Partitioning sources of soil‐respired CO2 and their seasonal variation using a unique radiocarbon tracer

Soil respiration is derived from heterotrophic (decomposition of soil organic matter) and autotrophic (root/rhizosphere respiration) sources, but there is considerable uncertainty about what factors control variations in their relative contributions in space and time. We took advantage of a unique whole‐ecosystem radiocarbon label in a temperate forest to partition soil respiration into three sources: (1) recently photosynthesized carbon (C), which dominates root and rhizosphere respiration; (2) leaf litter decomposition and (3) decomposition of root litter and soil organic matter >1–2 years old.


Introduction
A key area of uncertainty in the terrestrial C cycle concerns the processes, referred to collectively as respiration, that control the vast majority of C lost annually. Carbon is added to ecosystems through photosynthesis and lost either rapidly, by fueling plant metabolism and growth, or more slowly, through the decomposition of dead plant organic matter. The por-tion of total ecosystem respiration that takes place below-ground, soil respiration, has been estimated in forests to be as much as 69% of the total ecosystem respiration (plant and soil) and 55% of the carbon assimilated through photosynthesis annually (Janssens et al., 2001). Besides representing a large loss of carbon, soil respiration is also a major contributor to interannual variability in the net ecosystem balance (Goulden et al., 1996;Valentini et al., 2000).
Soil respiration is derived from autotrophic and heterotrophic sources. Autotrophic respiration includes CO 2 derived from root metabolism and the activity of microorganisms in the rhizosphere. Heterotrophic respiration includes CO 2 released during microbial decomposition of soil organic matter. While autotrophic respiration is linked to the supply of photosynthetic products from plants, heterotrophic respiration is derived from several different kinds of dead plant material (leaf litter, root detritus, soil organic matter) available to decomposers. Knowledge of the relative contributions of the two sources is relevant to studies of C cycling as each source returns C to the atmosphere on different time scales (years or less through root/rhizosphere respiration vs. decades to centuries through soil organic matter decomposition). Furthermore, we need to better understand all potential factors (environmental and/or phenological) controlling their seasonal variation to make accurate predictions of future C sequestration in forests. Important questions include: (1) how much of total soil respiration comes from autotrophic vs. heterotrophic sources?
(2) what portion of heterotrophic respiration comes from decomposition of different substrate types? and (3) how do changes in phenology, soil moisture and temperature, affect each of the sources of soil respiration? Different approaches have been used to find the relative contributions to soil respiration. Three widely applied methods include component separation, root removal and the use of isotopes, described in detail by Hanson et al. (2000). Component separation extrapolates measurements of CO 2 flux for components (e.g. incubations, excised roots) to the soil volume, but can be subject to artifacts associated with respiration rates measured under non-field conditions and ultimately limited by knowledge of factors like the quantity of roots in the soil. Root removal is commonly accomplished by trenching or girdling trees, and is limited in the short term by the time it takes for roots to die and in the long term by enhanced decomposition of severed or dead roots. Isotope studies offer the advantage that measurements are made in intact systems, but they tend to be relatively expensive and typically require lumping respiration sources that have similar isotopic signatures.
Estimates of the relative contribution of root respiration to soil respiration vary considerably. Kelting et al. (1998) estimated root respiration to account for 32% of soil respiration by trenching in a red oak stand (Quercus rubra L.). Also using trenching methods, Epron et al. (1999) estimated root respiration to contribute 60% of the annual total CO 2 from soils in a beech stand (Fagus sylvatica L.). A large scale girdling experiment in a boreal forest (Pinus sylvestris L.) was used to estimate that root respiration accounted for up to 56% of total soil respiration in the middle of the summer, but recognized that the contribution could be even higher given the longevity of starch reserves in recently-severed roots (Hö gberg et al., 2001). Another trenching experiment at a smaller scale in a Mediterranean forest (Quercus spp.), indicated an annual contribution of 23%, but other factors such as enhanced decomposition in trenched plots and recent logging of the site could have affected the results (Rey et al., 2002). The reporting of root respiration as a percentage of total respiration rather than an absolute quantity may be masking differences in root inventory, allocation, and total soil respiration among sites.
In this paper, we report on the application of a unique 14 C tracer to the characterization of soil respiration components. Large enrichments of the 14 C signatures of plants, litter and soil resulted from a release of 14 C in 1999 from one or more hazardous waste incinerators near our study site on the Oak Ridge Reservation, Oak Ridge, Tennessee. This isotope label was incorporated into living plants to varying degrees according to distance from the source (Trumbore et al., 2002). We used this unique 14 C-tracer to describe how respiration sources vary with season and plant activity.

