Carbon isotopic signature of CO₂ emitted by plant compartments and soil in two temperate deciduous forests

Abstract

• Context

The carbon isotope composition of the CO₂ efflux (δ¹³C_E) from ecosystem components is widely used to investigate carbon cycles and budgets at different ecosystem scales. δ¹³C_E, was considered constant but is now known to vary along seasons. The seasonal variations have rarely been compared among different ecosystem components.

• Aims

We aimed to characterise simultaneously the seasonal dynamics of δ¹³C_E in different compartments of two temperate broadleaved forest ecosystems.

• Methods

Using manual chambers and isotope ratio mass spectrometry, we recorded simultaneously δ¹³C_E and δ¹³C of organic matter in sun leaves, current-year twigs, trunk bases and soil in an oak and a beech forest during 1 year.

• Results

In the two forests, δ¹³C_E displayed a larger variability in the tree components than in the soil. During the leafy period, a pronounced vertical zonation of δ¹³C_E was observed between the top (sun leaves and twigs with higher values) and bottom (trunk and soil with lower values) of the ecosystem. No correlation was found between δ¹³C_E and δ¹³C of organic matter. Causes for these seasonal variations and the vertical zonation in isotope signature are discussed.

• Conclusion

Our study shows clear differences in values as well as seasonal dynamics of δ¹³C_E among different components in the two ecosystems. The temporal and local variation of δ¹³C_E cannot be inferred from organic matter signature or CO₂ emission rates.

1 Introduction

The carbon isotope composition (δ¹³C) of CO₂ is now commonly used at the ecosystem level as a natural tracer to investigate carbon processes and their responses to environmental conditions. Approaches coupling isotopic and mass balance are used to partition ecosystem CO₂ efflux (E _ECO) and photosynthetic fluxes (Zobitz et al. 2007). The δ¹³C of E _ECO (δ¹³C_EECO) is used to infer ecosystem or regional C sink strength by inversion modelling. Large uncertainties are remaining about the interpretation of δ¹³C_EECO, mainly due to the multi-source nature of E_ECO, and the temporal variability of δ¹³C values of E_ECO components and their contributions to E _ECO (Bowling et al. 2002; Hemming et al. 2005). Until recently, δ¹³C_E values used to interpret δ¹³C_EECO were considered to be similar among different ecosystem components (i.e., soil, roots, trunk, twigs and leaves) (Kodama et al. 2008).

So far, few studies have examined δ¹³C_E of several components concurrently: between leaf and soil (Mortazavi et al. 2005), trunk and soil (Kodama et al. 2008) or different plant organs (Eglin et al. 2009; Kuptz et al. 2011); they showed significant differences in δ¹³C_E among the targeted components. Moreover, the δ¹³C_E for a given component can exhibit a high seasonal or diurnal variability, up to 10 ‰ for leaves (Hymus et al. 2005; Prater et al. 2005), 4 ‰ for twigs (Damesin and Lelarge 2003), 3–5.5 ‰ for trunks (Maunoury et al. 2007; Ubierna et al. 2009) and 4.2 ‰ for soil (Ngao et al. 2005, Marron et al. 2009). However, only a few of these studies included wintertime measurements (Damesin and Lelarge 2003; Maunoury et al. 2007; Kuptz et al. 2011).

Seasonal variations in δ¹³C_E have been related to several factors such as environmental conditions several days before measurements (Bowling et al. 2002; Ekblad et al. 2005), canopy stomatal conductance (Mc Dowell et al. 2004), the nature and/or δ¹³C of respiratory substrates (Damesin and Lelarge 2003; Kuptz et al. 2011) or to other metabolic processes, such as variations in respiratory pathways (Tcherkez et al. 2003; Kuptz et al. 2011). Recently, a comprehensive study concluded that seasonal δ¹³C_E patterns in one ecosystem were similar for the different components of beech and spruce and between both species (Kuptz et al. 2011), also showing maximal differences for the trunk between summer and winter. The question now arising is how recurrent are seasonal patterns (1) for a given species at different environmental conditions and (2) between closely related species depending on leaf phenology.

