Carbon transfer, partitioning and residence time in the plant-soil system: a comparison of two 13 CO 2 labelling techniques

Various 13 CO2 labelling approaches exist to trace carbon (C) dynamics in plant-soil systems. However, it is not clear if the different approaches yield the same results. Moreover, there is no consistent way of data analysis to date. In this study we compare with the same experimental setup the two main techniques: pulse and continuous labelling. We evaluate how these techniques perform to estimate the C transfer time, the C partitioning along time and the C resi- dence time in different plant-soil compartments. We used identical plant-soil systems (Populus del- toides◊ nigra, Cambisol soil) to compare the pulse labelling approach (exposure to 99 atom % 13 CO2 for three hours, traced for eight days) with a continuous labelling (exposure to 10 atom % 13 CO2, traced for 14 days). The experiments were conducted in climate chambers under controlled en- vironmental conditions. Before label addition and at four successive sampling dates, the plant-soil systems were de- structively harvested, separated into leaves, petioles, stems, cuttings, roots and soil and soil microbial biomass was ex- tracted. The soil CO2 efflux was sampled throughout the ex- periment. To model the C dynamics we used an exponen- tial function to describe the 13 C signal decline after pulse la- belling. For the evaluation of the 13 C distribution during the continuous labelling we applied a logistic function. Pulse labelling is best suited to assess the minimum C transfer time from the leaves to other compartments, while continuous labelling can be used to estimate the mean trans- fer time through a compartment, including short-term stor- age pools. The C partitioning between the plant-soil com- partments obtained was similar for both techniques, but the time of sampling had a large effect: shortly after labelling the allocation into leaves was overestimated and the soil 13 CO2 efflux underestimated. The results of belowground C par- titioning were consistent for the two techniques only after eight days of labelling, when the 13 C import and export was at equilibrium. The C mean residence times estimated by the rate constant of the exponential and logistic function were not valid here (non-steady state). However, the duration of the accumulation phase (continuous labelling) could be used to estimate the C residence time. Pulse and continuous labelling techniques are both well suited to assess C cycling. With pulse labelling, the dynam- ics of fresh assimilates can be traced, whereas the continuous labelling gives a more integrated result of C cycling, due to the homogeneous labelling of C pools and fluxes. The logis- tic model applied here, has the potential to assess different parameters of C cycling independent on the sampling date and with no disputable assumptions.

Originally published at: Studer, Mirjam S;Abiven, Samuel (2013). Carbon transfer, partitioning and residence time in the plant-soil system: a comparison of two ¹³CO labelling techniques. Biogeosciences Discussions, 10(10):16237-16267. DOI: https://doi.org /10.5194/bgd-10-16237-2013 uous labelling. We evaluate how these techniques perform to estimate the C transfer velocity, the C partitioning along time and the C residence time in different plant-soil compartments.
We used identical plant-soil systems (Populus deltoides x nigra, Cambisol soil) to compare the pulse labelling approach (exposure to 99 atom% 13 CO 2 for three hours, 10 traced for eight days) with a continuous labelling (exposure to 10 atom% 13 CO 2 , traced for 14 days). The experiments were conducted in climate chambers under controlled environmental conditions. Before label addition and at four successive sampling dates, the plant-soil systems were destructively harvested, separated into leaves, petioles, stems, cuttings, roots and soil and the microbial biomass was extracted from the soil.

15
The soil CO 2 efflux was sampled throughout the experiment. To model the C dynamics we used an exponential function to describe the 13 C signal decline after pulse labelling.
For the evaluation of the 13 C distribution during the continuous labelling we suggest to use a logistic function. Pulse labelling is best suited to assess the maximum C transfer velocity from the 20 leaves to other compartments. With continuous labelling, the mean transfer velocity through a compartment, including short-term storage pools, can be observed. The C partitioning between the plant-soil compartments was similar for both techniques, but the time of sampling had a large effect: shortly after labelling the allocation into leaves was overestimated and the soil 13 CO 2 efflux underestimated. The results of 25 belowground C partitioning were consistent for the two techniques only after eight days of labelling, when the 13 C import and export was at equilibrium. The C mean residence time estimated by the rate constant of the exponential and logistic function was not valid 16238

