1. Quantifying pathways and temporal dynamics of carbon (C) flux between plants and soil is critical to our understanding of the long-term fate of C stored in soils. The potential priming of old organic matter decomposition by fresh C input from plants means that the impact of environmental changes on the interactions between plant C allocation and soil C storage need to be better understood. We used forest girdling to investigate the partitioning of total soil CO2 efflux (RS) into autotrophic (RA) and heterotrophic (RH) flux components and their interaction with litter decomposition.
2. The reduction in RS in girdled plots stabilized within two weeks at 65% of control plot values, indicating that RS is dominated by RH, and that only a small pool of available non-structural C remains in roots in late summer to sustain rhizosphere metabolic processes. RA contributions declined from 35% late in the growing season to about 25% in winter.
3. Our results indicate that actual root respiration (RR) and respiration by ectyomycorrhizas and other rhizospheric organisms (RM) contribute c. 50% each to RA between September and early November. During winter, RA remained significantly greater than zero despite frequent sub-zero air temperatures, with RM being a dominant component of RA during this period.
4. Forest girdling significantly increased the age of C in soil-respired CO2, consistent with the removal of contemporary C derived from RA. Partitioning of soil CO2 efflux on the basis of 14C results shows good agreement with the flux reduction observed between girdled and control plots.
5. Litter bag incubations indicate a promoting influence of an intact C supply to the rhizosphere on decomposition, indicating a positive rhizosphere priming effect.
6.Synthesis: Our results demonstrate significant contribution of mycorrhizas and other rhizosphere organisms to RS, and suggest a direct link between an intact rhizosphere and litter decomposition dynamics. These results highlight the tight coupling between autotroph activity and soil decomposition processes in forest soils, and add to the growing body of evidence that plant and soil processes cannot be treated separately.
Soil CO2 efflux (RS) represents the main pathway of assimilated carbon (C) from terrestrial ecosystems to the atmosphere (Janssens et al. 2001; Raich, Potter & Bhagawati 2002; Trumbore 2006). Changes in climate, pollution, or land use are set to alter both the input of C to soils and the rate of RS, resulting in changes in the amount of C stored in soils. The total flux of soil respired CO2 comprises a range of sources, which can be broadly divided into those originating from roots and organisms directly associated with them in the rhizosphere (autotrophic respiration, RA), and the decomposition of dead organic matter in soil (heterotrophic respiration, RH). The environmental conditions driving the rates of RA and RH are complex, and include both abiotic factors, such as temperature and soil moisture, and biotic factors including the amount and quality of leaf, woody, and root litter, as well as root growth and exudation rates (Subke & Bahn 2010). Over recent years, there has been an increased awareness of the role of plants in the decomposition of soil organic matter through rhizosphere priming processes (Kuzyakov 2002; Dijkstra & Cheng 2007; Fontaine et al. 2007). This means that the long-term storage of C in soils is not simply determined by the quantity and quality of litter entering the soil, or changes in abiotic soil conditions, but that there is a direct dependence on the activity of plants. In order to predict likely changes in soil C storage under changed environmental conditions, it is therefore necessary to identify different sources of soil CO2 efflux, and their dependence on environmental soil conditions, including the interaction with live plant roots and the rhizosphere.
Forest girdling has been applied successfully in a range of forest ecosystems as a means to eliminate the transport of assimilates in trees from the crown to roots, and thus remove substrate for RA, in order to partition total soil CO2 efflux (Högberg et al. 2001; Andersen et al. 2005; Olsson et al. 2005; Binkley et al. 2006; Scott-Denton, Rosenstiel & Monson 2006; Ekberg, Buchmann & Gleixner 2007). Following the girdling treatment, RS generally shows a gradual decline as root and rhizosphere metabolic processes decrease in the absence of fresh C supply, and remaining reserves of non-structural C (mainly starch) available for root metabolic processes diminish. The duration of this transient decline in RS of girdled plots compared with control plots differs according to the amount of non-structural C reserves in roots, which depends on tree species and ecosystem type, as well as the time of year when girdling was carried out (Högberg et al. 2001). Girdling experiments conducted late in the growing season indicate a stronger reliance of RA on recently assimilated C, owing to smaller root starch reserves and greater proportional allocation of assimilates belowground compared with earlier in the season (Högberg et al. 2001; Bhupinderpal-Singh et al. 2003). After available starch reserves in roots are depleted, fluxes in girdled plots have been reported to give robust estimates of RH, as evidenced by a constant ratio between girdled and control RS fluxes (Bhupinderpal-Singh et al. 2003; Subke et al. 2004).