In this context, we focused on the spatio-temporal δ¹³C_E dynamics of the main tree components (leaves, twigs and trunks) and soil in two temperate broadleaved forest ecosystems. The main objectives of this study were (1) to quantify differences in δ¹³C_E among components and (2) to characterize their seasonal variation, in particular between the two main phenological periods (i.e. leafy and winter period), and their potential link with environmental parameters (air and soil temperature, relative humidity, vapour pressure deficit), CO₂ efflux rates and the respective total organic matter δ¹³C. Synchronous in situ measurements of CO₂ efflux rates and δ¹³C_E were performed throughout 1 year in an oak (Quercus petraea L.) and a beech (Fagus sylvatica L.) forest using isotope mass spectrometry measurements. In order to validate CO₂ field-sampling methods adapted to leaf and twig components (Prater et al. 2005; Werner et al. 2007), we compared two methods of tissue incubation, with CO₂-free air or N₂ flushes usable under field conditions.

2 Materials and methods

2.1 Study sites and experimental setup

The study was conducted in two French forest sites belonging to the CARBOEUROPE IP network (http://www.carboeurope.org/). The Barbeau forest (cluster_FR1, FR–Fon site, 48°29′ N, 02°47′ E, Table 1) is a managed mature oak-dominated (Q. petraea) stand with an understory of Carpinus betulus L. Soil is a gleyic luvisol [World Reference Base (WRB) classification] of 80 cm depth, on millstone bedrock and covered with an oligomull humus type, named “Barbeau” in the following. The Hesse forest (cluster_FR1, FR–He site, 48°40′ N, 7°05′ E, Table 1), is a young beech-dominated (F. sylvatica) stand with a dystric cambisol (WRB classification) of 120 cm depth and an oligomull humus type. This site is subsequently named “Hesse”.

Table 1 Stand characteristics for Barbeau and Hesse in 2005

The study was conducted from March 2005 (before budburst) to January 2006 during nine field campaigns in Barbeau (06/04, 18/04, 02/05, 01/06, 21/06, 11/07, 09/09, 23/11, and 12/01) and four field campaigns in Hesse (16/03, 18/05, 05/07, and 14/09). Gas exchange measurements, CO₂, and organic matter sampling were always performed between 10:00 and 13:00 UT to avoid diurnal variations in δ¹³C_E as previously observed (Maunoury et al. 2007). Each four dominant oak and beech trees were randomly selected for the whole campaign. Soil measurements were performed around the sampled trees, within about 20 m².

2.2 Environmental and phenological parameters

At both sites and for each campaign, soil temperature was measured at 10 cm depth using a temperature probe (LM35CZ) in Barbeau and five copper–constantan thermocouples (Faculty of Agronomy of Gembloux, Belgium) in Hesse. Soil surface moisture (in the 0–6 cm layer) was measured using a capacitive ML2x Thetaprobe (Delta-T Device, Cambridge, UK) in Barbeau. Mean air temperature at 2 m height (sensor LM35CZ), rainfall and air relative humidity (to calculate water vapour pressure deficit, VPD) were determined every 30 min by meteorological stations installed on-site.

Budburst dates (Table 1) were determined by field observations as either 50 % of trees showing 50 % of bursted buds in Barbeau, or as the beginning increase in the leaf area index (LAI), measured at regular intervals before and during the leafy period (LI-COR LAI 2000) in Hesse. The leaf fall period was also recorded at both sites. The trunk growth period was determined by tape measurements of the radius at 1.30 m height every week in Barbeau and during each field campaign in Hesse (Table 1). The growing period is defined as the time of trunk growth, while the leafy period includes the time span between bud burst and leaf fall.

