Short-term dynamics of abiotic and biotic soil 13CO2 effluxes after in situ 13CO2 pulse labelling of a boreal pine forest

Authors


Author for correspondence:
Jens-Arne Subke
Tel:+44 (0)1904 432991
Email: js51@york.ac.uk

Summary

  • • Physical diffusion of isotopic tracers into and out of soil pores causes considerable uncertainty for the timing and magnitude of plant belowground allocation in pulse-labelling experiments.
  • • Here, we partitioned soil CO2 isotopic fluxes into abiotic tracer flux (physical return), heterotrophic flux, and autotrophic flux contributions following 13CO2 labelling of a Swedish Pinus sylvestris forest. Soil CO2 efflux and its isotopic composition from a combination of deep and surface soil collars was monitored using a field-deployed mass spectrometer. Additionally, 13CO2 within the soil profile was monitored.
  • • Physical (abiotic) efflux of 13CO2 from soil pore spaces was found to be significant for up to 48 h after pulse labelling, and equalled the amount of biotic label flux over 6 d. Measured and modelled changes in 13CO2 concentration throughout the soil profile corroborated these results. Tracer return via soil CO2 efflux correlated significantly with the proximity of collars to trees, while daily amplitudes of total flux (including heterotrophic and autotrophic sources) showed surprising time shifts compared with heterotrophic fluxes.
  • • The results show for the first time the significance of the confounding influence of physical isotopic CO2-tracer return from the soil matrix, calling for the inclusion of meaningful control treatments in future pulse-chase experiments.

Introduction

The use of carbon (C) isotopes as in situ tracers in pulse-labelling experiments is now a standard tool in ecophysiology, and has been instrumental to the unravelling of processes in assimilation, allocation and mineralization of C in plant–soil systems (Horwath et al., 1994; Ostle et al., 2003; Leake et al., 2006). Research in this field is important, because interactions between plants and soil organisms have been shown to have significant impacts on the decomposition of soil organic matter (Kuzyakov, 2002; Subke et al., 2004), thus providing a potential feedback between enhanced aboveground plant activity and elevated rates of soil decomposer activities. In particular, the short-term dynamics of C flow from leaves to belowground plant organs, and subsequently soil organisms, have been investigated in only a few studies (Simard et al., 1997; Ekblad et al., 2005; Carbone & Trumbore, 2007; Kastovska & Santruckova, 2007; Högberg et al., 2008; Kodama et al., 2008). Isotopic tracer studies will remain an important method for the investigation of processes of C flow from plants to soils and back to the atmosphere, and a thorough understanding of all biotic and abiotic processes affecting the transfer of the isotopic label in the plant–soil–atmosphere system is required to derive reliable conclusions.

To date, isotopic pulse labelling using either 13C or 14C has been limited mainly to small-scale experiments, using potted plants or mesocosms (Warembourg et al., 2003; Butler et al., 2004; Kuzyakov & Cheng, 2004), or in ecosystems with short vegetation only (Ostle et al., 2003; Rangel-Castro et al., 2004; Carbone & Trumbore, 2007; Kastovska & Santruckova, 2007; Wang et al., 2007). A significant part of our understanding of plant–soil interactions and C transfer is derived from controlled experiments in artificial settings, for example using potted plants. A recent experiment using a 13CO2 tracer experiment in a naturally regenerated forest stand (Högberg et al., 2008) provided unique insights into C allocation to plant organs and belowground respiration, including the time lag associated with phloem transport of assimilates. However, the interpretation of tracer enrichment data is prone to erroneous conclusions during the initial period following the pulse because of high enrichment in soil air following physical diffusion of tracer into soil pores. This effect of physical, and not biological, return of tracers was noted in the results of this earlier study on account of the high isotopic initial enrichment in soil CO2 efflux preceding the peak in enrichment in soluble carbohydrates in the phloem. To date, however, the impact of the abiotic flux of isotopic tracers from soil pores on estimates of the time and magnitude of belowground allocation of plant assimilates has not been explicitly investigated, despite its considerable potential for artefacts in return flux calculations.

