The diel imprint of leaf metabolism on the δ13C signal of soil respiration under control and drought conditions



This article is corrected by:

  1. Errata: Corrigendum Volume 202, Issue 4, 1412, Article first published online: 21 March 2014

Author for correspondence:
Matthias Barthel
Tel: +41 (0)44 632 8196


  • Recent 13CO2 canopy pulse chase labeling studies revealed that photosynthesis influences the carbon isotopic composition of soil respired CO213CSR) even on a diel timescale. However, the driving mechanisms underlying these short-term responses remain unclear, in particular under drought conditions.
  • The gas exchange of CO2 isotopes of canopy and soil was monitored in drought/nondrought-stressed beech (Fagus sylvatica) saplings after 13CO2 canopy pulse labeling. A combined canopy/soil chamber system with gas-tight separated soil and canopy compartments was coupled to a laser spectrometer measuring mixing ratios and isotopic composition of CO2 in air at high temporal resolution. The measured δ13CSR signal was then explained and substantiated by a mechanistic carbon allocation model.
  • Leaf metabolism had a strong imprint on diel cycles in control plants, as a result of an alternating substrate supply switching between sugar and transient starch. By contrast, diel cycles in drought-stressed plants were determined by the relative contributions of autotrophic and heterotrophic respiration throughout the day. Drought reduced the speed of the link between photosynthesis and soil respiration by a factor of c. 2.5, depending on the photosynthetic rate.
  • Drought slows the coupling between photosynthesis and soil respiration and alters the underlying mechanism causing diel variations of δ13CSR.


With climate change, understanding the carbon (C) cycle in terrestrial ecosystems has become increasingly important, as terrestrial ecosystems contribute – via photosynthesis and respiration – to the regulation of atmospheric CO2 concentrations. The largest CO2 source in terrestrial ecosystems is soil respiration (Lavigne et al., 1997; Janssens et al., 2001), which consists of heterotrophic (microbial) and autotrophic (root-rhizosphere) components. In the past, soil respiration dynamics were assumed to be solely controlled by temperature and soil moisture as these influence decomposition rates and microbial activity. However, as recent photoassimilates can contribute up to 60% of total soil respiration (Epron et al., 1999; Bhupinderpal-Singh et al., 2003), photosynthesis and its drivers should indirectly influence the magnitude and dynamics of soil respiration to a large extent.

13C natural abundance experiments showed that changes in recent aboveground weather conditions such as air temperature, vapor pressure deficit, air relative humidity and precipitation explain variations in the δ13C (carbon isotope composition) signal of soil respiration, indicating a direct, but time-delayed link between photosynthesis and soil respiration (Ekblad & Hogberg, 2001; Bowling et al., 2002; Barbour et al., 2005; Ekblad et al., 2005; Knohl et al., 2005; Kodama et al., 2008). Recently, a number of 13CO2 pulse chase labeling experiments clearly showed that photosynthesis is directly linked to root-rhizosphere respiration (Högberg et al., 2008), influencing its δ13C signal even on a diel timescale, as demonstrated by the diel pattern of δ13C in soil respiration (Bahn et al., 2009; Plain et al., 2009; Dannoura et al., 2011). However, the underlying mechanisms causing these short-term effects of recent photoassimilates on belowground biogeochemistry are poorly understood. Possible mechanisms include the diel pattern of transitory starch accumulation and remobilization at the leaf level, as hypothesized by Bahn et al. (2009). Transitory starch is formed in the leaf during the day and supplies C for plant metabolism during the night, including respiration (Zeeman et al., 2007). To date, it is unclear whether the isotopic imprint of this process is translated from above ground to below ground and hence is reflected in root respiration and thus soil respiration. Generally, the time that it takes for a C molecule to pass from above to below ground conveys highly important information about C use, plant physiology and plant–soil coupling (Kayler et al., 2010).

Recently it has been shown using 13CO2 canopy pulse labeling that drought reduces phloem transport velocities, thus reducing the supply of fresh assimilates supporting root respiration (Ruehr et al., 2009). Kuzyakov & Gavrichkova (2010) determined phloem transport rates as an important rate limiting step for plant–soil coupling. However, to date a detailed process understanding of plant–soil coupling and C use over short timescales is still lacking, particularly under drought conditions. The investigation of climatic extremes is generally underrepresented among climate manipulation experiments (Jentsch et al., 2007), although current climate change is increasing the probability and the intensity of extreme events (Sterl et al., 2008), where ecosystems are more susceptible to changes in precipitation rather than to changes in temperature (Reichstein et al., 2007; De Boeck et al., 2011). Because the biosphere mediates the C flow from the atmosphere to the soil system on a daily timescale, it is crucial to understand the key mechanisms determining C allocation, C transfer times and coupling strength under extreme conditions such as drought.

The aim of this study was to investigate the underlying mechanisms controlling the diel coupling between aboveground (canopy photosynthesis) and belowground (soil respiration) processes under control and drought conditions using a 13CO2 canopy pulse labeling approach. In particular, this study investigated the effect of drought on: plant C allocation; the time-lag between assimilation and (soil) respiration (plant C transfer times); plant C residence times; and the role of leaf mobile and immobile C pools in soil respiration on a diel timescale. It is hypothesized that leaf metabolism imprints a signal on temporal variations in the C isotopic composition of soil respired CO213CSR), reflecting changes in respiratory substrate. This imprint should be reduced under drought conditions because of a lower C supply from assimilation.