2.3 Trunk and soil CO₂ efflux rates

Trunk CO₂ efflux rate (E _T) was measured using a closed chamber system (see Damesin et al. 2005 for a detailed description). Briefly, a cylindrical polymethyl methacrylate (PMMA) chamber was temporarily fixed on the trunk, cleaned of mosses and lichens, with a rubber sealant (Terosta-7, Teroson, Ludwigsburg, Germany) and connected to an infrared gas analyser (IRGA, EGM4, PPSystems, Hitchin, UK). The installation was considered to be leak-free if blowing air along the seals caused no increase in the CO₂ level inside the chamber. A fan provided air mixing within the chamber. Once tightly fixed to the trunk, the chamber was purged from accumulated CO₂ by opening a 5-cm diameter lid. Once back to ambient CO₂ concentrations, the lid was closed to allow CO₂ accumulation. Three to four E _T measurements of 2 min during linear CO₂ increase were performed each time. The E _T values were determined from the slope of CO₂ concentration increase and expressed per unit volume of living tissue (i.e., phloem and sapwood; in μmol m⁻³ living tissue s⁻¹). For beech, the whole trunk volume was considered because living cells occur all along the trunk radius (Ceschia et al. 2002). For each oak tree, the living tissue cross-section was determined from a trunk core collected near the chamber at the end of the campaign. The chambers were reinstalled at the same place during each campaign.

Soil CO₂ efflux (E _S) was measured using two different closed dynamic systems (Ngao et al. 2006; Chemidlin Prévost-Bouré et al. 2009): In Barbeau, an EGM4 was connected to a homemade PMMA chamber (25.4 L, 12 cm height), while in Hesse, a Li-6200 (LI-COR Inc., Lincoln, NE, USA) IRGA was used with the Li-6000-9 chamber. In both cases, the soil chamber was put on collars previously inserted into the soil under the canopy (500 and 110 mm diameter in Barbeau and Hesse respectively, inserted 2–3 cm deep) at the beginning of the year, allowing measurements at the same place during field sessions. Two collars were installed at Barbeau and three at Hesse. E _S values were calculated from the slope of CO₂ increase and expressed per surface area (in μmol m⁻² soil s⁻¹).

2.4 CO₂ sampling for isotopic analysis

2.4.1 Incubation tests

The gaseous medium (N₂ vs. CO₂-free air) of an incubation setup may have an immediate impact on the CO₂ efflux rate by shortening the oxygen supply to living tissues and potentially influencing the isotope composition of emitted CO₂. We tested this effect using a system close to that described in Werner et al. (2007). Entire mature sun leaves and twigs were sampled from the top of the canopy of three oaks and three beeches at the end of summer. Each sample was immediately inserted into a 50-ml flask previously purged with either pure N₂ or CO₂-free air. Leaves (n = 44) and twigs (n = 20) were incubated in the dark under ambient temperature. In preliminary CO₂ efflux rate measurements (data not shown), we determined that an incubation time between 10 and 25 min was required to collect enough emitted CO₂ (800–900 ppm). The δ¹³C in air from the incubations was analysed by isotope ratio mass spectrometry (IRMS) as described below. These tests revealed no significant effect of the gaseous medium used, neither for beech or oak leaves nor for twigs (p = 0.72 for leaves and p = 0.53 for twigs, Fig. 1).

Fig. 1

Mean values of δ¹³C_E measured for leaves and twigs of oak and beech incubated after flushing with CO₂-free air (white) or N₂ (black). Error bars represent standard errors of the mean. n (leaf) = 27 for oak and 17 for beech, and n (twig) = 10 for oak and 10 for beech

2.4.2 Sampling of CO₂ emitted by ecosystem components

For CO₂ emissions at the leaf and twig levels, these components were incubated as described above. Sun leaves and twigs from the tree canopy were sampled using a rifle. Leaves and twigs were immediately inserted into 50-ml airtight syringes with valves (SGE, Australia) prepurged with pure N₂. Air from the incubation syringe was then transferred into an empty syringe in Barbeau or into a 12-ml Exetainer vial (Labco Ltd, High Wycombe, UK) in Hesse, according to the equipment available (see IRMS analyses below).