Site description
The study site is located in the Oak Ridge Reservation (ORR) of the Oak Ridge National Laboratory in East Tennessee, USA (35158 N and 84117 W). Mean annual precipitation is 1358 mm and mean annual temperature is 14.1 1C. Vegetation is dominated by Quercus spp and Acer spp (Johnson & Van Hook, 1989).
While the large 14 C release occurred in July/August 1999, leaves did not show a strong increase in radiocarbon content until the following spring. We observed a large gradient in radiocarbon content of this new leaf growth, with highest amounts of incorporated label in the western portion of the reserve (Trumbore et al., 2002). In the fall of 2000, leaf litter from two areas in the western and eastern portions of the ORR was collected, mixed (to homogenize), dried and stored. Repeated measurements of aliquots of the stored leaf litter showed that leaves collected from the western portion of the reserve had 14 C values ranging from 1 952% to 1 1055%, while the ones collected from the eastern reserve had values from 1 215% to 1 230%. Hereafter, these will be referred to as 'HL' (high labeled) and 'LL' (low labeled) litter, respectively.
The Enriched Background Isotope Study (EBIS) established a total of four sites, two on Inceptisols and two on Ultisols, along the gradient of 14 C label from the western to the eastern side of the reserve. The sites provided a gradient of differential labeling of plants, including roots, with high levels of label on the western side and low levels on the eastern side, hereafter referred to as 'HR' (high labeled root) and 'LR' (low labeled root) sites, respectively. After preliminary measurements showed no large differences in the isotopic signature of soil respiration with soil type, we limited our measurements to two sites: Tennessee Valley Authority, hereafter referred to as 'TVA' (HR site) and Walker Branch (LR site) both on the Ultisol soil type.
At each site, a total of eight 7 m Â 7 m plots were established to manipulate the 14 C content of leaf litter. Natural leaf litter was excluded by covering plots with landscape cloth from October to November, and ambient litter was replaced with HL or LL litter over three consecutive years. Litter exclusions occurred in the fall of 2000, 2001, and 2002, and leaf litter additions took place in May 2001, February 2002and February 2003 with 4 plots at each site receiving HL vs. LL litter. We used this factorial design, with combinations of HR-HL, HR-LL, LR-HL, LR-LL, together with the large 14 C differences in sources among plots and sites, to partition soil respiration into root/rhizosphere respiration and decomposition sources.

Moisture and temperature measurements
Litter-layer water content and soil water status were measured at long-term reference plots at each site. Continuous direct measurements of forest litter water content were based on the electrical resistance characteristics of wet vs. dry litter (Hanson et al., 2003). Buried frequency domain reflectometer waveguides (CS615; Campbell Scientific, Inc., Logan, Utah) were installed in the A and B horizons ( $ 30 cm). All data were logged as hourly means and stored on a data logger (CR10; Campbell Scientific, Inc., Logan, Utah). Additional soil temperature (between 0 and 5 cm depth into the mineral soil) and air temperature data were collected during field measurements.