In order to achieve a sufficient temporal resolution in δ13C measurements, laser spectroscopy technology was used to continuously monitor isotopic composition and mixing ratios of CO2 in canopy and soil gas exchange of drought-stressed beech (Fagus sylvatica) saplings. Additionally, a simple C allocation and growth model was used to gain a better mechanistic understanding of the processes.

Materials and Methods

Experimental design and set-up

The study was conducted in a climate chamber with controlled relative humidity (30–50%), temperature (18–22°C) and light conditions (maximal photosynthetic active radiation 560 μmol m−2 s−1) simulating a regular diel cycle with a light period of 15 h (Supporting Information Figs S1, S2). Isotopic gas-exchange measurements were performed on six 0.8-m-tall beech (Fagus sylvatica L.) saplings, grown in individual pots with potting soil (Containererde; Ökohum, Herrenhof, Switzerland). Three of the saplings were subjected to drought and three to nondrought conditions 3 wk before the 13CO2 canopy pulse label application. At the time of labeling, canopy conductance to water vapor was reduced by c.−87% in the drought-stressed plants compared with the control plants (Table 1) and the relative volumetric soil moisture content was reduced from 87 to 53% (Fig. S3). The drought and nondrought conditions were maintained at the same level after labeling by adding respective amounts of water lost daily as a result of evapotranspiration. In order to monitor soil and canopy isotopic gas-exchange fluxes separately, six combined canopy/soil chamber systems were used, where both the canopy and the pot of each beech tree were fully enclosed by a chamber compartment. The soil and canopy compartments of each chamber system were physically separated via two semicircular shaped discs placed around the trunk. By separating above- and belowground plant parts, contamination of the soil matrix with 13C tracer material was prevented during labeling and each compartment could be measured individually for isotopic composition and mixing ratios. In addition, each individual chamber system was equipped with sensors for air, soil, and leaf temperatures, soil moisture, and relative humidity. The soil chamber was further equipped with tubing in order to water the pots from the outside without deconstructing the chamber system. The six combined soil/canopy chambers were attached to a laser spectrometer (Aerodyne Research Inc., Billerica, MA, USA) which measured continuously (1 Hz) mixing ratios as well as the isotopic composition of CO2, alternating at chamber inlets and outlets. In parallel, water vapor mixing ratios were measured with a water vapor laser spectrometer (DLT-100; Los Gatos Research Inc., Mountain View, CA, USA) in order to calculate transpiration, stomatal conductance, and photosynthesis based on the total canopy leaf area (measured destructively after the experiment was completed).

Table 1.   Comparison of control and drought treatments for ecophysiological, environmental, biomass and 13Cexcess parameters
ParameterControlDroughtP-value% effect
(n = 3)(n = 3)
  1. A, canopy net photosynthesis; E, canopy transpiration; g, canopy conductance; SR, soil respiration; Ta, Ts and Tl, air, soil, and leaf temperatures; RH, relative humidity. *, P < 0.05; **, P < 0.01; ***, P < 0.001. Significant values based on the P value are set in bold.

The following parameters are means for the days before and after labeling
 A (μmol m−2 s−1)5.53 ± 0.811.42 ± 0.270.0172*−74
 E (mmol m−2 s−1)0.45 ± 0.070.13 ± 0.020.0273*−72
 g (mmol m−2 s−1)54.47 ± 14.267.34 ± 1.820.0554−87
 SR (μmol m−2 s−1)8.79 ± 0.234.68 ± 0.240.0006***−47
 δ13CSR (‰); natural abundance−23.80 ± 0.06−23.61 ± 0.190.4906−1
 Ta (°C)21.37 ± 0.2721.45 ± 0.210.8651±0
 Ts (°C)18.63 ± 0.2018.31 ± 0.200.4105−2
 Tl (°C)20.95 ± 0.3121.27 ± 0.280.5665+ 2
 RH (%)69.12 ± 1.1749.42 ± 1.250.0007***−28
The following parameters were assessed after the isotopic gas-exchange experiment was terminated
 Biomass (g)
  Tree total69.56 ± 2.8554.25 ± 1.590.0185*−22
  Leaf9.86 ± 1.9510.20 ± 0.420.8951−3
  Stem + twig29.50 ± 0.8925.01 ± 0.920.0452*−15
  Root30.20 ± 0.1719.04 ± 1.090.0012**−37
  Root : shoot0.78 ± 0.050.54 ± 0.030.0259*−30
 13Cexcess (mg 13C g–1 C plant)
  Added by labeling0.96 ± 0.020.62 ± 0.050.0099**−35
  Recovered in total biomass0.85 ± 0.030.28 ± 0.040.0008***−67
  Leaf0.03 ± 0.010.02 ± 0.010.5027−33
  Twig0.05 ± 0.010.02 ± 0.000.14311−61
  Stem0.14 ± 0.010.03 ± 0.000.0003***−78
  Root0.63 ± 0.010.21 ± 0.030.0005***−67
  Root : stem4.47 ± 0.206.77 ± 0.270.0050**+ 51
  Nighttime canopy respiration0.02 ± 0.000.01 ± 0.000.0050**−49
  Root respiration0.09 ± 0.010.08 ± 0.010.6213−12

Between one chamber outlet measurement and the next, inlet air was measured, framing each outlet measurement with a reference. The inlet and outlet were measured for 90 s each and the mean values were used for subsequent calculations. Within c. 1 h, 18 chamber measurements were completed; soil chambers were measured twice and canopy chambers once per replicate. After one sequence of 18 measurements had been completed, the laser spectrometers were automatically calibrated for 6 min. The measurement sequence of soil and canopy chambers as well as sensor recording was fully automated and controlled by a custom-written LabView program (National Instruments Corp., Austin, TX, USA). A detailed description of the set-up can be found in Barthel et al. (2011).

