Rapid mixing between old and new C pools in the canopy of mature forest trees



    Corresponding author
    1. Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland and
    Search for more papers by this author
    • *

      Current address: Department of Forest Ecology & Management, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Sweden.


    1. Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland and
    Search for more papers by this author

    1. Laboratory of Atmospheric Chemistry, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland and
    Search for more papers by this author

    1. Institute of Botany, University of Basel, Schönbeinstrasse 6, CH-4056 Basel, Switzerland
    Search for more papers by this author

S. G. Keel. Fax: +46 90 786 81 63; e-mail: sonja.keel@sek.slu.se


Stable C isotope signals in plant tissues became a key tool in explaining growth responses to the environment. The technique is based on the fundamental assumption that the isotopic composition of a given unit of tissue (e.g. a tree ring) reflects the specific C uptake conditions in the leaf at a given time. Beyond the methodological implications of any deviation from this assumption, it is of physiological interest whether new C is transferred directly from sources (a photosynthesizing leaf) to structural sinks (e.g. adjacent stem tissue), or inherently passes through existing (mobile) C pools, which may be of variable (older) age. Here, we explore the fate of 13C-labelled photosynthates in the crowns of a 30–35 m tall, mixed forest using a canopy crane. In all nine study species labelled C reached woody tissue within 2–9 h after labelling. Four months later, very small signals were left in branch wood of Tilia suggesting that low mixing of new, labelled C with old C had taken place. In contrast, signals in Fagus and Quercus had increased, indicating more intense mixing. This species-specific mixing of new with old C pools is likely to mask year- or season-specific linkages between tree ring formation and climate and has considerable implications for climate reconstruction using stable isotopes as proxies for past climatic conditions.


Carbon makes up roughly half of all plant dry matter. Most of this C is tied up in structural pools such as cellulose, hemi-cellulose and lignin, or in mobile compounds such as starch, sugars, lipids or amino acids. Daytime C fixation commonly exceeds the dissipation capacity of photosynthetic tissues, leading to transitory deposition of starch in chloroplasts as observed already by Sachs (1887), who used iodine to visualize starch in leaves. Carbon transfer away from the photosynthesizing leaf to active C sinks is the central prerequisite for continued photo-assimilation. In addition to instantaneous daytime phloem transport, these transitory starch pools in the leaf are reduced during the night, and C is exported to other parts of the tree (Zimmermann & Brown 1971; Dickson 1991). Our current understanding of C transfer in trees is based on seedling and sapling studies (using 14C markers). For mature trees, in a natural setting the timing of these C translocation processes is unknown. It is also unclear how nearby woody tissue is interlinked with canopy leaves, after shoot elongation has been completed in summer. Are mature branches representing neutral transfer pathways towards the large C sinks in the trunk or below ground? Is there lateral assimilate exchange between phloem and branch C pools?

Numerous C transfer experiments with seedlings or saplings have been carried out in the 1960s and 1970s using the then newly available radioactive markers (e.g. Balatinecz, Forward & Bidwell 1966; Rangnekar, Forward & Nolan 1969). Most of these seedling experiments focused on seasonal rates of signal dissipation, requiring high initial labels, achieved by several hours or even days of 14C labelling (e.g. Shiroya et al. 1966; Gordon & Larson 1968; Schier 1970; Smith & Paul 1988). We are not aware of works in which the immediate C transfer had been studied in tall trees and which had quantified the amount of labelled C, which entered woody tissue. Radiography, which was often used to detect the tracer, possibly constrained quantitative analysis of such studies.

As an example of such works with orchard species, rapid C uptake into woody tissue was found in 1-year-old apple trees, where labelled C was detectable in shoots immediately (i.e. after 6 h of labelling), regardless of the season (May, July, August, September or October) (Hansen 1967). One hour after a 3–5 h labelling, up to 3% of the applied C was retrieved in the axis of 8-year-old Pinus sylvestris L. trees (Hansen & Beck 1994). This suggests rapid and intense exchange of new C between leaves, phloem and adjacent stem tissue.

One of the advantages of studies with seedlings or young trees is that the whole plant, including roots can be sampled and analysed, and thus high recovery of C across the tissues can be achieved (Rangnekar & Forward 1972; Hansen & Beck 1994). In contrast, branchlets on a big tree represent ‘open systems’ that export C to a large plant body, wherethe ultimate fate of C remains undetectable. Inversely, the branch can also receive C from proximal stores, although it is unlikely on fully sun-exposed, mature branches in the upper canopy. Loss of recent C through respiration reduces C recovery in both seedling and branch labelling studies, and may contribute >50% to dark-respired CO2 in mature Fagus trees (Nogués et al. 2006). The emission of volatile organic C (in trees mainly isoprene, terpenes) represents an additional although small pathway through which C is lost. For temperate forest trees, isoprene emissions range from 4 × 10−8 to 3 × 10−4 µmol g−1 s−1 relative to leaf dry weight (Kesselmeier & Staudt 1999), which is two to five orders of magnitude smaller than photosynthetic uptake.