For trunk CO₂ emissions, the trunk chamber was purged after each E _T measurement with N₂ for 15 min until the CO₂ concentration dropped to near 0 ppm. Then, outlet and inlet tubes of the chamber were connected to allow the accumulation of CO₂ emitted by the trunk. After an increase of 700–800 ppm (during approximately 10 min in summer and 90 min in winter), the air in the chamber headspace, containing CO₂ originating only from E _T, was sampled using a 50-ml syringe and analysed by IRMS. Again, this N₂ flushing approach gave the same results as that using a CO₂ free-air flush or estimating the δ¹³C of emitted CO₂ with the Keeling plot method (Damesin et al. 2005).

For CO₂ emitted by soil, the Keeling plot method (Keeling 1958) was applied to determine the δ¹³C_E of E _S (Chemidlin Prévost-Bouré et al. 2009). The sampling setup consisted of 50-ml airtight syringes in Barbeau, or a homemade sampling device by-passing the Li-6200 air circuit (see Ngao et al. 2005 for details) in Hesse. After each E _S measurement, CO₂ concentration was allowed to increase again within the closed system. During this increase, five air samples were taken at steps of 50–100 ppm within a 400–1000 ppm range inside Exetainer vials or 50-ml syringe and analysed by IRMS. From the Keeling plots, the δ¹³C_ES for each collar was estimated using the ordinary least square regression model (Zobitz et al. 2007). δ¹³C_ES was determined as the intercept of the linear regression between the inverse of the CO₂ concentration and the δ¹³C of the air samples. δ¹³C_ES values having a standard error >5 % of the estimated value were discarded.

2.5 Sampling of plant and soil material for isotopic analysis

At both sites and for each campaign, trunk phloem samples of the four trees were taken using a core borer (0.5 cm diameter) at the chamber level or up to 30 cm above. Leaves and twigs used for the incubations and phloem samples were lyophilised and powdered using a ball mill (Type MM200, Retsch, Haan, Germany). Four soil cores (0–15 cm depth and 1.2 cm diameter) were sampled about 15 cm away from each collar. Soil samples did not include litter or roots.

2.6 IRMS analyses

To maintain airtight conditions, gas-filled syringes were processed within 12 h by laboratories close to the sites. Gas samples from the Barbeau were analysed with a NA-1500 elemental analyser (Carlo Erba, Milan, Italy) coupled to a VG Optima IRMS (Fison, Villeurbanne, France), as described by Maunoury et al. (2007). Those from Hesse were injected into a gas purification device (Gas–Bench II, ThermoFinnigan, Bremen, Germany) coupled to a Delta S IRMS (ThermoFinnigan, Bremen, Germany). All solid samples were analysed with the NA-1500/IRMS setup.

All δ¹³C values were expressed relative to the Vienna Pee Dee Belemnite international standard. Different laboratory working standards (glutamic acid, –28.06 ‰ for organic matter samples; air with 500 μmol mol⁻¹ of CO₂, –53.10 ‰ for air samples) were measured after each group of 12 samples to correct for any offset of the IRMS. The precision for isotopic measurements was ± 0.2 ‰, based on repeated measurements of the laboratory working standards. Both IRMS systems were inter-calibrated for gas analyses using the same reference gas as above, revealing a discrepancy of 0.7 ‰ that was removed to the values measured at Hesse to have comparable values between both sites.

2.7 Statistical analysis

Pearson’s correlation coefficients were calculated between δ¹³C_E of each component and climatic data or CO₂ efflux rates solely at Barbeau where measurements were more frequent. All climatic variables from the measurement day and the day before were tested.

Pearson correlations were also established between δ¹³C_E of different components. Non-parametric Kruskal–Wallis rank sum tests should reveal differences among components at each site, and during the two main phenological periods, followed by Mann–Whitney tests to compare one component to another.

A one-way ANOVA was applied to compare δ¹³C_E measurements in the incubation tests with N₂ or CO₂-free air flushings.

All analyses were performed using Statistica (version 7, Statsoft Inc., USA) and R 2.11.1 (R development core team 2010).