Labeling procedure

13CO2 canopy pulse labeling was performed by connecting all canopy chambers in parallel into a closed loop, into which 13CO2 (Cambridge Isotope Laboratories, Inc., Andover, MA, USA) could be released. A mass flow controller (Red-y Smart Controller GSC; Vögtlin Instruments AG, Aesch, Switzerland) regulated the 13CO2 label supply into the air stream, maintaining CO2 mixing ratios at c. 600 ppm. Additionally, two infrared gas analyzers (IRGAs) (Li840 and Li6262; Li-Cor Biosciences Inc., Lincoln, NE, USA), with different 13CO2 sensitivities, were connected to the loop in order to monitor the 13CO2 to 12CO2 ratio. After labeling the canopy for 30 min at 11 am, the system was flushed with label-free air for another 1 h. The flushing tubing outlet was directed into an air-conditioning exhaust, leaving no label within the climate chamber, hence avoiding potential secondary labeling. Subsequent to the label procedure, all label tubing was removed and the original measurement sequence resumed. The estimated amount of 13CO2 fixed by photosynthesis during labeling was based on the ratio and concentrations of 13CO2 and 12CO2 within the label loop. Given the mean photosynthesis measurements before and after the labeling (measured with the canopy chambers and the Aerodyne laser spectrometer for CO2), the total 13CO2 input for each individual plant could be calculated. As the CO2 concentration within the label loop was elevated compared with the normal measurement set-up (from c. 400 to 600 ppm), the photosynthesis rate was corrected for the higher CO2 concentration based on CO2 response curves which were obtained before the experiment was carried out (Fig. S4).

At the end of the label experiment, each tree was harvested and separated into stem, twig, leaf and root parts before drying at 60°C. Those samples were later milled, weighed and analyzed for their bulk δ13C content in order to estimate the recovered 13C content in the biomass.

Labeling for organic δ13C leaf sampling

In a subsequent labeling experiment, another six beech saplings were grown and labeled in a manner identical to the procedure described in the previous section, but these saplings had not been subjected to drought conditions. Instead of measuring gas exchange in the canopy, canopy chambers were removed after labeling and leaf samples taken in steps of 0, 3, 6, 9, 12, 15, 18, 21, 24, 48 and 72 h. Harvested leaf samples were immediately put in liquid nitrogen and then dried at 60°C for subsequent δ13C sugar and starch analysis. Sample preparation for isotope ratio mass spectrometry (IRMS) δ13C analysis was carried out according to Gessler et al. (2007), Gottlicher et al. (2006) and Wanek et al. (2001). A detailed description may be found in Methods S1. IRMS analysis was performed using a Flash EA 1112 Series elemental analyzer (Thermo Italy (formerly CE Instruments), Rhodano, Italy). For specific instrument details, see Methods S1. Carbon isotopic compositions are reported in the δ-notation, expressed relative to the international reference standard Vienna-Pee Dee Belemnite (V-PDB):

image(Eqn 1)

(R, the molar ratio of 13C : 12C for the standard and sample material.) In addition to the δ-notation, the 13C amount in solid samples added by pulse labeling (13Cexcess) was calculated as:

image(Eqn 2)



For Rstandard the value of 0.0111802 (V-PDB), reported by Werner & Brand (2001), was used. Atom %S is the sample atom percentage and atom %N is the natural background (set to −28‰). DW refers to the dry weight (g) and C to the C fraction (%) of the respective samples.

Laser spectroscopy measurements

CO2 isotope laser spectroscopy (QCLAS; Aerodyne Research, Inc., Billerica, MA, USA) enabled real-time simultaneous quantification of 13CO2 and 12CO2 isotopologs as well as corresponding CO2 mixing ratios by scanning across a small spectral window (near 2310 cm−1). A fully automated calibration system and an automated liquid nitrogen refilling device (Liquid N2 Microdosing System # 906; Norhof, Maarssen, the Netherlands), needed for the infrared detector, reduced system servicing to a minimum. The QCLAS was calibrated using two calibration gases with known isotopic composition and mixing ratios (low standard: 12CO2 = 361.90 ppm, 13CO2 = 3.99 ppm; high standard: 12CO2 = 529.07 ppm, 13CO2 = 5.79 ppm). The 1σ precision was 0.25‰ at 1 s and 0.05‰ at 90 s averaging time (estimated from Allen variance plots). In order to account for nonlinearity effects, a third calibration gas was dynamically diluted with CO2-free air, resulting in a CO2 range of 300–850 ppm with constant isotopic composition. The 12CO2 mixing ratio obtained for the dilution calibration was then used to normalize the measured isotope ratios to a reference mixing ratio of 400 ppm. In addition, a quality control standard was repeatedly measured to check for long-term stability and resulted in an accuracy of c. 0.25‰.


