Seasonal variations of belowground carbon transfer assessed by in situ 13 CO 2 pulse labelling of trees

Introduction Conclusions References


Seasonal variations of belowground carbon transfer assessed by in situ 13CO2 pulse labelling of trees
Soil CO 2 efflux is the major biospheric source of carbon from terrestrial ecosystems to the atmosphere and it accounts for a large fraction (40-70 %) of total ecosystem respiration (Janssens et al., 2001). It includes the respiration of microorganisms and soil meso-or macroorganisms involved in the mineralisation of soil organic matter ("heterotrophic respiration"), and the use of recently assimilated substrates that fuels root metabolism and rhizospheric microorganisms that Published by Copernicus Publications on behalf of the European Geosciences Union.
are either associated to roots (symbionts) or that feed on root exudates ("rhizospheric" respiration, Subke et al., 2006). The transfer of photosynthetic products belowground therefore fulfils energy requirements of metabolic processes occurring in the root and associated rhizosphere.
Up to recently, temporal variations of soil CO 2 efflux were mostly ascribed to environmental drivers (e.g. Bahn et al., 2010;Epron et al., 1999;Raich et al., 2002), but there is now a growing amount of evidence that soil CO 2 efflux is strongly linked to plant activities (Högberg and Read, 2006;Ryan and Law, 2005;Trumbore, 2006). Beyond the relationships between ecosystem productivity and soil CO 2 efflux at a regional scale (Janssens et al., 2001), temporal variations in soil CO 2 efflux were recently ascribed to short-term changes in microclimatic conditions (e.g. photon flux density, vapour pressure deficit, air temperature) that affect canopy photosynthesis (Craine et al., 1999;Gaumont-Guay et al., 2008;Tang et al., 2005). Not only the total flux of soil CO 2 efflux but also its 13 C composition was correlated to short-term fluctuations in microclimate confirming a close coupling between photosynthesis and soil CO 2 efflux (Bowling et al., 2008;Ekblad et al., 2005;Marron et al., 2009;Mortazavi et al., 2005).
Root and root-derived carbohydrates contribute to 50-60 % of soil CO 2 efflux in forest ecosystems (Subke et al., 2006) highlighting the effect of photosynthesis on soil CO 2 efflux. Soil CO 2 efflux will therefore vary according to temporal change in canopy photosynthesis but the coupling will depend on both the velocity of carbon transfer belowground (Kuzyakov and Gavrichkova, 2010;Mencuccini and Hölttä, 2010;Plain et al., 2009) and on the partitioning of photosynthates among several pools and organs within the tree, i.e. on carbon allocation (Litton et al., 2007). The contribution of recent photosynthate to soil CO 2 efflux further depends on the partitioning of carbon allocated belowground to growth, storage and respiration of both root and microorganisms feeding on root-derived carbohydrates (Ryan and Law, 2005). This will affect the residence time of carbon in the tree-soil system before it returns to the atmosphere.
Total annual belowground carbon allocation can be successfully derived from soil CO 2 efflux and litterfall using a mass balance equation (Giardina and Ryan, 2002). It varies with stand age and resource availability (Litton et al., 2007). It increases with mean annual temperature at global scale (Litton and Giardina, 2008). However, there is still little information on the seasonal variations of belowground carbon allocation in forest ecosystems. The strength of the different carbon sinks is thought to change seasonally and among species (Körner, 2003). Differences in carbon allocation patterns in spring between diffuse porous and ring-porous species are well-known . The latter should start their radial stem growth prior leaf expansion to restore hydraulic conductivity after winter embolism of large vessels (Bréda and Granier, 1996). Similarly, evergreen and deciduous trees exhibit different seasonal patterns of the contribution of stored and recent carbon to respiration (Kuptz et al., 2011). Besides phenology, a direct control of allocation by environmental drivers cannot be excluded. Drought reduced the contribution of recent photosynthate to soil CO 2 efflux and the transfer of carbon into the microbial biomass in young potted beech (Ruehr et al., 2009). Both soil water content and temperature affected carbon allocation to microbial biomass in Ericaceae species (Gorissen et al., 2004) and the velocity of carbon transfer in the trunk of oak and pine (Dannoura et al., 2011). In the latter study, we demonstrated that pulse labelling of field-grown trees with 13 CO 2 and high frequency of measurement of 13 C in respiratory efflux during the chase period using laser-based infrared gas analysers is a powerful approach to study seasonal variations of carbon transfer (see also Bahn et al., 2009, on grassland).
The aim of this study was to assess the belowground transfer of recently assimilated carbon in two deciduous broadleaved species (beech and oak) and one coniferous evergreen species (maritime pine) at distinct phenological stages during the growing season. 13 CO 2 recovery in soil CO 2 efflux was recorded for 3 weeks by tuneable diode laser absorption spectrometers (TDLAS) after pulse labelling of 8-10 m tall trees and compared to 13 C recovery in both root and microbial biomass and their respective respiratory CO 2 . We hypothesised that (i) the transfer time would vary between broadleaved and coniferous species due to difference in anatomy that affect the velocity of carbon transport in the phloem (Dannoura et al., 2011) and growth rhythm (deciduous versus evergreen species), (ii) the amount of 13 C recovered in soil CO 2 efflux would vary along the season reflecting changes in carbon allocation related to phenology, temperature and soil water content, and (iii) 13 C would be first recovered in the roots and later in the microbial biomass.