The termination of fresh C supplies not only limits respiration by roots, but also removes root exudations which normally form the supply of C to mycorrhizal fungi and bacteria on root and hyphal surfaces. There is evidence that several ectomycorrhizal fungi (EM) rely not only on C directly supplied by the host plant, but are also capable of producing extracellular enzymes involved in the decomposition of soil organic matter (SOM) (Read, Leake & Perez-Moreno 2003; Read & Perez-Moreno 2003; Talbot, Allison & Treseder 2008). Tree girdling is therefore likely to limit decomposition of soil organic matter as primed by the metabolism of the labile C supplied by plant roots under un-girdled conditions. Soil priming may result in an enhancement as well as an inhibition of heterotrophic decomposition (‘positive’ or ‘negative’ priming effect, respectively), which has been demonstrated compellingly in both potted plant/soil systems, and under field conditions for a range of ecosystems (Kuzyakov 2002). The magnitude of soil priming under field conditions, however, remains poorly quantified, and more studies are needed in order to estimate the potential impact of changes in C supply to the soil by plants on the mineralization of C from both recent litter and older SOM. The role of mycorrhizal fungi in the transfer of fresh C from plant roots to the rhizosphere is so far not well understood. Previous results suggest a contribution of between 5% and 25% of total RS originating from respiration by mycorrhizal hyphae in forest soils (Heinemeyer et al. 2007; Moyano, Kutsch & Rebmann 2008), and more data are needed to better quantify this flux fraction. We use forest girdling in a mature Western Hemlock (Tsuga heterophylla) stand in order to address three hypotheses: (1) root starch stores in late summer are small, resulting in a rapid decline in RS in girdled forest plots, (2) a significant portion of RA is respired by ectomycorrzal (EM) hyphae and associated rhizospheric organisms, rather than roots per se, and (3) decomposition of recent and old SOM is accelerated in forest plots with intact assimilate supply to roots.
Materials and methods
The study site is a 33-year-old Western Hemlock (Tsuga heterophylla) stand located about 8 km SE of York, England (53º 55′ N, 0º 59′ W; c. 20 m a.s.l.). Mean annual temperature is 9.0 °C, and mean annual precipitation c. 627 mm (Hulme & Jenkins 1998). The tree density is about 675 trees ha−1, with a tree height of c. 18 m. There was virtually no woody debris or ground vegetation on the forest floor of this managed forest. The soil is a sandy gley podzol (Holme Moor series in the classification of the Soil Survey of England and Wales), with poor soil drainage, and ground water level near the surface in this area is controlled by ditches. Fine roots were concentrated in the organic soil horizon (average depth of 5 cm) and near the interface with the mineral soil.