3 Results

3.1 CO₂ efflux rates

At both sites, the trunk CO₂ efflux E _T showed a pronounced seasonal evolution and ranged from 10 (April) to 130 (June) μmol CO₂ m⁻³ of living tissue s⁻¹ in Barbeau (Fig. 2a), and from 10 (March) to 88 (July) μmol CO₂ m⁻³ of living tissue s⁻¹ in Hesse (Fig. 2b). Variations of soil CO₂ efflux were also marked, especially in Barbeau where E _S ranged from 0.7 (November) to 5.1 (July) μmol CO₂ m⁻² s⁻¹. In Hesse, it ranged from 0.7 (March) to 1.8 (September) μmol CO₂ m⁻² s⁻¹ but summer efflux data, which are assumed to be highest, are missing. The maximum observed values of trunk and soil CO₂ efflux at Barbeau (Fig. 2a) occurred during the trunk growth period when air and soil temperatures were high (Fig. 3). The lowest CO₂ efflux rates occurred during winter.

Fig. 2

Seasonal changes in trunk and soil respiration rate in Barbeau (oak forest) (a) and Hesse (beech forest) (b), in δ¹³C of emitted CO₂ in Barbeau (c) and in Hesse (d), and in δ¹³C of total organic matter in Barbeau (e) and in Hesse (f), of leaf (filled circles), twig (filled triangles), trunk (filled squares) and total soil (open circles). Vertical dashed lines delimitate the budburst date and the leaf fall period. Before budburst, δ¹³C_E and δ¹³C_OM of leaves and twigs were measured on buds and previous year-twig. The trunk growth period is indicated in gray. Error bars represent standard errors of the mean

Fig. 3

Seasonal changes in rainfall (gray histogram), soil (open triangles) and air (filled triangles) temperature, soil moisture (open diamonds) and air relative humidity (filled diamonds), in Barbeau (oak forest) (a) and Hesse (beech forest) (b). Error bars represent standard errors of the mean

3.2 Carbon isotope composition of emitted CO₂

In oak forest, the carbon isotope composition of leaves (δ¹³C_EL) showed the largest seasonal variations (4.6 ‰, Fig. 2c). It was always higher than δ¹³C of CO₂ emitted by buds in spring, i.e. the first measurement before budburst (−22.3 ‰). It increased during the whole growing season, with a small decrease in June and then decreased from September onwards. δ¹³C_EL showed a similar pattern in Hesse albeit at a smaller observed range (2.3 ‰) and lower values (Fig. 2d).

At both sites, δ¹³C_E of twigs (δ¹³C_ETG) clearly increased between budburst and May (Fig. 2c, d). In September, the values were near those before budburst for oak or those in May for beech. During the trunk growth period, the δ¹³C_ET was lower than the δ¹³C_E of the two canopy components at both sites (Fig. 2c, d). In Barbeau, δ¹³C_ET increased during winter and reached a maximum of −18.9 ‰ in January 2006 (Fig. 2c).

Soil δ¹³C_ES was rather similar at both sites, and seasonal variations were low (Fig. 2c, d). E _S and δ¹³C_ES were not correlated (Table 2).

Table 2 Pearson’s correlation coefficient (r) matrix between carbon isotope signatures of emitted CO₂ in leaves, twigs, trunk, soil and soil and climatic conditions in oak forest or δ¹³C_E of leaves, twigs, trunk and soil

In Barbeau, δ¹³C_ETG was not related to any climatic variable, whereas δ¹³C_EL was related to soil moisture (p = 0.041, Table 2) and δ¹³C_ET was related to air and soil temperature (p = 0.011 and p = 0.012, Table 2) and vapour pressure deficit (p = 0.033). δ¹³C_ES values were only related to soil moisture (p = 0.006). Moreover, δ¹³C_ET was negatively related to E _T and to E _S (r = −0.715; p = 0.030 and r = −0.667; p = 0.050, respectively, Table 2). If only the values from the leafy period were kept, correlations between δ¹³C_EL and δ¹³C_ETG (r = 0.955; p = 0.011), δ¹³C_ES (r = 0.917; p = 0.029) and soil moisture (r = 0.960; p = 0.041), and between δ¹³C_ES and soil moisture (r = 0.966; p = 0.034) were remaining.