Labelling was done three to four times covering the whole leafy season for all species. Three trees were selected for each labelling date in all sites and two of them were labelled. Tree height ranged between 8 to 10 m in all stands (DBH are given in Table 1). A 0.5-0.6 m deep trench was dug around each tree at least 5 months before labelling to limit possible Biogeosciences, 8, 1153Biogeosciences, 8, -1168Biogeosciences, 8, , 2011 www.biogeosciences.net/8/1153/2011/ Table 1. Diameter at 1.3 m height (DBH, cm), cumulative photosynthetic photon flux density (PPFD, mol m −2 d −1 ) over 24 h following the beginning of labelling, air temperature (T A , • C) averaged over 24 h following the beginning of labelling, volumetric soil water content (SWC, m 3 m −3 ) on the day of labelling. Soil CO 2 efflux (F S , mgC m −2 h −1 ), root mass per unit area (M R , gDM m −2 ), specific root respiration (R R , mgC kg −1 DM h −1 ), microbial carbon per unit area (C M , gC m −2 ) and microbial respiration (R M , mgC kg −1 soil h −1 ) were averaged over the entire chase period (20 days). Time lag in 13 C recovery in soil CO 2 efflux (L, day), asymptotic value of cumulative label recovered in soil CO 2 efflux (CLR FS (∞) in mg 13 C) and half residence time (t1/2, day) were estimated by fitting cumulative label recovered in soil CO 2 efflux to Eqs. (9-11).
Date of labelling confounding effects of root cutting on carbon allocation. The trench was lined with a polyethylene film and filled back. All roots and root exudates within this soil volume therefore originated from the isolated tree, and were contained in this trench volume. The areas delimited by the trench were about 3 m 2 in the beech (2.9-3.8) and oak stands (2.2-5.7), and 6 m 2 in the pine stand and were larger than the average areas per tree (inverse of tree density). In early spring 2009 and at least three months prior to the labelling to allow enough time for ingrowth of mycorrhizal hyphae, root exclusion cores were installed in those trench plots that were selected for the labelling experiment. A root auger (Ø 8 cm; depth of corer 15 cm) was used to extract intact soil cores to a depth of 15 cm. Upon removal of the auger, the soil was gently pushed out with a crank and wrapped in a 30 µm nylon mesh bag of similar dimensions (BUISINE, Clermont de l'Oise, France) and then put back in the core hole. The 30 µm mesh excludes root growth in the cores so that the main passage for labelled carbon into these root exclusion cores is by hyphae of mycorrhizal fungi. Of course, passive diffusion or motile bacteria mediated transfer of labelled carbon cannot be excluded. During the chase period, the root exclusion cores were retrieved and compared to normal cores collected freshly with the same auger.
In April 2009, rainfall exclusion roofs, made with polyethylene film and supported by a woody frame (3×3 m), were installed 1.5 m above the forest floor on two beech trees to divert rainfall from the soil delimited by the trench. Predawn leaf water potential was measured once on 3 to Photosynthetic photon flux density (PPFD; Delta T-BF2, Cambridge, UK for beech, home-made quantum sensor using a gallium arsenide photodiode for oak and DBE, Solems, Palaiseau, France for pine), air temperature (T A ; Vaisala HMP45, Helsinki, Finland) and soil temperature at 10 cm depth (T S , home-made copper-constantan thermocouples) were recorded half-hourly at each site. Values were averaged for 24, 48 or 72 h after labelling. Soil water content (SWC) in the vicinity of each labelled tree inside the area delimited by the trench was recorded half hourly at 30 cm depth for pine with TDR probes (CS616, Campbell Scientific, Logan, UT, USA) or at 10 cm depth for oak with an impedance probe (ML2x ThetaProbes, Delta-T Device, Cambridge, UK), or measured once a week at 15 cm depth for beech using TDR probes (Trase, SoilMoisture Equipment Corp., Santa Barbara, CA, USA).

Crown labelling
The whole crown of the tree was inserted into a 20 to 40 m 3 chamber made of 200 µm polyane film (celloflex 4TT; Prosyn polyane, Saint-Chamond, France) and held by two 12 m height stainless steel scaffoldings put up on both sides of the tree . Air temperature inside the crown labelling chamber was recorded, controlled and maintained at the outside air temperature. Evolution of [ 12 CO 2 ] and [ 13 CO 2 ] inside the chamber was monitored with a 12 CO 2 / 13 CO 2 infrared gas analyser (S710, SICK/MAIHAK, Reute, Germany, accuracy of 5 %). A total amount of 50.4 g of pure 13 CO 2 (99.299 atom %, Eurisotop, Cambridge Isotope Laboratory Inc., Andover, MA, USA) was progressively injected for 1 to 5 h at a flow rate adjusted to maintain the 13 CO 2 mixing ratio in the chamber between 300 and 400 µmol mol −1 , except in November and February for pine where higher 13 CO 2 mixing ratios (800-1000 µmol mol −1 ) were used to compensate a lower photosynthetic activity (see Dannoura et al., 2011;Plain et al., 2009 for more details). By only inserting the crown into the labelling chamber, we avoided contamination of the soil atmosphere by diffusion of 13 CO 2 into the soil pores that would later have back-diffused into the atmosphere leading to an artefact in soil CO 2 efflux isotopic composition (Subke et al., 2009).

