek Corp #24188, PA, USA) or activated molecular sieve traps (Gaudinski et al., 2000).To prepare canisters, we pre-cleaned them with N 2 and heat following the manufacturer's instructions, evacuated them to ≤ 1 mTorr, and capped the valves with rubber septa prior to overnight shipping to the field site.In the field, we connected a syringe needle to the gas well tubing and filled the canisters by piercing the septa.To sample with molecular sieve traps, we used the Licor to pull soil air through the trap in a flow-through configuration.During trapping, we maintained a flow rate of 60 mL min −1 , and timed trapping to collect 2 mg C (total trapping time ranged 30 s to 15 min, depending on concentration).The molecular sieve (13X 8/12 beads, Grace) was washed, and then pre-conditioned by baking at 750 • C under vacuum for 12 h.Molecular sieve traps were activated using the same procedure for extraction, below.
C. L. Phillips et al.: Biological and physical influences on soil 14 CO 2 seasonal dynamics Atmospheric samples from the eddy covariance tower were also sampled from just above the forest canopy at 21 m above ground level into glass flasks, using a programmable flask package and compressor (Andrews et al., 2013).These whole-air samples were collected approximately every 6 days at 12:30 a.m.local time, so that they reflected respiration not influenced by photosynthesis.

Root and soil incubations
We collected roots from 0-5 cm in three locations in August 2011 to determine the 14 C of R a .In the field, roots were rinsed in distilled water and placed in sterilized Mason jars.Atmospheric CO 2 was removed from the jar headspace by recirculating air through a soda lime trap and IRGA.The jars were shipped overnight to the Center for Accelerator Mass Spectrometry (CAMS) at Lawrence Livermore National Laboratory, and CO 2 was extracted within 48 h, as described below.
Soils were incubated to compare laboratory measurements of R h with observations from the trenched plot.Soil cores were sampled from each plot during well installation, and shipped on ice to CAMS.We removed the majority of roots by hand-picking, and allowed the remainder to senesce by resting the soils for two weeks before sealing the incubation jars.The closed jars were purged with CO 2 -free air, and incubated at 25 • C until at least 0.5 mg C-CO 2 could be extracted from the headspace.Incubation time ranged from 4 to 126 days, depending on the activity of each sample.

14 C sample processing
CO 2 from canisters, flasks, and incubation jars was purified cryogenically at CAMS using a vacuum line, and CO 2 trapped on molecular sieves was released by baking at 650 • C under vacuum for 30 min while condensing CO 2 cryogenically.Purified CO 2 was reduced to graphite on iron powder in the presence of H 2 (Vogel et al., 1984).Subsamples of CO 2 were analyzed for δ 13 C at the UC Davis Stable Isotope Laboratory (GVI Optima Stable Isotope Ratio Mass Spectrometer), and were used to correct 14 C values for mass-dependent fractionation.
Radiocarbon abundance in graphitized samples was measured on the Van de Graff FN Accelerator Mass Spectrometer (AMS) at CAMS, is reported in 14 C notation with a correction for 14 C decay since 1950 (Stuiver and Polach, 1977).In 14 C notation, values > 0 ‰ indicate the presence of "bomb" C that was fixed after 1950, whereas values ≤ 0 ‰ indicate C that was fixed prior to 1950.AMS samples had an average precision of 2.5 ‰.Total uncertainty associated with AMS plus sampling and CO 2 extraction was estimated to be 8.7 ‰ for molecular sieve traps, and 3.2 ‰ for air canisters, based on the standard deviation of contemporary atmosphere process standards (N = 5 for each sample type).

Data analysis
The analysis of field data had three components: (1) calculating 14 CO 2 of surface flux from profile measurements, (2) estimating CO 2 and 14 C production by soil horizon, and (3) partitioning total soil respiration into R h and R a .Each component is discussed below.