A forest-scale, full-season C-labelling experiment (Keel, Siegwolf & Körner 2006) revealed surprising delays in the rate at which new C appeared in new structural tissue. For instance, it took up to four full seasons for a complete replacement of ‘old’ C signals in leaves to take place. It was concluded that new assimilates are rapidly mixed with old mobile C pools in branch wood, and that it is this mix which is retranslocated to the active structural sinks.

Although new assimilates would suffice (quantitatively) to support all new growth (Hoch, Richter & Körner 2003), it is likely that they are first mixed into the old mobile C pool leading to a continuous ‘dilution process’. For leaves of tropical forest trees, 6 d sufficed to turn over the mobile C pool often enough to complete the dilution process and all non-structural C was new C (Körner & Würth 1996).

In this study, we aimed at exploring such ‘local’ transfer processes of recently assimilated C in nine European forest tree species using the Swiss canopy crane (Pepin & Körner 2002; Körner et al. 2005). We used three conifers and six broadleaved species of which we pulse labelled 1-year-old branchlets in the upper canopy using a bagging technique. One of the conifers is deciduous (Larix decidua Mill.) as all the broadleaved species are, and two conifers are evergreen. Evergreen conifers and broadleaved deciduous trees differ in many leaf and wood traits. Conifers have sclerophyllous leaves, which assimilate C for several years, whereas broadleaved trees have softer and mostly thinner leaf tissue with a longevity of only a few months. Furthermore, conifer phloem lacks sieve plates. Within the broadleaved trees, there are also significant differences between species. Quercus petraea (Matt.) Liebl. has thicker leaves with higher assimilation rates than the other taxa. All these differences could affect the speed of short term C translocation and the size of the pool into which new C is mixed and thus ‘diluted’.

Based on the short replacement time of old by new C found in Tilia (Keel et al. 2006), we hypothesize that very low mixing of new C with old C takes place before C is reinvested into new tissue in this species. As a result, we hypothesize that less new C is found in the branch wood of Tilia, because after shoot maturation C is flowing through the branch with a lower lateral C exchange. In contrast, Quercus and Fagus are expected to show a higher exchange of carbohydrates between phloem and woody tissue because of the long replacement time of old by new C that we found in these species (Keel et al. 2006). Therefore, more pronounced signals in woody tissue of labelled branchlets are expected in Quercus and Fagus. The broad comparison across a suite of typical European forest taxa should provide insight into the immediate and longer-term (4 months) fate of C at the branch level after assimilation in contrasting tree types.


Site description

The experiment was carried out in a diverse mixed forest located near Basel, Switzerland [47° 28′ N, 7° 30′ E; 550 m above sea level (asl)] with tree heights of 30–35 m. The forest is situated on a silty–loamy rendzina, and is characterized by 15-cm-deep, rock-free topsoil and 15- to 30-cm-deep rocky subsoil underlain by fragmented limestone bedrock. In the upper 10 cm, the soil has a pH of 5.8 (measured in distilled water extracts).

A 45 m free-standing tower crane equipped with a 30 m jib (crane arm) and a working gondola provided access to 62 dominant trees in an area of ∼3000 m2. The forest is species rich and is dominated by Fagus sylvatica L., Q. petraea, L. decidua and Picea abies L. with Carpinus betulus L., Tilia platyphyllos Scop., Acer campestre L., Prunus avium L., P. sylvestris L. and Abies alba Mill. as companion species. Abies was not included in the study because none of the individuals reaches the upper canopy. Since spring 2001, 14 broadleaved trees of this forest were exposed to elevated CO2 (Körner et al. 2005), an experiment that is not associated with the current analysis, but reduced the number of trees available for the present study.

13C pulse labelling

We labelled fully sun-exposed, 1-year-old, fruitless branchlets of nine species in the upper canopy. However, branchlets were too big in Prunus (only the upper part of 1-year-old branchlets was labelled) and too small in conifers (additionally part of the previous year's branch segment was included), and carried fruits in Tilia (fruits were removed with scissors before labelling). Because the amount of assimilating leaf tissue should be comparable within as well as between species, we tried to select branchlets with similar leaf area (i.e. fewer leaves in the case of species with big leaves such as Quercus). On average, 7.4 leaves in angiosperm species were labelled. Branchlets on the same main branch were selected as references for isotope analysis, but within safe distance to prevent any contamination with labelled CO2. The base level signals (natural abundances of 13C) are relatively robust within a given species as shown by a survey carried out on the same site (Chevillat et al. 2005). Therefore, untreated reference branches were collected only once in July 2004 to minimize the impact on trees. The robustness of our results was tested by manipulating the isotope ratios of the references by ±2%, which did not significantly affect our study.