13 C in soil CO 2 efflux
The isotope composition of soil CO 2 efflux (δ 13 C FS ) was computed from [ 12 CO 2 ] and [ 13 CO 2 ] measured at the inlet and outlet of flow-through soil CO 2 efflux chambers (Marron et al., 2009;Plain et al., 2009) by tuneable diode laser absorption spectroscopy with a trace gas analyzer (TGA 100A; Campbell Scientific, Logan, UT). Two or three collars were installed for each tree at 60 cm from the trunk. Air was pumped continuously through the chamber at a flow rate ranging from 18 to 72 dm 3 h −1 . The chamber design ensured that differences in pressure between the inside and the out-side of the chamber remained below 10 −2 Pa. A manifold was used to switch between working standards and the chamber inlet and outlet lines. A mean mixing ratio was recorded over 20 s after a 30 s stabilisation. The isotope composition of each working standard that was 0.5 % certified for CO 2 mixing ratios (Air Products, Paris, France for beech, Air Liquide, Paris, France for oak and DEUSTE Steininger GmbH, Mühlhausen, Germany for pine) was measured by an isotope ratio mass spectrometer (IRMS, Delta S, ThermoFinnigan, Bremen, Germany). The ranges of available mixing ratios were respectively from 293.2 to 1281.2 µmol mol −1 for 12 CO 2 and from 3.17 to 13.75 µmol mol −1 for 13 CO 2 . The precision of the instrument at reproducing calibration tank values was 0.3 µmol mol −1 and 0.007 µmol mol −1 for 12 CO 2 and 13 CO 2 (Dannoura et al., 2011).
The isotopic composition of soil CO 2 effluxes (δC FS , ‰) was calculated as: where R VPDB is the isotopic ratio of Vienna Pee Dee Belemnite (Coplen et al., 2002). Total CO 2 mixing ratio ([CO 2 ], µmol mol −1 ) was calculated from the mixing ratios of individual isotopologues by: where f other is the fraction of CO 2 containing all isotopologues other than 12 C 16 O 16 O and 13 C 16 O 16 O, and assumed to be 0.00474 (Griffis et al., 2004).

13 C in respiratory CO 2 of roots and soil microorganisms
Two soil cores (20 cm depth) were collected 1, 3, 6, 9 and 15 days after labelling and sliced into 3 parts (0-5, 5-10 and 10-20 cm). The soil was immediately sieved through a 2mm mesh and living fine root fragments (diameter <2 mm) were picked up and rinsed. Coarse root fragments were discarded. For each core, the fine roots collected at all depths were pooled together. Oak and pine roots of the two cores were further pooled to obtain a sufficient sample size while beech roots were kept separately for each core. Each root sample was then enclosed in a 125 ml glass flask. A pump was used to circulate gas from the flask to an IRGA (EGM1, PPSystems, Hitchin, UK, LI840, Licor, Lincoln, USA or S710, SICK/MAIHAK, Reute, Germany). The CO 2 initially inside the flask was removed using a soda lime trap. After 10 min, the soda lime trap was by-passed and the increase in CO 2 mixing ratio in the flask was recorded until reaching values above 400 ppmv. Air was then sampled in a 10 ml Exetainer glass-vial (Labco limited, High Wycombe, UK) or 25-ml glass flask for IRMS analyses. Root-free soil subsamples (approx 150 g) from 0-5 and 5-10 cm depth were incubated similarly in 250 ml flask. Soil from the two cores Biogeosciences, 8, 1153-1168, 2011 www.biogeosciences.net/8/1153/2011/ was pooled for oak. The isotopic composition (δ) of CO 2 in sampled air was measured within a week IRMS (Delta-S, Finnigan-Mat, Thermoquest Corp., San Jose, CA, USA, for beech and pine samples or VG Optima, Fison, Villeurbanne, France, for oak samples) coupled to a gas purification device (Gas-Bench II, ThermoFinnigan, Bremen, Germany). Incubations were done on all beech and oak trees but for pine only those labelled in June. Temperature during incubation was 19-22 • C for beech and oak, and 22-24 • C for pine.

13 C in root and microbial carbon
After incubation, roots were immediately plunged in liquid nitrogen, transferred into a −80 • C freezer and finally freezedried. Freeze-dried root samples were ground, weighed and analysed for carbon isotope composition and total carbon using an elemental analyser (NA 1500 NCS, Carlo Erba, Milan, Italy) coupled to IRMS. After incubation, soil subsamples were stored at 4 • C for less than 4 days before processing for microbial carbon. The δ 13 C of soil microbial carbon was determined from δ 13 C of soluble carbon extracted from two 40 g of soil subsamples. One subsample was fumigated for 24 h with chloroform vapour, while the other was not fumigated (Vance et al., 1987). Extraction was performed using 0.05 M of K 2 SO 4 for one hour under vigorous shaking. The extracts were filtered, analysed for carbon content using a total organic carbon analyser (Shimadzu TOC-VCSH, Tokyo, Japan), either freeze-dried or oven-dried at 60 • C (we observed no difference between the two methods) and then analysed for carbon isotope composition IRMS. Microbial carbon was also extracted from the deepest part of the soil cores in beech and pine (10-20 cm).