Total soil respiration
Soil respiration was measured concurrently with sampling for isotopes in the months of May, July and September of 2002, March, May, September and October of 2003 and March of 2004 at the two sites. The isotopic signatures measured in March 2004 at the TVA site were not used as they were confounded by a unexpected new and localized release of 14 C on the date of our measurements. This new release event did not affect the isotopic signatures measured at the Walker Branch site. Measurements were carried out in 6 plots per site including 3 plots with HL and 3 with LL litter treatment.
Closed dynamic chambers were used to measure total soil respiration and to collect CO 2 for radiocarbon analysis as described by Gaudinski et al. (2000). Briefly, an infrared gas analyzer (LI-800, LiCor, Inc. Lincoln, Nebraska) was attached to a pumping system that circulated air to and from the chamber, and the CO 2 concentration increase was monitored for approximately 5 to 10 min. During this time, CO 2 concentrations rose from 380-400 ppm to 500-900 ppm depending on the respiration rates. The rate of increase was usually linear with a few exceptions due to either pump failure or to the effect of placing the chamber lid on the collar which sometimes caused an initial perturbation of the CO 2 flux that lasted for 40 seconds to 1 minute. Nonlinear rates of CO 2 increase were not considered in the estimation of fluxes. Once the flux was measured, the air in the headspace of the chamber was scrubbed of CO 2 initially present (which has a different radiocarbon signature from that of soil respiration) by passing the air coming from the chamber through soda lime. The scrubbing was complete when 3 times the volume of the chamber passed through soda lime which took $ 40 min at flow rates of $ 0.8 L min À1 . Low flow rates (o1 L min À1 ) were desirable for preventing disturbance of the CO 2 concentrations in the soil profile beneath the chamber. The concentration of CO 2 in the chamber reached a steady state value at the end of the scrubbing period that depended on the soil respiration and air circulation rates, but usually remained higher than ambient CO 2 concentrations (between 500 and 600 ppm). When scrubbing was complete, the flow was redirected to a U-shaped tube filled with molecular sieve (mesh size 13X; Advanced Speciality Gas Equipment, South Plainfield, NJ, USA) for trapping of the CO 2 emitted inside the chamber. A container filled with drierite (calcium sulfate) was placed before the trap to remove water that would decrease the efficiency of the molecular sieve (Gaudinski et al., 2000). Activated molecular sieve traps CO 2 at ambient temperatures and releases it when heated at 475 1C (Bauer et al., 1992). The trapping time depended on the respiration rates, with the goal of collecting sufficient carbon for 14 C and 13 C analysis (about 2-3 mg C total). CO 2 collection from all 6 plots at each site took place between 10 am and 5 pm.
Molecular sieve traps were taken to the University of California, Irvine (UCI) where CO 2 was released and the traps reactivated by baking at 610 1C. The released CO 2 was purified cryogenically and converted to graphite using the Zn reduction method (Vogel, 1992). An aliquot of each 14 C sample was analyzed for 13 C using continuous flow isotope ratio mass spectrometry at UCI. Approximately 0.1 micro liters of purified CO 2 were removed from the vacuum line with a syringe and injected into a He-flushed septum-capped vial. The isotopic signature of the CO 2 was measured using a Gas bench II inlet to a Delta-plus stable isotope mass spectrometer.
Graphite targets were sent to Lawrence Livermore National Laboratory (LLNL) for 14 C analysis by accelerator mass spectrometry. The results are expressed in D 14 C the deviation from a standard (oxalic acid) in per mil (%): where the 14 C/ 12 C ratio of the sample is corrected for mass-dependent isotope fractionation to a common value of À25% in d 13 C (Stuiver & Polach, 1977). Overall accuracy of our Zn-reduced targets measured at LLNL, based on repeated measures of secondary standards (oxalic acid 2, ANU, Tiriwood), is AE 5 per mil for modern samples.
At each site, we obtained the 14 C signature of total soil respiration (hereafter referred to as D 14 C total respiration ) from each of the 6 plots. The values per treatment reported here and used in the calculations are the means ( AE standard deviation) of three measurements. We tested for significance of differences of these means between treatments using one-way ANOVA.