The labelling treatment was applied on cloudless days between 0800 and 1100 h on 8 d in summer 2003 (7–16 July; only broadleaved species) and again in 2004 on 7 d (30 June to 30 July). To avoid a few cloudy days, some labelling experiments had to be postponed to end of July, and thus, the overall treatment across the nine taxa in 2004 was spread over a period of 30 d. Because branch elongation at the site was terminated by the end of May, this small temporal shift of labelling in midsummer most likely had no influence on the results. In a separate labelling campaign, Fagus and Quercus were also pulse labelled in September 2003 to test whether recently assimilated C is allocated to woody tissue at roughly the same rate in early fall as compared to midsummer. We always labelled two adjacent branchlets (three branchlets in July 2003) simultaneously and considered these as one and the same replicate. On average, 12 branchlets (six pairs) could be labelled per day. Three to five replicates per species were used which were, whenever possible, on separate trees. However, in some cases (Acer, Prunus, Tilia), only one or two trees were available, and hence all replicates were on the same tree. For this kind of experiment and basic physiology involved, the results should not be particularly affected by individual characteristics of a tree, but rather by variations in photosynthetic rates of the involved leaves. Therefore, the variability in branchlet labels within as well as between tree crowns is expected to lie in the same range for a given species. We thus used branchlets as replicates. The order in which branchlets were labelled was randomly chosen, and so were the trees within species where we had a choice.

Branchlets were enclosed in 2 L transparent plastic bags (Minigrip, Seguin, TX, USA), which were sealed tightly around the stem with synthetic modelling clay. After removing the air inside the bag, it was filled with CO2-free air through a soft silicon tube. At the same time as the CO2-free air was flushed through the tube, 2 mL highly 13C-enriched CO2 (>98 atom %, Cambridge Isotope Laboratories, Andover, MA, USA; denoted as 13CO2 in the following) was injected into the tube using a syringe resulting in a ca. 1000 ppm 13CO2 atmosphere inside the bag. After 15 min, 2 mL additional 13CO2 was injected with a syringe directly through the plastic bag, resulting in a total injected amount of 2.1 mg 13C. The tiny punched hole in the plastic bag was sealed immediately with sticky tape.

Because regular infrared gas analysers (e.g. Li-820, Li-Cor, Lincoln, NE, USA) employ filters, which prevent measurement of 13CO2, we were not able to monitor the CO2 concentrations achieved in the bag. Therefore, a constant labelling time of 45 min was chosen, which ensured [by calculating consumption based on volume and known photosynthetic characteristics from Poorter et al. (2006)] that most of the 13CO2 had been assimilated. When experiments were carried out in July 2004, we collected samples of the bag air for determination of CO2 concentrations by mass spectrometry, just before the 45 min labelling was terminated. Using a syringe, 12 mL bag air was sampled and filled into evacuated glass vials, which were brought to the laboratory for analysis. The isotope ratio of the bag air was determined with a gas bench II linked to a mass spectrometer (Delta Plus XL, Thermo Finnigan, Bremen, Germany). The CO2 concentration of every analysed gas sample was calculated from the calibration line with standard gas samples of known CO2 concentrations (340, 5015 ppm). The area of the voltage signal peak of the mass spectrometer for CO2 (masses 44, 45, 46) was integrated over time and proportional to the CO2 concentration of the air sample. Reference gas samples were included with each series of measurements. Up to 20 000 ppm, the CO2 concentrations agreed well (y = 1.04x; r2 = 0.99) with infrared gas analyser readings (Innova 1312, Ballerup, Denmark).

Immediately after gas samples were collected, the bags were removed and three small leaf discs (12 mm in diameter) from different leaves were punched to obtain the maximum initial 13C label. Leaf samples were shock-frozen in liquid nitrogen to deactivate enzymes and were kept in a cool box until noon. After approximately 2 h (at around 1230 h; t = 2 h), the first set of branchlets was removed and shock-frozen as described earlier. The branchlets were separated into leaf, wood, bark and bud tissue before oven-drying at 80 °C for 2 d. Shortly before sunset, roughly 9 h after labelling (at around 1930 h; t = 9 h), the second set of branchlets was collected and treated as described earlier. In experiments carried out in July 2004, the axis 5–10 cm below the section of the branchlet that had been enclosed in the bag was also collected for analysis.