The isotope composition of soil microbial carbon (δ 13 C M ) was calculated as: where C fum and C nonfum (mgC kg −1 soil ) refer to fumigated and non fumigated extracts.
Soil microbial carbon was computed as: where 0.45 is a correction factor used to account for the nonextractable fraction of microbial carbon (Vance et al., 1987). However, the fumigation efficiency in the sandy soils of the pine stands can be higher since the microbe protection is probably low in these unstructured soils (Achat et al., 2010). The isotope composition of microbial carbon was also estimated in the two root exclusion cores that were collected 3 and 6 days after labelling in beech and oak, and 6 and 9 days after labelling in pine, and sliced into two parts (0-5 and 5-10 cm).
Gravimetric soil water content was determined by comparing the mass of approx. 40 g of soil before and after drying at 105 • C.

Calculations
The relative abundance of 13 C in any efflux or compartment (Ab) was calculated from the isotope composition as: Ab was corrected for the background 13 C content measured on unlabelled trees or on the same tree one day before labelling (Ab UN ), and multiplied by X, X being either the total carbon content in a compartment (root, C R , or microbial carbon, C M ) or the CO 2 flux to estimate the excess amount of 13 C (Excess 13 C) in that efflux or compartment: Root and microbial respiration (R R and R M , mgC kg −1 h −1 ) were calculated from the linear slope of the increase in CO 2 mixing ratio in the incubation flasks (d[ where P is the atmospheric pressure (Pa), V is the total volume including flask, tubes and IRGA cells (m 3 ), M is the dry mass of the root or of the soil (kg), T is the temperature (K), and 8.314 J mol −1 K −1 is the ideal gas constant. Soil CO 2 effluxes (F S , mgC m −2 h −1 ) were calculated as: where F is the air flow through the chamber (m 3 h −1 ) and S is the surface of soil chamber (m 2 ). The cumulative label recovered in soil CO 2 efflux (CLR FS ) was calculated by summing daily averages of 13 C in excess in the soil CO 2 efflux.
The kinetics of the label recovered in soil CO 2 efflux were described using a four-pool model similar to those describing the supply of carbon for leaf growth (Lattanzi et al., 2005) or for root respiration (Lehmeier et al., 2008). The model was fitted on the observed CLR FS values (Fig. 1). A quantity of the label that will be respired belowground (Q) leaves the crown pool (C) at a rate following a first-order kinetics and arrives after a time lag (L) into an active belowground pool (B1) where it can be transferred and back transferred to an inactive pool (B2) at rates following also a first-order kinetics.  Open circles are measured data and the curves represent the best fits of Eqs. (9) and (11), respectively. The example shown is for pine labelled on 3 August 2009. Asymptotic value of CLR FS (CLR FS (∞)), half residence time of soil respired 13 C in the whole plant-soil system (t1/2) and time lag (L) between the start of the labelling and the first appearance of 13 C in soil CO 2 efflux are depicted on the graphs. In this example, the best-fitted rate constants are k CB1 = 0.022, k B1B2 = 0.001, k B2B1 = 0.130 and k B1A = 0.022.
Finally, the label leaves the substrate pool in a similar way and accumulates in the atmosphere (A). The rates of change in the different pools are: where k CB1 , k B1B2 , k B2B1 and k B1A are the rate constant of transfer between the crown to the B1 pool), between B1 and B2 pools (transfer and back transfer), and between B1 and the atmosphere (respiration). The cumulative label recovered in soil CO 2 efflux at any time t is therefore: and the asymptotic value, CLR FS (∞) is equal to Q C (0). CLR FS (∞) multiplied by the trenched surface represents the amount of carbon allocated to belowground respiration. The B1 pool is a metabolically active pool that contributes to soil CO 2 efflux while the B2 pool is metabolically inactive, and does not contribute to soil CO 2 efflux. We also unsuccessfully tested a two-metabolically-active-pool model that would have represented roots and microorganisms. We are aware that our model is oversimplified and that it does not describe the complexity of the fate of carbon belowground. We consider this model as an empirical one acknowledging that the B1 and B2 compartments are virtual pools, and that calculating half residence time of label within each pool from the adjusted rate constants would have been misleading. We therefore used the model for estimating the seasonal variation in the amount of carbon allocated to soil CO 2 efflux and we calculated the half residence time of soil respired 13 C into the whole plant-soil system (t1/2) as the time needed to reach 50 % of CLR FS (∞).
The time lag was not estimated from Eq. (9) but determined by fitting a quadratic function to the relationships between Excess 13 C in F S and the time after labelling ( Fig. 1) using non linear least-squares regression (PROC NLIN of SAS software, Marquardt-Levenberg method, SAS Institute Inc., Cary, NC, USA): The model predicted satisfactorily CLR FS values at 7, 14 and 20 days after labelling across species and seasons (R 2 = 0.99, p < 0.001).