Surface flux 14 CO 2
Due to recent reports of isotopic disequilibria caused by surface chambers (Albanito et al., 2012;Midwood and Millard, 2011;Nickerson and Risk, 2009a), for this study we focused on profile measurements, which may be less prone to sampling artifacts.We estimated 14 C of surface flux from profile measurements using a gradient approach following Nickerson et al. (2013).The gradient approach is often used to calculate surface CO 2 flux from subsurface concentrations by applying Fick's first law of diffusion: where F is the CO 2 flux density (µmol m −2 s −1 ), D(z) is the soil CO 2 diffusivity (m 2 s −1 ) at depth z (m), and C is the CO 2 concentration (µmol m −3 ).As described by Nickerson et al. (2013), if we assume the isotopologues of CO 2 ( 12 CO 2 , 13 CO 2 , and 14 CO 2 ) diffuse independently of one another, we can use Eq. ( 1) to model fluxes of each.The isotopic ratio of 14 C to 12 C in surface flux can thus be modeled as the quotient of Eq. ( 1) applied to 14 CO 2 and 12 CO 2 : where F 14 and F 12 are the fluxes of 14 CO 2 and 12 CO 2 , respectively, and D 14 (z) and D 12 (z) are the depth-specific diffusivities for each isotopologue.The quotient of diffusion coefficients for a rare and common isotope is also the inverse of the fractionation factor, α, which is 1.0044 for 13 CO 2 diffusion through soil (Cerling et al., 1991), and is estimated to be approximately 1.0088 for 14 CO 2 (Southon, 2011).Using this relationship, we can simplify and discretize Eq. ( 2) to yield: where α 14 is the fractionation factor for 14 C, and z 1 and z 2 are arbitrary depths with increasing CO 2 concentration.Similarly, the 13 C / 12 C ratio in surface flux can be calculated by replacing 14 C with 13 C values.Note that Eq. ( 4) indicates the isotopic ratio of surface flux can be calculated without knowing the diffusivity of CO 2 in soil which is difficult to measure well and uncertain to model (Pingintha et al., 2010).
To convert between values (for reporting purposes) and absolute 14 C / 12 C ratios (for flux calculations), we used the following equations: where notation (‰) is calculated by standardizing fraction modern (FM) to the year 1950 to allow inter-comparison of samples from different analysis years (Yr), and 8267 yr is the 14 C mean decay rate.FM was related to the sample 14 C / 12 C ratio following the derivation in Southon et al. (2011), where it is shown that 14 C activity ≈ 14 C / 12 C.
In the equation above [ 14 C / 12 C] S is the sample 14 C ratio, δ 13 C is the sample 13 C abundance in ‰ notation, which is used to normalize the 14 C ratio for mass-based fractionation to δ 13 C = −25 ‰, and 0.95 • [ 14 C / 12 C] OX1 is the normalized 14 C ratio of the oxalic acid I standard.We calculated the 13 C and 14 C composition of surface fluxes at Willow Creek using Eq. ( 3) with data from the soil surface (z 1 = 0 cm) and the shallowest gas wells (z 2 = 7 or 8 cm).On two sampling dates, however, there were missing observations in plot 4 at the 7 cm depth, and we instead used data from gas wells at 14 cm.To assess errors from this gap-filling approach, we compared flux calculations for days when both the shallowest well and next depth were available (N = 28) and found the gap-filling approach caused a small positive bias in estimated surface flux (mean difference in 14 CO 2 = 2.5 ‰, σ = 7.3 ‰), which was similar in magnitude to the combined AMS and sampling error.Observations for the soil surface were only available for about half the sampling dates; for missing dates we assumed δ 13 C = −9.5 ± 1 ‰ and 14 C = 30 ± 5‰, based on an average of available data.To estimate uncertainty for surface flux isotopic ratios, we applied Monte Carlo simulations (1000 iterations) to propagate the uncertainty associated with each measurement in Eq. (3).

CO 2 and 14 CO 2 production by soil horizon
To vertically partition the production of CO 2 , we again applied Fick's law (Eq.1) to determine fluxes from subsurface soil layers.After experimenting and finding no functional types that satisfactorily fit the CO 2 profiles through time, we chose to calculate dC / dz across soil layers by discrete difference.We used the following discretized form of Fick's law: where F (z 1 ) is the flux at the top of a soil layer, D(z 1 , z 2 ) is the average diffusivity within the layer (following Turcu et al., 2005), and C z 1 and C z 2 are CO 2 concentrations in gas wells at the top and bottom of the soil layer.We modeled soil diffusivity following Moldrup et al. (2004) based on soil water content, porosity, and moisture release characteristics.
Because the four soil plots had similar vertical profiles for physical variables, we compiled porosity and moisture release data from all plots and applied a loess fit to interpolate between measured depths.Diffusivity was modeled with soil moisture data specific to each plot, and moisture between measured depths was estimated by linear interpolation.Diffusivity was corrected using soil temperature measurements from each plot, as in Pingintha et al. (2010).Good agreement between surface flux rates calculated with Eq. ( 7) and direct measurements with the Licor 8100 supported the accuracy of this approach (Slope = 0.95, R 2 = 0.89, N = 46).The production of CO 2 in each soil layer was estimated as the difference between fluxes entering the bottom and leaving the top of the layer (Davidson et al., 2006;Gaudinski et al., 2000), as follows: where P (z 1 , z 2 ) is the production in the soil layer between depths z 1 and z 2 .The 14 C of production in each layer was calculated as in Gaudinski et al. (2000).
where indicates 14 C of production and flux in ‰ units.Uncertainty of production rates and isotopic composition were estimated with Monte Carlo simulations, randomly sampling errors to add to each component measurement within its range of analytical uncertainty, for 1000 iterations.

