Allocation of carbon to fine root compounds and their residence times in a boreal forest depend on root size class and season

Authors


Author for correspondence:
Sonja G. Keel
Tel: +1 609 258 8064
Email: skeel@princeton.edu

Summary

  • Fine roots play a key role in the forest carbon balance, but their carbon dynamics remain largely unknown.
  • We pulse labelled 50 m2 patches of young boreal forest by exposure to 13CO2 in early and late summer. Labelled photosynthates were traced into carbon compounds of < 1 and 1–3 mm diameter roots (fine roots), and into bulk tissue of these and first-order roots (root tips).
  • Root tips were the most strongly labelled size class. Carbon allocation to all size classes was higher in late than in early summer; mean residence times (MRTs) in starch increased from 4 to 11 months. In structural compounds, MRTs were 0.8 yr in tips and 1.8 yr in fine roots. The MRT of carbon in sugars was in the range of days.
  • Functional differences within the fine root population were indicated by carbon allocation patterns and residence times. Pronounced allocation of recent carbon and higher turnover rates in tips are associated with their role in nutrient and water acquisition. In fine roots, longer MRTs but high allocation to sugars and starch reflect their role in structural support and storage. Accounting for heterogeneity in carbon residence times will improve and most probably reduce the estimates of fine root production.

Introduction

The study of fine roots of trees in natural settings is a challenge. Not only are they small and hidden in the soil, their structure is also fragile and easily destroyed during sampling. In recent years, great advances have been made, thanks, in part, to the introduction of techniques such as minirhizotrons and carbon (C) isotope approaches.

Although fine roots have been treated previously as one homogeneous pool that turns over rapidly, it has now become clear that fine root populations consist of short-lived, low-order roots and more slowly cycling roots of higher orders (Pregitzer et al., 2002; Gaudinski et al., 2010; Xia et al., 2010). Significant differences in root chemistry and morphology suggest that the most ephemeral roots (first and second order) are much like leaves on a branch system. Compared with higher order roots, they are characterized by higher nitrogen (N) concentrations, higher specific root lengths and higher respiration rates (Pregitzer et al., 1998, 2002; Xia et al., 2010; Burton et al., 2012). Their primary function is to take up water and nutrients, whereas higher order roots provide structural support for lower order roots and transport water and nutrients to the rest of the plant. The presence of root scars indicates that higher order roots outlive ephemeral root modules, which are shed like leaves (Pregitzer et al., 2002).

Despite this improved understanding, our knowledge of fine root C dynamics still remains rudimentary. Are the functional differences described above also expressed in terms of C allocation patterns or C residence times? And how would these depend on season? Better knowledge of the below-ground C cycle of trees is necessary to advance forest ecosystem modelling. The model Radix is among the most sophisticated fine root models for forests (Riley et al., 2009). It simulates, for example, short- and long-lived fine roots, accounts for stored C, and incorporates seasonal growth and respiration patterns. However, uncertainties in turnover time predictions are associated with a lack of information on short-term C dynamics and, so far, starch and sugars have been treated as one single pool (total nonstructural carbohydrates), although their residence times could vary widely.

Here, we aim to fill this gap in the experimental data. Our goal is to determine the mean residence times (MRTs) in fine root C pools by combining large-scale 13CO2 pulse labelling with compound-specific isotope analysis of root tissue. In contrast with previous studies that used long-term free-air CO2 enrichment systems for 13C labelling (Matamala et al., 2003; Körner et al., 2005; Handa et al., 2008), we exposed the tree canopy to a 3000 times higher 13CO2 label by enclosing young boreal forest patches in closed chambers for a few hours only. This made it possible to trace recently assimilated C into fast as well as slow cycling compounds. For our analysis, we selected three C pools: soluble sugars, which are precursors for starch and cellulose synthesis, function as the main transport form of C, can store C, and feed respiratory processes; starch as the dominant form of C reserve/storage in roots, which can be remobilized during times of low C supply; and structural C (i.e. the backbone of root tissue), which is mainly cellulose, hemicelluloses and lignin and is only recycled to a very limited extent (Schädel et al., 2009). We compared two different fine root size classes (< 1 and 1–3 mm in diameter) to determine whether they are functionally different in terms of their C compound dynamics. In addition, we analysed the bulk tissue of first-order mycorrhizal root tips, which were too small to allow studies of labelled sugar and starch pools. For comparisons between first-order roots and roots selected by diameter (< 1 and 1–3 mm), we refer to them as ‘root tips’ and ‘fine roots’, respectively.