In July 2003, an additional third set of branchlets was labelled, which was sampled in November 2003 after leaves had been shed. Of these branches, we collected also tiny (2 mm) wood cores at 50, 80 and 110 cm distance from the tip using an increment puncher (WSL-puncher, Swiss Federal Research Institute WSL, Birmensdorf, Switzerland). After oven-drying, all tissues were weighed.

C isotope analysis of organic samples

Leaf discs were ground using pestle and mortar; bigger samples were ground with a steel ball mill (Mixer Mill & Retsch MM 2000, Haan, Germany) or with a centrifugal mill (ZM 1000, Retsch), and 0.6–0.8 mg dried powder was packaged in tin capsules for δ13C analysis. Samples were then combusted in an elemental analyser (EA-1110, Carlo Erba Thermoquest, Milan, Italy). Via a variable open split interface (Conflo II, Finnigan MAT), gas samples were transferred to the mass spectrometer (Delta S, Thermo Finnigan MAT), which was operated in continuous flow mode. The precision for δ13C analysis was <0.1‰. Amounts of assimilated C were calculated according to the following formula:


where [C] is the C concentration of the sample calculated from the calibration line with reference samples of known C concentrations. The area of the voltage signal peak of the mass spectrometer for CO2 (masses 44, 45, 46) was integrated over time and was proportional to the C concentration of the sample. The dry weight of the sample is annotated as DW. RL and RC refer to the 13C/12C isotope ratios of the labelled and the control plant tissue, which were calculated by solving the following equation for Rsample:


RPDB is the 13C/12C ratio of the international PeeDee Belemnite standard (0.011237), and δ13Csample is the measured isotope ratio in ‰.

Leaf signals measured immediately after the 45 min labelling (t = 0) were assumed to reflect the total amount of C assimilated, because substantial export to woody tissue can be neglected within this short time. Therefore, these immediate leaf signals were regarded as the maximum signal of every labelled branchlet, and their values were set to 100%. Signals in other tissues and at later sampling times are expressed in % of the maximum leaf signals, and are referred to as relative signals. Using relative signals has the advantage that any difference in signal strength caused by variation in uptake of labelled CO2 is eliminated.

Because we took samples within a few hours after labelling, we assumed that labelled C was in essence tied to non-structural carbohydrates (NSCs) (mainly starch and sugars) and not to structural components. Using published data of the same trees studied here (Hoch et al. 2003), we roughly estimated the amount of NSC in milligrams and calculated the time needed to replace NSC pools in leaves and woody tissue by new photo-assimilates. We assumed that trees assimilate CO2 for 10 h a day (of a total day length of 14–15 h) at the same rate as during the 45 min labelling procedure.


For each tissue type and labelling campaign, we performed repeated measures analysis of variance (anova) with species as a fixed factor and sampling time as a repeated factor. Differences among functional traits (needle leaved versus broadleaved) were analysed as planned orthogonal contrasts within the factor of species. One-sided Student's t-tests were carried out to test whether relative isotope signals were greater than zero. All errors refer to standard errors. Statistical analysis was carried out using R version 2.0.1 (R Development Core Team 2004).


After branchlets had been enclosed in bags for 45 min of labelling, the CO2 concentrations had decreased from the 2 × 1000 ppm injected to 166 ppm on average, indicating that >90% of the applied 13CO2 had been removed (only measured in July 2004).

Maximum leaf signals (in mg C) measured immediately after labelling (t = 0) varied greatly between species (Table 1), and were mostly below 90% of the calculated uptake (<1.95 mg). By using signals relative to the maximum leaf signals, we eliminated differences in signal size, which were caused by variation in 13C uptake. Throughout the rest of the text, we therefore refer to relative signals whenever possible.

Table 1.  Absolute amounts of labelled C recovered in leaves of nine tree species immediately after a 45 min pulse labelling with 13CO2 (= maximum signals) are shown in the first column (‘labelled C’)
SpeciesLabelled C (mg)Leaves (%)Petiole (%)Bark (%)Wood (%)Buds (%)Bark (%) (below)Wood (%) (below)Sum (%)
  1. Signals relative to maximum leaf signals are shown for all tissues 9 h after labelling. In addition to results for tissues which had been enclosed in the bag, bark and wood approximately 0–10 cm below the enclosed section (annotated as ‘below’) are shown. In the last column, the sum of labelled C, which could be retrieved in all tissues together relative to maximum leaf signals, is shown. Mean ± SE for n = 3–5 branchlets.