Contributions of R h and R a
Although trenched plots have several known limitations for estimating heterotrophic soil activity (e.g., increased soil moisture, root senescence, and potential changes in microbial composition), we used comparisons of the trenched and intact plots to partition total soil respiration (R tot ) by two methods: bulk surface fluxes, and isotopic mixing.We compared both these approaches, first computing R h / R tot as the quotient of surface CO 2 flux from the trenched plot and the average of the intact plots, and second by applying a twoend-member isotopic mixing equation: where R h and R tot are the 14 C of surface flux from trenched plot and intact plots, respectively, and R a was estimated from root incubations.Uncertainty associated with isotopic partitioning estimates was calculated following Phillips and Gregg (2001).

Diffusional model simulations
We adopted the model described in Nickerson and Risk (2009b) to simulate diffusion of 14 CO 2 in addition to other isotopologues.Our modeled soil profile was 1 m deep with 100 layers, and at each time step gas transport between neighboring layers was calculated with a 1-D discrete version of Fick's law, using isotopologue-specific diffusivities.Diffusivity of 12 CO 2 was calculated from soil physical variables following Moldrup et al. (2004), and the diffusivity of 13 CO 2 and 14 CO 2 were calculated by multiplying the Moldrup diffusivity by fractionation factors of 1.0044 and 1.0088, respectively.For all simulations we initialized the CO 2 concentration profile with an analytical steady-state solution (Nickerson and Risk, 2009b).We iterated the model with a 1 s time step until the concentration and isotopic composition of soil profiles were stable for at least 3 model days.The default soil physical and biological variables reflect values observed at Willow Creek, and are shown in Table 1.

General patterns
The 14 CO 2 of soil air in intact profiles was intermediate between the atmosphere and the trenched plot profile (Fig. 2), with 14 CO 2 in intact profiles averaging 48 ‰ (S.D. = 9 ‰, N = 85), trenched plot observations averaging  3).The total range in soil 14 CO 2 over the sampling period was about two to three times greater than in air samples from the tower, indicating atmospheric variation was not the primary factor driving soil 14 CO 2 variability.The computed 14 CO 2 of surface fluxes (Fig. 3) indicated microbial soil respiration was more enriched in 14 C than total respiration by a seasonal average of 34 ‰ (95 % CI = 23-44 ‰).This is approximately equivalent to a mean age six to eight years older, based on the recent rate of decline of atmospheric bomb 14 C of 4 to 5.5 ‰ yr −1 (Graven et al., 2012).In intact plots, respired 14 C decreased over the course of the 2012 growing season, from a high value in March of 77 ‰ (only Plot 1 sampled) to a low in October of 37 ‰ (Plots 1-3, averaged).This 40 ‰ seasonal decrease was also significantly correlated with soil moisture (Fig. 4).In contrast to the intact plots, microbially respired 14 C from the trenched plot remained comparatively elevated through the growing season.
Other impacts of trenching included a substantial decrease in surface CO 2 flux, by an average of 39 % over the course of the 2012 growing season (Fig. 5a), and elevated summer soil moisture compared to the intact plots (Fig. 5c).The decrease in CO 2 flux rate and the lack of soil drying, which was likely due to cessation of plant transpiration, both provided strong indications that trenching was successful at excising live roots.We observed no impacts of trenching on soil temperature (Fig. 5b).
Surface CO 2 flux measurements using the novel FD chambers compared favorably with periodic side-by-side LI-8100 measurements, producing highly linear relationships (overall R 2 = 0.6).FD chambers tended to measure lower fluxes than the LI-8100, however, in contrast to earlier laboratory comparisons (Risk et al., 2011).While it is difficult to know under field conditions which instrument is most correct, we  chose to normalize all the FD chambers to the common LI-8100 instrument, by adjusting each FD using a linear multiplier and offset obtained from side-by-side measurements made on > 20 different days.The corrected fluxes are presented here (Fig. 5).FD measurements also had a considerable number of high magnitude fluxes, or spikes, which have also been noted in FD measurements in other field studies (Lavoie et al., 2012;Risk et al., 2013).No attempts were made to smooth the soil flux data, because the spikes may be natural high-frequency phenomena (i.e., pressure pumping) that are not normally detected by the comparatively slowsampling, dynamic soil chambers like the LI-8100 system.Alternatively, the spikes may be due to sampling error, due to transient deviations between the FD's two internal CO 2 sensors that simultaneously measure soil and atmospheric concentrations.
Isotopically, microbially respired fluxes from the trenched plot did not have identifiable seasonal trends, but they had similar total variation to fluxes from the intact plots.For most days surface fluxes from the trenched plot fell within a 20 ‰ range for 14 CO 2 , but one observation exceeded the minimum by almost 50 ‰.It is important to note, however, that this high value was calculated using the 14 cm gas well depth to gap-fill missing data from the 7 cm depth, which may have induced a positive bias in calculated surface flux 14 CO 2 .On the other hand, the 14 cm depth was not uniquely elevated in 14 C on that particular sampling day.High 14 CO 2 levels exceeding 100 ‰ were found in both shallow and deep gas wells from this profile (Fig. 2, bottom panel).