Statistical analyses
Data were analysed separately for each species using SAS software (SAS Institute Inc., Cary, NC, USA). Two-way ANOVA (tree and sampling date, nested within tree) were used for Excess 13 C in F S , C R and R R and three-way ANOVA (tree, depth and sampling date nested within tree and depth, including interaction between tree and depth when the two cores were not pooled) were used for 13 C in excess in C M and R M . A four-way ANOVA was used to compare 13 C in excess in C M of normal cores and root exclusion cores (tree, depth, core type nested within tree and depth and sampling date nested within tree, depth and core type, including Biogeosciences, 8, 1153Biogeosciences, 8, -1168Biogeosciences, 8, , 2011 www.biogeosciences.net/8/1153/2011/ interactions between tree and depth when the two cores were not pooled). The residual errors of each model are shown as vertical bars in the figures. Linear correlations between the different parameters and environmental variables were tested with a two-tailed test for significance.

Excess 13 C in soil CO 2 efflux
Soil CO 2 efflux (F S ) varied within a similar range in the three species according to seasons, with highest values in July for beech and oak and lowest values in February for pine (Table 1). A significant Excess 13 C in F S was rapidly observed in beech and oak (Fig. 2) after a time lag (L, Table 1) of less than 0.5 to 1.5 days. In oak, L was negatively correlated to cumulative PPFD on the day of labelling (R = −0.86, p < 0.05). Low cumulated PPFD were observed in September. L was higher in pine than in the two broadleaved species during the growing season, including October labelling (1.6 to 2.5 days) and increased markedly in November and February, up to 4.4 days, and this increase was clearly related to the drop in temperature. In pine, L was indeed negatively correlated to both air and soil temperature, with the highest correlation between L and soil temperature averaged for 4 days since the start of the labelling (R = −0.84, p < 0.01).
In pine, the lowest CLR FS (∞) values were observed for trees labelled in November and February, reflecting lower F S values in these seasons (Table 1). CLR FS (∞) was indeed positively related to both air and soil temperature, with the highest R between CLR FS (∞) and air temperature averaged for 24 h since the start of the labelling (R = 0.84, p < 0.01). The marked influence of temperature on the shape of the time courses of Excess 13 C in soil CO 2 efflux (Fig. 2) may explain the negative correlation between CLR FS (∞) and L in this species (R = −0.92, p < 0.01), which was not observed for the two broadleaved species. In beech and oak, CLR FS (∞) was not related to any of the measured environmental factors. Despite the lack of relationships with climatic factors, there were clear seasonal differences in CLR FS (∞) in beech with the higher CLR FS (∞) observed in July (Table 1). Despite lower soil water content under rainfall exclusion roofs, this had no marked influence on CLR FS (∞) in beech.
Half residence time of soil respired 13 C in the tree-soil system averaged 6.9 and 6.5 days in beech and oak respectively with no clear seasonal trend. It increased markedly in pine in November and February compared to June, August and October (Table 1), with the highest correlation between t1/2 and air temperature averaged for 72 h since the start of the labelling (R = −0.77, p < 0.05). In oak and in pine, t1/2 and L are correlated (R = 0.87, p < 0.05 and R = 0.84, p < 0.01, respectively) but not in beech.

Excess 13 C in root and microbial respiration and root and microbial carbon
Fine root biomass (0-20 cm depth) was higher in beech (255 gDM m −2 , values for each tree are in Table 1, with DM for dry matter) than in oak (96 gDM m −2 ) and pine (50 gDM m −2 ), while specific root respiration was similar for beech and oak (87 mgC kg −1 DM h −1 ) and much higher in pine (411 mgC kg −1 DM h −1 , measured only in June 2009). Microbial carbon (0-10 cm depth) was higher in oak (68 gC m −2 ) than in beech (39 gC m −2 ) and pine (22 gC m −2 ). Microbial respiration was higher in oak (5.1 mgC kg −1 soil h −1 on average) than in beech (3.6 mgC kg −1 soil h −1 ), with lower values in August especially for beech trees under rainfall exclusion roofs, and lower in pine (2.2 mgC kg −1 soil h −1 , measured only in June 2009). An excess amount of 13 C was already found one day after labelling in the root and microbial compartments (both carbon and respiration) in oak and beech. It occurred later in pine, i.e. three days after labelling during the growth season (June to October) and six days after labelling in November and February. Excess 13 C in the microbial compartment was higher in the top soil (0-5 cm) than at deeper soil depths for all species.
The patterns of Excess 13 C in microbial (Fig. 3) and root (Fig. 4) respiration and in microbial (Fig. 5) and root biomass (Fig. 6) were on the whole consistent with TDLAS data (Fig. 2). The mean value of Excess 13 C in F S calculated over the chase period was well related with mean Excess 13 C in R R in beech (R = 0.86, p < 0.01), better than with Excess 13 C in R M (R = 0.77, p < 0.05). A similar correlation was observed between Excess 13 C in F S and Excess 13 C in C R (R = 0.91, p < 0.01) while no correlation was observed between Excess 13 C in F S and Excess 13 C in C M . In oak, Excess 13 C in F S was only related to Excess 13 C in R R (R = 0.85, p < 0.05) while no correlation was observed between Excess 13 C in F S and Excess 13 C in R M , C M or C R . In pine, we found a weak correlation between Excess 13 C in F S and Excess 13 C in C R (R = 0.72, p < 0.001).
Excess 13 C in C M and R M was significantly lower in root exclusion cores compared to normal cores for beech, especially in July while CLR FS (∞) is maximal (Fig. 7). A similar but not significant difference between normal and root exclusion cores was observed for oak and pine.