The model used is a very simple representation of the plant C balance and was originally published by Rasse & Tocquin (2006) for Arabidopsis (Fig. 1). As the model was used for modeling the δ13C signal of soil respiration (δ13CSR), all fluxes and pools of the model were run separately for 12C and 13C. Thus, each flux consisted of a 12C and a 13C ‘branch’ (indicated by double arrows in Fig. 1). The 13C ratio of soil respiration was calculated from the sum of 12C fluxes from root maintenance respiration, root growth respiration and heterotrophic soil respiration and the sum of the corresponding 13C fluxes. Finally, a time-lag function was incorporated to account for C transportation times and 13C-signal dispersion effects from the leaves via stem, roots, and soil to the measured soil respiration signal. Associated parameters are tlm, tlmax and tlerror, representing the mean and maximal transportation times and a standard deviation around that mean, respectively.

Figure 1.

Schematic overview of the model structure illustrating fluxes between various plant carbon pools. The model was adapted and modified from Rasse & Tocquin (2006); gray boxes denote pools; double arrows indicate simultaneous carbon fluxes of 12C and 13C. For abbreviations, see Table 3.

The model calibration was based on a Bayesian approach using the Differential Evolution Adaptive Metropolis (dream) algorithm (version 1.1; Vrugt et al., 2009) encoded in Matlab (The MathWorks Inc., Natick, MA, USA). For a detailed description of the model structure, calibration and parameters, see Table 3 and Methods S1.


All δ13CSR values exceeding four times the standard deviation of the mean of natural abundance values were classified as labeled. The assessment of peak times was performed using a least squares optimization of four parameters of a co-spectral model by Massman & Clement (2004). This model was able to fit the δ13CSR data around peak time very well for all six trees (all root mean square error ≤ 8.99). Peak times were derived individually for each tree from the optimized peak time (frequency) parameters (fx). All exponential decay fits were performed from the second peak [correction added after online publication 18th August 2011: in this sentence, the text ‘first maximum’ has been corrected to read as ‘second peak’] in δ13CSR, where the half-life time (HLT) and mean residence time (MRT) were calculated as 1/τ and ln(2)/τ, respectively, from the equation: inline image. All values in the Results section denote the mean ± SE (= 3) for control and drought treatments.


Ecophysiology, biomass and 13Cexcess

In general, canopy photosynthesis and soil respiration showed significantly reduced values under drought and followed a pronounced diel cycle (Fig. 2a,b; Table 1). In addition, a reduction in canopy transpiration and conductance (g) was observed, which was accompanied by a decrease in relative humidity (69 ± 1% in control and 49 ± 1% in drought replicates; Table 1). Under control conditions, the diel cycle of soil respiration was linearly dependent on soil temperature (TS; R2 = 0.76,  0.001), whereas in drought replicates such a relationship could not be found (Fig. 2c). Instead, a clear step between day and night was visible, which was not determined by changes in TS. Soil respiration was elevated during the daytime (probably as a result of assimilate input) by 0.43 μmol m−2 s−1 and decreased to a baseline value at night.

Figure 2.

(a) Canopy photosynthesis (A) and (b) soil respiration (SR) in control (closed symbols) and drought-stressed (open symbols) beech saplings (Fagus sylvatica) shown for the first day after canopy 13CO2 label application; the black line denotes soil temperature; error bars are ± SE of the mean. (c) Soil respiration dependence on soil temperature for drought-stressed (open circles) and control (closed circles) plants; gray, daytime data; black, nighttime data.

Moreover, Table 1 shows differences between control and drought samples in biomass and 13Cexcess data for various plant tissues, where 13Cexcess values were used as a measure of C allocation within the plant. Both parameters showed a reduction in all plant parts under drought, with the greatest effects in roots and stem tissue in absolute terms. However, according to the 13Cexcess data, relatively more C was partitioned to the roots than to the stem under drought conditions.

A complete plant C balance was determined by calculating how much 13C was fixed during the labeling and how much of it was finally recovered in biomass and respiration. The total amount of 13C added by photosynthesis was estimated to be 0.91 mg 13C/g C plant in control plants and 0.62 mg 13C g−1 plant in drought-stressed plants. In control plants, 89% of the added 13C was recovered in biomass (0.85 mg 13C g−1 plant, as presented in Table 1), 9% in root respiration (0.09 mg 13C g−1 plant) and 2% in nighttime canopy respiration (0.02 mg 13C g−1 plant). By contrast, in drought-stressed plants only 45% was recovered in biomass (0.28 mg 13C g−1 plant), 13% in root respiration (0.08 mg 13C g−1 plant) and 1% in nighttime canopy respiration (0.01 mg 13C g−1 plant), leaving 41% unexplained (0.25 mg 13C g−1 plant). Comparing control and drought samples, a significantly lower proportion of recovered 13C in plant biomass (−67%) and nighttime canopy respiration (−49%) was found. However, with only 59% of the initially added 13C recovered in the drought-stressed samples, interpretation of the data is limited.

Time-lags and residence times of δ13CSR

Contamination of the soil atmosphere with tracer material was avoided during the canopy 13CO2 pulse labeling because of the physical separation between canopy and soil chamber compartments. Hence, no 13CO2 tracer artifacts (potentially diffusing from soil pore spaces) influenced the δ13C measurements of soil respiration (δ13CSR). Apparently, the later observed enrichment in δ13CSR could be exclusively attributed to direct transport of labeled photoassimilates to the root system and their subsequent utilization for respiration (Fig. 3b). The coupling speed between above-ground (photosynthesis) and below-ground (root respiration) processes was reduced under drought conditions. That is, the first appearance of 13C tracer material in soil respiration after label application (time-lag1 (tl1)) increased from 2.73 ± 0.03 to 6.71 ± 0.66 h between control and drought replicates, and the time to the first δ13CSR maximum (time-lag2 (tl2)) from 9.82 ± 0.35 to 21.04 ± 0.14 h. Both time-lags (tl1 and tl2) correlated strongly with the photosynthetic rate of the respective plant canopy (Fig. 4a). Furthermore, the time-lags described by the model time-lag function were 6.3 ± 2.1 and 11.4 ± 3.6 h for the control and drought replicates, and are well within the measured ranges of tl1 and tl2 for the two treatments (calculated from model parameter tlm; Table 2).