Heterotrophic respiration
We collected leaf litter (Oi plus Oe/Oa horizons) and soil cores representing the top 5 cm of mineral soil from each of the six plots (3 plots with HL and 3 with LL litter treatment) at each site. We did not sample deeper soil layers because heterotrophic respiration sources decrease sharply with depth and only $ 25% of total fluxes originated in the mineral soil comes from layers between 8 to 70 cm depth (E and B horizons) at this site (Gaudinski and Trumbore, 2003). Soil and leaf litter samples were incubated separately as to obtain independent radiocarbon signatures for each horizon. To calculate the isotopic signature of heterotrophic respiration (D 14 C heterotrophic ), we weighted the isotopic signature of litter and soil layers according to their contribution to the total decomposition flux (see below).
Leaf litter was sampled by collecting all material above the mineral soil in a 15 Â 15 cm area. The sample was placed in an airtight plastic bag and refrigerated during transport and storage. Soil samples were collected using a core (4.7 cm diameter by 5 cm long) inserted vertically into the soil. The core was carefully removed and the soil extruded with a minimum of disturbance into a flask, which was capped and refrigerated for transport and storage. Samples of soil and leaf litter were collected on all sampling dates except in July 2002. The isotopic signature of heterotrophic respiration used for July was calculated as the mean of the signatures obtained in May and September at each site. No data are available for October 2003 at both sites.
The leaf litter and soil samples were refrigerated for at least 1 week (up to a month) prior to further analysis. While this amount of time is dictated by the time of transport from the field, it also allows time for fine roots that can be present in the samples to die so that their autotrophic respiration no longer contributes significantly to the CO 2 evolved in the incubation. One liter jars (Mason) were used to separately incubate soil cores and leaf litter samples. Humidity inside the jar was maintained by pouring approximately 100 mL of water over glass beads placed in the bottom of the jar. The flasks containing the soil cores were uncapped, weighed (for gravimetric determination of moisture content after the incubation was finished) and placed inside the 1 L jars over the glass beads. Leaf litter samples were weighed and a sub-sample was placed in perforated aluminum foil packets before sealing in the Mason jar.
For approximately 1 week after sealing the jars, we measured the rate of accumulation of CO 2 in the jar headspace using an infrared gas analyzer (LI-6252, Inc. Lincoln, Nebraska). After CO 2 fluxes were determined, air inside the jars was scrubbed of CO 2 , and CO 2 was allowed to re-accumulate until there was sufficient concentration (1-2% CO 2 ) for 13 C and 14 C analysis. CO 2 samples for isotope analyses were collected by connecting the jars to evacuated stainless steel canisters (of 0.5 liters in volume). The 14 C and 13 C signatures were determined as described above. At each site, we report soil and leaf litter decomposition signatures (D 14 C soil decomposition and D 14 C leaf litter decomposition respectively) as the mean ( AE standard deviation) of three plots with identical litter addition treatments.

Autotrophic respiration
To determine the radiocarbon signature of autotrophic respiration (D 14 C autotrophic ), we incubated live roots in the field. Since this 14 C signature is dependent on photosynthetic carbon supplies, and is independent of leaf litter carbon inputs, roots for analysis were collected outside the experimental plots to avoid withinplot disturbance. At two or more randomly chosen locations, we excavated roots mostly in the upper 5 cm of mineral soil, though some may be from slightly deeper as root networks were pulled out without distinction of depth, size (most wereo5 mm diameter), or species.
After sufficient roots for three replicates were collected, they were shaken and washed free of soil, divided into 3 sub-samples and placed immediately in airtight containers. Roots were normally collected at 5-6 pm local time. Isotopes were sampled using the same methodology described for total soil respiration, though we did not measure CO 2 flux rates as we assumed manipulation of the roots may have altered the fluxes (Rakonczay et al., 1997). Air inside the containers was scrubbed of CO 2 , CO 2 was allowed to reaccumulate, and was collected by trapping on molecular sieve. The accumulation of CO 2 for trapping started within an hour after roots had been removed from the soil and lasted for an amount of time from 10 min to an hour depending on the respiration rate of the subsample. Molecular sieve traps were transported to UCI and the 14 C and 13 C signatures of the CO 2 collected determined as described above. At each site we report the mean ( AE standard deviation) of three replicates.