Acer1.38 ± 0.0869.6 ± 1.31.75 ± 0.11.32 ± 0.20.32 ± 0.10.13 ± 0.11.87 ± 0.20.44 ± 0.075.5
Carpinus1.51 ± 0.2151.1 ± 3.00.42 ± 0.10.94 ± 0.30.42 ± 0.20.07 ± 0.01.02 ± 0.20.32 ± 0.254.3
Fagus1.90 ± 0.0540.9 ± 5.20.97 ± 0.21.14 ± 0.40.53 ± 0.20.15 ± 0.10.97 ± 0.00.40 ± 0.145.1
Prunus1.70 ± 0.0465.4 ± 2.41.00 ± 0.10.78 ± 0.20.23 ± 0.10.20 ± 0.00.99 ± 0.50.35 ± 0.069.0
Quercus1.78 ± 0.0751.3 ± 1.50.43 ± 0.01.2 ± 0.20.40 ± 0.10.23 ± 0.11.35 ± 0.40.25 ± 0.155.1
Tilia0.97 ± 0.3562.8 ± 191.40 ± 0.74.92 ± 1.80.57 ± 0.10.54 ± 0.24.24 ± 1.70.68 ± 0.175.1
Larix1.15 ± 0.0661.2 ± 2.710.6 ± 1.91.87 ± 0.50.25 ± 0.12.61 ± 0.70.30 ± 0.176.8
Picea1.34 ± 0.1069.8 ± 7.96.14 ± 0.50.88 ± 0.10.59 ± 0.20.43 ± 0.00.07 ± 0.077.9
Pinus0.97 ± 0.1079.0 ± 4.12.72 ± 0.51.31 ± 0.60.07 ± 0.10.86 ± 0.50.31 ± 0.184.3

Leaf signals

Nine hours after labelling, the sum of signals found in all tissues together was still dominated by leaf signals (Table 1). In six out of nine species, leaf signals were larger than 60% and reached a minimum of only 41% in Fagus. At the same time, this species reached highest maximum labels (in mg C) measured immediately after labelling (t = 0). Relative leaf signals varied between species in the experiment carried out in the year 2004 (P = 0.007), but not in 2003 (P = 0.63), where only broadleaved species were included in the study (Fig. 1). Between noon (t = 2 h) and evening (t = 9 h), leaf signals decreased in all species and at similar rates.

Figure 1.

The amount of labelled C found in leaves of 1-year-old twigs 2 and 9 h after a 45 min pulse labelling in nine tree species. Signals relative to maximum leaf signals measured immediately after labelling (t = 0) are shown. Experiments were carried out in July 2003 (upper left panel), in September 2003 (upper right panel) and in July 2004 (lower panel). Mean ± SE values are shown (n = 3–5 branches). Ac, Acer campestre; Cb, Carpinus betulus; Fs, Fagus sylvatica; Pr, Prunus avium; Qp, Quercus petraea; Tp, Tilia platyphyllos; Ld, Larix decidua; Pa, Picea abies; Ps, Pinus sylvestris.

The calculated time needed to replace NSC pools in leaves by new photo-assimilates was 9 d in deciduous species (Table 2) whereas in evergreen conifers, replacement times reached up to 63 d in Picea.

Table 2.  The total estimated time (days) required to replace the entire non-structural carbohydrate (NSC) pool of leaves and woody tissue by new photo-assimilates
SpeciesLeaves (d)Wood (d)
  1. Replacement times were calculated based on leaf signals measured immediately after labelling (t = 0), wood signals measured 9 h after labelling (t = 9) and published NSC concentrations of the corresponding trees (Hoch et al. 2003). Because NSC concentrations can vary considerably within the growing season, the numbers shown are rough estimates. We assumed that leaves assimilated CO2 for 10 h a day.


Transport to woody tissue and bark

Within 2 h after labelling was completed, 13C labels were detected in woody tissue (P < 0.001) and bark (P < 0.001) of the nine study species (Fig. 2). Signals in woody tissue measured 9 h after labelling ranged from 0.2 to 1.9% relative to maximum leaf signals (Table 1), which refers to 4–22 µg C depending on the species. In general, signals were higher in bark compared to woody tissue (note the different scales in Fig. 2), and roughly agreed within the different species. In bark as well as wood, signals increased with time and were significantly higher in conifers compared to broadleaved species (bark: P < 0.001; wood: P < 0.001). Wood signals in broadleaved trees were significantly stronger in July 2003, which was an exceptionally dry summer, compared to July 2004 (P < 0.001). At the same time, the mass of current year's shoots was lower in 2004 compared to 2003 (Table 3). In July 2004, wood and bark below the branch section, which had been enclosed in the bag, were also analysed, and signals were found within 2–9 h after labelling, indicating rapid basipetal translocation (Table 1). In all species (except Fagus), no labelled C had been invested beyond 110 cm down the branch as suggested by mini-cores collected along the branch (Fig. 3). We did not sample any control branches but to get an idea of the natural 13C abundance in branches, we refer to published data for branch wood δ13C of Fagus after leaf fall (Damesin & Lelarge 2003).