In ecosystems with a pronounced seasonality, the size of the fine root starch pool varies in response to season (Ericsson & Persson, 1980). It is therefore critical to assess the residence time of C in root stores during different times of the year. Thus, we labelled early (June) and late (August) in the season.

The following questions were addressed. First, what are the residence times of C in sugars, starch and structural components of fine roots? Second, how do residence times in these pools vary seasonally? And, finally, do residence times depend on root size?

We hypothesized that MRTs in sugars, starch and structural C would be on the order of days, months and a few years, respectively. The residence time of C in sugars was not expected to vary seasonally, whereas the residence time in starch was predicted to increase late in the season as a result of the build-up of C stores. We also hypothesized that the MRT in structural C would decrease in late summer because root longevity is typically shorter in winter. Residence times in starch and structural C were predicted to increase with root size, as larger roots are typically longer lived and store more C. By contrast, the residence time of sugars was not expected to vary with size class.

Materials and Methods

Site description

The labelling experiments were carried out at Rosinedalsheden in a c. 14-yr-old, naturally regenerated boreal Pinus sylvestris L. (Scots pine) forest in Sweden (64°09′N, 19°05′E, at 145 m above sea level). The soil profile is a podzol, with a 2–3 cm thick organic mor-layer (C : N = 33, pH 4.5). There is a sparse cover of the ericaceous dwarf shrubs Calluna vulgaris L. and Vaccinium vitis-idaea L., and a ground layer of Cladonia spp. lichens. In the upper 30 cm of the soil layer, the fine root biomass of pine (< 3 mm in diameter) amounted to 313 ± 16 g m−2 as determined on the unlabelled reference plots in September 2008. Roots with a diameter of < 1 mm contributed 136 g m−2 and roots of 1–3 mm in diameter contributed 177 g m−2, 37 and 47 g m−2 of which were in the organic layer, and 99 and 130 g m−2 in the mineral soil, respectively. These measurements are based on 16 30-cm-long cores taken with a 4-cm-diameter corer on four different plots. Each core was split into organic and mineral soil, and live root biomass was determined for three size classes (< 1, 1–3 and > 3 mm in diameter) and for all three species (Pinus, Calluna, Vaccinium) separately.

13CO2 labelling

In early and late summer 2007, eight 50 m2 octagonal shaped plots were 13CO2 labelled as described previously (Högberg et al., 2008, 2010). Transparent plastic chambers were erected enclosing, on average, 62 trees, which were up to 3–5 m tall. At each labelling event, two plots were labelled simultaneously (one of which was fertilized with N, see this section) and paired with an unlabelled reference plot. After sealing the chamber, 25 l of pure CO2 with > 96 atom%13C (Spectra Gases, Alpha, NJ, USA) were released into the chamber, increasing the 13C concentration from 1.1 atom% or − 8‰δ13C (natural abundance) to 22 atom% or > 24 000‰. During the course of the 1.5–3.5 h labelling events, the atom%13C of the CO2 in the chamber decreased from 22 to 11 atom% as a result of the photosynthetic uptake of 13CO2, diffusion of 13CO2 into soil pores (Subke et al., 2009) and respiratory release of unlabelled CO2. Similar amounts of 13C per plot were taken up in June and August: 9 ± 1 and 10 ± 1 g, or 0.017 and 0.018 mg 13C m−2 s−1, respectively (Högberg et al., 2010). Two weeks before each labelling event, two of the four plots were fertilized with 100 kg N ha−1 (Högberg et al., 2010). Within the duration of the present study, the fertilizer had no effect on 13C allocation to fine roots (repeated measures ANOVA for bulk tissue, P = 0.43) or on the MRT of C in sugars (ANOVAs for < 1 mm and 1–3 mm diameter roots: P = 0.63, P = 0.39), starch (P = 0.99, P = 0.21), structural compounds (P = 0.93, P = 0.33) or bulk tissue (tips, < 1 mm and 1–3 mm diameter roots, P = 0.31, P = 0.16, P = 0.82). Therefore, we treated the four plots labelled in each season as replicates.

Root sampling

Allocation of 13C to fine roots was monitored over 1 yr, with a higher sampling intensity shortly after labelling. Roots were collected 4–10 d before and 2, 4, 6, 14 ± 1 and 28 ± 1 d after labelling using a 15-cm-diameter soil corer. Additional samples were collected in October of the same year (c. 70 or 130 d after labelling), and in June and July of the following year (c. 300–400 d after labelling). To avoid roots from surrounding unlabelled trees, sampling was restricted to the central 10 m2 of the 50 m2 plots (Göttlicher et al., 2008; Högberg et al., 2008). At each sampling event, four cores were collected per labelled plot including samples from an unlabelled reference plot. Only the organic mor-layer was sampled. Soil cores were kept on ice blocks in a cooler and brought to the laboratory within 2.5 h of sampling, where samples were placed in a refrigerator (4°C).