Explanation 1: changing R h and R a contributions
To account for seasonal declines in respired 14 CO 2 from the intact plots, we first examined changes in relative contributions from heterotrophic and autotrophic CO 2 sources.We expected that increasing contributions from 14 C-deplete root respiration could lead to decreases in total soil respired 14 CO 2 .Root-respired 14 CO 2 measured from incubations of roots from 0-5 cm depth was 39 ‰ (S.D. = 4 ‰, N = 4).Consistent with expectation, root-respired CO 2 had less 14 C than microbially respired (i.e., surface flux from the trenched plot), with a seasonally averaged difference of 46 ‰ (95 % CI = 33-60 ‰).In terms of C age, CO 2 respired from the trenched plot was 8 to 12 yr older than root respiration.We estimated contributions from heterotrophic and autotrophic sources by two methods.Our first approach was to compare the quotient of surface CO 2 fluxes from the intact and trenched plots.This approach produced a U-shaped seasonal pattern for R h / R tot (Fig. 6).Heterotrophic contributions descended from 100 % in March to a minimum of about 30 % in mid-summer, and returned to 100 % by mid-October.Note that the quotient of surface fluxes often exceeded 1 outside the growing season because rates in the trenched and intact plots were similar to each other and near zero.
Estimates of R h / R tot using the second approach, an isotopic mixing equation, provided similar estimates as surface fluxes from March through July, but then diverged and remained close to zero through the remainder of the growing season.Two 14 C measurements from the intact plots were actually more deplete in 14 C than the autotrophic endmember, providing negative estimates of R h contributions, and these are shown on the zero line in Fig. 6.Essentially, the two partitioning approaches diverged because flux rates in the intact plots returned to levels similar to the trenched plot by the end of the growing season, but 14 C did not.Both partitioning approaches pointed towards decreasing heterotrophic contributions in the first half of the summer as a possible explanation for the decrease in respired 14 CO 2 from intact plots, but other mechanisms are needed to explain the continued 14 C decrease in late summer.

Explanation 2: changing vertical CO 2 contributions
We next investigated whether the seasonal decline in respired 14 CO 2 from intact plots was related to changes in the vertical distribution of CO 2 production in the soil profile.Because deep soil carbon is older and has less 14 C than shallow substrates, we expected seasonal warming and drying of the soil profile could cause deep C to become destabilized and respired.We found, however, only weak evidence that variation in the vertical distribution of CO 2 production influenced the 14 C signature of surface respiration.Vertical partitioning calculations indicated approximately 40 to 80 % of total production originated from the uppermost 8 cm (Fig. 7).The 14 C of surface flux tended to increase with the fraction of CO 2 produced in the uppermost soil layer (slope p = 0.002, R 2 = 0.3), but the relationship was only significant when all four plots were analyzed.When the trenched plot was excluded, the slope of this relationship had a p value of 0.07.
Vertical partitioning exhibited some seasonality (Fig. 7a), and we found a weak correlation between the fraction of CO 2 produced by the top layer and soil moisture, but only when all four plots were analyzed (slope p = 0.01, R 2 = 0.12).Furthermore, in contrast to our expectation of deep CO 2 containing less 14 C, we found the 14 C of soil air did not show consistent patterns with depth (Fig. 2).Gradients were especially variable in the intact soil plots, sometimes increasing with depth and sometimes decreasing.To investigate vertical CO 2 gradients in more detail, we also calculated the 14 C of CO 2 produced in each subsurface horizon (Fig. 8), in contrast to examining only the 14 CO 2 gradients in soil air, which are attenuated by diffusion.Unfortunately, we found that 14 C production estimates were prone to error in deep soil where bulk CO 2 production rates were low, because the bulk production term occurs in the denominator of 14 C calculations and tends to inflate isotopic errors in the numerator (Eqs.8 and 9).We therefore present only a subset of the calculated production 14 C results, filtering out values where production rate was ≤ 0.2 µmol m −2 s −1 for the soil layer.The remaining observations, where were focused between 0 and 20 cm, indicated no vertical trends in 14 C of production.The lack of vertical gradient in 14 C of CO 2 production may also indicate that CO 2 in this layer is root derived.In contrast to soil organic matter, roots have limited age gradients with soil depth (Schrumpf et al., 2013) From the vertical partitioning analysis, we did not find a compelling explanation for the correlation between respired 14 CO 2 and moisture.Although the vertical distribution of CO 2 production varied substantially through time, correlations with soil moisture and 14 C were weak, and we lacked evidence that 14 CO 2 abundance decreases with depth.