Eight plots were established within the stand (total stand dimensions 50 × 200 m) in early June 2008. Each plot constituted between 10 and 12 trees within a 10-m radius, and all measurements were carried out within a radius of 2 m from the centre of each plot. Soil surface CO2 flux was monitored from three surface collars (20-cm diameter) in each plot before the girdling treatment. Surface soil collars were inserted gently into the Ol horizon only (typically <1 cm), without severing roots in the deeper organic layers. The smooth forest floor surface meant that even at such a shallow insertion depth the collars formed a good seal with the soil. Collars were secured with three steel pegs (0.5-mm diameter, 20-cm length) and inspected visually for physical disturbance before measurements were taken. On the 6th August 2008, two different mesh collars were installed in each plot. These collars were inserted 27 cm (± 0.95 cm) into the soil, thus cutting through the main rooting zone concentrated in the top 5 cm. Whilst we did not quantify fine root distribution throughout the soil profile specifically, only negligible amounts of fine roots were observed below 5–10 cm, owing presumably to the high water content in the gley soil. Four windows (4 × 6 cm) cut into the wall of each collar and covered by nylon mesh with a gauge of either 41 or 1 μm were situated immediately below the soil surface. These treatments are designed to either allow the in-growth of fungal hyphae whilst excluding roots, or excluding both hyphae and roots within the collars, following the design of (Heinemeyer et al. 2007). Soil CO2 efflux measured on the different collar types therefore includes different flux fractions of RS: (i) 1-μm mesh collars give measurements of RH, (ii) CO2 flux from 41-μm collars includes both RH and CO2 derived from EM hyphae (RM), and (iii) fluxes measured on surface collars represent estimates of total RS, i.e. including RH and RA. The flux separation on this basis relies on the effective exclusion of plant-derived C via roots and mycorrhizal hyphae in the 1-μm mesh collars. We did not carry out direct verification of either root biomass, hyphal abundance, or DNA analysis in the exclusion collars. However, we use a combination of different flux measurements, relying either on root exclusion collars or on flux reduction due to girdling and 14C abundance in soil respired CO2 to cross-validate the effectiveness of root and mycorrhizal exclusion by 1-μm collars, and thereby the validity of the approach (see below).
All plots were ranked and blocked on the basis of repeated soil CO2 efflux surveys on surface collars between June and August 2008, and four girdling and four control treatments were assigned accordingly to ensure no bias in soil CO2 efflux. On 5th September 2008, a 15–20-cm wide strip of the bark was removed around the circumference of all trees within girdling plots at about 1.5-m height. This resulted in the complete disruption of phloem transport, whilst care was taken not to damage the xylem, so that the treatment did not affect water transport within the trees.
Soil CO2 efflux measurements
Soil CO2 efflux was measured from the three collar types in each plot between 20th August 2008 and 26th January 2009. Sampling was carried out three days following the girdling treatment, and then weekly until 12th November, followed by monthly measurements. After 13th January 2009, measurements were carried out twice a week to capture short term variation in flux components during winter. On each sampling day, fluxes were measured at around midday (between 10 am and 2 pm) within the space of 1 h. Flux measurements terminated shortly before the trees at the site were felled on 28th January 2009. Measurements were performed using a Li8100 infrared gas analyser (IRGA) with 20-cm diameter survey chamber (Li-Cor Bioscience, Lincoln, NE, USA), and fluxes were corrected for each collar individually according to the exact collar height above the soil surface (3.00 ± 0.85 cm).
Following the girdling treatment, we consider total soil CO2 efflux in girdled plots to represent RH, whilst the difference between flux measurement in girdled and control plots represents RA. We apply this partitioning from two weeks following tree girdling, when the reduction in RS measured in girdled plots had reduced to a constant fraction of the RS measured in control plots. Using the flux measurements on mesh-collars, RA is further divided into CO2 flux derived from autotrophs, and respired in the absence of roots by mycorrhizal hyphae and associated organisms (RM), and that respired by roots per se (RR). RM is calculated as the difference between 41-μm mesh collars and 1-μm mesh collars, whilst RR is calculated from the difference between surface collar fluxes and RM.
Measurement of 14C content in soil CO2 efflux
Two samples of soil respired CO2 were collected in each plot on 21 and 22 October 2008. Additional collars (19-cm diameter, 10-cm high) were placed on the soil surface and secured by two metal pegs (30-cm long) inserted into the soil and fastened to the collars by a strong rubber band. Collars were not inserted into the soil in order to avoid severing roots or mycorrhizal networks, and sand was placed around the outside of each collar and onto the rubber band to form a good seal with the soil surface. Chamber covers (19-cm diameter, 36-cm height; total volume of chambers and collars: 13.04 l) were placed on these collars, with a large rubber seal which both secured the chamber covers and provided a seal between collar and chamber.