Calculation of respiration components
Fraction of total soil respiration coming from leaf litter decomposition (FLD) Contrasting the levels of 14 C in CO 2 respired from HL and LL treatments at a given site allowed us to uniquely determine the contribution of leaf litter decomposition to total soil respiration using isotope mass balance: where FRR, FLD and FSD are the fractional contributions of autotrophic respiration, leaf litter decomposition and soil decomposition, respectively, to total soil respiration. The D 14 C autotrophic , D 14 C leaf litter decomposition and D 14 C soil decomposition values were derived from our field and laboratory incubations, as described above. Since we incubated not only the labeled leaf litter added to plots, but the entire litter layer (Oi 1 Oe/Oa), FLD as calculated above reflects the contribution from the decomposition of the entire O horizon. At a given site, D 14 C autotrophic and FRR should be the same for both treatments. The difference between HL and LL treatments in D 14 C soil decomposition was generally very small compared to the difference in D 14 C leaf litter decomposition (see the results section). Therefore, we could assume that the differences in 14 C between HL and LL treatments were overwhelmingly caused by leaf litter decomposition. Applying Eqn (1) to each treatment and subtracting LL from HL we solved for FLD as follows: FLD ¼ ðD 14 C total HL À D 14 C total LL Þ=ðD 14 C leaf litter decomposition HL À D 14 C leaf litter decomposition LL Þ We report one single value of FLD for each site (HR and LR) and sampling date. The associated uncertainty was estimated by error propagation.
Fraction of total soil respiration coming from autotrophic respiration (FRR) To estimate the fraction of total respiration derived from autotrophic respiration using 14 C, we applied the mass balance approach described in Trumbore et al., (2002): Here, FRR includes metabolic respiration from live roots as well as anything with a similar 14 C signature, such as symbiotic mycorrhizal fungi and the activity of microorganisms decomposing recent photosynthetic products. While we will hereafter refer to FRR as the fraction coming from root respiration, it should be recognized that it refers to the sum of all of these processes.
The radiocarbon signature of heterotrophic respiration (D 14 C heterotrophic ) was determined as the weighted contribution of CO 2 derived from leaf litter and soil organic matter decomposition: where L is the fractional contribution of leaf litter decomposition to total heterotrophic respiration. One way to determine L is to compare the relative rates of CO 2 evolution from our incubations of leaf litter and soil. However, the temperature and moisture conditions of our incubations differed from field conditions, and if we attempt to scale the rates of heterotrophic respiration derived from laboratory incubations to conditions in the field, we overestimate observed soil respiration by at least a factor of two. Hence, we could not use our laboratory incubations to derive weighting factors with any confidence.
A second method of determining L is obtained by assuming that FRR and L are the same for both HL and LL treatments for a given site. Since L and FLD represent the contribution of decomposing litter to heterotrophic and total respiration, respectively, they are related by: Substituting Eqns (4) and (5) into Eqn (3) results in an equation with two unknowns, FRR and FLD. Since we have calculated FLD using Eqn (2) above, we can solve this equation for FRR, with the requirement that FRR and FLD be identical for both HL and LL treatments within a given site. We only report values of FRR that satisfied this assumption within the error we calculated for FLD using Eqn (2). We report the uncertainty in FRR by treatment calculated using the standard deviations of total soil respiration, root and heterotrophic respiration according to Phillips & Gregg, (2001).

Total soil respiration fluxes
Although there was high spatial variability in soil respiration fluxes, we saw no systematic differences in CO 2 fluxes between the two sites ( Fig. 1; data for March 2004 at the TVA site were not reported). The lower fluxes measured at TVA in July 2002 and Walker Branch in September 2003 were most likely due to moisture differences between the sites. Since the two sites were sampled on different days and rainfall from isolated summer precipitation events is very unevenly distributed across the landscape, the two sites do not always replicate each other. Soil moisture varied from 20% to 35% volumetric water content (at approximately 10 cm depth) across sites during the three sampling periods in 2002, except in July, where moisture at TVA site fell to 11%. In September 2003, we have no data for soil moisture content, but leaf litter at Walker Branch was extremely dry and crunchy, while soil respiration measurements at TVA were carried out the day after a heavy rain event.
14 C signatures of total soil respiration, root respiration and heterotrophic sources We observed a marked difference in the radiocarbon signatures of leaf litter decomposition between HL and LL treatments at both TVA (HR) and Walker Branch (LR) sites (Fig. 2). The 14 C-labeled litter addition imparted a higher radiocarbon signature to total soil respiration on all dates except those where litter was extremely dry (July 2002 andSeptember 2003). The isotopic signature of decomposing soil organic matter increased slightly with time, though differences between HL and LL treatments remained small compared to those associated with leaf litter decomposition. Across sites and litter treatments, the radiocarbon signatures were overall higher at TVA (HR) than at Walker Branch (LR), reflecting the continued influence of 14 Cenriched carbon acquired by plants during the 1999 labeling event.
The 14 C signature of root respiration varied with time ( Fig. 2). Temporal variations between sites followed very similar patterns, although values in the HR site remained $ 100% higher than the LR site throughout 2002-2003. The very depleted values that we observed at the Walker Branch site in July 2002 (À0.28%) and May 2003 (2%) were attributable to a local release of fossil fuel derived CO 2 , which has been verified by measurements of reduced radiocarbon content in ambient air sampled at this site since 2000.