We present data from a second pulse-chase experiment conducted at the same site as the previous stand-scale pulse experiment (Högberg et al., 2008). In the present study, for the first time, we monitored the abiotic return of an atmospheric 13CO2 pulse label in real time from soil in deep collars. These deep collars deliberately exclude plant assimilate input as a result of severing of roots and distinguish this flux from the biotic flux measured from undisturbed soil (surface collars; including soil organic matter (SOM) and plant-derived CO2). Temporal changes of 13C in CO2 were sampled from permanently installed soil gas sampling chambers. These approaches provide novel insights into the temporal dynamics of heterotrophic and autotrophic C-flux components, as well as supplying for the first time a direct measure of the magnitude of abiotic 13CO2 flux and its relation to the time lag of photoassimilate transfer belowground.

Materials and Methods

Site description

The experiment was carried out in June 2007 at the same site as a previous stand-scale 13C-CO2 pulse-labelling experiment described by Högberg et al. (2008) at Rosinedalsheden, northern Sweden (64°09′N, 19°05′E, 145 m asl), in a young (c. 15 yr old) naturally regenerated Pinus sylvestris L. stand. The understory was a sparse cover of Calluna vulgaris L. and Vaccinium vitis-idaea L., while the ground was densely covered by lichens (dominated by Cladonia spp.). The soil is a weakly developed podsol with a thin mor-type organic layer (c. 3 cm), formed on a postglacial alluvial sand. More details of plant height and biomass within the pulse chambers are given in Table 1.

Table 1.  Plant biomass data for the vegetation within the two pulse chambers
Plot12
  1. Data on tree age, height and basal area do not include stems < 1.3 m height. Tree age and height values represent mean ± standard deviation. Tree dry biomass is calculated from site-specific allometric equations, using tree diameter at breast height and height. Understory biomass is estimated from values of percentage plot cover and biomass per unit area.

Tree number
Total 51 49
< 1.3 m height  0 7
Tree age (yr) 14 ± 3 14 ± 3
Tree height (m) 2.9 ± 0.9 3.3 ± 1.0
Basal area (cm2)384476
Foliar biomass (kg)
Trees > 1.3 m height 6.9 6.7
Trees < 1.3 m height 0.0 0.3
Vaccinium vitis-idaea. 0.1 0.2
Calluna vulgaris 1.5 1.2
Total 8.5 8.4
Other above-ground biomass (kg)
Trees > 1.3 m 46.9 56.9
Trees < 1.3 m 0.0 0.3
Cowberry < 0.1 0.1
Heather 0.9 0.7
Lichen 10.0 8.5
Total 57.8 66.5

13CO2 labelling method

Each pulse chamber had an octagonal basal area covering 50 m2, and a height of 4 m, resulting in a chamber volume of 200 m3. A supporting frame of aluminium scaffolding pipes (10 cm diameter) was erected and covered by a plastic film, which was sealed around the base of the chamber with sand. Air inside the chamber was cooled using an industrial chiller capable of circulating twice the chamber volume of air every minute, which controlled air temperature inside the enclosure to within 1°C of ambient air. Isotopic pulse labelling was carried out on 15 June 2007 by introducing 25 l of 95% 13C-atom enriched pure CO2 shortly after sealing the chambers, resulting in an increase in CO2 concentration by c. 150 µmol mol−1 (ppm) and an initial 13C content of 21 atom% (average of three samples taken from each chamber starting 1 min after release of the tracer). The internal CO2 concentration was monitored continuously using an infrared gas analyser (IRGA; Li8100; Li-Cor, Lincoln, NE, USA), and the enclosure remained sealed for the period of time (c. 100 min) required for the CO2 concentration to decrease from c. 530 to 315 ppm. As IRGA measurements only recognize a fraction of the 13CO2 in air (Lee et al., 2006), measurements at high 13C enrichments are prone to considerable underestimations of the actual CO2 concentration. In addition to the continuous measurements, therefore, periodic gas samples were taken from within the chamber to monitor the changes in actual CO2 concentration and its isotopic signature over time by isotope ratio mass spectrometry.

Measurements of soil CO2 efflux and isotopic composition

Soil CO2 efflux was measured continuously using flow-through chambers (PVC; 20 cm diameter, 10 cm height), which were either on the soil surface (referred to as ‘surface’; including plant-derived CO2) or placed on PVC collars of the same diameter which had been inserted 30 cm into the soil (referred to as ‘deep’ collars). The lower edge of deep collars is well below the main rooting depth of these forests (Plamboeck et al., 1999), and fluxes measured from these collars, which were inserted 24 h prior to 13CO2 pulse labelling, therefore exclude CO2 derived from assimilation during the pulse period as a result of severing of roots. Air was drawn continuously through the chambers at a constant rate of 0.3 l min−1, with ambient air entering the chamber through an inlet with an orifice of 1 cm to maintain an air seal at this flow rate and ensure no leakage of CO2 from the chamber space.