Explanation 3: changes in 14 C of heterotrophic respiration
As stated in the general trends, surface fluxes from the trenched plot varied in 14 C by as much as 50 ‰ through the 2012 growing season, but remained comparatively high and did not seem to explain the decrease in respired 14 CO 2 from intact plots.Observations from the trenched plot provided a unique opportunity to examine R h in a more dynamic environment than traditional laboratory incubations.To place these trenched plot results in context, here we compare the trenched plot observations, which are essentially an in situ incubation, to more commonplace in vitro incubations in static laboratory conditions.We found that for both laboratory incubations and trenched plot measurements, the vertical distribution of soil CO 2 production was similar (Fig. 9b).Both approaches had the highest production rates between 0-20 cm, and very little production in deeper soil.This similarity conferred some confidence that manipulating the soil either by trenching or by more disruptive coring did not alter the relative microbial activity of deep versus shallow soil.We found striking differences, however, between 14 CO 2 produced in laboratory incubations and 14 CO 2 in the trenched plot (Fig. 9a).In laboratory incubations, respired 14 CO 2 had a similar vertical gradient as bulk solid soil.Below 15 cm, CO 2 from incubations did not contain bomb C (i.e., 14 C < 0 ‰) and reflected the old C substrates present in deep soil.In contrast, CO 2 in the trenched plot was greater than 0 ‰ at all depths, containing bomb C throughout the profile.Although in situ soil air is somewhat impacted by atmospheric CO 2 invasion, atmospheric effects were unlikely to have substantial impact, because soil CO 2 concentrations ranged five to 20 times greater than atmospheric CO 2 .Following the same incubation procedure used by many others (Cisneros-Dozal et al., 2006;Gaudinski et al., 2000;Schuur and Trumbore, 2006), we picked out the majority of roots from soil cores before incubating them, and this root removal may have dramatically altered respired 14 CO 2 in comparison to the trenched plot.This comparison between in vitro and in situ microbial respiration suggests that C substrates for respiration are very different in lab incubations from the field, particularly below 15 cm.In the field, C from decaying roots was an important microbial substrate in the trenched plot, throughout the profile.The 14 C of microbial respiration from the trenched plot was influenced not only by the quantity and quality of soil organic matter pools, but perhaps more importantly by the availability of root C. In the lab incubations release of old C due to disturbance of ped structure may have augmented release of old C, as in Ewing et al. (2006).

Dynamic simulations
Because incubation 14 CO 2 measurements are used in many studies to assess the age of C that is actively utilized by microbes, and to characterize heterotrophic end-members for respiration source partitioning, we wanted to confirm the apparent discrepancy between field and laboratory microbial 14 CO 2 production.We used a dynamic CO 2 diffusion model as an alternate tool to constrain the 14 C of production in the trenched plot.We prescribed a range of production 14 C profiles to assess if microbial production of old 14 C-deplete CO 2 at depth could give rise to modern soil air CO 2 gradients (i.e., 14 C > 0 ‰), like we observed in the trenched plot.For these simulations we assumed that the vertical distribution of bulk CO 2 production was the same as observed in the incubations, and we parameterized all other soil variables to match actual soil conditions as much as possible (Table 1).For the first simulation (Fig. 10a), we started with 14 CO 2 production profiles that were observed in the laboratory incubations.With each subsequent simulation we included more 14 C at depth, progressing towards a vertically constant isotopic profile with 14 C production = 86 ‰ (the 14 C produced by the 0-5 cm depth incubation).In other words, if microbial production in the trenched plot had the same 14 C abundance as in lab incubations, we would expect steadystate soil CO 2 in the trenched plot to look similar to the black line in Fig. 10a.This set of simulations demonstrated two important points.First, it highlighted that the 14 C soil air CO 2 profiles differ somewhat from 14 C CO 2 production profiles, due to diffusive mixing and infiltration of atmospheric CO 2 .Second, it showed that the 14 C produced in lab incubations was much too old in deep soil to give rise to the CO 2 profiles observed in the trenched plot.In order to obtain 14 CO 2 soil air profiles in the range we observed in the trenched plot (50-120 ‰), the 14 C of production would have to exceed 0 ‰ through the length of a 1 m profile (as in Fig. 10e or f).