Introduction
While carbon (C) cycling within terrestrial ecosystems has been extensively studied 10 in the last decades, many processes and plant-soil-atmosphere C fluxes are still not well understood. For example, it is still under discussion, how single plants or whole ecosystems will respond to changes in climate (temperature, water availability and atmospheric CO 2 concentration). Of special interest is the change in the velocity of C cycling, in the C allocation patterns and in the C residence time within different com-15 partments of the plant-soil system. Stable isotope tracing is a powerful tool to study the C fluxes and pools within the plants and the soil with minor disturbance (Dawson et al., 2002;Brüggemann et al., 2011;Werner et al., 2012). The use of natural labelling approaches (based on isotopic fractionation occurring during biochemical reactions in plant and soil) is valuable 20 in many cases, but is inappropriate if more than two sources are involved or if the difference in the isotopic composition of the sources is too small (Bowling et al., 2008;Werth and Kuzyakov, 2010). Artificial labelling techniques (using stable or radioactive isotopes) can overcome these difficulties (Amelung et al., 2008;Epron et al., 2012;Glaser, 2005 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | sampling intervals. Two main techniques can be distinguished to label organic matter by exposure of the plant to labelled CO 2 : pulse and continuous labelling (field of applications reviewed in Meharg, 1994;Kuzyakov and Domanski, 2000). In a pulse labelling (PL) experiment, highly 13 C enriched CO 2 (usually 99 atom% 13 CO 2 ) is added in a pulse, i.e. over a short period of time (a few hours) and the label is traced in the plant- 5 soil system in the following days (Epron et al., 2012;Leake et al., 2006). In continuous labelling (CL) experiments, the plant is continuously exposed to less strongly labelled CO 2 (generally < 10 atom% 13 CO 2 ) or 13 C-depleted CO 2 over the whole experimental period and samples are taken during and/or at the end of the labelling (e.g. in Esperschütz et al., 2009;Yevdokimov et al., 2006). With continuous labelling C dynamics 10 can be studied over larger time periods, as for example in Free Air Carbon Exposure (FACE) experiments, where whole ecosystem areas are exposed to elevated CO 2 ( 13 Cdepleted) for several years (e.g. in Grams et al., 2010;Keel et al., 2006). However, the continuous labelling technique has also been applied in the same time scales as the pulse labelling technique (days-weeks), but it is not clear if these approaches yield the 15 same results regarding C cycling within plant-soil systems and how we can interpret them. While there are generally approved approaches to analyse the 13 C dynamics in plants after pulse labelling (exponential model), no consistent approach exists for the continuous labelling technique.
To make best use of the two 13 CO 2 labelling techniques and their results a proper 20 evaluation of the techniques is essential. A comparison based on literature is hardly possible, since they have been applied to numerous plant species and soil types and under a variety of environmental conditions. Studies based on identical plant-soil systems grown under controlled environmental conditions are needed in order to elucidate the potential of these labelling techniques to assess C dynamics and to evaluate how 25 one can compare them. To our knowledge, only one study has made such a direct comparison so far, whereby the focus lay on the effect of labelling duration on belowground C partitioning (Warembourg and Estelrich, 2000). In this study we compare the results for above-and belowground plant-soil compartments, obtained by pulse and (short-term) continuous labelling and discuss their potential to estimate C transfer velocity, C partitioning and C residence time. We suggest a new approach to assess the C dynamics based on the 13  The soil consists of a clay loam texture (20 % sand, 45 % silt, 35 % clay), with a pH of 4.8 and an organic C and N content of 2.2 % and 0.2 %, respectively. The soil was sieved by hand through a square sieve (2.5 × 3.5 cm) leaving the soil structure largely intact, but we large pieces of organic material and coarse gravel were removed. The plant pots were filled with 7.5 dm 3 moist soil (average dry weight of 2642 ± 402 g).

15
The poplar trees, 15 per experiment, were grown indoors under artificial light from stem cuttings for five weeks and were then transferred into the labelling chambers (described below), where they were left for one week to acclimatize prior to labelling. One day before labelling the dry weight of fresh biomass (without the cutting) was 4.0 ± 1.2 g and the total leaf area 692 ± 113 cm 2 per plant. During the PL and CL experiment the Introduction ) wrapped around the cuttings, to prevent the diffusion of the labelled gas from the plants' atmosphere into the soil. The plant roots are in individual soil pots, which are also hermetically separated from the room atmosphere. The pots are aerated individually, with ambient air (flow rate = 0.8 L min −1 ), to prevent anaerobic conditions. Further, each pot 15 has a separate in-and outlet for watering. The environmental conditions in the chamber (CO 2 concentration, air humidity and light) are automatically regulated (valve system programmed with LabVIEW, National Instruments Switzerland Corp.). The 12 CO 2 and H 2 O concentrations in the chamber atmosphere and of the pot in-and outlets are monitored online with infrared gas ana-20 lyzers (LI-840A, LI-COR Inc.). In addition, gas samples can be taken manually from up to nine individual pots for further analysis of the soil 13 CO 2 efflux.

Labelling procedure
To label the plants we added CO 2 enriched in 13 C to the shoots (upper chamber system). In the pulse labelling (PL) experiment, the CO 2 concentration in the chamber 25 was reduced first to 250 ppm, then 99 atom% 13  Inc.) was injected up to a concentration of 1000 ppm CO 2 and kept on this concentration level (CO 2 saturation) for 2.5 h. After flushing the chamber with ambient air, the plant shoots were exposed to CO 2 with isotopic signatures close to ambient air (δ 13 C = −3 % ) from a CO 2 gas cylinder till the end of the experiment (8 days). In the continuous labelling (CL) experiment, 10 atom% 13 CO 2 (Cambridge Isotope Laborato-5 ries, Inc.) was added continuously to the upper chamber system (for 14 days). Due to technical restrictions the light intensity within the labelling chambers was low (79 ± 25 µmol m −2 s −1 ) and the temperature high (31 ± 3 • C). Day-night cycles of twelve hours allowed for a positive C balance. To ensure optimal C assimilation at the given light availability, the CO 2 concentration was held on a high level (495-540 ppm), the soil 10 was kept moist (close to 100 % field capacity) and the plants were grown in a humid environment (65-74 % relative air humidity) throughout the experiment.