Figure 3.

Enrichment in 13C of (a) canopy and (b) soil respiration after canopy 13CO2 label application (vertical solid black line); (c) control samples shifted for time-lag 2; closed circles, controls; open circles, drought-stressed plants; gray areas, nighttime.

Figure 4.

(a) Time-lag (tl1 and tl2) dependence on photosynthetic rate for both the first appearance of 13C tracer within the soil (circles; R2 = 0.96, n = 6,  0.01) and the first peak of δ13C (squares; R2 = 0.87, n = 6,  0.01); (b) half-life time (HLT; circles) and mean residence time (MRT; squares) dependence on photosynthesis (both R2 = 0.98, = 6,  0.001); controls, closed symbols; drought-stressed plants, open symbols. A, canopy net photosynthesis.

Table 2.   Overview of optimized parameters used in the model for control and drought-stressed plants run with upper and lower bounds for each parameter, as well as mean and maximum likelihood (mle) parameter estimates
  1. GRmax, maximum growth rate; MSP, mobile sugar pool; RSal, root to shoot allocation ratio; STbr, baseline starch production coefficient; tlm, modeled mean time-lag of δ13C soil respiration after labeling; tlerror, standard deviation of tlm; tlmax, modeled maximum time-lag of δ13C soil respiration after labeling; α, growth respiration coefficient; β, stem growth coefficient.


In addition, the HLT and the MRT (Fig. 4b), obtained from the exponential decay δ13CSR function, showed a similar relationship to photosynthesis as found for the time-lags. The average MRT in control plants was calculated as 27.87 ± 1.02 h and in drought plants as 59.46 ± 5.37 h, which was more than twice as long as in control plants.

Diel dynamics of δ13CSR in control samples

The measured δ13CSR signal in control treatments followed a pronounced diel cycle in the first 24 h after label application (Fig. 3b), including a strong secondary increase after the initial peak. Interestingly, the δ13C of canopy respiration (δ13CCR) increased also during the first night after label application but showed little change during subsequent nights (Fig. 3a). When δ13CSR was shifted for the time-lag assigned to the time from leaf C uptake to soil efflux – which was assumed to be reflected by the first maximum in δ13CSR (tl2) after label application – variations of δ13CSR coincided with daytime and nighttime periods (Fig. 3c). After the first δ13CSR maximum (caused by the canopy labeling), new, nonlabeled assimilates (from photosynthesis after the labeling) diluted the C pool with 12C atoms, which in turn caused a depletion in the δ13CSR signal in the remaining daytime. Coincidental with nightfall, a second increase in the δ13CSR was observed which lasted almost for the entire night. At ‘sunrise’ on the next day, δ13CSR started to decay permanently without any further increase. For interpretation of the data, we adapted a simple model of plant physiological processes (Rasse & Tocquin, 2006) to model mobile and immobile C pools. The model captured the measured δ13CSR diel dynamics despite its simplicity (Figs 1, 5a; Tables 2, 3). A combination of a mobile sugar pool (MSP) and a transient immobile starch pool (TISP), feeding maintenance and growth respiration fluxes in roots was able to explain the δ13CSR behavior in control plants. That is, transitory starch, which was labeled within the labeling time window during the daytime, was re-mobilized at night, thus re-enriching δ13CSR and δ13CCR. Therefore, it was hypothesized that leaf metabolism directly drives the diel rhythm of δ13CSR.

Figure 5.

Mean ± SD for δ13C of soil respiration and the corresponding model output for (a) control and (b) drought-stressed plants; white line, measured data; dark gray area, ± SD; black line, model output; ± 99% confidence intervals resulting from parameter uncertainties are not visible; MSP, mobile sugar pool; TISP, transient immobile starch pool; light gray areas, nighttime; vertical solid line, canopy 13CO2 label application.

Table 3.   List of abbreviations used in the paper
  1. V-PDB, Vienna-Pee Dee Belemnite.