Influences on 14 CO 2 seasonal variation
We found a monotonic decrease in 14 C of surface flux from intact plots through the 2012 growing season, which was consistent with the seasonal decline found by Gaudinski et al. (2000) at Harvard Forest, and the decline in ecosystemrespired 14 CO 2 at an Alaska tundra site by Hicks Pries et al. (2013).We examined three possible explanations for this seasonal decline: shifts in autotrophic versus heterotrophic contributions, deep versus shallow contributions, and variability in 14 C of heterotrophic respiration.We found substantial seasonal variation in all of these potential explanatory variables, but each had a weak or no relationship with respired 14 CO 2 .Although our trenched plot treatment was not spatially replicated, the 14 C of respiration from the trenched plot was consistently greater than intact plots following the first spring sampling event.Based on this shift in respired CO 2 towards older, 14 C-enriched bomb C when roots were cutoff, as well as the shift in microbial respiration towards even older pre-bomb C when roots were picked out from incubated soils, we believe one of the more compelling explanations for the growing-season decline in respired 14 CO 2 was an increasing dependence through the summer on newly photosynthesized plant C by both roots and microbes.
The typical pattern for gross photosynthesis at Willow Creek based on several years of eddy covariance measurements has been a parabolic curve peaking in June-July (Cook et al., 2004;Desai et al., 2005).This pattern mirrored our estimates of R h / R tot based on surface flux rates, suggesting that heterotrophic relative contributions reached a minimum when plant growth peaked.When we used an isotopicmixing approach to partitioning, however, it suggested that heterotrophic contributions continued to remain low until the fall.A possible explanation of this discrepancy is that microorganisms in the intact plots switched during the growing season to substrates such as root exudates and new root litter that were more deplete in 14 C than the substrates initially available following spring thaw.The CO 2 respired from intact plots in late summer may have been produced by microbes but carried the 14 C signature of new roots.If microbes in intact plots switched to newly available substrates, then the trenched plot would have no longer provided a good measure of heterotrophic 14 C for mixing-model partitioning.Hopkins et al. (2013) have also shown that 14 C abundance in root respiration declines over the course of the growing season.While we measured root respiration at only a single time point and did not explicitly assess root respiration seasonal variability, the analysis by Hopkins et al. suggests that the root 14 C end-member is non-static through time.Their findings support our observation that soil respiration 14 CO 2 declined in the presence of roots, and that more recent photosynthates tended to dominate respiration as the growing season passes.
We initially found that 14 C of surface flux from intact plots correlated with soil moisture; however, supporting analyses did not indicate a clear cause-and-effect relationship.We had expected that moisture might alter 14 C by changing vertical partitioning of soil respiration sources.We expected seasonal soil drying might cause shallow soils to become less active, due to water stress, and deep, seasonally saturated soils to become more active, due to improved oxygenation.This expectation was not substantiated, however, by the vertical partitioning analysis.Although we calculated that the percentage of CO 2 produced in the top 8 cm varied seasonally between 40-80 %, we did not find a significant correlation with moisture, unless we included observations from the trenched plot.Observations from the trenched plot tended to have high leverage on regression analyses, because they grouped at the wet end of the soil moisture spectrum and at the high abundance end of the 14 C spectrum.This points to the general challenge of parsing-out environmental drivers in soil respiration analyses.Because moisture in the trenched plot remained high through the summer, we could not assess the impacts of soil moisture in the absence of root inputs.Conversely, because root inputs covaried with moisture in the intact plots, it was also not entirely possible to assess the impacts of changing plant C allocation, independent of soil moisture changes.
In fact, one possibility is that estimates of heterotrophic soil respiration contributions (R h / R tot ) based on bulk surface flux rates are overestimated during late summer, because of higher soil moisture levels in the trenched plot than in the intact plot.This could partially explain why isotopic partitioning estimates diverged from the bulk flux estimates.If microbial contributions were over-estimated in the late summer, however, this does not provide a satisfying explanation for the low 14 CO 2 abundances respired from intact plots into the fall, as trees entered winter dormancy.The effects of moisture and changing C inputs may overprint one another, with late summer drought suppressing microbial respiration, and increasing contributions of new C both explaining the decline in 14 CO 2 from intact plots from spring into fall.