The CO2 concentration and 13C:12C isotopic ratio in sample lines from soil chambers were measured directly in the field using a mobile laboratory equipped with a mass spectrometer and a custom-built continuous flow interface. During one measuring cycle, the flow interface samples sequentially from up to 15 sample lines and one additional line allocated to a reference gas cylinder. The sample lines were allocated to six chambers in plot 1 (three surface chambers and three measuring deep collars) and three surface chambers in plot 2. Additional lines were used for ambient air from each of the plots (sampled from next to a chamber inlet), and additional surface soil chambers located outside the labelling plots for 13C natural abundance determinations.

The continuous flow isotope ratio mass spectrometer (CF-IRMS) consisted of a gas sampling interface connected to a 12-cm-radius magnetic sector mass spectrometer (SIRAS Series 2; Micromass, Manchester, UK), a non-ionizing electromagnetic radiation (NIER) type ion impact source, a triple Faraday collector system, and a rotary/turbo-molecular pumping vacuum system, interfaced to a Microsoft Windows™ data system (model name ‘PVS 12’, built by Pro Vac Services, Crewe, UK). The CF-IRMS, together with a temperature control system and gas supplies for the CO2 reference and helium (He) carrier, was custom-built to fit into a regular horse box trailer, thus providing the opportunity to conduct isotope ratio measurements in real time and under field conditions, with no need for off-line sampling and the associated problems of gas handling and storage.

Soil CO2 efflux (F) was calculated from CO2 concentration in soil chamber air according to:

image(Eqn 1)

(CSample and CAir, the CO2 concentrations in the chamber headspace and ambient air, respectively; flow, the flow rate of air through the chamber; A, the area covered by a soil chamber.) The isotope ratio of soil CO2 efflux was calculated using a two-source mixing model:

image(Eqn 2)

(C and δ, respectively, the CO2 concentration and 13C/12C isotopic mixing ratio of a gas; the suffixes relate to soil-derived CO2 (‘Soil’), CO2 in ambient air (‘Air’), and CO2 in the sample line (‘Sample’), containing a mixture of the other two sources.)

CO2 concentration and isotopic composition within the soil

Preliminary soil gas measurements were performed during the June 2007 labelling experiment, and repeated in a follow-up experiment in August 2008. The pulse-labelling methodology in 2008 was identical to that used in 2007, and we present data from the more thorough 2008 soil profile CO2 monitoring here.

Soil gases were sampled in situ at five depths below the soil surface, using permanently installed gas sampling chambers (Magnusson, 1989). These ‘artificial soil pores’ consist of 70–95-cm-long polyethylene tubes, sealed at the protruding upper end by a combined butyl rubber and silicon rubber septum, and at the perforated lower end by a 5-cm-long gas-permeable polytetrafluoroethylene (PTFE) membrane (expanded Gore-Tex™ membrane). The internal pore volume of sampling chambers varied between 7 and 11 cm3, depending on length.

Soil gases were sampled at depths of 0, 5, 20, 40 and 60 cm, where zero depth is the mineral soil surface, that is, below the c. 3-cm-thick mor layer. Nests of sampling chambers were replicated four times within the isotope-labelled plot, and replicated three times at an external control plot c. 30 m away from the labelling area. The latter was used as a background reference regarding soil CO2 and 13CO2 concentrations. All chambers were installed 3 d before the labelling, and sampling was carried out at regular time intervals over 48 h after labelling.

The procedure for gas sampling included the extraction of a gas volume equal to the internal chamber volume, which was discarded. After 5 min, a 6-ml gas sample was taken with a gastight valved syringe, and transferred to pre-evacuated gas containers. The CO2 concentration and the 13C abundance of the CO2 were determined by continuous-flow isotope ratio mass spectrometry (CF-IRMS; ANCA TG-system, 20-20 Analyzer; Europa Scientific Ltd, Crewe, UK).