From May to September (leafy period), δ¹³C_E exhibited a vertical zonation in both ecosystems, revealing significant differences between components (Kruskal–Wallis ANOVA, p < 0.001 for each site, Fig. 4).δ¹³C_EL values were globally the highest, and δ¹³C_ETG values were lower in Barbeau (p = 0.003) but not in Hesse (p = 0.380). δ¹³C_ETG values were always higher than those of trunk (p = 0.001, both systems) and soil (p = 0.001 in Barbeau and p = 0.002 in Hesse). At both sites, δ¹³C_ET and δ¹³C_ES values were not significantly different (p = 0.07 in Barbeau and p = 0.16 in Hesse) and represented the lowest δ¹³C_E values.

Fig. 4

Mean values (±standard errors of the mean) of δ¹³C_E (bold) and δ¹³C_OM (italic) observed for each component (leaf, twig, trunk and soil) during the leafy period and winter, in Barbeau (oak forest) (left) and Hesse (beech forest) (right)

During leaf fall (October) and winter (January), the vertical zonation was not maintained (Fig. 4). This period was characterised by an increase in δ¹³C_ET.

3.3 δ¹³C of total organic matter

In contrast to δ¹³C_E, δ¹³C of total organic matter (δ¹³C_OM) showed weak temporal variations. At both sites, the most pronounced variations occurred in leaves and twigs with a slight decrease in July and a slight increase towards September (Fig. 2e, f). Variations in δ¹³C_OMT during the year were very narrow, with the same annual average of −26.5 ‰ for the two forests (Fig. 2e, f). δ¹³C_OMS values were stable and averaged −27.01‰ ± 0.06 in Barbeau and −26.21‰ ± 0.08 in Hesse.

When comparing organic matter to CO₂ efflux δ¹³C, differences between δ¹³C_EL and δ¹³C_OML were up to 9.3‰ (Barbeau) and 8.5‰ (Hesse). Differences between δ¹³C_ETG and δ¹³C_OMTG for oak and beech were lower, with maxima of 5.6 and 6.1 ‰, respectively. During spring and summer, the difference between δ¹³C_ET and δ¹³C_OMT was maximum 1.7 ‰ (Barbeau) and ranged between −2.4 and 1.3 ‰ (Hesse). After leaf fall in Barbeau, this difference reached values as high as 7.6 ‰ in January. Differences between δ¹³C_ES and δ¹³C_OMS were always positive in Barbeau (1.0–2.5 ‰) but could be negative in Hesse (−1.6–1.0 ‰.).

Finally, the differences in δ¹³C_OM between organs were lower compared to those for δ¹³C_E. Even if δ¹³C_OMTG was generally less negative than the δ¹³C_OM of other components, no stable vertical zonation was apparent (Fig. 4).

4 Discussion

4.1 Incubation methods for leaf and twig in the field

Field protocols to collect CO₂ efflux from tissue at the top of the canopy showed that δ¹³C_E of attached leaves was similar to that of detached ones (Prater et al. 2005). Furthermore, an in-tube incubation method using CO₂-free air flushes had been validated for δ¹³C_E measurements and showed no difference to online gas exchange measurement (Werner et al. 2007). Here, we complete the methods debate on which approaches are compatible to field conditions, by testing two variants of tissue incubation with CO₂-free air or N₂ flushes. Our results clearly demonstrate that the different gases used do not change the measured δ¹³C_E of leaves or twigs. Tissue incubation in vials previously flushed with N₂ or CO₂-free air can thus be used in the field to sample the emitted CO₂ from current-year branches.