Destructive harvests
The plant-soil systems were destructively harvested at five sampling dates with three 15 replicates each. The first sampling was done one day before the labelling experiment started and represents the natural isotopic background signature (thereafter referred to as sampling date t = 0). Subsequently plant-soil systems were sampled 0.1 (2 h), 1, 2 and 8 days after the pulse labelling and after 1, 2, 8 and 14 days during the continuous labelling experiment. The sampled bulk materials were dried in the oven (24 h at 60 • C) 20 for later δ 13 C analysis.
At each sampling date, the total leaf area was measured with a handheld area meter (CID-203 Laser leaf area meter, CID Inc.) and the plant-soil systems were separated into leaves, petioles, stems, cuttings, roots (washed with deionised water and carefully dabbed with tissue) and bulk soil (visible roots were removed with tweezers). The soil microbial biomass was extracted from fresh soil by chloroform fumigation extraction (CFE). The extraction was performed according to Murage and Voroney (2007) 1 M KCl without removal of excess salts. Subsamples of the CFE extracts were stored in a freezer for later elemental analysis. The remaining CFE extracts were freeze-dried for δ 13 C analysis.

Soil respiration
Soil CO 2 efflux samples were collected one day before the beginning of the labelling, 5 five times during the first day (2, 4, 6, 8, 21 h) and after 1, 2, 3, 4, 5 and 8 days in both experiments. During the CL, additional samples were collected after 11 and 14 days. The gas samples were taken from three pots corresponding to the last sampling date. To analyse the soil 13 CO 2 efflux, each pot was connected to a closed loop.
A pump circulated the air (flow rate of 0.8 L min −1 ) from the pot to a T-piece, equipped 10 with a septum for manual gas sampling, through the gas analyzer (LI-840A, LI-COR Inc.) and back to the same pot. At each sampling date, three gas samples per pot were taken for δ 13 C analyses. First, the air used to aerate the pots was sampled (atmospheric background). Then the pot was cut off from the aeration and linked to the loop. Two samples with a span of 100 ppm CO 2 were taken. The soil respiration rate 15 was assessed by the slope of the linear regression line of the increase in the CO 2 concentration measured between the two sampling dates. The isotopic signature of the soil respiration was then estimated by the Keeling plot approach (Keeling, 1958;Pataki, 2003). The approach is based on a two end-member mixing model (assuming preservation of mass), whereas the two end-members are 20 the atmospheric background and the soil 13 CO 2 efflux. The isotopic signature of the sampled CO 2 (in the pot) shows a linear relationship to the inverse of its concentration. The intercept of the linear regression line yields the isotopic signature of one endmember (soil 13 CO 2 efflux). In a recent publication, Brand and Coplen (2012) have demonstrated the non-linearity of the δ notation and that δ values should consequently 25 not be used to assess mass balances when the differences in the δ-values are large (as it is usually the case in labelling experiments). Therefore we used 13 C atom fraction, BGD 10,2013 Carbon transfer, partitioning and residence time in the plant-soil system instead of the δ-values, to calculate the signature of the soil respiration based on the Keeling plot approach.

Procedure
The dried plant and soil samples were milled to a fine power with a steel ball mill and 5 weighed into tin capsules. Elemental C content of the solid samples was analysed in an elemental analyzer (CHN-900, Leco Corp). The elemental C analysis of liquid CFE extracts was performed by a TOC-500 analyzer (Shimadzu Corp.). The isotopic analyses were done by isotope ratio mass spectrometry (IRMS). To estimate the isotope ratios, the solid samples were combusted in an elemental analyser (EA 1110, Carlo Erba) and the resulting CO 2 was transferred in a helium stream via a variable open-split interface (ConFlo II, Finnigan MAT) to the mass spectrometer (Delta S, Thermo Finnigan; Werner et al., 1999). The precision of the δ 13 C solids analyses was ±0.12 % (CL) and ±0.09 % (PL). The gaseous soil CO 2 efflux samples were transferred in a helium stream directly from the gasbench (Gasbench II, Thermo 15 Finnigan) to the mass spectrometer (Delta Plus XL, Thermo Finnigan). The precision of the gaseous δ 13 C analyses was ±0.44 % (CL) and ±0.51 % (PL). The precisions indicated here are the standard deviations of working standards (leaf biomass, commercially available CO 2 ) measured frequently along with the experimental samples.