Transfer time of carbon belowground
Time lags estimated for beech and oak ranged from 0.5 to 1.3 d and Excess 13 C peaked 2 to 3 days after labelling, a value similar to those observed on 3 m tall poplars (Horwath et al., 1994;Mikan et al., 2000). The 13 C peak was observed after one day in 0.5 m tall beech seedlings (Ruehr et al., 2009) and after 0.5 d in silver birch seedlings (Pumpanen et al., 2009). Considering the path length from tree leaves to the soil in our experiment, the transfer times we report are shorter. This might be due to a better time resolution of measurements based on tuneable diode laser absorption spectrometry compared to those based on off-site isotope ratio mass spectrometry.
No other labelling experiments on tall trees were reported so far in natural conditions in the field. The transfer time of photosynthates to ecosystem respiration has been indirectly estimated by tracing fluctuation in natural abundance of 13 C related to climate-induced variations in discrimination during photosynthesis. Time lags of three to five days observed on 8-9 m tall Douglas fir (Bowling et al., 2002;McDowell et al., 2004) and in a loblolly pine plantation and a mixed hardwood forest (Mortazavi et al., 2005), which is higher than those observed in this study during the active growing season (1.6-2.7 days for pine).
In contrast with beech and oak for which recovery of 13 C was observed in soil CO 2 efflux after a time lag of less than 1 day on average during the growing season, the time lags for pine ranged between 1.6 to 2.7 days from June to October, and more than four days in the resting season (November and February). The tracer peaked after 4 to 6 days in the first period, a range that is in agreement with those reported for 2-m tall Scots pines (Högberg et al., 2008) and 4-m tall black spruce , after considering the difference in tree size. Similar differences were reported after 14 C labelling of small potted seedlings in the laboratory between coniferous species (Norway spruce and Scots pine) and silver birch (Pumpanen et al., 2009). The markedly different transfer time of 13 C belowground between the two broadleaved species and the pine highlights differences in the velocity of photosynthate transport via the phloem sap between angiosperm and gymnosperm that were related to differences in phloem anatomy (Dannoura et al., 2011;Wingate et al., 2010). The doubling of time lag and peak time between summer and winter time, as well as the delayed peak time (more than 10 days in the resting season) were consistent with an effect of temperature on phloem loading, especially on retrieval after leaching (Peuke et al., 2006), on the viscosity of phloem sap (Hölttä et al., 2009) or on the carbohydrate sink strength (Wingate et al., 2010) that may affect the velocity of phloem sap (Dannoura et al., 2011).
The soil tortuosity and moisture might have influenced the transfer time of CO 2 from roots to the soil surface (Mencuccini and Hölttä, 2010;Stoy et al., 2007). However, this putative effect would have led to an underestimation of the contrast found between broadleaved and coniferous species since the pine trees of our study were on a sandy soil with higher porosity than the soils of the two broadleaved trees. In addition, the close coupling between soil surface efflux and incubation of either root or root-free soil indicates that time lags related to transport in the soil air space are negligible.  Fig. 3. Time courses of Excess 13 C in microbial respiration (R M ) after whole crown pulse labelling of beeches (top), oaks (middle) and pines (bottom) at 0-5 cm depth (left) and 5-10 cm depth (right). Each tree is identified by its date of labelling. There are significant differences between trees for beech (F 7,78 = 23.1, p < 0.001) and oak (F 5,23 = 13.3, p < 0.001), between depths (F 1,78 = 44.9, F 1,23 = 31.7 and F 1,20 = 15.8 for beech, oak and pine respectively, p < 0.001) and between sampling dates (F 64,78 = 4.6, F 23,23 = 3.5 and F 16,20 = 3.2 for beech, oak and pine respectively, p < 0.001). The verticals bars represent the root mean error. Data are presented on two different panels according to depth but depth was included in the ANOVA.
The rainfall exclusion experiment conducted on two beech trees in August did not reveal any change in transfer time in contrast to what was observed on potted beech seedlings (Ruehr et al., 2009). The authors did not measure leaf water potential but both stomatal conductance and net photosynthesis of sapling leaves were markedly decreased in their exper-iment. Despite marked differences in soil water content (0-15 cm depth), there was however no difference in predawn leaf water potential (−0.4 MPa) in our study indicating that we did not impose any substantial drought that would have affected the rate of phloem transport (Hölttä et al., 2009 Fig. 4. Time courses of Excess 13 C in root respiration (R R ) after whole crown pulse labelling of beeches (top), oaks (middle) and pines (bottom). Each tree is identified by its date of labelling. There are significant differences between trees for beech (F 7,42 = 4.1, p < 0.01) and oak (F 5,15 = 3.0, p < 0.05) and between sampling dates for beech (F 32,42 = 4.7, p < 0.001). The verticals bars represent the root mean error.