Results and Discussion

Temporal dynamic of 13C label return

Soil-derived CO2 showed strong enrichment in 13C immediately following the pulse-labelling period (Fig. 1). This initial high enrichment was observed for all collar types (deep and surface), indicating that the origin of 13C is not related to transfer of assimilated labelled CO2 to below ground, as this would be absent in deep collars, which did not include intact roots. The isotope ratio of CO2 sampled with deep collars returned to approximately natural abundance values within c. 48–72 h and remained at this level until the end of the monitoring period. The surface collars initially showed the same decline in δ13C values as the deep collars, but started to show an increase in isotope ratio c. 48 h after labelling. Maximal values of the isotope flux, averaged for the surface collars of the two plots separately, were recorded c. 3.5 d (plot 1) and 4 d (plot 2) after the pulse. The magnitude of this time lag is consistent with results reported in the literature for coniferous forests (Ekblad & Högberg, 2001; Bowling et al., 2002), and indeed with measurements of soil CO2 efflux in the experiment at the same site the year before (Högberg et al., 2008).

Figure 1.

δ13C of soil CO2 efflux following stand-level 13CO2 pulse labelling. Symbols are mean values of surface collars in plot 1 (black circles) and plot 2 (open blue circles), and deep collars in plot 1 (red triangles). Error bars are standard errors, and for the two surface flux means are indicated in one direction only to aid clarity. The dashed line represents mean isotopic composition of natural abundance soil CO2 efflux (mean of collars before pulse labelling). The y-axis is scaled to a maximum of 150‰ for δ13C, but isotopic values following the pulse reached values of up to 644‰ (see in-set graph; axis labels are the same as in the main graph).

As a result of the high temporal resolution provided by the field-deployed mass spectrometer, these data allow an unprecedented separation of physical and biological 13CO2 flux following pulse labelling immediately after tracer application. The results clearly show that, for the initial 24 h following labelling, abiotic return of labelled CO2 from the soil dominates isotope flux. By the time the respired CO2 became labelled (biotic label return), deep collar δ13C signals had almost returned to pre-pulse levels. While high initial 13C isotopic values have been noted previously (Högberg et al., 2008), and were attributed to abiotic label return, our data unambiguously demonstrate that initial peaks commonly observed when soil is exposed to an atmospheric tracer during an isotopic pulse are caused by direct diffusion from the atmosphere, not by respiration of photosynthates that had been allocated below-ground (biotic label return).

The physical penetration of the label from the atmosphere into the soil profile during the pulse-labelling event is clearly demonstrated by the measurements of 13CO2 in the soil over 48 h following the pulse (Fig. 2a). High enrichment to a depth of 20 cm in the soil profile was evident immediately after the pulse chamber was removed. Over the following hours, the 13CO2 in the superficial soil layers diffused both back to the atmosphere and to deeper soil layers, where a gradual trend towards higher isotopic enrichment was observed (Fig. 2b). The dissipation of this physically entrained tracer ended after c. 12 h, when the surface soil mineral layers reached a minimum in δ13C (Fig. 2a), and isotopic enrichment at depth (Fig. 2b) was stable. The observed changes in 13CO2 abundance over the 12-h period following the pulse are fully compatible with a one-dimensional vertical CO2 transport model, which is capable of resolving both the temporal dynamics and the magnitude of 13CO2 concentrations at different depths (see Supporting Information Notes S1, Table S1 and Figs S1 and S2 for a complete model description and modelling results).

Figure 2.

Time series of isotopic abundance in the soil pore space during the 2008 pulse-labelling experiment. The indicated soil depth is measured from the interface of organic and mineral layers. (a) 0 cm (triangles), 5 cm (circles), and 20 cm (squares). (b) 40 cm (squares) and 60 cm (triangles). Values are means of three replicates, and error bars indicate 1 SE.

In comparison, 13CO2 concentrations in the soil of the external control plot were very stable over time, with δ13C values of −20‰ (SE = 0.5‰, n = 3) at the mineral soil surface and −23.2‰ (SE = 0.2‰, n = 3) in the mineral soil layers (not shown). Throughout the 48-h monitoring period, the absolute CO2 concentrations were stable over time and increased with depth from c. 1000 ppm at the mineral soil surface to c. 4000 ppm below 20 cm in both the labelled and the control locations. The rapid decrease in δ13C over the first 5 h at the most shallow depths and the timing of the increase in isotopic signature following this decline corroborate the initial (up to 48 hours) flux results in Fig. 1 as being driven by physical diffusion. The increase in 13CO2 abundance in the upper soil profile at 24 and 48 h after the pulse indicates a new source of 13C which is likely to be the respiration of labelled assimilates allocated below-ground. The increase in δ13C between 24 and 48 h seen in Fig. 2(b) indicates that, also at the lower depths, below-ground allocation of pulse-derived 13C is detectable. This is also evident in the flux results, in which deep collars remained slightly elevated with respect to natural abundance measurements throughout the monitoring period by between 5 and 10‰ on average (Fig. 1). The magnitude of this contribution is, however, insignificant, and does not therefore affect the calculation of biologically derived flux estimates. Taken together, these results confirm the flux results shown in Fig. 1.