At the start of sampling, all chambers were first scrubbed to remove atmospheric CO2 by circulating the equivalent of seven chamber volumes of air through a cartridge containing soda lime (NaOH; Fisher Scientific, Leicestershire, UK). CO2 was then allowed to build up in the chambers and concentrations monitored using the IRGA. When sufficient CO2 had accumulated (5–15 mL), it was collected using the molecular sieve sampling system (see Supporting Information for a detailed description of molecular sieves for CO2 trapping and release).
Chamber concentrations were found to stabilize at relatively low levels in some chambers (between c. 650 and 800 ppm), resulting in absolute CO2 amounts in many chambers that were insufficient for 14C determinations. Where this was the case, CO2 was sampled onto molecular sieve traps after c. 6 h, and the chambers were left to allow CO2 build up again without having been opened. Following the second enrichment period, CO2 was once again trapped onto the same molecular sieve cartridges used for the first period of sampling. We also collected CO2 samples from ambient air and soil incubations (see Supporting Information for detail). 14C analyses were carried out at the Scottish Universities Environmental Research Centre, East Kilbride, UK (Freeman et al. 2007). All 14C results were normalized to a δ13C of –25.0‰ to account for mass-dependant fractionation, and expressed as %modern (Stuiver & Polach 1977). Measurement uncertainties associated with isotope concentrations are expressed as standard deviations, following convention. A detailed description of CO2 sampling on molecular sieves, sample preparation, corrections, and calculations of 14C content in soil respired CO2 is given in the Supporting Information.
We use a simple end-member mixing model to calculate the fraction of RH on RS according to:
where fH is the fraction of heterotrophic CO2 in total soil CO2 efflux, and 14C is the 14C activity of the respective sources. It is assumed that the autotrophic C contribution affected by girdling is recent and that its 14C content is therefore identical to that of the contemporary atmosphere at 105.0%modern, based on extrapolation of a long-term atmospheric 14C record (Levin et al. 2008).
We collected T. heterophylla needle material from branches at the same site in July 2008. Needles were air-dried and c. 1 g of dry litter material was placed in polyester mesh bags (0.3-mm mesh aperture size) fitted with weather resistant tags for identification. Needle material collected from branches was still green and, whilst not identical to litter on the forest floor, it was of very similar chemical composition. Since the objective of the experiment was not to derive ‘natural’ decomposition rates of local litter, but to investigate relative rates of heterotrophic decomposition following the termination of aboveground C supply, this material is a reasonable substrate for the purpose of this experiment. On the 28 August 2009, the surface litter layer (Ol horizon) was carefully removed over an area of 0.1 m2 at the centre of each of the experimental plots. Thirteen litter bags were placed on top of the Of horizon, and covered by additional litter material from the litter collected previously. All litter bags were moistened prior to placement in the plots, and one set of sample bags was collected immediately to control for effects of litter moistening on mass loss. Litter bags were collected at fortnightly intervals, dried at 60 °C for 48 h, and any root material that had grown into the bags was removed. Remaining litter dry mass was expressed as the fraction of the initial litter mass corrected for the mass loss following litter moistening at the beginning of the experiment.
Methodological validation of soil CO2 efflux partitioning methods
Since the experimental approach includes three different methods for separating RH and RA (flux comparison between girdled and control plots, deep collar insertion vs. surface collars in control plots, and radiocarbon analysis of soil CO2 efflux), we use the comparison of all these methods to compare and thereby cross-validate the individual approaches.
Measurements of environmental parameters
Continuous measurements of soil temperature at 2-, 5-, and 20-cm depth, as well as soil moisture in the organic layer were recorded continuously in one plot. Additionally, soil moisture in all plots was recorded on soil CO2 measuring days using SM200 soil moisture sensors (Delta-T Devices Ltd, Cambridge, UK), and air temperature inside the soil chamber was automatically recorded by the soil CO2 efflux equipment.