Figure 2.

The uptake of labelled C in bark (upper three panels) and wood (lower three panels) of 1-year-old twigs 2 and 9 h after a 45 min pulse labelling with 13CO2 in nine tree species. Signals relative to maximum leaf signals measured immediately after labelling (t = 0) are shown. Note the different scales for wood and bark tissue. Mean ± SE values are presented (n = 3–5 branches). For further details, see Fig. 1.

Table 3.  Dry weight (g) of 1-year-old twigs (mean ± SE), which were pulse labelled and separated into leaf, wood and bark tissue after sampling
SpeciesJuly 2003July 2004
  1. n = number of branchlets.

Acer53.01 ± 0.230.27 ± 0.060.36 ± 0.0662.02 ± 0.140.08 ± 0.020.13 ± 0.02
Carpinus101.77 ± 0.090.17 ± 0.020.22 ± 0.0360.87 ± 0.150.06 ± 0.010.08 ± 0.01
Fagus101.43 ± 0.080.33 ± 0.050.30 ± 0.0461.30 ± 0.130.10 ± 0.010.17 ± 0.01
Prunus63.45 ± 0.230.18 ± 0.020.38 ± 0.0362.26 ± 0.210.06 ± 0.010.17 ± 0.03
Quercus62.29 ± 0.300.36 ± 0.040.45 ± 0.0462.90 ± 0.340.22 ± 0.030.32 ± 0.04
Tilia62.69 ± 0.270.43 ± 0.110.93 ± 0.1861.62 ± 0.160.11 ± 0.010.33 ± 0.03
Larix62.80 ± 0.850.59 ± 0.191.63 ± 0.39
Picea69.25 ± 0.620.69 ± 0.064.36 ± 0.39
Pinus65.46 ± 1.540.37 ± 0.130.77 ± 0.30
Figure 3.

Wood δ13C (‰) of ‘mini-cores’ in November 2003 4 months after labelling (July 2003). Samples were collected within the branch section which had been enclosed during labelling (referred to as ‘in bag’), and along the branch at different distances from the tip (indicated as 50, 80 and 110 cm). As a reference, we refer to published branch wood δ13C of Fagus after leaf fall represented by the dotted line (Damesin & Lelarge 2003). Mean ± SE values are shown (n = 3–5 branches). Ac, Acer campestre; Cb, Carpinus betulus; Fs, Fagus sylvatica; Pr, Prunus avium; Qp, Quercus petraea; Tp, Tilia platyphyllos.

Four months after pulse labelling, the remaining wood signals relative to maximum leaf signals were strongly species specific (P < 0.001), with high signals in Fagus and Quercus and very low ones in the other four broadleaved species (not measured in conifers; Fig. 4). Although only significant at the 1% level (P = 0.066), we observed pronounced differences in the signal change over time between species. In Fagus, signals had doubled whereas in Tilia, they were nearly gone after 4 months.

Figure 4.

Remaining 13C signals in woody tissue of six deciduous tree species 4 months after labelling (t = 4 months) compared to 9 h after pulse labelling with 13CO2 in July 2003 (t = 9 h). Signals relative to maximum leaf signals measured immediately after labelling (t = 0) are shown. Numbers above the graph indicate the change in signal size in percent. Mean ± SE values are shown (n = 3–5 branches). Ac, Acer campestre; Cb, Carpinus betulus; Fs, Fagus sylvatica; Pr, Prunus avium; Qp, Quercus petraea; Tp, Tilia platyphyllos.

NSC pools in wood would be replaced in roughly 138 d averaged over all species excluding Quercus, which showed almost four times longer turnover times (Table 3) most likely because of higher NSC concentrations in this species (Hoch et al. 2003). Note that in wood, these replacement times reflect the ‘lateral’ net allocation of assimilates transported in the phloem. If no lateral exchange of labelled C had occurred, the time needed to replace the pool could become infinite, irrespective of the actual rate of longitudinal phloem transport.


Using in situ pulse labelling, we showed that recently assimilated C enters woody tissue of 1-year-old twigs in mature trees within 2–9 h after labelling. Because labelled C had not only been recovered within but also below the section, which had been labelled (bagged), we have strong evidence for rapid basipetal translocation. Because bark assimilation in essence is refixation of respired CO2 (Wiebe 1975; Foote & Schaedle 1976; Cernusak & Marshall 2000; Aschan, Wittmann & Pfanz 2001; Pfanz et al. 2002), which is unlikely to be strongly labelled within 45 min, we assume that direct assimilation of labelled C contributed very little to null to bark signals in our study. In general, the results of our labelling study confirm findings of earlier C allocation experiments performed with seedlings or young trees which had found labelled C in the axial tissue shortly after labelling (Balatinecz et al. 1966; Hansen 1967; Hansen & Beck 1994).