20
The isotopic ratios measured were expressed in the delta (δ) notation relative to the international standard Vienna Pee Dee Belemnite (V-PDB, 13 C/ 12 C = 0.0111802). The significance of the 13 C enrichment was tested by t tests (unpaired, two-sided, R statistics) at the individual sampling dates (t = x), compared to the natural isotopic background measured before labelling (t = 0). The excess atom fraction x E ( 13 C) P/reference BGD 10,2013 Carbon transfer, partitioning and residence time in the plant-soil system within a plant-soil compartment (P), was calculated according to Coplen (2011) in order to assess mass balances (reference is t = 0). The total mass of label recovered in excess m E ( 13 C) (in mg 13 C) within the plant tissues (PT), the soil (S), the microbial biomass (MB) and the soil respiration (SR) was then calculated by multiplying the excess atom fraction with the C pool size or C flux present and taking into account the 5 change in molar C weight due to the 13 C tracer addition (Eq. 1-3), as suggested by Brand and Coplen (2012).
where m(C) PT, S is the mass (in mg) of C present in the plant-soil compartment (PT) or the soil organic matter (S), x( 12 C) PT, S and x( 13 C) PT, S is its 12 C and 13 C atom fraction, 10 respectively, and M( 12 C) and M( 13 C) the molar weight (mg mol −1 ) of 12 C and 13 C.
where m S is the mass of soil (in mg dry weight) and c(C MB ) is the microbial (MB) C concentration (fraction of total soil dry weight). The later was assessed by elemental analysis of the fumigated vs. non-fumigated CFE extracts, applying a conversion factor 15 of 0.45, as suggested by Joergensen (1996).
where F (C) SR is the soil respiration (SR) rate (in mg C day −1 ) extrapolated to 24 h.
The cumulative loss of 13 C by soil respiration (in mg) was estimated by the integral of the curve fits, for the three measured pots separately. To fit the curve in the PL Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | experiment, we used the model proposed by Warembourg and Estelrich (2000). The increase at the beginning was described by a logarithmic function and the decline of the signal after the label peak with an exponential function. In the CL experiment we used a logistic function to fit the curve, as described below.

2.5
Modelling the 13 C distribution to assess C dynamics 5

The 13 C distribution dynamics
The dynamics of 13 C recovered after PL in the plant-soil compartments are characterized by three phases (Fig. 1a). An initial lag phase (no detectable signal), a phase dominated by the import of 13 C from other compartments, until a maximum label strength is reached (peak) and a phase of 13 C export, controlled by 13 C transfer to other compart-10 ments and respiratory losses. Thus the import vs. export of 13 C determines the shape of the signal peak (discussed in Epron et al., 2012).
There is no consistent approach to describe the 13 C dynamics in the plant-soil system during CL (of pre-existing plants). Warembourg and Estelrich (2000) used a logarithmic function to describe the tracer dynamics in experiments with different labelling 15 durations. However, when plants were exposed continuously to the label, they observed sigmoidal -shaped curves. We tested the logistic and the logarithmic curve fit on our 13 C mass excess data. In all plant-soil compartments the logistic model yielded a better fit than the logarithmic model and it proved to be quite robust (Supplement). Therefore we suggest using a logistic function to describe the tracer dynamics within plant-soil 20 compartments during CL experiments. Logistic growth functions have their origin in ecology (population growth), but they have also been used to describe the accumulation of specific compounds and nutrients in plant tissues (e.g. in Bonvehi et al., 1997;Moustakas and Ntzanis, 2005;Iwahashi et al., 2012;Gutierrez-Gonzalez et al., 2013). We propose that the different phases (Fig. 1b) represent the development towards 25 homogeneously labelled C import and export. The initial lag phase reflects the 13 C transfer time, i.e. the time needed for the 13 C to be transported from the chloroplast to the particular plant-soil compartment, analogous to the lag phase in the PL. A phase of exponential (net) 13 C accumulation follows thereafter, which slows down after the inflection point, due to increased labelling of the C export (respiratory losses, transfer to other compartments). In the final stage (stationary phase) the C import and export are homogeneously labelled, i.e. the 13 C, which is introduced into the compartment, 5 is in equilibrium with the 13 C exported. If the system is in a non steady state, the stationary phase would only be temporary. E.g. during plant growth, the amount of 13 C would steadily increase after the steady state.