In the laboratory, pine roots were immediately picked by hand and kept moist on damp paper in plastic bags. Dead roots and roots from understorey shrubs were discarded. On the day of field sampling, 8–10 mycorrhizal root tips (first order sensuPregitzer et al., 2002), with an average diameter of ≤ 0.3 mm and a total dry weight (DW) of 0.07–1.7 mg, were extracted per soil core and cleaned under a dissecting microscope. Samples were immediately frozen. If the DW of a sample was below the threshold of 0.07 mg, two samples from the same soil core were pooled before analysis. A total of c. 10 000 root tips were extracted. The remaining samples could not be processed within a day and were placed in a freezer until washed and sorted. Coarser roots were grouped into < 1 mm and 1–3 mm diameter size classes. The < 1 mm size class included roots that had root tips. Root samples were freeze–dried and ground to a fine powder before stable isotope analysis on an ANCA-NT solids/liquids preparation module coupled to a Europa Scientific 20–20 isotope ratio mass spectrometer (IRMS; Europa Scientific Ltd., Crewe, UK). Results are reported as 13C excess, that is, 13C enrichment of labelled roots relative to roots from unlabelled reference plots.

Sugar and starch extractions

Sugars were extracted from < 1 mm and 1–3 mm diameter roots with 1.5 ml of de-ionized water per 60 mg of powdered, freeze–dried root material at 85°C for 30 min. After centrifugation, the supernatant was transferred to ion-exchange cartridges (OnGuard II H and A 1cc cartridges; Dionex Corporation, Sunnyvale, CA, USA) to remove ionic components. The neutral fraction was collected and analysed by high-performance liquid chromatography-IRMS (for a description of the system, see Wild et al., 2010) equipped with either a HyperREZ XP Ca2+ column (Thermo Electron, Bremen, Germany) at 85°C with 0.5 ml water min−1 as eluent, or a HyperREZ XP Pb2+ guard column (Thermo Electron) at 80°C with 0.25 ml water min−1 as eluent. Total sugar values represent the sum of sucrose, glucose, fructose, pinitol, raffinose and stachyose. Starch was extracted from 100 mg of dried material by α-amylase digestion, and the resulting glucose was analysed for carbon isotopes (elemental analyser coupled to a Delta Plus IRMS; Thermo Electron) (Göttlicher et al., 2006; Richter et al., 2009). An earlier, simpler version of the digestion method (enzymatic hydrolysis) showed no significant 13C fractionation and had a calculated recovery of 99–105% (Wanek et al., 2001).

13C enrichment in structural C was calculated using a three-pool isotope-mixing model based on the concentrations (c) of each pool and their respective δ13C:

image
image

Estimation of C residence times based on exponential decay functions

Assuming that temporal changes in the 13C excess of roots can be described by first-order kinetics, the general equation for 13C excess, at time t after labeling, can be written as: inline image, with one exponential decay function for each C pool. The MRT of C corresponds to 1/λi and inline image at t = 0, represents the initial difference (at peak 13C labelling) in δ13C between labelled and unlabelled roots. In sugars, bulk root tissue and some starch samples, the observed decrease in 13C excess was consistent with a two-pool metabolic model, where one pool had a turnover time much longer than the time scale of the measurement period and the time constant λ was zero, resulting in decay functions best described by inline image. One-pool models (inline image) had a much lower r2 and, based on visual comparisons, it was very obvious that the one-pool models did not capture the actual 13C dynamics as well as the two-pool model (Supporting Information Figs S1, S2). In the case of sugars, a probable explanation for the presence of two sugar pools with very different turnover rates is given by different forms/functions of sugars. Transport sugars are likely to have a much shorter residence time and most allocation of photosynthates is to this fraction in the shorter term. Sugar molecules that result from starch breakdown occur much later after labelling, have a longer residence time and, in total, represent a much smaller fraction of the total labelled sugar pool. As our data complied with these expectations/criteria, we believe that it is appropriate to fit two-pool models, although one-pool models are more commonly used (e.g. Matamala et al., 2003). Because 13C labels in bulk root tissue were dominated by the sugar pool (Table 1; Figs 1, 2), we also fitted exponential decay functions with a constant (i.e. a two-pool model). Similarly, starch is a C pool that can be remobilized and could be built up with recycled C later in the season. Hence, we fitted exponential decay functions with a constant, but only in cases in which r2 was higher relative to functions without a constant. By contrast, the fraction of structural C that can be recycled is probably rather small (e.g. Schädel et al., 2009), and we fitted one-pool models for this compound. In cases in which two-pool models were fitted, the MRTs presented in Table 1(a) are for the second metabolic pool (i.e. with a shorter turnover rate). Functions were fitted to the data after a maximum in 13C excess had been reached (bulk root tissue, days 2–6; sugars, day 2; starch, days 4–60; structural C, days 23–280). If less than three time points remained, no functions could be fitted.