Amount of carbon allocated to soil CO 2 efflux
It was shown that a simple exponential decay function failed to describe complex kinetics of the label recovered in soil CO 2 efflux which was better described by the sum of two exponential functions . Compartmental models have proven to be an effective and elegant tool to analyse the supply of carbon to root respiration in grass cultivated in quartz sand in a growth cabinet (Lehmeier et al., 2008(Lehmeier et al., , 2010a. The compartment model (Eq. 9) used to estimate the amount of 13 C allocated to soil CO 2 efflux assumed that the labelled substrate leaving the crown pool will fill a metabolically active pool that contributes to soil CO 2 efflux after a time lag where it can be transferred and back transferred to a metabolically inactive pool. In grass, more than half of respired carbon cycled through a short lived storage pool before being respired (Lehmeier et al., 2008). The situation encountered by trees in the field under fluctuating climatic conditions is much more complex than in the Lehmeier's studies and these two virtual pools have no sound physiological bases. At several occasions, this inactive pool was not required in pine (i.e. the best fit was obtained with k B1B2 = 0). A rough estimation of pool size from rate constants (Lehmeier et al., 2010a, supporting information) indicates that both pools were of similar size in beech while the inactive pool was twice the size of the active pool in oak. The model accounted for labelled substrates that are available for root and microbial respiration during the chase period. It does not account for any subsequent mobilisation of stored labelled carbon nor for decomposition of labelled roots, mycelia and microorganisms. Our experiment highlighted the species specific pattern in the seasonality of the carbon allocation to soil respiration. In pine, a four-times lesser amount of 13 C was recovered in soil CO 2 efflux after labelling done in November and February compared to those done in June, August and October, indicating that carbon is differently retained in other plant compartments depending on the season. Inhibition of root growth and rhizosphere activity by low soil temperature in late autumn and winter may explain the seasonality of carbon allocation belowground (Kagawa et al., 2006). During the resting season in pine, the amount of 13 C allocated to soil CO 2 efflux accounted for less than 1-4 % of amount of 13 CO 2 taken up by the crown estimated as the difference between the amount of 13 C injected into the labelling chamber and the residual amount of 13 CO 2 in the chamber at the end of the labelling. It raised to 7-10 % during the active growing season, with no strong difference between June, August and October labelling, which contrasts with recent observations on Scots pine (Högberg et al., 2010).
The mild drought imposed by rainfall exclusion did not decrease the contribution of new photosynthates to soil CO 2 efflux in contrast to what reported in potted beech saplings (Ruehr et al., 2009). The relative amount of 13 C allocated to belowground respiration varied markedly throughout of the . Data for 10-20 cm depth in beech and pine are not shown but they are included in the statistical analyses. Each tree is identified by its date of labelling. There are significant differences between trees (F 7,120 = 22.0, F 5,23 = 10.1 and F 7,120 = 3.8, for beech, oak and pine respectively, p < 0.001), between depth (F 2,120 = 22.0, F 1,23 = 47.1 and F 2,120 = 14.5 for beech, oak and pine respectively, p < 0.001) and between sampling dates (F 96,120 = 2.5 for beech, p < 0.001, F 23,23 = 2.4 for oak, p < 0.05 and F 96,120 = 3.0 for pine, p < 0.001). The verticals bars represent the root mean error. Data are presented on two different panels according to depth but depth was included in the ANOVA.
leafy season with a maximum of 18-21 % of 13 C recovered in soil CO 2 efflux in July against only 3 % in May and 1-6 % in August. This was also evident from both fine root and microbial respiration, and from both fine root and mi-crobial biomass. This finding agrees well with the high rhizosphere respiration found in early July as inferred from a trenched plot experiment in the same beech forest . Low allocation of 13 C to soil CO 2 efflux in May  Fig. 6. Time courses of Excess 13 C in root carbon (C R ) after whole crown pulse labelling of beeches (top), oaks (middle) and pines (bottom). Each tree is identified by its date of labelling. There are significant differences between trees (F 7,42 = 5.9, F 5,9 = 10.7 and F 7,18 = 24.1, for beech, oak and pine respectively, p < 0.001) and between sampling dates (F 32,42 = 2.0 and F 9,15 = 2.69 for beech and oak respectively, p < 0.05, F 31,18 = 7.4 for pine, p < 0.001). The verticals bars represent the root mean error. might be related to the sink competition with active aboveground growth that occurs essentially in the few weeks between early May and the end of June. This timing is obvious from leaf phenology observations and recordings of tree diameter growth of this forest stand (Granier et al., 2008). At the end of the growing season until leaf fall, the main carbon sink may be the build-up of storage (Barbaroux et al., 2003) that diverts potentially most carbohydrates resources away from belowground metabolism. In oak, there was low variation in the amount of 13 C allocated to soil CO 2 efflux among seasons (7-11 %), which was contrasting with the pattern in beech. This may suggest that stored carbon acts as a buffer between CO 2 uptake and C allocation belowground in this species. Indeed, oak is known to have higher non structural carbohydrate concentration in stem and coarse roots than beech, with also more pronounced seasonal variations in these non structural carbohydrates (Barbaroux et al., 2003). Despite these differences in the carbon allocation patterns among the three species, we found no difference in residence time of 13 C allocated to belowground respiration that averaged 6-7 days during the active growing season. This value is longer than the 4-5 days reported for shrub species (Carbone and , and much more than the 35 h value reported for small Scots pines (Högberg et al., 2008) but it still represents a fast turnover. In pine, the residence time of 13 C allocated to belowground respiration was much higher in November and February, confirming that carbon is retained in other plant compartments during the resting season.