The experiment followed a randomized block design, with plot allocation based on ranked soil CO2 efflux measurements prior to girdling. Differences in fluxes between girdling treatments and soil moisture differences were tested by two-way anova for repeated measures (with girdling treatment and time as independent variables). Flux differences within one sampling date were tested using Student’s t-test. Litter mass loss rates were compared by fitting exponential decay functions (y = y0 e-bx) to litter bag data, and comparing values of the decay function (-b) using fitted parameter outputs for error estimates in a two-tailed t-test. Significance levels for all tests were P < 0.05. All tests were carried out using spss Statistics 17.0 software (SPSS Inc., Chicago, IL, USA).
Impact of tree girdling on soil CO2 efflux
Tree girdling significantly reduced soil CO2 efflux within two weeks of the girdling date (Fig. 1a). This reduction in flux rate remained significant for all subsequent sampling dates, with the exception of one date in mid-January. Autotrophic flux contributions showed a constant decline from 35% in late September to c. 25% in early January. Integrated over the period from 18 September 2008 to 26 January 2009, soil CO2 efflux averaged 117 and 169 g C m−2 from girdled and control plots, respectively, indicating an overall autotrophic contribution of 30% in this period. Wintertime fluxes showed a direct variation of RH with temperature, whilst RA remained constant during this period (Fig. 1a).
Partitioning the autotrophic flux on the basis of the mesh-collar fluxes indicated approximately equal contributions from RR and RM to RA in control plots for the period September to November, when flux fractions are also significantly different from each other (Fig. 1b). Following that period, both flux contributions remain relatively constant at c. 0.1 and 0.25 μmol m−2 s−1 (RR and RM, respectively), with spatial variations resulting in no significant differences between flux fractions (see Fig. S1a in Supporting Information). In the girdled plots, calculated fluxes of RR and RM were close to 0 throughout the experiment, indicating that in these plots, fluxes measured from all three collar types did not differ significantly and represent RH only, with no influence of RA (see also Fig. S1b).
Soil moisture at the experimental plots showed a considerable degree of variation owing to different sand/clay mixing ratios following site preparation before the plantation of the forest. This resulted in a slight but consistent difference in soil water content (SWC) that persisted from before the girdling date throughout the experiment, but was not significant at any of the sampling dates (Fig. 1c). During the entire measuring period, soil moisture was generally very high, with no evidence of soil moisture limitation of soil CO2 efflux. Since the small difference was consistent from pre-girdling conditions throughout the monitoring period, it cannot be assigned to a change in water uptake following tree girdling. Given the blocking of plots according to pre-girdling soil CO2 efflux, a confounding influence of soil moisture differences on soil CO2 efflux impact by the treatment can be ruled out.
The litter bag experiment showed two distinct phases of mass loss: a rapid initial decline (first three sampling dates in Fig. 2), followed by a reduced rate of mass loss throughout the remainder of the experiment. This change in slope likely indicates two distinct phases of litter mass loss, where the rapid initial decline is governed by leaching of soluble organic constituents (Berg 2000), and lasts approximately 1 month from litter placement in the field. The second phase shows a slower rate of mass loss and is likely to rely more on the rate of organic matter processing by decomposer organisms. During this second phase, litter bags in control plots show a greater rate of mass loss compared with those in girdled plots. Fitting an exponential decay function to the last eight sampling dates (Fig. 2) confirms a significant difference in the decay constant over this period (1.7 10−3 ± 9.9 × 10−5 day−1 and 1.3 10−3 ±1.0 × 10−4 day−1 for control and girdled plots, respectively; P < 0.05).
14C activity on soil CO2 efflux and methods comparison
There was significantly greater 14C activity in soil respired CO2 sampled from girdled plots compared with control plots (108.9 ± 0.54 and 107.5 ± 0.55%modern, respectively; P = 0.05), consistent with the removal of recently assimilated C with a 14C activity of 105.0%modern.