The time lag of between 48 and 72 h needed for the physically diffused tracer to approach pre-pulse levels in this experiment may be affected by soil temperature and moisture conditions, and is likely to differ for other soil types. Our results clearly emphasize the need for appropriate controls in pulse-chase experiments in order to separate physical diffusion fluxes from truly biological fluxes of tracers within soils. Pulse-labelling systems where the pulse chamber is in continuum with the soil (Ostle et al., 2003; Leake et al., 2006; Carbone & Trumbore, 2007; Kastovska & Santruckova, 2007; Högberg et al., 2008), or where the tracer is applied to the atmosphere without any enclosure (Talhelm et al., 2007) do not always adequately consider this flux, and it is likely that soil return fluxes of the isotopic marker are overestimated under these conditions. Where possible, a physical barrier between canopy and soil may be used to prevent direct ingression of the tracer into the soil, for example where ground vegetation structure allows this, or when using whole-tree chambers (Medhurst et al., 2006). While the scale of our experiment was conducive to the separation of biotic and abiotic tracer returns, the confounding impact of the physical diffusion in and out of soil pores on the measurement of below-ground allocation of assimilated tracer applies to experiments on all scales, from potted plants to whole ecosystems. In systems where the time scale of biotic tracer return by soil-respired CO2 is shorter than in our forest system (e.g. in grasslands), a meaningful control to separate biotic and abiotic fluxes becomes even more urgent. In the light of our results, the time lags for below-ground allocation of photosynthates are likely to have been underestimated, and the absolute amount of assimilates returned via soil respiration overestimated in previously published results from short-stature vegetation. Based on the 13CO2 flux from deep collars in this study, we estimate a cumulative abiotic 13CO2 flux of 116 ± 5.4 µmol m−2 over the first 24 h alone, that is, an amount comparable to the cumulative biotic flux over the 6 d following the pulse labelling (Fig. 3). Abiotic 13C return after 48 h was 130 ± 8.0 µmol m−2, with no significant increase following this period. This amount is equivalent to c. 0.1 mg 13C for the 50-m2 chamber area, and represents c. 2.3% of the estimated uptake of labelled CO2 by the vegetation (4.0 and 4.2 g for plots 1 and 2, respectively). These results clearly indicate that in all plant–soil systems, where the initial physical return is likely to mask any tracer return following below-ground allocation, the experimental methodology has to include a true control that allows a separation of physical and biological fluxes, in order for conclusions to be meaningful. However, for the purpose of using the isotopic draw-down during the pulse for estimating tracer assimilation by photosynthesis, the amount of 13CO2 diffusing into soil pores represents a minor pool, and thus only a small error.

Figure 3.

Cumulative pulse-derived 13CO2 flux from surface collars (black circles) and deep collars (triangles), and the 13CO2 flux resulting from the below-ground allocation of assimilated 13C tracer by plants (squares), calculated as the difference between surface and deep collars. Values are mean 13C fluxes in excess of natural abundance (n = 6 for surface and n = 3 for deep collars) with error bars indicating 1 SE.