In situ versus in vitro heterotrophic 14 CO 2
The variation we observed in 14 CO 2 respiration from the trenched plot indicated that the "active" C pool utilized by microbes is dynamic through time, varying at least 20 ‰.Although the factors driving this variation could not be entirely discerned from this study (i.e., we did not find significant correlations between 14 C from the trenched plot and temperature or moisture), we had indirect evidence that microbes responded readily to changes in substrate availability.
We showed that 14 CO 2 from soil incubations decreased with depth, reflecting the 14 C of bulk soil, whereas in situ CO 2 was modern through the soil profile.This discrepancy suggests that microbes at depth in the field were not consuming soil carbon from depth, but rather modern substrates that may have come from decaying roots (which were mostly picked out of the incubated soil cores), or from dissolved carbon transported from the shallow subsurface.Other field studies have previously noted modern 14 CO 2 in soil air at depth (Gaudinski et al., 2000;Hirsch et al., 2003); however, previous studies were unable to rule-out root respiration as a source of this CO 2 .Because our trenching treatment cut off live roots, we were able to show that microbial activity can also produce modern CO 2 at depth in intact soil columns.Advective transport of substrates from the soil surface has been shown to create infillings of modern organic matter that serve as an important component of the "active" microbial C pool at depth in other ecosystems (Marin-Spiotta et al., 2011).Future work at Willow Creek that examines 14 C of dissolved organic carbon could help determine whether the source of modern carbon at depth is root inputs or surface carbon that is translocated.The large seasonal range in soil-respired 14 CO 2 found in this study points to exciting possibilities for using 14 C as a sensitive indicator of changing soil metabolism.Based on our findings and recent analyses by Hopkins et al. (2013), which show that root respiration from several forest sites becomes more similar to the atmosphere in 14 C content over the course of the growing season, we believe one of the more pronounced signals that can be detected using 14 CO 2 measurements is the relative contributions of current photosynthates.A persistent challenge, however, is detecting the amount and type of C respired by soil microbes.At Willow Creek, it appears that recent photosynthates are an important substrate for microbial respiration, but in systems with large pools of old soil C, summer microbial activity may utilize much older C as well (Hartley et al., 2012).
Going forward, our study and others suggest more complexity involved in partitioning root and microbial respiration with 14 CO 2 than was previously appreciated, as both end-members appear to be highly dynamic.We have several recommendations for others studying soil 14 CO 2 .
Use caution in extrapolating laboratory incubations to field conditions.Using laboratory incubations as an approximation for heterotrophic activity could compound, rather than simplify, interpretation of respired CO 2 sources.Laboratory incubations are useful for comparisons between disturbed soil cores, and within the context of understanding soil organic matter dynamics they can be used to assess the turnover time of the "active" C pool, or the pool that is most readily destabilized by microbial activity.Within the context of understanding in situ microbial activity, however, it becomes important to consider the more complete spectrum of microbial associations, including not only soil organic matter associations but also close associations with intact roots (Kuzyakov, 2006).For deep soils in particular, in situ microbial respiration is likely much more impacted by root-derived C, and younger in terms of 14 C age, than is represented by soil incubations.
Consider an alternative scheme for partitioning sources of soil respiration.Partitioning soil respiration into root and microbial sources has been a persistent challenge for many years.Using 14 C as a tracer (Schuur and Trumbore, 2006), or a combination of 14 C and 13 C (Hicks Pries et al., 2013) are both approaches that have been used to isotopically partition root and microbial end-members.Such measurements usually depend on one-time measurements of the root and microbial end-members, because the sampling process is destructive, and 14 C measurements are costly.However, in light of the finding that root and microbial end-members may vary through time with inputs of new photosynthates, an alternative approach should be considered that focuses instead on partitioning respiration into present-year and older C stores.Such partitioning could be done without any destructive sam-pling or extrapolation from incubations, and may be equally useful for studies that seek to examine coupling between above-and below-ground activity.Instead of measuring root and microbial end-members, a very early-season measurement of respired 14 CO 2 could be used to represent the baseline condition, or the end-member for C sources from previous years, and atmospheric CO 2 could be measured as the end-member for new photosynthates.Repeated measurements through the growing season of respired 14 CO 2 could be partitioned into present year and previous C sources using a two end-member mixing model.Such an approach may not be appropriate in all ecosystems, for example, if summer root inputs stimulate priming of old soil C pools (Hartley et al., 2012).
Dynamic models are a useful complement to static, steadystate models for interpreting soil gas data.In studies where deep soil C dynamics are of interest, analyses that go beyond directly measured values of surface flux 14 CO 2 or soil air 14 CO 2 to calculating flux and production profiles can also reveal useful insights about underlying sources of CO 2 that contribute to surface emissions.The steady-state Fickian models that are often used to calculate production profiles (e.g., Eqs.7-9) are useful for this purpose but can have very large uncertainties, particularly if steady-state assumptions are violated.Dynamic models, like the Nickerson and Risk model demonstrated here, provide a useful alternative to constrain production profiles, and are also useful for investigating 14 CO 2 responses to dynamic changes in soil environment.
Measure soil respiration 14 CO 2 at the beginning, middle, and end of the growing season.For researchers primarily interested in an average growing season 14 C respiration value, this study corroborated previous work suggesting that seasonal variation in respired 14 CO 2 is substantial (Hicks Pries et al., 2013;Hirsch et al., 2003;Hopkins et al., 2013;Schuur and Trumbore, 2006).At a minimum, sampling time points at the beginning, middle, and end of the growing season are ideal to capture the seasonal progression of new C additions.