Because only 10% of the labelled CO2 remained in the bag after labelling, but rarely 90% of the calculated 13C was found in branchlet tissue, we conclude that in most cases the bags were slightly leaking. Because we used signals relative to initial leaf signals for each branchlet, the differences in label uptake were not a problem. By using relative signals, we not only remove differences in signals caused by leaks, but also differences in initial signal strength, which could have resulted from species-specific uptake as a result of variation in photosynthetic rate. However, because highest absolute leaf signals measured immediately after labelling did not agree with highest photosynthetic rates or stomatal conductance measured on the same site (Zotz, Pepin & Körner 2005; Keel et al. 2007), there was no indication for species-specific uptake rates. We believe that the amounts of the CO2, which leaked out of the bag and were assimilated by adjacent leaves, are negligibly small, because there is a constant breeze in the upper canopy of these tall trees, which immediately dilutes the labelled CO2 with ambient air.

Given the evolutionary older and supposedly less efficient phloem in conifers, we were surprised to find a more rapid propagation of the wood and bark signals in conifers than in broadleaved species. Phloem transport rates are difficult to measure, but estimates using changes in natural abundance of C isotope ratios in response to changes in climatic conditions suggest rates that are similar in pine and beech (Keitel et al. 2003; Brandes et al. 2006). Technological advances in sap flow measurements by monitoring the propagation of laser-induced heat signals in phloem by infrared cameras might yield the needed information in the near future (Helfter et al. 2007).

The mass of current year shoots was lower in July 2004 compared to July 2003, an exceptionally dry summer (Leuzinger et al. 2005). Because shoot elongation is terminated towards the end of May, the drought did not affect axial growth in the same, but rather in the following year, which is supported by data for other trees on the same site (Asshoff, Zotz & Körner 2006). The longer shoots in 2003 might explain why wood signals were somewhat more pronounced in that year compared to 2004. The branch is an open system, and if we assume that the label is evenly distributed within the branch (not restricted to the current-year shoot), the amount of labelled C in a shorter current-year branchlet will be smaller compared to a longer current-year branchlet.

Four months after labelling, wood signals in Tilia were reduced, whereas in Fagus and Quercus, wood signals had strongly increased compared to the day of labelling suggesting significant species specificity in C allocation only in the long term. By this time, any divergence in signal dilution caused by differences in transport rates is assumed to be negligible. Our data therefore suggest a low mixing of old with new C in branch wood of Tilia, in line with the rapid (within one season) replacement of old by new C reported for the whole-canopy labelling experiment (Keel et al. 2006). On the contrary, the slow replacement of old by new C measured in Fagus and Quercus in the canopy labelling experiment supports the obviously high mixing of old with new C as shown by sustained signals over 4 months.

Among the six broadleaved species studied, only Tilia stores C as lipids (triacylglycerols) in addition to NSC (Hoch et al. 2003). Because the lipid pool most likely undergoes less C cycling because of more costly synthesis as well as degradation compared to carbohydrates such as starch, the exchangeable C pool is likely to be smaller in Tilia than in other species. The occurrence of lipids might therefore explain the lower mixing of old with new C observed in this species (Fig. 5). If this proves to be true, we would expect that low mixing of old with new C also applies for the three conifers studied here, because they store lipids too (Hoch et al. 2003). A 300 year tree ring chronology showing stronger and more consistent correlations of δ13C in cellulose with temperature in P. sylvestris (a ‘fat tree’) compared to Q. petraea (a ‘starch tree’) supports our findings (Reynolds Henne, Villigen, personal communication). In contrast, a similar experiment carried out at the treeline (2080–2230 m asl) showed surprisingly high amounts of old C in new structural tissue of shoots in L. decidua and Pinus uncinata (both ‘fat storers’), which does not support our hypothesis (von Felten et al. 2007). In the light of our results, it seems very likely that the findings by von Felten et al. (2007) reflect the mixing process as we documented it here. The treeline data for Larix and Pinus thus reflect a similar delay in signal appearance as was shown by Keel et al. (2006) for lowland forest trees, continuously exposed to 13C-depleted CO2 during four growing seasons.

Figure 5.