Table 1.   Mean residence time (MRT) of carbon (C) in different fine root compounds/fractions and root size classes of boreal Pinus sylvestris L. trees estimated by fitting exponential decay functions (a) or by inverse modelling (b).
Fraction/compoundRoot sizeLabelling in JuneLabelling in August
MRT (d)aSEMRT (d)aSE
  1. aMean ± 1SE of four plots labelled in June and August (early and late summer). MRTs were estimated on the basis of the observed decline in 13C excess after pulse labelling. In some cases MRTs could only be estimated for one plot (SE not applicable (n/a)). See text for further details.

  2. bRoot tips were too small for sugar and starch extraction. As an alternative to an isotope-mixing model, we approximated MRTs in structural C on the basis of 13C in bulk root tissue measured > 30 d after labelling. At that time, most C had cycled through short-turnover pools.

  3. c13C excess in structural C was calculated using an isotope-mixing model and data of bulk roots, sugars and starch.

(a)
Bulk rootTips41511n/a
Bulk root< 1 mm197136
Bulk root1–3 mm16843
Sugars< 1 mm7151
Sugars1–3 mm4120.3
Starch< 1 mm1009047660
Starch1–3 mm19860417n/a
Structural CbTips2802832842
Structural Cc< 1 mm1046164714n/a
Structural C1–3 mm726147526n/a
  Labelling in JuneLabelling in August
Fraction/compoundRoot sizeMRT (d)SEMRT (d)SE
(b)
Bulk rootTips193325
Bulk root< 1 mm45138428
Bulk root1–3 mm3512103
Sugars< 1 mm6140.3
Sugars1–3 mm4130.1
Starch< 1 mm1111918066
Starch1–3 mm80422442
Structural C< 1 mm6146351147
Structural C1–3 mm5324367076
Figure 1.

Temporal change in 13C excess of bulk root tissues of three fine root size classes (tips, < 1 and 1–3 mm in diameter) after 13CO2 labelling in June (left) and August (right). 13C excess was calculated as the increase in δ13C relative to unlabelled reference roots, which were sampled simultaneously on nearby plots. Means ± 1SE of four plots are shown. Note the different y-axes.

Figure 2.

Temporal change in 13C excess of fine root sugars (upper panels), starch and structural carbon (lower panels) after 13CO2 labelling in June (left) and August (right). Means ± 1SE of four plots are presented for < 1 mm diameter roots (open symbols) and 1–3 mm diameter roots (closed symbols). The insets show 13C excess of sugars measured > 60 d after labelling on a different y-axis. Only data after 13C peaks have been reached are presented.

To test the effect of potential changes in sugar or starch concentrations over time, we estimated MRTs based on the change in 13C excess multiplied by the concentration.

Estimation of C residence times with inverse modelling

We also used the modified terrestrial ecosystem (TECO) model (Zhou et al., 2010) to estimate MRTs and their uncertainties. This inverse modelling approach focuses on the performance of fine root C dynamics at the whole-system level by differentiating C flux pathways (Luo et al., 2001). The temporal change in 13C excess of each fine root C pool is used to estimate the rate at which C leaves each pool. This rate can be described by a constant (i.e. the C transfer coefficient) which is directly related to the MRT of C in the respective pool.

Mathematically, the model is represented by a first-order differential equation:

image(Eqn 1)

where X(t) is the 13C excess of each C pool and c is the C transfer coefficient, which is the inverse of the C residence time (MRT). X0 represents the 13C excess of the initial pool, which is also estimated by the model. A Bayesian probabilistic inversion approach was employed to optimize the model parameters (Xu et al., 2006). Bayes’ theorem states that the posterior probability density function (PPDF) p(c|Z) of parameters can be obtained from prior knowledge of parameters (prior probability density function p(c)) and the information contained in the datasets (represented by the likelihood function p(Z|c)). The prior probability density functions p(c) were specified by setting limiting intervals for parameters assuming a uniform distribution over the parameter space and by constructing the likelihood functions assuming that errors in the observed data follow Gaussian distributions. The likelihood functions p(Z|c) were specified according to distributions of observation errors (e(t))