Conclusions
By examining soil 14 CO 2 with high vertical and temporal resolution we showed that respired 14 CO 2 is strongly influenced by recently assimilated carbon; however, we could not fully resolve the mechanisms underlying low levels of 14 C late in the growing season and the correlation between 14 C and soil moisture.Our results indicated that heterotrophic 14 C is dynamic and sensitive to immediate substrate availability, and that experimental manipulations to isolate heterotrophic and autotrophic activity can substantially impact estimates of heterotrophic 14 C. Inputs of new photosynthates over the growing season, which have been shown to decrease the 14 C content of root respiration (Hopkins et al., 2013), may also lead to decreases in the 14 C content of microbial respiration.Studies that make use of 14 CO 2 measurements for examining disturbance or climatic change impacts should be interpreted with an understanding that respired 14 CO 2 can fluctuate seasonally by 40 ‰, and that this variability may reflect not only changes in root contributions, but possibly root impacts on 14 C of heterotrophic respiration as well.

Fig. 1 .
Fig. 1.Schematic of soil plot layout and below ground sensor installation.

Fig. 3 .
Fig.3.Computed 14 CO 2 of surface flux (R tot for intact plots and R h for trenched plot) and atmospheric 14 CO 2 (21 m above ground level) for the same period.Note that for the trenched plot, fluxes on 2012.42 and 2012.49were calculated using measurements from 14 cm depth rather than 7 cm, due to missing data.

Fig. 5 .
Fig. 5. Time series of (a) soil CO 2 flux measured with forceddiffusion probes, (b) soil temperature at 5 cm, and (c) volumetric soil moisture at 4 cm.The mean of the three intact plots is shown for clarity.

Fig. 6 .
Fig. 6.Heterotrophic contributions to total soil respiration, estimated by two methods.Grey points show hourly R h / R tot estimated from the quotient of surface fluxes from the trenched and intact plots (all intact plots averaged).Solid black line shows mean quotient estimated by loess fitting.Black points show 14 C partitioning estimates for each plot.

Fig. 7 .
Fig. 7. Vertical partitioning, expressed as fraction of CO 2 produced in uppermost soil layer (top 7 to 8 cm).Errors bars were calculated from Monte Carlo simulations to propagate uncertainties from gas well measurements.(A) variation in vertical partitioning through time, with soil water content shown for seasonal context, and (B) vertical partitioning versus 14 C of surface flux.The grey regression line includes plot 4 (slope p < 0.01, R 2 = 0.29) and the black regression line excludes plot 4 (slope p = 0.07, R 2 = 0.19).

Fig. 8 .
Fig. 8. Variation in estimated 14 CO 2 production profiles over the sampling period.Sampling days are distinguished by shade, from dark (late 2011 and early 2012) to light (late 2012).Because estimated errors are inflated by low production rates (see Eq. 9), we omitted ∼ 20 % of observations where soil layer CO 2 production rate was ≤ 0.2 µmol m −2 s −1 .

Fig. 9 .
Fig. 9. (a) 14 C of bulk solid soil, CO 2 respired in laboratory incubations, and soil air CO 2 from trenched plot (note the respiration units differ for field and laboratory samples).(b) CO 2 production rate in incubations and in trenched plot.Error bars for bulk soil and laboratory incubations are the standard deviation of replicate cores (N = 3), and for the trenched plot are the standard deviation of sampling dates (N = 10).

Fig. 10 .
Fig. 10.Comparison of production and soil air 14 CO 2 profiles from dynamic simulations of 1-D diffusion.