Spatial variation in pulse return

There was considerable variation in the magnitude of the peaks in 13C enrichment among single surface collars, indicating a considerable heterogeneity in C allocation from trees to chamber locations only metres apart. 13C signals in soil CO2 efflux from surface collars became significantly higher than 13C signals in deep collars between 44 and 61 h after labelling (Fig. 4a). Following a consistent increase in 13CO2 flux from all surface measurements, return fluxes reached a plateau between 4 and 5 d after the pulse, with a slight trend to decreasing isotope fluxes following this period. The cumulative flux of label-derived 13C (Fig. 4b) from individual chambers over 5 d showed a 4-fold variation (ranging from 76 to 320 mmol 13CO2 m−2), with no significant difference between collars grouped by plot (P = 0.75, Student's t-test). This variation may be caused either by differential dilution of plant-derived C flux by roots of trees outside the labelling enclosure, or by different relative contributions from plant-derived C to the total soil CO2 efflux. Roots have not been found to extend significantly beyond 4 m from tree stems in this stand (Högberg et al., 2008), so that a significant dilution through root-derived CO2 from unlabelled trees is unlikely for these collars, which were all located at the edge of the central 10-m2 area within the 50-m2 chamber area. We hypothesize instead that this variation relates to the rooting density beneath individual collars, which in turn determines the relative predominance of plant-derived CO2 in total soil CO2 efflux. As the plots are used for continued experimental work, we were not able to destructively sample root density underneath the chamber locations. Instead, we tested this hypothesis by correlating the isotopic enrichment with the proximity of surrounding trees taller than 1.3 m for each of the collar locations. Isotopic enrichment for each collar correlated significantly with the number of stems in a 2-m radius (Fig. 5) as well as with the summed circumference (which scales linearly with the cross-sectional area of the phloem) of all trees in a 2-m radius (data not shown). The good correlation of isotopic enrichment with number and size of trees in the vicinity of the collars further supports the hypothesis that root density, via autotrophic contributions to C input, accounts for the variability in isotopic enrichment observed.

Figure 4.

(a) Instantaneous fluxes of pulse-derived 13CO2 (i.e. after subtraction of natural abundance 13CO2 concentrations) for surface and deep collars in both plots. Thin lines: three surface chambers in plots 1 (solid blue lines) and 2 (broken red lines); bold line: average of three deep collars in plot 1 (with 1 SE indicated by error bars). The symbols identify the point in time for each chamber (blue circles for plot 1; open red squares for plot 2) from which the isotopic flux is significantly different from the deep collar fluxes (identified by z score; P < 0.05). (b) Cumulative 13CO2 return from surface collars in both plots (line assignment as in panel a) after subtracting isotopic values of deep collars. White circles indicate the mean cumulative flux (±1 SE) from all six surface collars (shown for 12-h intervals only).

Figure 5.

Correlation between the amounts of 13C returned from chambers in all plots with the number of stems located within 2 m of a chamber. The line indicates a quadratic regression (y = 1.53x2 −13.1x + 105.2; R2 = 0.993, F = 39.8, P < 0.001).

Diurnal soil CO2 flux variations

The continuous monitoring system indicates distinct diurnal fluctuations in the soil CO2 efflux in all collar treatments, with diurnal maxima at about twice the flux of diurnal minima. The existence of these marked diurnal fluctuations is in contrast to earlier results from a 40-yr-old Picea abies stand in the same area (Olsson et al., 2005; Betson et al., 2007), where mean fluxes were at least twice as high as reported here, and showed negligible diurnal variations. Our results suggest that, even at the relatively high temporal resolution in Betson et al. (2007) with five measurements taken over 24 h, significant diurnal dynamics went undetected. However, the absolute difference in magnitude of the flux is likely to relate to the difference in plant cover between sites (relatively sparse cover of Pinus sylvestris with understory and ground layers in this study, compared with a closed canopy of older and larger Picea abies trees).

The deep collar CO2 flux showed a declining trend over the first 3 d after the pulse. This is likely to have been caused by the installation of these collars only 24 h before pulse labelling. In order to compare the CO2 flux dynamics between the two collar types, we therefore normalized the flux values from each individual collar by dividing every reading by the average of all readings in a moving 24-h time window. This revealed a distinctive diurnal pattern for the relative magnitude of soil CO2 efflux from the different collar treatments (Fig. 6): Deep collars showed a good correlation with air temperature, with daytime peak fluxes between 14:00 and 17:00 h, whilst fluxes measured by surface collars peaked at c. 21:00 h with no discernable direct correlation with temperature. The good correlation between air temperature and (not normalized) soil CO2 efflux from deep collars, in contrast to the absence of any correlation for surface collars (Fig. 7a), further illustrates the distinct difference between collar treatments. Changes in air temperature propagate into the soil only slowly as a result of the soil's thermal inertia, resulting in an increasing time lag in the diurnal amplitude between air temperature and soil temperatures at increasing soil depth. The fact that there is a good correlation between deep collar fluxes and air temperature therefore indicates that the main source of CO2 is at shallow depth, where air temperature represents a good surrogate for soil temperature. For the surface collars, temperature correlations improve after introducing a time lag, with the best correlation obtained for fluxes correlated with air temperatures recorded 6 h previously (Fig. 7b). The thermal inertia of soils means that temperature fluctuations at increasingly deeper depth have a time lag of increasing length, and the observed pattern may be indicative of the majority of CO2 being produced at greater depth in surface collars compared with deep ones. However, deep collar insertion did not affect the distribution of soil organic matter in the soil profile, and the dynamics of the heterotrophic flux (deep collars) would be expected to form part of the surface collar flux also, with autotrophic contributions forming an additive component. Time shifts in soil respiratory activity with respect to temperature have been reported in previous studies across a range of ecosystem and vegetation types (Tang et al., 2005; Vargas & Allen, 2008a,b), and have been linked to the rate of plant assimilation and therefore assimilate supply to roots.