image(Eqn 2)

where constants σ2 are the error variances of the observed data, Zi(t) are the observed isotopic data and ϕiX(t) are the modelled values, which are a product of X(t) from Eqn 1 and c representing the C transfer coefficients. Then, with Bayes’ theorem, the PPDF of parameters c is given by: p(c|Z)∝p(Z|c)p(c). Parameterizations for local conditions are usually not necessary in inverse modelling, as MRTs were estimated based on observational data. The probabilistic inversion constructs parameter distributions and assesses parameter uncertainties by quantifying maximum likelihood estimates (MLEs), means and confidence intervals, and offers much richer information contained in the data, model structure and prior knowledge on parameters than do exponential decay functions (data not shown; Xu et al., 2006).

Statistical analysis

Analysis of variance was used to test for differences in size classes, season (June vs August) or method (exponential decay vs inverse modelling) of the MRTs of C, sugar or starch concentrations. All analyses were performed with R (version 2.11.1; R Development Core Team, 2011).

Results

Allocation of recent C to bulk root tissue

Within 2 d of labelling, the bulk root tissue of all size classes was 13C labelled in both June and August (Fig. 1). Labels were highest in root tips and peaked 4–60 d after labelling. In fine roots, 13C excess reached maximum levels 2–4 d after tracer addition and decreased rapidly over the first few weeks. This resulted in short MRTs of 4–19 d based on exponential decay functions (Table 1a), but somewhat longer MRTs of 10–84 d based on inverse modelling (Table 1b; F1,12 = 11.3, P = 0.006 for < 1 mm diameter roots; F1,8 = 1.34, P = 0.28 for 1–3 mm diameter roots). Surprisingly, the MRT of bulk C in root tips was in a similar range to that in fine roots. Most likely, this is a result of the smaller fraction of short-turnover pools (i.e. sugars and starch) in tips. In this root category, 13C did not consistently decrease after labelling in August (Fig. 1), explaining why we were only able to fit an exponential decay function to data from one of our four plots (Table 1a). This was not a problem for inverse modelling, where MRTs could be estimated for all four plots.

Allocation of recent C to root sugars, starch and structural C

Root sugars were the most rapidly and strongly labelled C pool, reaching up to 450‰13C on average (Fig. 2, upper panels). Peaks in 13C of sugars occurred after 2–4 d, and MRTs were around a few days, independent of which approach was used for estimation (Table 1a,b) or when labelling took place (F1,12 = 3.22, P = 0.098 and F1,10 = 4.02, P = 0.073 for < 1 mm and 1–3 mm diameter roots, respectively). Although 13C allocated to sugars turned over rapidly, a small, labelled fraction was still present > 1 yr after labelling (Fig. 2, inset). In line with this, declines in 13C of sugars were best described by exponential decay functions with two C pools, one of which has a turnover time longer than the time scale of the measurement period (> 1 yr; see the Materials and Methods section). The MRTs presented in Table 1(a) represent the second sugar pool with the shorter turnover rates. Exponential decay functions without a constant (i.e. ignoring the potential presence of a slow turnover pool) had an average r2 of 0.44, compared with r2 of 0.93 for functions with a constant. Residence times of C in root sugars were significantly shorter in 1–3 mm diameter (3.36 d) than in < 1 mm diameter (5.36 d) roots (F1,26 = 9.77, P = 0.0043). This was true in early as well as late summer and based on both methods used to determine MRTs (Table 1a,b). Sugar concentrations were c. 2% of the DW and were significantly higher in 1–3 mm diameter roots (2.2% of DW) relative to < 1 mm diameter roots (1.7% of DW; F1,19 = 26.5, P < 0.0001). Patterns in sucrose were very similar to those of total sugars (data not shown).

Allocation of recent C to root starch and structural C was slower, and 13C peaks were observed 4–60 d after labelling, with a few exceptions, where 13C peaked in the following year (Fig. 2, lower panels). In contrast with sugars, 13C labels were much weaker and reached c. 22‰ for starch and 7‰ for structural C. After labelling in June, MRTs for starch were around a few months (80–198 d), but increased to more than a year (or 476 d) after labelling in August in < 1 mm diameter roots (F1,11 = 10.4, P = 0.008) as well as 1–3 mm diameter roots (P = 0.014). For single plots, however, MRTs ranged from 7 d up to 588 d (explaining the large SE). In line with the results of sugars, starch concentrations were also significantly higher in 1–3 mm diameter roots (4.5% of DW) relative to < 1 mm diameter roots (2.9% of DW; F1,19 = 11.6, P = 0.003).