Fraction of total soil respiration coming from leaf litter decomposition (FLD)
According to Eqn (2), the fraction of total soil respiration derived from leaf litter decomposition (FLD) was proportional to the difference in the 14 C signature of total soil respiration between HL and LL treatments.

Fraction of total soil respiration coming from root respiration (FRR)
Our approach of solving for FRR by requiring FLD and FRR to be identical in HL and LL plots at a given site produced consistent results for the Walker Branch site on most dates (Table 1). We had greater problems with the TVA site and our constraints were met only on May 2003. When the isotopic differences among root respiration, heterotrophic respiration and total respiration were small compared to their respective standard deviations, HL and LL treatments did not yield consistent

Partitioning of total soil respiration
A summary comparing the absolute contributions (in mg C m À2 h À1 ) of root respiration and leaf litter decomposition for the Walker Branch site is shown in Fig. 4. Fluxes labeled as 'other' were calculated from mass balance (total respiration minus root respiration minus leaf litter respiration), and have isotopic signatures close to those evolved in soil incubations (FSD; Eqn 1). Heterotrophic respiration, especially leaf litter decomposition, constituted a large portion of the total CO 2 respired in the growing season (May and September), though it was also the most variable component in space and to 40 AE 16 mg C m À2 h À1 on the different sampling dates with the exception of September 2003 (88 AE 35 mg C m À2 h À1 ). Heterotrophic respiration was larger than root respiration on the days we sampled during the peak of the growing season, but in early spring (March 2003 and, root respiration and heterotrophic sources contributed equally to total soil respiration.

Discussion
Large differences in the radiocarbon signatures of leaf litter between treatments allowed us to successfully apply a 14 C mass balance approach to separate sources of soil respiration into components derived from root/ rhizosphere respiration, leaf litter decomposition, and other heterotrophic sources (Fig. 4). We were able to partition total soil respiration at different times in the growing season and identify the effect of moisture on heterotrophic sources. We achieved consistent results across sites in many cases, indicating that our methods of determining isotopic signatures for respiration components do not have significant systematic errors (though our methods for determining fluxes, especially in incubations, do). The partitioning approach gave consistent results even when the isotopic signature of root respiration varied by over 100% from one sampling date to the next. Our method does have the problem that if the separation between isotopic signatures of respiration components is not larger than their respective standard deviations (Fig. 2), we cannot solve for FRR in any meaningful way.

Leaf litter decomposition
Estimates of the relative contribution of decomposing leaf litter to total soil respiration (FLD) showed good agreement across sites when moisture was not a limiting factor (Fig. 3). The greater standard deviations of D 14 C leaf litter decomposition characteristic of HL treatments (Fig. 2) translated into large uncertainties in the estimates of FLD on some dates. In March 2004, greater uncertainty for FLD was expected as isotopic differences in D 14 C leaf litter decomposition between treatments were reduced by dilution of the 14 C label in leaf litter in the fall of 2003 (the first fall season that native litter fall was allowed into the plots after 3 years of 14 C labeled litter additions). The strong dependence of FLD on moisture conditions was evidenced by the marked difference between sites in July 2002 and September 2003. The link between leaf litter decomposition and moisture has been documented by Lee et al. (2004) who found a linear decrease in the relative contribution of the litter layer with lower litter water contents in a manipulated study in a mixed forest. In a tropical forest, an indication of the dependence of FLD on litter moisture content was observed by Goulden et al. (2004) when patterns of soil respiration rates varied in concert with the moisture content of the surface leaf litter. Numbers in brackets represent the standard error according to Phillips & Gregg, (2001). Dates when the isotopic signatures did not satisfy the mass balance are indicated as ''na''; blank entries indicate times when HL and LL treatments did not satisfy the assumptions upon which the calculations were based. In our study, times of lower total soil respiration fluxes (July 2002 at TVA andSeptember 2003 at Walker Branch;Fig. 1) coincided with times of lower relative contributions from leaf litter decomposition (Fig. 3). As a result of the dependence on moisture conditions, and in contrast with root respiration, fluxes from leaf litter decomposition were highly variable in time. This suggests that litter decomposition fluxes can dictate the temporal variation in total soil respiration in response to changing moisture conditions, in agreement with findings from other studies (Hanson et al., 2003;Borken et al., 2005). Hanson et al. (2003) developed a model to predict total soil respiration validated against repeated measurements of litter fluxes and moisture. Model predictions were found to improve when the response of litter fluxes to variations in litter moisture contents were taken into account. In addition of being a considerable fraction of soil respiration (up to 42%), the response to moisture conditions emphasizes the relevance of this component of total soil respiration.