4.2 Vertical zonation of δ¹³C_E from canopy to soil during the leafy period

In both forests, our measured δ¹³C_E values were comparable to those obtained for leaves in deciduous (Hymus et al. 2005) or coniferous forests (Prater et al. 2005), twigs (Damesin and Lelarge 2003), trunks (Damesin et al. 2005; Maunoury et al. 2007; Kuptz et al. 2011) and soil of deciduous forests (Ngao et al. 2005), coniferous forests (Ekblad et al. 2005) or rainforests (Buchmann et al. 1997).

We revealed that during the leafy period, δ¹³C_E values significantly differed among the ecosystem components, overall decreasing from the top of the canopy to the soil (Fig. 4). We observed this zonation at both sites, two different deciduous tree forests in distinct climatic conditions. Interestingly, no such zonation was found by Kuptz et al. (2011) in spruce and beech forests during the leafy period.

When there was a difference in δ¹³C_E between twigs and leaves, it systematically consisted of a ¹³C impoverishment of the CO₂ emitted by twigs relative to leaves. This difference cannot be explained by differences in substrate δ¹³C because leaves have generally a significantly lower δ¹³C for starch and soluble sugars than twigs (Damesin and Lelarge 2003; Eglin et al. 2009). The gap between δ¹³C_EL and δ¹³C_ETG might be related to a difference in the balance of CO₂ released by either pyruvate decarboxylation (resulting in ¹³C-enriched CO₂) or by the Krebs cycle (resulting in ¹³C-depleted CO₂, Tcherkez et al. 2003; Gessler et al. 2009).

The main hypotheses explaining differences in δ¹³C_E between twigs and trunks are (1) an isotope discrimination during the assimilate transport in the phloem along twigs and trunk (Damesin and Lelarge 2003; Gessler et al. 2007), (2) changes in PEPc activity (Gessler et al. 2009; Kuptz et al. 2011), known to discriminate against ¹²C during carbon fixation (Cernusak et al. 2009), (3) or a substantial contribution of belowground-evolved CO₂ brought by the xylem sap stream to the upper part of the tree, i.e., in the trunk (Aubrey and Teskey 2009; Grossiord et al. 2012). Other processes may contribute to the ¹³C impoverishment in trunk compared to twigs, especially the progressive mixing along metabolite translocation from sun and shade leaves via twigs to trunk (Eglin et al. 2010). It will be interesting to address in future studies in more detail the reasons for the differential isotope discrimination in twigs and trunk found in our study.

Comparable values of isotope composition of CO₂ efflux in trunk and soil may be explained by the coupling of both components via C assimilates. Carbon substrates are rapidly transferred in broadleaved species from trunk to roots and via root exudates also to soil microorganisms (Dannoura et al. 2011; Epron et al. 2011). The comparable values for trunk and soil also suggest that contrary to substrate translocations from leaves to twigs, there is no apparent isotope discrimination during carbon translocation from trunk to soil.

4.3 Seasonal variations in δ¹³C_E

Seasonal ranges of emitted CO₂ δ¹³C were in agreement with ranges previously observed for beech leaves (Eglin et al. 2009), beech twigs (Damesin and Lelarge 2003), oak trunks (Maunoury et al. 2007) and hardwood forest soil (Mortazavi et al. 2005; Marron et al. 2009). Our maximum δ¹³C_E values of leaves and twigs have been obtained during summer, in agreement with δ¹³C_E of a coniferous forest ecosystem (Bowling et al. 2002; Mortazavi et al. 2005) but oppositely to δ¹³C_E of a deciduous forest ecosystem (Mortazavi et al. 2005).