In structural components of < 1 and 1–3 mm diameter roots, MRTs were at least 500 d (1.4 yr). The highest values were found in < 1 mm diameter roots based on exponential decay functions, with significantly lower values resulting from inverse modelling (F1,9 = 12.1, P = 0.007). To estimate MRTs in structural C of root tips, we used an alternative approach, whereby exponential decay functions were fitted to data collected at least 30 d after labelling. At this time, most C had already cycled through short-lived pools. This approach had to be used because we lacked compound-specific data for tips. As expected, calculated residence times were shorter (280–328 d) than in the larger size classes.

Accounting for changes in sugar or starch concentrations over time did not result in significantly different MRTs for sugars or starch (data not shown), but showed a slightly different temporal dynamic in 13C allocation (Fig. 2 vs Fig. 3).

Figure 3.

Change in allocation of 13C-labelled photoassimilates to different fine root compounds over time expressed as the percentage of tracer present. Upper panels represent < 1 mm and lower panels 1–3 mm diameter roots. Bar areas: sugars, white; starch, grey; structural carbon (C), black.

Seasonal change in C allocation to fine roots

In general, 13C labels were stronger in August than in June (Figs 1, 2; note the different y-axis in Fig. 1). This was most pronounced in bulk root tips (day 2, P = 0.027) and in starch of < 1 mm diameter roots (day 30, P = 0.028). Carbon residence times in 1–3 mm diameter roots decreased significantly after labelling in August as opposed to in June (F1,8 = 8.4, P = 0.020). A similar trend was observed in the sugar pool, but the differences were only marginally significant (< 1 mm, P = 0.098; 1–3 mm, P = 0.073). In starch, higher 13C labels in August were generally associated with longer MRTs (Table 1) and, 1 yr after labeling, more 13C was present (Fig. 3). Sugar concentrations increased slightly between June and August (1.6–2% vs 1.7–2.5% of DW; F1,19 = 9.6, P = 0.006), whereas starch concentrations were found to decrease (3.6–5.5% vs 2.1–3.5%; F1,19 = 12.7, P = 0.002).

Plot-level C allocation to fine roots

We combined our fine root biomass estimates with the 13C data to scale up 13C allocation to fine root biomass for the entire 50 m2 plots. Averaged over all eight plots, 3.7% of the assimilated tracer was recovered in fine roots.

Discussion

All fine root size classes and specific compounds studied in this young boreal pine forest were rapidly supplied with recently assimilated C. We observed distinct differences in C allocation patterns between root tips and fine roots, which suggest that these two groups have different primary functions within the fine root population. Furthermore, our results highlight that season controls the residence times of C within fine root compounds, as well as the amount of recent C allocated to fine roots.

Allocation to sugars

Photosynthates are transported from the canopy to below-ground tissues as sugars. This, together with the fact that sugars only make up 1–2% of the tissue DW, explains why we measured strongest 13C labels in this pool. As the primary function of sugars is to supply respiratory processes and provide ‘building blocks’ for storage and structural compounds, they turn over rapidly. However, we also detected a small fraction of sugars that had a very slow turnover time and remained labelled for more than a year (Fig. 2, inset, Fig. 3). In line with this, exponential decay functions suggested the presence of two sugar pools, one of which had a turnover time > 1 yr. This was indicated by the much higher r2 for functions that included a constant (i.e. represented a two-pool rather than a one-pool model) and also by the two slopes when plotted on logarithmic scales (Fig. S3). The occurrence of two identical C pools with different turnover rates is not uncommon and has been described, for example, for sucrose in leaves, where the two pools are associated with different compartments (vacuole vs transport sucrose) (Farrar & Farrar, 1986). Most likely, the slow turnover pool represents sugar molecules that result from starch breakdown, whereas the fast turnover pool reflects transport sugars. Contamination of the sugar samples by slow turnover compounds, such as terpenes or phenolic compounds, is unlikely.

Alternatively, the presence of sugar pools with different turnover rates could be related to the temporary deposition of labelled C in adjacent tissues along the transport pathway (Kagawa et al., 2006; Keel et al., 2007). Carbon is moved through leaky tubes formed by sieve elements and companion cells, and some exchange of C with adjacent tissues takes place (Van Bel, 2003). This mixing of different C pools could explain why new root tissues include some older C (Handa et al., 2008; Bader et al., 2009; Sah et al., 2011).