Other heterotrophic sources
Across sites and same litter treatments, the radiocarbon signature of soil decomposition was always higher at TVA (HR) than at Walker Branch (LR) (on average by $ 120%, Fig. 2). In contrast, the difference in the radiocarbon signature of soil decomposition between HL and LL treatments within a given site remained small ( $ 40%). This indicates that a major source of decomposing material in the 0-5 cm soil layer is derived from carbon fixed in the original 1999 14 C release, rather than the labeled litter treatment applied subsequently. Dead roots are the likely source since differences in soil organic matter in this layer remained small in (J. Jastrow, 2004. In theory, we should have been able to perform a similar analysis to that of FLD to calculate the fraction of heterotrophic respiration coming from labeled root decomposition from the difference between HR and LR sites in D 14 C soil decomposition . However, since root respiration radiocarbon values differed between sites, we could not independently solve for the root decomposition end member as we could for FLD.

Root respiration sources
The accurate estimation of D 14 C heterotrophic is obviously important in the calculation of FRR (Eqn 3) and this in turn depends strongly on the value of 'L' (Eqn 4). Overestimation of 'L', as would have occurred with the use of our incubation data as weighting factors, would have led to an overestimation of D 14 C heterotrophic, especially in HL treatments (Fig. 2). This in turn would have translated into an overestimation of FRR. We are confident that the isotopic signatures measured in incubations (D 14 C leaf litter decomposition and D 14 C soil decomposition ) are unlikely to be a cause of error as radiocarbon values of CO 2 derived from incubated material have been shown to be independent of temperature or moisture variation even with very large changes in CO 2 evolution rate (Dioumaeva et al., 2003;Cisneros-Dozal unpublished data).
The estimates of the fractional contribution of root respiration to total respiration (FRR) reported here are not easily compared to the annual means commonly reported in literature. We have not attempted to estimate annual root respiration contributions because of the limited nature of our sampling. However, FRR ranged from a low of 16% to a high of 64% in our study, within the range of published estimates. The annual mean contribution of root respiration to soil respiration has been estimated at $ 60% from trenching experiments (Epron et al., 1999;Hö gberg et al., 2001). A lower estimate by Kelting et al. (1998) of 32% in early June was recognized to be a minimum since the sampling did not take place during the period of maximum root growth. In addition, the use of excised roots to measure rates of root respiration remains questionable, because their function has been observed to decrease rapidly after excision (Rakonczay et al., 1997). An earlier study carried out in the ORR estimated that respiration from live roots accounted for 35% of total soil respiration annually for Liriodendron tulipifera L. spp. (Edwards & Harris, 1977). Rey et al. (2002) reported an annual contribution from root respiration of 23% although previous disturbance at the site were thought to be the cause of this low estimate. It is important to note that in our study, times of highest relative contributions (FRR) were not necessarily times of high root respiration fluxes and thus both quantities should be reported.  Table 1), however the absolute fluxes were not different (Fig. 4).
In contrast to other studies (Epron et al., 2001;Hö gberg et al., 2001;Rey et al., 2002), we did not observe marked temporal variation in root respiration fluxes with the exception of the increase in September 2003 (Fig. 4). Several studies indicate a link between soil respiration and plant activity through the phenological control on root respiration fluxes (Hö gberg et al., 2001;Janssens et al., 2001;Reichstein et al., 2003;Curiel et al., 2004). For this forest, a correlation between root growth and leaf area expansion was observed by Joslin et al. (2001), with the highest root elongation intensity taking place after the completion of leaf area expansion. In terms of root respiration fluxes however, we did not see an increase during the periods of maximum root growth observed in May, June and July for Quercus spp . We did observe an increase from May to September in 2003; however an earlier increase from March to May would have been expected in accordance with the timing of root growth. More frequent measurements (which may obviate some of the differences in soil moisture and temperature conditions between widely spaced sampling intervals) may be required to capture the relationship between root growth and root respiration patterns.
The observed large differences in the 14 C signature of root respiration between sampling dates (Fig. 2) was an unexpected result. These temporal changes were very similar at both sites, which would seem to rule out sampling or measurement errors as the cause. Other causes could be related to changes in the internal sources of C being utilized, or to fluctuations in the radiocarbon content of atmospheric CO 2 (i.e., perhaps further local releases). Trumbore et al. (2002) showed that nonstructural carbohydrate pools were labeled in the 1999 event. Depending on the carbohydrate turnover rate, the 14 C content of this storage pool could be greater than that of more recent photosynthetic products. For example, leaf buds grown each spring have higher 14 C values than atmospheric CO 2 , and parasitic plants growing in early spring also have 14 C signatures close to our root respiration 14 C measurements for the same time period (Trumbore, unpublished data). The observed declined in the 14 C of CO 2 respired by roots at both HR and LR sites from May to July in 2002 and March to May in 2003 indicates the use of stored C as an energy source prior to leaf-out, with a change to current photosynthetic C after the forest canopy matures and the root growth peaks in July. A further indication that roots use more recently fixed carbon during the summer is the very low 14 C values for root respiration observed at Walker Branch in July 2002 and May 2003, (i.e., a site under the influence of recent local fossil fuel burning). However, root respiration values remain higher at the HR site (TVA) than the LR site (Walker Branch) throughout the summer, indicating that either some portion of root respiration is derived from storage pools or that differences in atmospheric 14 CO 2 persist between sites. We have taken bi-weekly time averaged samples of air at 1 m height above-ground at both HR and LR sites to monitor for events like 14 C releases, but it is difficult to link these to values we observe in root respiration measured on one day. Further analysis of radiocarbon in carbohydrates of root tissues is needed in order to better elucidate the source of carbon respired throughout the growing season.