ACanopy net photosynthesisμmol m−2 s−1
AgCanopy gross photosynthesisg C s−1
DWDry weightg
δ13CBULKδ13C of leaf bulk material (V-PDB scale)
δ13CCRδ13C of canopy respiration (V-PDB scale)
δ13CNOCFδ13C of neutral organic carbon fraction in leaf (V-PDB scale)
δ13CSRδ13C of soil respiration (V-PDB scale)
δ13CSTARCHδ13C of leaf starch (V-PDB scale)
δ13CWSOCFδ13C of water-soluble organic carbon fraction in leaf (V-PDB scale)
ECanopy transpirationmmol m−2 s−1
GCarbon used for plant growth (net)g C s−1
gCanopy conductancemmol m−2 s−1
GainModeled parameter associated with maintenance respiration
GlCarbon used for leaf growthg C s−1
GlrCarbon used for leaf and root growthg C s−1
GRCarbon used for plant growth (gross)g C s−1
GrCarbon used for root growthg C s−1
GRmaxMaximum growth rateg C s−1
GsCarbon used for stem growthg C s−1
HLTHalf-life timeh g
LFLeaf carbon poolg
LRRatio of LF pool to RT pool 
MRTMean residence timeh g
MSPMobile sugar poolg
MSPminMinimum level of mobile sugar poolg
MSPminfluxCarbon flux ensuring MSPming C s−1
OffsetModeled parameter associated with maintenance respiration
PartFraction of heterotrophic respiration to SR
RgGrowth respirationg C s−1
RgcCanopy growth respirationg C s−1
RgrRoot growth respirationg C s−1
RHRelative humidity%
RmcCanopy maintenance respirationg C s−1
RmrRoot maintenance respirationg C s−1
RSalRoot to shoot allocation ratio
RTRoot carbon poolg
SFCarbon allocated to MSPg C s−1
SRSoil respirationμmol m−2 s−1
SRautAutotrophic soil respirationg C s−1
SRhetHeterotrophic soil respirationg C s−1
STStem carbon poolg
STbaseBaseline starch productiong C s−1
STbrBaseline starch production coefficient
STcNight-time starch breakdowng C s−1
TISPTransient immobile starch poolg
TaAir temperature°C
TlLeaf temperature°C
tl1Time-lag to first appearcance of δ13C SR after labelingh
tl2Time-lag to first peak of δ13C SR after labelingh
tlmModeled mean time-lag of δ13C SR after labelingTime step 18 min
tlmaxModeled maximum time-lag of δ13C SR after labelingTime step 18 min
tlerrorStandard deviation of tlmTime step 18 min
TsSoil temperature°C
αGrowth respiration coefficient
βStem growth coefficient

In addition to the measured and modeled data, diel δ13C dynamics in leaf carbohydrates compounds were assessed. As explained in the Materials and Methods section, another set of well-watered beech trees (= 6) were labeled and leaf samples were taken in steps of 3 h over the course of a day. The bulk leaf material showed a strong enrichment in 13C after label application which steadily decayed thereafter (Fig. 6a). A more detailed analysis of the leaf material, however, revealed pronounced diel dynamics in the water-soluble organic C fraction (organic acids, amino acids and soluble sugars; δ13CWSOCF; Fig. 6b), the neutral organic C fraction (soluble leaf sugars only; δ13CNOCF; Fig. 6c), and the starch fraction (δ13CSTARCH; Fig. 6d). δ13CWSOCF and δ13CNOCF decayed during the remaining daytime, whereas δ13CSTARCH values remained at a steady level. During the nighttime, by contrast, δ13CSTARCH rapidly decreased to natural abundance levels, whereas the corresponding δ13CWSOCF and δ13CNOCF values increased simultaneously. The concurrent increase in δ13C leaf sugars during the night probably stemmed from transient starch degradation within the leaf. Note that, as δ13C in leaf sugar and leaf starch returned to natural abundance levels after the first night, it seems unlikely that any 13C tracer material remained within the leaf mobile C pool thereafter. Those results corresponded well to the steady exponential decay in δ13CSR after the ‘starch’ peak and to the singular increase of δ13CCR.

Figure 6.

Diel dynamics of δ13C in leaf carbohydrates after canopy 13CO2 label application. (a) Bulk leaf material, (b) water-soluble organic carbon fraction (WSOCF), (c) neutral organic carbon fraction (NOCF), and (d) leaf starch. Dotted horizontal line, zero; gray areas, nighttime; error bars are ± SE of the mean (= 6).

Diel dynamics of δ13CSR in drought samples

As observed in control plants, the δ13CCR of drought-stressed plants also showed a singular increase within the first night, which was, however, less marked (Fig. 3a). Further, no strong secondary increase in the δ13CSR signal was observed in the drought-stressed samples (Fig. 3b). Instead, weaker diel cycles were measured over the entire label period, which were not captured by the plant physiological model (Fig. 5b). The model was nonetheless able to fit the prolonged time lag and the smaller intensity of the δ13CSR very well. Furthermore, the obtained optimized parameters from the model output delivered additional information on physiology (Table 2). The proportion of the assimilative flux incorporated into TISP (STbr) was reduced from c. 30 to 3% under drought, whereas parameters relating to storage were identical to those of control plants (β, stem growth coefficient; GRmax, maximum growth rate; and RSal, root to shoot allocation ratio). Furthermore, according to the model parameters (gain, modelled parameter associated with maintenance respiration; α, growth respiration coefficient), the absolute and relative contributions of maintenance respiration to total respiration (compared with growth respiration) were higher in drought-stressed trees. In addition, the contribution of SRhet (heterotrophic soil respiration) to total SR was also reduced from 60 to 30% under drought (part, fraction of heterotrophic respiration to SR).


Carbon allocation

The 13Cexcess data indicated that the internal C allocation and utilization changed under drought conditions. At the very least, the short-term data suggest that, proportionally, more C was allocated to the roots at the expense of stems. Stem and roots contain the highest quantities of reserves in adult beech (84%; Barbaroux et al., 2003) and therefore provide a good measure for C storage strategies. A higher C dislocation to roots under drought stress, as found in the 13Cexcess data, is known as the ‘optimal partitioning theory’ after Bloom et al. (1985), which states that plants allocate more nutrients and C to belowground parts when they are limited by a water or nutrient shortage. This theory has been validated for beech by monitoring stimulated fine-root production in stands along a precipitation gradient (Meier & Leuschner, 2010) as well as for seedlings at sufficient irradiance intensities (Lof et al., 2005). Furthermore, a greater allocation of recent assimilates to roots under drought has also been found for Triticum aestivum (Palta & Gregory, 1997) and Populus tremuloides (Galves et al., 2011).