Allocation to starch and structural C

Starch and structural compounds are freshly synthesized once photosynthates have reached fine root tissues, causing delays in maximum 13C labels relative to sugars. This, together with the larger size of these pools, explains why 13C labels were much weaker relative to sugars (note the different y-axis in Fig. 2). On average, our MRT in starch of 0.6 yr agrees well with the estimate of Endrulat et al. (2010) of 0.5 yr based on pulse labelling of single and smaller trees. It is also in line with the results from a study that combined root modelling with an inadvertent 14C labelling (Gaudinski et al., 2009). However, we detected a previously unexplored and strong effect of seasonality or phenology on the MRT of starch that is discussed in more detail below.

As very little recycling of C takes place in structural compounds, we were not surprised to find the slowest turnover rates in this pool. Interestingly, our MRTs of 1.4–2.9 yr for fine roots (Table 1a,b) agree well with estimates of root lifetimes based on minirhizotrons for several temperate forests (Strand et al., 2008). Yet, our estimates are short compared with other C isotope studies that applied a continuous C isotope labelling or used the radiocarbon bomb peak approach (Gaudinski et al., 2001, 2010; Matamala et al., 2003; Keel et al., 2006; Handa et al., 2008). As the only other pulse labelling experiment we know of also documented short MRTs for structural C (Endrulat et al., 2010), we conclude that the differences are associated with the method used. Most likely, the stronger labels that are introduced by pulse labelling compared with, for example, continuous labelling allow the study of C dynamics at a higher temporal resolution and permit the detection of shorter lived fine roots. Our data provide another piece of evidence suggesting that the fine root population is composed of short- and long-lived roots that are captured differently by different methods (Gaudinski et al., 2010; Xia et al., 2010). Minirhizotrons, C isotope pulse labelling and sequential soil coring are techniques suited to the observation of the finest, most dynamic fraction. By contrast, 14C analysis and continuous C isotope labelling capture more slowly cycling roots. Alternatively, our MRTs could be slightly biased towards fast cycling roots, as our sampling was restricted to the O-horizon, and roots are often more long lived in the mineral soil (e.g. Joslin et al. 2006; Gaul et al. 2009). Climatic factors are unlikely to explain the shorter MRTs found, as MRTs are typically greater at low temperatures or high latitudes (Hendrick & Pregitzer, 1993; Gill & Jackson, 2000).

Effect of season

Increased below-ground allocation of photosynthates later in the season has been demonstrated in previous studies (Shiroya et al., 1966; Hansen & Beck, 1994) as well as in our own experiment (Högberg et al., 2010). In agreement with this, we found stronger 13C labels in fine roots after labelling in August relative to June (note the different y-axis in Fig. 1). Our data suggest that, in late summer, more photosynthates are invested into building new, longer lived tips (Fig. 1). Furthermore, the allocation of recent C to longer term storage pools is enhanced, as indicated by the dramatic increase in MRTs in starch. In August, we measured up to six times higher MRTs in starch than in June (Table 1) and, on a single plot, an MRT of 588 d was reached. The observed changes between early and late summer are probably linked to tree phenology. In June, needle elongation has not been completed and a large fraction of recent C is directed into the growth of above-ground tissues. Later in the summer, more C is available for below-ground growth and C is stored for the following season (Ericsson & Persson, 1980) as indicated by the longer MRTs.

In contrast with the dynamics of the starch pool, MRTs in bulk tissue of 1–3 mm diameter roots were shorter in late summer relative to early summer, and similar trends were observed in sugar pools. In addition, sugar concentrations were increased, indicating a stronger allocation of labelled C to respiratory pools. This is supported by the higher soil CO2 efflux rates measured in this study late in summer (Högberg et al., 2010) and suggests higher fine root activity. This could be the result of the aforementioned effects of phenology in addition to effects of season, as soil temperatures at 5 cm depth were slightly higher in August than in June (11.5 vs 9.1°C; Högberg et al., 2010).

Exponential decay vs inverse modelling

In most cases, MRTs estimated by exponential decay functions were in line with inverse modelling (Table 1a,b). However, for C compounds for which 13C peaks were reached late (e.g. structural C) or for which 13C did not consistently decrease (e.g. in bulk tissue of tips), MRTs were shorter based on inverse modelling. Although differences in MRTs between methods were sometimes large, they were only statistically significant for bulk tissue and structural C in the < 1 mm diameter size class. This discrepancy can be explained by the fact that only data collected after the maximum 13C enrichment had been reached were used to fit exponential decay functions, whereas all data could be used for inverse modelling. Furthermore, for some plots, only a few data points remained to fit exponential decay functions. In such cases, no MRT could be estimated and the mean MRTs presented in Table 1(a) are thus for fewer replicates (n ≤ 4) relative to the mean MRTs based on inverse modelling (n = 4; Table 1b). The main advantage of using inverse modelling in our study was that we could estimate MRTs in cases in which the exponential decay approach failed as a result of data loss. Important patterns, such as the increase in MRT in starch between early and late summer, could be confirmed, which strengthens our findings.