C transfer velocity and C partitioning
The C transfer velocity is usually assessed by the first significant 13 C signal detection 10 ("lag time"), but the period to the maximum has also been used as indicator for the C transfer velocity in PL studies (Kuzyakov and Gavrichkova, 2010). We used the lag time to assess the minimum transfer time of fresh assimilates from the leaves to other plantsoil compartments. The mean transfer times of C within the plant-soil compartments were estimated by the time of the signal peak (PL) or the time of inflection (CL) minus 15 the lag phase (which was negligible in this study with small tree seedlings). The mean C transfer time reflects the time needed until the majority of the labelled compounds are transferred into a plant-soil compartment and the export of the labelled compounds gains importance. The C partitioning was assessed with both techniques by the relative 13 C distribu-20 tion within the different plant-soil compartments at a sampling date. The fraction of 13 C within the leaf, petiole, stem, cutting, root and microbial biomass (in %) was calculated as total amount of 13 C in the compartment, relative to the sum of 13 C in all compartments. The belowground C partitioning was estimated analogues, for the roots, the microbial biomass and the cumulative respiratory C loss. The bulk soil was excluded 25 due to the lack of significant signal detection. The effect of sampling date and labelling technique on the estimation of C partitioning was tested with a two-way ANOVA (R statistics). Thus the C partitioning reflects the amount of C from fresh assimilates re-16248 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | maining in a plant-soil compartment relative to the amount within the other compartments (and does not refer to a proportion of net C assimilation allocated into a plant tissue or soil compartment). As an alternative to assess the C partitioning, we tested the use of the 13 C peak amount (PL) and the amount of 13 C at the stationary level (CL) for the calculation of the relative 13 C distribution into the single plant-soil compartments. 5

C residence time
The mean C residence time (MRT) is the time that a carbon atom remains on average in a compartment and is defined as the ratio of the holding capacity (pool size) and the (net) C flux through the pool. The MRT is assessed in tracer studies by measuring the changes in the label strength in a pool over time and deducing C fluxes by mathematical models fitted to the data points. We used R statistics (R Core Team, 2013) to fit the models by nonlinear least squares (function nls). The MRT was calculated as the inverse of the rate constant (MRT = 1 k −1 ) of the exponential model (Eq. 4) and the logistic model (Eq. 5) in the PL and CL, respectively. Thus, the rate constant in the PL is based only on the 13 C efflux, while in CL it is based on the net 13 C flux, making the 15 later more reasonable to estimate the C residence time as defined above. However, both models are only valid to describe one kinetic pool, with constant pool size (steady state) and proportional fluxes (first order kinetics).
where a is the amount of 13 C at the peak, k is the rate constant of the tracer loss after 20 the label peak (Fig. 1a, export phase), t is the time of sampling and b is the peak time.
where a is the amount of 13 C at the stationary level (Fig. 1b), k is the rate constant of In addition we used the duration of the accumulation phase (CL) as an indicator for the C residence time. It is the time needed (after the time lag) to reach equilibrium between the 13 C import and export (Fig. 1b). The length of the accumulation phase was assessed by the time the derivative of the logistic curve (mg 13 C day −1 ) was less than 1 % of the stationary level.

5
3 Results and discussion

13 C detection and distribution
The fresh plant tissues (leaves, petioles, stems, roots) were enriched in 13 C by hundreds of per mil δ 13 C in both experiments (Table 1), indicating a substantial assimilation and incorporation of 13 C. The variability of total 13 C assimilated was quite high between 10 the plant replicates, reducing the significance of the isotopic enrichments measured. In the compartments with a large C pool (cuttings, microbial biomass, soil organic matter), the increase in δ 13 C signal was only a few per mil and it was mostly not statistically significant. The signal strength of the labelled assimilates was diluted by mixing with the present carbon pool, and in case of the PL, additionally by new unlabelled assimilates 15 transferred into the plant-soil compartment, resulting in isotopic enrichments close to the IRMS detection limit in large C pools. The expression of mass excess 13 C (Fig. 2) takes into account the present pool size and demonstrates the total amount of 13 C distributed in the plant-soil system. After pulse labelling, the leaves showed the highest peak in 13 C (13.7 ± 2.1 mg), followed by 20 the stems (3.2 ± 0.9 mg), the cuttings (0.9 ± 0.2 mg), the petioles (0.7 ± 0.2 mg) and the roots (0.4±0.1 mg). Even in the microbial biomass a small label peak could be observed (0.02 ± 0.01 mg) in parallel to the peak in the soil 13 CO 2 efflux (0.39 ± 0.22 mg day −1 ).
The same distribution pattern was detected in the continuous labelling experiment.
After 14 days of labelling 19.6 ± 5.8 mg 13 C was recovered in the leaves, 7.7 ± 3.5 mg in 25 the stems, 2.0 + 0.7 mg in the petioles, 1.5 + 0.5 mg in the cuttings, 0.8 ± 0.5 mg in the Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | roots, 0.05 ± 0.03 in the microbial biomass and 0.32 ± 0.11 mg 13 C day −1 was respired belowground.