The absolute difference in tissue biomass did not confirm the 13Cexcess results. The higher root biomass in control plants was probably caused by unconstrained growth conditions, which led to a much faster biomass increase compared with drought-stressed plants. According to McDowell (2011) and references therein, growth declines before photosynthesis in drought-stressed plants, which results in an excess of nonstructural carbohydrates. Accordingly, a switch from growth to root reserve storage occurs (Galves et al., 2011). Therefore, it is probable that the growth effect on the biomass data did override the actual allocation difference. Generally, C availability and thus storage depend largely on the duration of drought (McDowell, 2011) – after a sufficiently long drought period (< 30 d) photosynthesis cannot compensate maintenance respiration, which causes an overall decline in nonstructural carbohydrates.

Time-lags and residence times of δ13CSR

Both time-lags (tl1 and tl2) correlated well with photosynthetic rate, indicating that the 13C label transport velocity from leaves (site of assimilation) to roots (site of respiration) and thus the speed of the link between above- and belowground processes depends on canopy photosynthesis, and hence on C supply and demand. A similar time-lag dependence on photosynthetic rate was discussed in a recent review by Kuzyakov & Gavrichkova (2010), where it was stated that high photosynthesis and water availability accelerate assimilate transport, altering the time-lag between photosynthesis and soil respiration. They argue that one of the bottleneck processes responsible for the time-lag is the transport rate of assimilates in the phloem. Physiologically, these findings are explained by enhanced phloem loading at the source based on higher assimilate supply and the subsequent water influx into sieve cells from the apoplast (Münch, 1930). Active phloem loading sets up a gradient in osmotic pressure, which in turn increases the hydrostatic pressure gradient between the two phloem ends, thus increasing phloem transport velocities (Nobel, 2009). Phloem transport velocities are generally in the range of 0.5–1 m h−1 in tree species (Zimmermann & Braun, 1971), which is comparable to the control values in the present study (0.4 m h−1, estimated from tl1). However, the transport rate of 0.09 m h−1 in drought-stressed plants is higher than the recently published rate of 0.01 m h−1 in beech saplings (Ruehr et al., 2009). Generally, the present results are not consistent but supportive with those of Hall & Milburn (1973), who found a reduction in phloem exudation, accompanied by a rise in sap concentration, in Ricinus under drought stress. Similarly, Hölttäet al. (2009) showed that phloem viscosity may play an additional role in determining transport velocities. Interestingly, no significant relationship with time-lag could be found for either transpiration or conductance, although xylem and phloem are interlinked (Münch, 1930; Helfter et al., 2007). Whether a much faster information transfer via pressure-concentration waves, as suggested by Mencuccini & Hölttä (2010), existed could not be determined as no rapid changes in photosynthesis were induced. Nevertheless, this did not hinder the interpretation of the results, as the time needed for a C molecule to pass to belowground was assessed and thus represents a direct measure of above- and belowground plant coupling and its integrated information on physiology (Kayler et al., 2010).

In addition to canopy photosynthesis being responsible for the C transport within the plant, it was shown that photosynthesis also determines the C turnover within the plant. The MRT and HLT dependences on photosynthesis suggest that a higher C fixation rate drives faster C utilization within the plant. Under control conditions, assimilated C was quickly turned over (MRT = 27 ± 3.1 h) as a result of respiration, growth or storage in nonlabile pools, for example structural biomass, which in turn caused a shorter residence time of label within the labile C pool. This result is consistent with a study conducted by Carbone & Trumbore (2007), who calculated an MRT of c. 13 h at shrub and grassland sites within the first 6 d after labeling. For a temperate grassland, an MRT value of 57 ± 2.7 h has been reported (Bahn et al., 2009), which is almost twice as high as the control values reported in this study. Overall, in the control plants, fresh assimilates were rapidly respired and converted to starch or structural biomass. Drought-stressed plants, however, required all available assimilates for maintenance processes.

Diel dynamics of δ13CSR

The results obtained from isotopic gas-exchange measurements in canopy and soil, δ13C analysis of leaf carbohydrate compounds, and the model output strongly suggest a transitory leaf starch pool to be responsible for the observed diel dynamics in δ13CSR of control plants. Transitory starch, synthesized in chloroplasts during the day, provides a steady and sufficient supply of C for growth, sucrose synthesis and respiration throughout the subsequent night, where the starch degradation rate is dependent on the photoperiodic length (Zeeman et al., 2007). Strong diel cycles in δ13CSR after canopy labeling were also reported by Bahn et al. (2009), who found an enrichment of δ13CSR in the late night and early morning hours for alpine grassland. Recent studies in which 13CO2 pulse label was applied to entire canopies of beech, oak (Quercus petraea), and pine (Pinus pinaster) showed similar diel patterns in δ13CSR as found in the present study (Plain et al., 2009; Dannoura et al., 2011). Furthermore, the dependence of respiration on various plant C pools has been recently reported for Lolium perenne by Lehmeier et al. (2008), where respiration consists of two fast pools and one short-term storage pool through which C cycled at least once before being respired. However, none of those pools represented a single biochemical compound or a specific location within the plant. Werner & Gessler (2011) reviewed three major mechanisms assignable to diel variations in the natural C isotope composition of respired CO2: substrate-driven variations; fractionation-driven variations; and flux-biased variations. Transitory starch was also discussed as a prominent player in substrate-driven natural variations of δ13C of respiration fluxes. As a result of isotope fractionation processes during the production of starch in the chloroplast, transitory starch is naturally enriched in 13C compared with cytosolic sucrose (Gleixner & Schmidt, 1997; Tcherkez et al., 2004). Accordingly, diel starch storage and remobilization were shown to be responsible for natural diel variations in phloem exudates of Eucalyptus delegatensis (Gessler et al., 2007), with up to 2.5‰13C enrichment during the night and 13C depletion during the day. Also, for a beech forest, natural diel variations in δ13CSR have been explained by starch day : night dynamics (Gavrichkova et al., 2011). However, natural abundance studies have produced contradictory results, as a multitude of processes interact with each other (Werner & Gessler, 2011 and references therein).