Effect of size

Because tips were not suberized (i.e. nonwoody), we assume that a larger fraction of labelled C was taken up relative to the C already present, leading to stronger and longer lasting 13C labels in the bulk tissue of tips relative to fine roots in general (Fig. 1). Alternatively, the slow decline in 13C labels could indicate the continuous import of labelled C. The temporal pattern of 13C labels in the bulk tissue of tips is somewhat misleading. It gives the impression that C in tips has a longer residence time than in fine roots, contrary to our observations; in line with expectations, MRTs in structural C were shorter in tips relative to fine roots (Table 1a).

The distinct difference in 13C allocation to the different size classes (Fig. 1) demonstrates that the conventional classification of fine roots on the basis of diameter (e.g. < 3 mm) includes roots that are functionally and physiologically very different (Pregitzer et al., 2002). The strong allocation of 13C to root tips is probably associated with their high activity in terms of nutrient and water uptake. In line with this, previous studies have shown that respiration rates are highest in the finest roots (Pregitzer et al., 1998). At the same time, foraging for water and nutrients requires rapid growth, and the strong labels could thus be associated with greater cell wall deposition relative to larger fine roots. As we only selected ectomycorrhizal tips, some labelled C found in this size class was probably also allocated to the fungal mantle that covers the tip. The high sugar and starch concentrations in the largest size class, however, reflect their function as C stores. The longer MRT in structural C of fine roots indicates that these roots are longer lived relative to tips, and provide structural support for the lower order roots.

Accounting for more heterogeneity in fine root C dynamics, in terms of fast and slow turnover pools, is likely to reduce estimates of fine root production, as suggested by previous studies. Using the root model Radix that simulates slow as well as fast cycling roots, below-ground net primary production was up to 70% lower relative to estimates based on a single fine root pool (Gaudinski et al., 2010). In the Swedish Coniferous Forest Project (SWECON), annual fine root (< 2 mm in diameter) production was originally estimated to make up > 50% of the entire C budget in a very similar P. sylvestris forest as that studied here (Persson, 1978; Ågren et al., 1980). This was later corrected downwards to 16% (Högberg et al., 2002) based on the finding that a large fraction of photoassimilates supported the respiration of fine roots and associated microbiota instead of being invested into the growth of fine roots (Högberg et al., 2001). If these patterns are common to all forests, global estimates of C allocation to fine root production (Jackson et al., 1997) may also be too high. Global vegetation models usually simulate only one fine root pool with turnover rates of ≤ 1 yr−1 (e.g. Sitch et al., 2003; Gerber et al., 2010; Zaehle & Friend, 2010). Accounting for an additional fine root pool that turns over more slowly is likely to reduce predictions of fine root production for forests.

We conclude that MRTs in fine root compounds span a range from days to several years or even decades. Even within a specific compound, up to six-fold differences in MRTs were observed, as illustrated by the strong increase in the MRT of C in starch between early and late summer. Our data add another piece of evidence, indicating that the fine root population is composed of fine roots that are functionally different. Root tips are mainly responsible for water and nutrient uptake. This is reflected by more rapid turnover relative to larger roots and the pronounced allocation of recent C, which is necessary to support high tissue activity, growth and the mycorrhizal symbiosis. Larger fine roots act as C stores, transport water and nutrients to other parts of the plant and provide structural support. This explains why they are longer lived than tips. Current ecosystem or global vegetation models rarely account for fine root pools that turn over at slow rates. The resolution of part of this heterogeneity in C residence times is likely to reduce predictions of below-ground primary production.

Acknowledgements

We are indebted to M. Blackburn and S. Schaffner for invaluable help with root sampling, washing and grinding; D. B. Metcalfe, S. Göttlicher, J. Parsby, T. Hörnlund, E. Powers, A. Schindlbacher, P. Ineson, J.-A. Subke and H. Vallack for support with labelling; and A. Olsson, E. Nordin, R. Siegwolf and M. Saurer for help with stable isotope analysis. We thank C. Schädel, S. Battermann, S. Newell, L. Morales, S. Rabin, D. Stanton, and three anonymous reviewers, for helpful comments on the manuscript. Funding was provided by the Kempe Foundation (to P.H., T.N., S.L. and V.H.), Swedish University of Agricultural Sciences and the Swedish research councils FORMAS and VR (to P.H.).

Ancillary