C transfer velocity
The soil respiration was significantly enriched in 13 C already five hours after pulse labelling and nine hours after the continuous labelling started. Such a fast minimum C 5 transfer time from the leaves to the soil has already been reported for young poplars (Horwath et al., 1994) and other tree seedlings (Barthel et al., 2011;Pumpanen et al., 2008). The signal detection in the CL experiments was delayed due to the weaker label strength of the fresh assimilates (10 atom% vs. 99 atom% 13 C in the PL). The same amount of labelled compounds in a compartment yields a lower signal in the CL than in the PL and more time is needed to reach the lower detection limit. The individual plantsoil compartments were enriched in 13 C already on the first sampling date. Hence we missed the lag time to the specific compartments due to the low sampling frequency of the destructive harvests. The mean transfer times ( Table 2, parameter b) were two days shorter in the PL (0-2 15 days) than in the CL experiment (2-4 days). In the PL experiment, the mean transfer time increased with the distance to the assimilating leaves, e.g. it was one day in the aboveground plant tissues and two days in roots. Thus the mean transfer time assessed by the label peak in PL reflects mainly the minimum transfer time of the labelled assimilates from the leaves to the other plant-soil compartments, due to a preferential 20 labelling of labile compounds with PL (Meharg, 1994). On the opposite, the mean transfer times assessed by CL are the shortest in the belowground soil compartments (SOM, microbial biomass) and the longest in the stems, roots and leaves, which are the plant organs known to store most C (Barbaroux et al., 2003). This indicates that continuous labelling leads to a more homogeneous labelling, including transient C storage pools, periment is rather an indicator for the C transfer through the compartment (short-term C cycling) than into the compartment (C transfer from other tissues).

C partitioning
The patterns of the relative 13 C distribution within the plant-soil compartments obtained by the two labelling techniques were similar throughout and equivalent at the end of the 5 experiments (Table 3). The differences in the proportion of C allocated into plant-soil compartments at the specific sampling dates were up to 6.6 % between the two labelling techniques (Table 3), as for example in the leaves and stems at sampling date one. However, the only significant difference observed was a slightly higher allocation to the petioles (+0.2-1.7 %) and to the microbial biomass (+0.1 %) with CL compared 10 to PL. The results of the last destructive sampling reveal, that most of the assimilated C remained in the leaves (62.5 ± 0.5 %), followed by the stems (23.4 ± 0.1 %), petioles (6.3 ± 0.1 %), cuttings (4.7 ± 0.1 %), roots (2.9 ± 0.6 %) and microbial biomass (0.1 ± 0.1 %). Thus the bigger part (> 90 %) of net assimilated C was recovered in the aboveground plant tissues. We assume that the dominant allocation into leaf biomass 15 was promoted by the low light availability in the climate chambers, which was limiting for C assimilation and by the high soil water and nutrient availability in the pots, reducing the need for root production. Increased shoot vs. root allocation has been observed in poplar trees also by other authors, who grew plants under high N and water availability (Coleman et al., 2004;Pregitzer et al., 1990) or under light limitation in the understory 20 (Landhäusser and Lieffers, 2001). The time of sampling had, like the labelling approach, a minor effect on the results of C partitioning in plant-soil compartments (Table 3). Between the sampling on the first day and day eight, a significant difference could be observed in the petioles (+1.4 %) and the cuttings (−2.3 %), but the changes in all other plant-soil compartments were 25 not significant. In contrast, the C partitioning observed directly after PL (0.1 days) was largely different. The allocation into the leaves was overestimated by approximately 20 % compared to the following sampling dates. Similarly, a trend of increased C al-16252 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | location into the leaves (by 5 %) at the early sampling dates can be observed in the CL experiment. The overestimation of the C allocation to leaves within the first days of labelling or directly after pulse labelling is due to the time lag in tracer distribution. As shown in the previous section, the mean 13 C transfer time from leaves to roots was two days in the PL and the data from the CL indicates, that a steady state between tracer 5 import and export in the plant compartments was reached approximately after six days (discussed in the next section).
To assess the belowground C partitioning, the time of sampling is of much greater importance (Table 4). At the end of the experiments, 13.3 ± 1.3 % of the 13 C recovered was detected belowground. Most of it was released as CO 2 (81.2 ± 0.9 %), and a small 10 amount remained in the root (18.0 ± 0.3 %) and microbial biomass (0.7 ± 0.6 %). The results obtained at a specific sampling date are similar for the two labelling techniques, except for the generally higher proportion of 13 C detected in the microbial biomass with CL. However, the results at the different sampling dates were strongly distinct. At the first sampling dates, the estimated C allocation to roots was more expressed or even 15 dominating (43-75 % in the PL and 31-65 % in the CL). This might be due a time lag in the tracer distribution at the plant-soil interface. As discussed above, the first labelled assimilates were detected belowground within a few hours. However, much more time (6-8 days) was needed to reach an equilibrium (stationary state) in the belowground C fluxes (Table 2). This is in line with the one week allocation time proposed by Warem-20 bourg and Estelrich (2000) and the time delay of 5-6 days in the steady labelling of root exudation observed by Thornton et al. (2004). Accordingly, a time lag between the label strength in the roots and the rhizomicrobial respiration might have caused the underestimation of the proportion of respired C to total belowground C at the beginning of the experiment. 25 Another way to estimate the C partitioning between the plant-soil compartments independent on the sampling time, is the use of the amount of 13 C at the label peak or the stationary level, i.e. by the parameter "a" of the exponential and logistic model fit, respectively ( values of C partitioning. The differences are less than 1 % in the compartments. But the results of the relative amount at the label peaks in PL overestimate the allocation to the leaves (73.0 %) and underestimate the allocation to petioles (3.6 %) and stems (16.9 %). This might be due to lack in label peak detection. The leaves were sampled directly after labelling, while the next sampling date was one day later. The peak in 5 the petioles and stems might have occurred before and thus the peak amount was underestimated.