C fluxes through root and microbial compartments
In all species, Excess 13 C in F S was more related to Excess 13 C in the root compartment (respiration and biomass) than in the microbial compartment. Root respiration that was measured on excised fine roots retrieved from soil cores might have been affected by wounding (Marsden et al., 2008). Flushes in microbial respiration can occur after sieving, due to aeration and increased substrate for respiration (Achat et al., 2010). Both potentially overestimate and lead to high variability in measured fine root and microbial respiration in the short term, making comparisons with soil CO 2 efflux difficult. However, root and microbial respiration seem to be well scaled and observations were repeatable through seasons and between labelled trees. Higher specific fine root respiration in pine compared to beech and oak could not be explained by a higher specific root length since gymnosperm species are known to have thicker fine roots than angiosperm species (Ostonen et al., 2007;Steele et al., 1997). Indeed, reported ranges of specific root lengths for our species in non-fertilized stands were 5.7-31.5 for beech (Ostonen et al., 2007), 7.2-29.1 for oak (Bakker, 1999) and 4-12 m g −1 for pine (Bakker et al., 2006). Maritime pine is a fast growing species in contrast to beech and oak, and one fast growing pine species was found to have a higher specific root  . 7. Excess 13 C in microbial carbon (C M , left) or Excess 13 C in microbial respiration (R M , right) between normal cores and root exclusion cores. Data are pooled by date of sampling and depth for presentation but the statistical analyses were done on non-pooled data. Each tree is identified by its date of labelling. For C M , there are significant differences for beech and pine between trees (F 7,64 = 7.1 and F 7,64 = 8.2, p < 0.001), between depth for beech and oak (F 1,64 = 4.5 and F 1,62 = 3.9 respectively, p < 0.05) and between sampling dates for beech and pine (F 32,64 = 2.0 and F 32,62 = 2.2, p < 0.05). For R M , there are significant differences between trees and between sampling dates for beech (F 7,63 = 10.9 and F 32,63 = 3.7, p < 0.001), between depths for all species (F 1,63 = 14.9, F 1,16 = 18.2 and F 1,24 = 13.5 for beech, oak and pine respectively, p < 0.001). The dashed bars represent the root mean error.
respiration rate than other slow-growing Pinaceae and Fagaceae species (Comas et al., 2002). Of course, root sorting in not trivial, and we are aware that our root samples could have included different proportion of low-active old roots or even-dead roots depending on the species.
The tracer was recovered at the same time in the fine root and the microbial compartments and peaked after 3 to 6 days, indicating a fast transfer of carbon between roots and microorganisms. Large variability among collected soil samples confounds spatial with temporal variations of Excess 13 C in soil compartments, precluding attempts to partition Excess 13 C between roots and microbial biomass and to estimate the residence time of carbon within each compartment. Microbial biomass is composed of a large number of different microorganisms and includes the extraradical mycelium of mycorrhizal fungi. A rapid transfer of photosynthate to ectomycorrhiza has indeed been reported (Esperschütz et al., 2009;Högberg et al., 2010;Leake et al., 2001) and was recently observed from truffle sporocarps after pulse-labelling of chestnut trees (B. Zeller, unpublished data). It has been suggested that a decrease in carbon transfer to microbial biomass can be an early indicator of drought effect on carbon allocation belowground in shrublands (Gorissen et al., 2004). This contrasts with the lack of effect of rainfall exclusion in our experiment in a beech forest. The fact that the Excess 13 C in microbial biomass was less in the root exclusion cores than in normal cores indicates that root exudation also contributes to the fast transfer of photosynthates to soil micro-organisms, among which gram negative bacteria that were found to use recent plant derived carbon (Kramer and Gleixner, 2008). However, the Excess 13 C in non fumigated soil extracts that would have reflected the dissolved organic carbon released from roots remained very low expect in few oak samples (data not shown). This might be caused by a www.biogeosciences.net/8/1153/2011/ Biogeosciences, 8, 1153-1168, 2011 rapid absorption of the released carbon by the bacteria and/or by an important dilution into a large pool of soluble carbon (Esperschütz et al., 2009). Spatial variability of 13 C distribution after pulse-labelling is indeed expected at a millimetre scale (Grieve et al., 2006). We didn't attempt to separate bulk soil from root adhering soil where more Excess 13 C is expected in the vicinity of the roots both in non fumigated soil extracts and in microbial biomass because of root exudation and rhizodeposition (Esperschütz et al., 2009) and the rhizospheric soil might not be well represented in our fumigated soil extracts.

Conclusions
Our study has highlighted important seasonal variation in carbon transfer and allocation belowground. Beech appeared to rely more on recent carbon uptake while allocation/mobilisation of stored carbon may damp seasonal variations of carbon allocation belowground in oak. Despite a lower velocity of carbon transfer belowground in pine compared to the two angiosperm species during the active growing season, a similar amount of carbon is allocated to soil CO 2 efflux, with a strong seasonal variability. This study highlighted the species specific seasonal pattern of the carbon allocation to fulfil energy requirements of metabolic processes occurring belowground but it failed to confirm any effect of drought on belowground carbon allocation. Additional in situ water manipulation experiments are required to confirm previous results on potted saplings. This study also indicated a fast transfer of photosynthate to the microbial biomass including mycorrhizal mycelia and microorganisms feeding on root exudates. However, further analyses will be needed to better describe the fate of carbon in the complex soil microbial community, a prerequisite for improving our understanding of soil carbon dynamics in forest ecosystems and their contribution to the global carbon cycle.