A schematic representation of carbon uptake (arrows) and mixture with old mobile carbohydrate pools (rectangles) in different tree species. Unlabelled carbon (C) is shown in white, labelled C in black and mixtures of unlabelled and labelled C are hatched. The upper two rows represent an experiment where labelled C (black arrow) is mixed into old C pools (white) resulting in species-specific 13C labels. The two lower rows illustrate the situation after a labelling experiment where unlabelled C (white arrow) is mixed into labelled pools (hatched) resulting in different degrees of signal dilution. The presence of lipids as mobile C stores in Tilia and their lack in Fagus and Quercus is a likely explanation for the species-specific mixing of old and new C. Synthesis of lipids is more costly than synthesis of carbohydrates such as starch, and could explain the smaller size of the exchangeable C pool in Tilia.

Carbon or oxygen isotope ratio analysis in tree rings is a widely recognized tool for climate reconstruction (Libby & Pandolfi 1974; Pearman, Francey & Fraser 1976; Mazany, Lerman & Long 1980; Leavitt & Long 1991). Isotope ratios of concurrent assimilation are assumed to reflect mean climatic conditions, which in turn leave their one-to-one fingerprint in woody tissue (Schleser et al. 1999). However, isotope ratios of 1 or 2 subsequent years are often found to correlate (so called ‘autocorrelation’) (Monserud & Marshall 2001). A recent study indeed showed that assimilates from summer and autumn are allocated to early wood of the following year (Kagawa, Sugimoto & Maximov 2006). It is not only the plausible transfer of reserves to structural growth, but also intrinsic mixing of all sequentially assimilated C within the mobile pool which dampens (flattens) the structural tissue–climate linkages in terms of isotope signals, according to our data.

Most likely, plant-respired CO2 represents a mixture of old and new C as well, which is supported by a recent labelling study with mature Fagus trees, where dark respiration contained 40% unlabelled C (Nogués et al. 2006). In addition, as suggested by these authors, signals were dampened because of respiration of different substrates (carbohydrates and fatty acids, which are not equally labelled because of different turnover times), which is supported by previous studies (Tcherkez et al. 2003; Hymus et al. 2005).

In our experiment, no labelled C was detected in branch wood tissue beyond 1 m distance from the point of labelling (except for Fagus). The most likely explanation for the rapid decrease is that labelled C is diluted with a basipetally increasing pool of unlabelled (old) C. Alternatively, C allocation is spatially constraint and much of the photo-assimilates remain in the branch where they contribute to the spring flush in the following year. This would explain why our full canopy labelling took so long to produce entirely labelled foliage in branchlets (Keel et al. 2006), although new emerging foliage seems to be largely C autonomous by the time of unfolding (Keel, unpublished results).

The roughly estimated replacement times for NSC pools in woody tissue indicated that the pool is replaced nearly twice during a 200 d growing season. The net transfer of new C from phloem to axial woody tissue within the same day of labelling was therefore considerable, although shoot elongation had been completed in July and NSC concentrations in branch wood of deciduous species remain rather stable throughout the growing season (Hoch et al. 2003).

The classic view of C allocation where C is directly transferred from leaves to pools that require C for growth, storage or cell maintenance seems too simplistic. Our data suggest that in some species, recent C is first mixed with a given pool of mobile carbohydrates before it is invested into new tissue or can passively enter a nearby pool as shown for woody tissue. Hence, the occurrence of labelled C in a tissue does not necessarily indicate that the tissue was a true C sink, but rather that C could have reached the tissue passively. Although this might be considered self-evident, the data set we present appears to be the first in situ proof under natural canopy conditions.

In conclusion, we found rapid allocation of recent C to woody tissue of 1-year-old branchlets across nine European forest tree species. Four months after labelling, the remaining 13C signals were strongly species specific reflecting a high degree of mixing of old with new C in Fagus and Quercus resulting in stronger, persistent signals in branch wood. In contrast, signals were markedly weaker in Tilia because of a low mixing of new with existing C. The low mixing in Tilia is likely associated with the presence of lipid stores in woody tissue, which undergo less C cycling. The exchangeable carbohydrate pool is therefore smaller, resulting in a lower mixing of new with old C pools. Based on our results, climatic conditions should therefore be more precisely recorded in tree rings of Tilia and perhaps other lipid storing species (e.g. Pinus), whereas the continuous C pool mixing blurs the structure–climate linkage in tree rings of Fagus and Quercus.


We thank Erwin Amstutz and Olivier Bignucolo for crane operations, Maria Rosa Guerrieri for support with sample preparation and Matthias Saurer for his support in operating the mass spectrometer. Thomas Fabbro is acknowledged for statistical advice, Sebastian Leuzinger and Günter Hoch for discussion of the results and two anonymous referees for valuable inputs on an earlier version of this manuscript. The CO2 enrichment experiment was funded by the Swiss National Science Foundation projects no. 3100-059769.99, no. 3100-067775.02, no. 5005-65755 (NCCR Climate) and the Swiss Canopy Crane by the Swiss Agency for the Environment, Forest and Landscape.