Figure 6.

Normalized soil CO2 efflux measured from collars inserted 30 cm into the soil (a, deep collars) and measured on the soil surface (b, surface collars) in plot 1. Flux averages from three replicates each were normalized using a running mean over 24 h around a given measurement. Air temperature is also presented (c), and vertical dotted lines mark midnight.

Figure 7.

Temperature correlation of soil CO2 efflux from deep and surface collars. (a) Correlation of deep (triangles) and surface (circles) collars with air temperature, which is here used as a surrogate of soil temperature. The solid line indicates the exponential regression for deep collars, with broken lines indicating the 95% confidence band of the regression (y = 0.615 exp0.0438x; R2 = 0.71, F = 273, P < 0.001). (b) Correlation of surface collar fluxes with time-lagged temperature of 6 h. Solid and broken lines as in panel a (y = 0.740 exp0.0272x; R2 = 0.45, F = 84, P < 0.001).

Soil trenching has been used widely to partition soil CO2 efflux into autotrophic and heterotrophic components (Hanson et al., 2000; Subke et al., 2006), and, arguably, our deep collar treatment constitutes a small trenched plot in which root input to the soil is physically inhibited. We are aware of only one study in which continuous soil CO2 efflux measurements have been taken on trenched and controlled plots (Gaumont-Guay et al., 2006), and there also a slightly less pronounced, but none the less consistent shift in diurnal flux dynamics was observed. It follows that a trenching study that relies on sporadic soil CO2 efflux measurements is prone to a significant bias depending on the time of day at which measurements are taken. As the use of either deep or surface collars has no influence on the soil temperature below the chambers, we suggest that here also the flow of substrates from roots to associated organisms is linked to plant activity above-ground. The origin of these dynamics is not clear from our present level of understanding of below-ground processes. We note that, in this study as well as in the work by Gaumont-Guay et al. (2006), water uptake by roots has been experimentally disrupted, and further work into drivers of autotrophic respiration (including root-associated respiration) in relation to substrate supply and water uptake by roots is needed to help resolve the mechanisms underlying the observed patterns.

In contrast to the total soil CO2 efflux rate, the isotopic value of this flux showed virtually no diurnal dynamics (Fig. 1), which implies that the relative contributions of plant-derived C and C originating from soil organic matter are constant. This observation is in apparent contradiction to the decoupled flux contributions observed for the CO2 flux from surface and deep collars (Fig. 6). It is possible that the turnover of C through the microbial biomass buffers any short-term changes in isotopic values of substrates metabolized in the soil, so that the magnitude of observed isotope flux represents a time-averaged value of all substrates consumed in the previous hours.

Conclusions

We have been able to isolate the isotopic flux caused by physical diffusion into the soil during isotopic pulse labelling for the first time, with direct implications for any isotopic labelling experiment using an atmospheric tracer. Further, the continuous soil CO2 efflux data show an unambiguous shift in C flux dynamics in relation to the presence of roots. The high temporal resolution of isotope ratio measurements as well as total soil CO2 efflux provided by the field-deployed mass spectrometer has been critical for the evaluation presented in this paper. Following the proof of concept that a 13CO2 pulse results in traceable enrichment throughout all relevant ecosystem compartments in the previous year (Högberg et al., 2008), the high data density and real-time data collection meant that better insights could be gained into the nature of flux dynamics both for the isotopic tracer and for the flux dynamics in relation to C input to the rhizosphere by roots.

Acknowledgements

This study was supported by grants from SLU, the Kempe Foundations, the research councils VR and FORMAS (to PH), and the UK Natural Environment Research Council (NERC). We are grateful to other members of the CANIFLEX team for assistance in the field, and Anders Ohlsson for support with isotope analysis.

Ancillary