Conclusions
We have used 14 C labeled leaf litter and roots together with a mass balance approach to quantify heterotrophic and autotrophic (root and rhizosphere) respiration sources in a temperate forest. Heterotrophic sources, especially leaf litter decomposition, account for a large fraction of total soil respiration in this temperate deciduous forest. Heterotrophic respiration in the surface mineral soil layer (0-5 cm) was dominated by the decomposition of C fixed 3-5 years earlier (during the 1999 labeling event). Leaf litter decomposition is a major contributor to overall heterotrophic CO 2 fluxes, but it is also the most variable in space and time in response to changing moisture conditions. The instantaneous relative contributions from leaf litter decomposition ranged from $ 1 AE 3% to 42 AE 16%, corresponding to absolute fluxes of 6 AE 3 to 96 AE 38 mg C m À2 h À1 , and decreased to near zero when leaf litter was extremely dry. As a result of its dependence on moisture conditions, leaf litter decomposition is apparently the main source of temporal variation in total soil respiration in this forest.
In contrast, root/rhizosphere respiration fluxes did not exhibit marked temporal variation, ranging from 34 AE 14 to 40 AE 16 mg C m À2 h À1 with a single exception of increased fluxes (88 AE 35 mg C m À2 h À1 ) in September 2003. The relative contributions from root/rhizosphere respiration ranged from 16 AE 10 to 64 AE 22%. Times of highest relative contributions did not necessarily coincide with times of highest absolute fluxes, underscoring the importance of reporting quantities. The radiocarbon signatures of root respiration provided preliminary information on the source of C respired by roots and indicated that sources of C respired may shift from stored C pools in early spring to recent photosynthetic products as the summer progresses.
The 1999 14 C-pulse being studied in the EBIS project provided a unique opportunity for rapid evaluation of the components of soil respiration. The same isotope mass balance approaches can presumably be applied in other ecosystems where the sources of soil respiration differ substantially in age (such as in boreal forests). While similar studies can not be done for all established forest ecosystems, other situations where anthropogenic activities may have perturbed the background isotopic 14 C status may exist and should be investigated.