Substrate-driven variations (sugar vs starch) could not explain the small diel variations in drought-stressed samples, as no distinct TISP peak in δ13CSR was identified in drought-stressed plants and the observed low diel variations were not captured by the model. The model output showed that much fewer assimilates from photosynthesis were branched into the TISP (small STbr), which might explain why a strong peak attributable to starch degradation was not distinguishable in the δ13CSR of drought replicates. The only response that might be related to TISP degradation was the small increase in δ13CCR in the first night. A down-regulation of transient starch production in leaves under drought stress has been previously shown for other species, such as Betula pendula (Paakkonen et al., 1998), Picea abies (Kivimaenpaa et al., 2003; Zellnig et al., 2010), and Spinacia oleracea (Zellnig et al., 2004). Consequently, substrate-driven variations could not explain the very small, but distinct, δ13CSR diel changes under drought. However, according to the model and the gas-exchange measurements, drought-stressed plants show a distinct diel flux-biased change in soil respiration. As drought reduces soil microbial activity, the ratio between SRaut and SRhet should be largely biased toward SRaut under drought stress (Muhr & Borken 2009; Borken et al., 2006). Hence, the δ13CSR signal of drought replicates should be less influenced by the δ13C released by soil microbes, and therefore changes in the source more visible. Consequently, diel variability in the contribution of SRaut with a distinct isotopic composition might drive δ13CSR in drought-stressed plants, which is consistent with the potential mechanisms driving short-term dynamics proposed by Werner & Gessler (2011). For this reason, the small diel variations in δ13CSR are attributed to the dilution of the δ13C pool by fresh, unlabeled assimilates from daytime photosynthesis. As photosynthesis stops at night, dilution of δ13CSR ceases. This aside, the possibility cannot be completely dismissed that TISP is also involved in diel δ13CSR dynamics under drought stress. As there were no carbohydrate samples available for drought-stressed plants, it was assumed that if there was any leaf starch produced, the δ13C dynamics in leaf carbohydrates should be similar to those observed in the control plants. However, both our modeling and the SR data suggest that the enrichment would be in much smaller intensity.

Further, as already discussed above in the ‘Carbon allocation’ section, a surplus of nonstructural carbohydrates should be apparent under drought stress. Such a surplus would aid diel dynamics in δ13CSR as labeled material is not allocated to growth but to storage. Stored carbohydrates are then potentially redistributed, depending on the duration of the drought period. To examine this, however, further experiments focusing on the diel δ13C assessment of biochemical compounds in various tissues (leaf, stem and roots) and in different drought phases are needed.

Overall, the results demonstrate the relevance of leaf-scale processes for understanding the dynamics of soil respiration flux to the atmosphere, as basic concepts such as soil respiration dependence on temperature fail under drought because of the lack of transient starch. Further research needs to be carried out to investigate the leaf-level effects on soil respiration under field conditions and to test its relevance for terrestrial C cycle models.


Drought facilitated higher C allocation to roots and reduced the speed of the link between photosynthesis and soil respiration where photosynthetic rate was the actual rate-limiting step. In control plants, soil (root) respiration underlay a diel substrate shift, which was determined by leaf metabolism. The switch between a mobile sugar and a transient immobile C pool (starch) determined the interaction between assimilation and soil respiration on a diel timescale. The strong imprint of leaf metabolism on the diel pattern of δ13CSR was verified by a mechanistic C allocation model and additionally supported by δ13C measurements of leaf carbohydrate compounds. By contrast, diel dynamics in the δ13CSR of drought-stressed plants were not substrate driven. Instead, the model suggests that diel cycles were based on the diel variability regarding the influence of root respiration on δ13CSR, which led to different flux contributions between day and night. The fast decay of δ13CSR within several days in the control plants suggested that the utilization of recent assimilates started immediately after allocation and was rapidly completed. The results indicate that control plants invest C more quickly in storage, growth, or respiration than drought-stressed plants as the supply of assimilates does not limit these processes. The observed differences between control and drought-stressed beech saplings in the coupling between above- and belowground processes suggest that there are limitations in current C cycle models representing terrestrial ecosystem C cycling under a changing climate.


An earlier version of the manuscript was greatly improved by the helpful comments of three anonymous reviewers. Sincere thanks also to Minh Ngo for help with English grammar. Further, the authors thank Annika Ackermann and Roland A. Werner for the IRMS measurements at the Isolab, ETH Zurich. In addition, we would like to thank Peter Plüss and Patrick Flütsch for their excellent technical support as well as the Grassland Group at ETH Zurich for hosting this project. This study was financed by a Marie Curie Excellence Grant from the European Commission to A.K. (Project No: MEXT-CT-2006-042268).