C residence time
The estimates of the mean residence time (MRT) of the PL technique are longer than the one by the CL (Table 2). In the PL experiment, the longest MRT was detected in 10 roots (34 days), then the MRT decreased in the order of petioles (21 days), stems (13 days), cuttings (9 days), microbial biomass (6 days) and leaves (3 days). These residence times are in the range of values reported in literature. For example, mean residence times of 16-41 days have been reported for fine roots (Keel et al., 2012), 3.2 days for the total microbial biomass (Yevdokimov et al., 2007) and 2.4 days for leaves 15 of beech seedling (Ruehr et al., 2009). In the CL experiment, the MRTs were around one day in all plant-soil compartments. The cuttings, leaves and stems had the highest MRT (1.1-1.3 days), followed by the petioles and microbial biomass (0.9 days) and the lowest MRTs were detected in the roots and SOM (0.8 days). In this study the system was not at a steady state, but characterized by plant growth. 20 Due to 13 C accumulation, the 13 C efflux and thus the signal decline rates were underestimated in the PL experiment (Fig. 1a). Furthermore the exponential model assumes that the dynamics are governed by 13 C efflux, but it has been shown that remobilisation of stored 13 C is leading to 13 C influx even after the signal peak (Barthel et al., 2011;Endrulat et al., 2010;Epron et al., 2011). This is relevant for compartments, which are 25 farther away from the assimilating leaves and are characterized by a broad label peak (e.g. in roots, microbial biomass). The prolonged import of 13 C even after the signal peak leads to a further underestimation of the decline rate. Therefore we can assume 16254 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | that the MRTs assessed by the PL technique overestimate the C residence time, especially in the belowground compartments. With the logistic model of the CL technique the net 13 C flux is observed. Therefore this model is better suited to estimate the C residence time. However, if the system is not at steady state (change in pool size), as in this study, the rate constant is over-5 estimated (illustrated in Fig. 1b) and consequently the MRT underestimated and not valid.
The length of the accumulation phase in the CL could be applied to assess the mean C residence time, even if there is a change in pool size (Fig. 1b). It reflects the time between first label appearance and steady state of the 13 C import and export of 10 a compartment. In the present setup this residence time was 4-8 days in the different plant-soil compartments ( Table 2). The longest residence time was estimated for the roots and stems (7.6 days), followed by the leaves (7.2 days), petioles (6.5 days), cuttings (6.2 days), microbial biomass and soil respiration (5.9 days) and in the SOM (4.2 days). We think that this estimation of the C residence time is the most reasonable, 15 except for the SOM. In the SOM we would expect a longer or at least equal residence time as in the microbial biomass. The poor estimation in the SOM is due to the fact that the isotopic enrichment was too close to the detection limit (no significant enrichment).

Comparisons of techniques
The pulse labelling technique is most suitable to detect the minimum C transfer time 20 from the leaves to the roots. The complete labelling of the fresh assimilates facilitates a fast signal detection in the plant compartments. However, the amount of assimilates labelled during the relatively short labelling period is not sufficient to achieve a detectable signal in large C pools, such as the soil organic matter. Consequently the investigation of C partitioning and C residence time is restricted to those pools, 25 which allow clear signal detection (e.g. at least twice the magnitude of the background noise). A further disadvantage of the pulse labelling technique is, that the key parameter to consider is the decline of the 13 C signal. Thus the estimation of C allocation is based on what remains in a compartment (and not on what is allocated to it). The calculation of the mean residence time is based on the assumption that the system is at steady state, but such conditions hardly exist in nature. Thus the calculation of the mean residence time based on the rate constant of the exponential model provides at best an approximation (as it is the case for the logistic model in continuous labelling 5 experiments). Continuous labelling labels the compounds not as strong, but for longer durations and more homogeneously. Therefore this technique has the potential to detect 13 C dynamics (allocation priorities) in all plant-soil compartments, and can be applied to determine even large C pools. The parameters of the logistic model used to describe 10 the tracer dynamics lead to more specific information on C cycling. The time lag is an indicator for the maximum transfer velocity, however its assessment is poorer than with the PL technique. The time of inflection (minus the lag time) marks the mean C transfer velocity through a compartment and thus illustrates the short-term C cycling including transient storage pools. The length of the accumulation phase is an indicator for the 15 mean C residence time in a compartment and the level of the steady state reflects the amount of C allocated into it.