δ¹³C_E seasonal variations in leaves might be related to changes in respiratory substrate δ¹³C due to variable ¹³C discrimination during C assimilation. The latter is itself linked to environmental conditions like soil moisture (Mortazavi et al. 2005). The correlation between δ¹³C_EL and δ¹³C_ETG (Table 2) suggests that δ¹³C_ETG variability is linked to the same mechanisms as that of δ¹³C_EL. The increase in δ¹³C_ET in winter might be explained by a switch of respiratory substrates from photosynthesis-derived sugars (lower δ¹³C) during the leafy period to stored carbohydrates, i.e. starch with higher δ¹³C, in the dormancy period as suggested before (Maunoury et al. 2007; Kuptz et al. 2011). Another explanation for these winter values is that, during winter time, transpiration is null and δ¹³CE_T was no more influenced by a possible contribution of belowground-evolved CO₂. Surprisingly, we did not observe any increase in δ¹³C_E for twigs, which, like trunks, probably use starch as main respiratory substrate. This unexpected difference between trunk and twig δ¹³C_E may be explained by a metabolic discrepancy such as, e.g. differences in PEPc activity, during winter. Correlations between δ¹³C_ET and E _T or air temperature have already been observed by Maunoury et al. (2007) in the same oak forest, with comparable correlation coefficients, suggesting that the respiration rate, which is influenced by air temperature, affects δ¹³C_ET.

During the leafy period, respiratory substrates are partially derived from C recently assimilated by leaves and transported by the phloem towards the trunk base and roots (Dannoura et al. 2011; Epron et al. 2011) and may finally end up via root exudates as organic matter in the soil. Such a substrate similarity between trunk and soil may result in δ¹³C_ES values close to those measured for trunk during the leafy period. In contrast, during winter, a lack C assimilation lowers dramatically this recent C source for soil respiration, while trunk stored compounds with higher δ¹³C values supply substrates for E _T. Furthermore, carbon supplies to the soil via root exudates are much lower or nil during winter. In addition, the heterotrophic component, i.e. microbial respiration, contributes more to the soil CO₂ efflux and to δ¹³C_ES than during the leafy period (Epron et al. 2001). This decoupling between trunk and soil respiratory substrate pools therefore may explain the significant differences in δ¹³C_E of both components in winter.

At the soil and ecosystem level, temporal variations of δ¹³C_E have been explained by environmental conditions with or without time lag (from 1 to 10 days) before measurement (e.g. Bowling et al. 2002; Ekblad et al. 2005; Marron et al. 2009). The present study showed a differential impact of environmental conditions on the δ¹³C_E of the different components studied., Specifically, in contrast to other studies, we found no correlation between δ¹³C_EL or δ¹³C_ETG and climate (Mc Dowell et al. 2004; Mortazavi et al. 2005), which might be explained by the lower measurement frequency in our study. Yet, we found several correlations of environmental factors with δ¹³C_ET, suggesting that the trunk data integrate the overall climate effect on the tree. This argument can be supported by the facts that (1) a time lag exists for substrate transport in phloem from leaves to soil (Barnard et al. 2007) and (2) a progressive mixing of several different substrates (e.g. recent vs. stored compounds, top vs. bottom of tree canopy) along this C translocation.

Lastly, the general enrichment of δ¹³C_E in comparison with δ¹³C_OM has been classically highlighted in previous studies (Damesin and Lelarge 2003; Klumpp et al. 2005; Maunoury et al. 2007). Our results show a mismatch between the δ¹³C_E of each component and δ¹³C_OM of the mature leaf, but these discrepancies were not stable throughout the year. Thus, our results invalidate the hypothesis of Bowling et al. (2008) that bulk leaf δ¹³C can be used as a reference value to predict differences between δ¹³C_E and δ¹³C_OM of plant components and ecosystem.

5 Conclusion

Our study highlights isotopic differences of CO₂ emitted by the top (sun leaves and twigs) and the bottom of the forest (trunk base and soil) both with regard to higher δ¹³C values during the leafy period (higher at the top) and of seasonal dynamics (higher at the top). Variations in substrate δ¹³C (via the use of stored compounds or a mixing effect) might be the major—but not the only—explanation for these differences. Our study confirmed that δ¹³C of CO₂ emitted by the forest components cannot be deduced from the δ¹³C of the total organic matter of the component, or from CO₂ efflux intensity. Nowadays, high frequency measurements by tunable diode laser spectroscopy allow the analysis of temporal dynamics in δ¹³C_E (Marron et al. 2009), especially during winter time, and offer thus better possibilities to understand inherent variability its link to metabolic processes.