Conclusions
The C transfer velocity, C partitioning and C residence time can be assessed with both labelling techniques. The C transfer velocity of fresh assimilates from the leaves, 20 through the plant and to the belowground compartments is best assessed by the pulse labelling technique. However, PL is restricted to smaller C pools, due to the dilution of the tracer signal in large C pools. The plant-soil C partitioning pattern obtained by PL and CL technique are very similar, but the time of sampling is crucial. One has to account for the time lag in C transfer from the leaves to other compartment and for the 25 mean residence of the C within it. In the current study on young poplar trees, 4-8 days BGD BGD a developing Populus deltoides plantation, Tree Physiol., 24, 1347Physiol., 24, -1357Physiol., 24, , 2004. Coplen, T. B.: Guidelines and recommended terms for expression of stable-isotope-ratio and gas-ratio measurement results, Rapid Commun. Mass Sp., 25, 2538-2560. Dawson, T. E., Mambelli, S., Plamboeck, A. H., Templer, P. H., and Tu, K. P.: Stable isotopes in plant ecology, Annu. Rev. Ecol. Syst., 33, 507-559, 2002 Endrulat, T., Saurer, M., Buchmann, N., and Brunner, I.: Incorporation and remobilization of 13 C within the fine-root systems of individual Abies alba trees in a temperate coniferous stand, Tree Physiol., 30, 1515-1527 Lata, J. C., Priault, P., Barthes, L., and Loustau, D.: Seasonal variations of belowground 25 carbon transfer assessed by in situ 13 CO 2 pulse labelling of trees, Biogeosciences, 8, 1153-1168, doi:10.5194/bg-8-1153 Pulse-labelling trees to study carbon allocation dynamics: a review of methods, current knowledge and future prospects, Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | water relations across scales, Biogeosciences, 9, 3083-3111, doi:10.5194/bg-9-3083-2012, 2012. Werner, R. A., Bruch, B. A., andBrand, W. A.: ConFlo III -an interface for high precision delta(13)C and delta(15)N analysis with an extended dynamic range, Rapid Commun. Mass Sp., 13, 1237Sp., 13, -1241Sp., 13, , 1999 Werth, M. and Kuzyakov, Y.: 13 C fractionation at the root-microorganisms-soil interface: a review and outlook for partitioning studies, Soil Biol. Biochem., 42, 1372-1384 a, b, k ) of the exponential and the logistic model used to describe the 13 C dynamics in the pulse (PL) and continuous (CL) labelling, respectively. Parameter a is the total amount of 13 C [mg] at the signal peak (PL) or at the stationary level (CL) and is an indicator for the amount of C allocated and retained in the plant-soil compartments (the proportion of the total is given in brackets). Parameter b marks the time of the signal peak (PL) or the time of inflection (CL) and mean reflects the C transfer velocity. Parameter k is the rate constant describing the decrease (PL) and increase (CL) of the 13 C abundance in a compartment, which is the basis for the mean residence time (MRT) calculation. The time of the stationary level is a further indicator for the C residence time in the compartments.  Table 3. Effects of labelling technique and sampling date on the estimation of C partitioning between plant-soil compartments. Relative 13 C distribution (in %) between plant-soil compartments measured at different sampling dates after pulse and during continuous labelling. The effects of labelling technique ("labelling") and sampling date ("sampling") were tested for the sampling dates in common (1, 2 and 8 days) by two-way analysis of variance (ANOVA). No significant interaction effect was detected between the independent variables. The significance levels indicated are P < 0.05 (*) and P < 0.01 (**).  Table 4. Effects of labelling technique and sampling date on the estimation of belowground C partitioning. Relative 13 C distribution (in %) between belowground pools and fluxes measured at different sampling dates after pulse and during continuous labelling. The effects of labelling technique ("labelling") and sampling date ("sampling") were tested for the sampling dates in common (1, 2 and 8 days) by two-way analysis of variance (ANOVA). No significant interaction effect was detected between the independent variables. The significance levels indicated are P < 0.05 (*) and P < 0.01 (**).  1. 13 C tracer dynamics after label addition. Visualisation of the 13 C dynamics in plant-soil compartments after pulse labelling (a) or during continuous labelling (b) given for a system at steady state or at growth with an increase in pool size. The dynamics can be described by three phases: (1) lag phase (time needed for C transfer), (2) phase dominated by 13 C import or net accumulation and (3) phase dominated by 13 C export or stationary phase (equilibrium between 13 C import and export).

16266
BGD 10,2013 Carbon transfer, partitioning and residence time in the plant-soil system  Fig. 2. Dynamics in the 13 C distribution. 13 C label recovered after the pulse and during the continuous exposure of the plants to 13 CO 2 in leaves (a), petioles (b), stems (c), cuttings (d), roots (e), microbial biomass MB (f), soil organic matter SOM (g) and in the soil respiration SR (h) expressed as mass of 13 C in excess m E [mg 13 C and mg 13 C day −1 ]. Error bars indicate ± one standard deviation of the three plant individuals. The best fits (nonlinear least squares) are given for the exponential function after pulse labelling and for the logistic increase during continuous labelling. The coefficient of determination (R 2 ) and the root-mean-square-deviation (RMSD) were calculated with the individual measurement points. A sensitivity analysis of the logistic model fit can be found in the Supplement.