For the 22 October 2008, when 14C samples were taken, we are in a position to compare three independent soil CO2 efflux partitioning methods. Figure 3 shows the relative flux fractions of RH/RS on that sample date based on (i) flux partitioning on the basis of 14C results, applying the two end-member mixing model, (ii) the comparison of soil CO2 efflux measured from shallow collars in girdled and control plots, and (iii) the deep collar results obtained by the comparison of 1-μm mesh collars to surface collars in control plots only. The fraction of heterotrophic flux calculated on the basis of the 14C results was close to the value calculated on the basis of soil CO2 efflux measurements for the same date, whilst the mesh-collar method resulted in slightly lower, but not significantly different flux estimates. The comparison further shows that, for measurements on this particular date, flux fractions calculated on the basis of the girdling results are associated with the smallest spatial error.
The rapid decline in soil CO2 efflux following tree girdling confirms our first hypothesis and indicates that rhizosphere respiration relies predominantly on C recently assimilated by the vegetation. Carbohydrate reserves in roots that are available for metabolic processes within roots or that are exuded into the rhizosphere appear to be depleted within two weeks in this mature coniferous forest. This is consistent with the results of earlier girdling experiments, where the decline in RS in girdling plots lasted between 2 and 4 weeks depending on the season during which girdling was carried out (Högberg et al. 2001, 2009; Subke et al. 2004). Our results, which were obtained in late summer, correspond well to the results of Högberg et al. (2001) obtained in August in a mature spruce forest in N Sweden. Lower root reserves of carbohydrates at peak season and warmer soil temperatures are likely to cause the much shorter time lag for CO2 reduction compared with girdling earlier in the season as noted by Högberg et al. (2001).
The flux results in Fig. 1a further indicate that there is a small contribution of RA that persists throughout the winter period covered by this experiment. Measurements during January were carried out at a higher frequency, and show variability directly correlated with soil temperature. This variation is caused exclusively by changes in RH, with RA remaining at a constant flux value. Rather than being dormant, significant autotrophic processes appear to persist throughout most of winter in temperate forests located in a relatively mild oceanic climate as at our study site. The fact that there is no short-term response of RA to changes in temperature suggests that RA is limited predominantly by the supply of substrates, which does not respond to the short-term changes in temperatures. The variation in temperature observed during January appears to have no impact on the availability of labile C for respiration by the rhizosphere. With air temperatures frequently below freezing during the night, and short day lengths during winter, the actual amount of any C assimilated during these months is likely to be small, and the RA contributions observed during this period are likely to stem from root reserves of C which would have been depleted in the girdled sites. Whilst the slight increase at the end of the measuring period was not significant, it may be indicative of a renewed supply of assimilates following a relatively warm period in the previous weeks (Fig. 1c).
Results in Fig. 1b show that about half of RA in the soil at our forest site originates from respiration in the extramatrical mycelium of ectomycorrhizas. In accordance with our second hypothesis, this is a significant contribution to total soil CO2 efflux and corresponds to c. 15% of total RS. Field measurements of the contribution of mycorrhizal mycelia to RS have been reported from only a few studies (Heinemeyer et al. 2007; Moyano, Kutsch & Rebmann 2008). Heinemeyer et al. (2007) report that RA was dominated by RM for measurements from late October to December in a pine stand located close to our experimental site, and using the same mesh collar methodology. This contrasts with the findings of Moyano et al. (2007), who report only minor contributions of RM to RA in pine and beech forests in Germany. Whether mycorrhizal mycelia are the actual source of respired CO2 is questionable, as other organisms living in association with hyphae in the rhizosphere may be involved in hyphal turnover and co-metabolism of soil organic matter as a consequence of plant derived labile C ‘leaking’ from the hyphal network. We none the less choose to refer to this flux contribution as mycorrhizal flux, as these organisms are clearly central to this portion for total RS, and in our experimental setup, the in-growth of fungal hyphae is the critical difference between the two deep soil collar types. The results of our and the cited studies clearly show the reliance of EM fungi on C supply from plant roots, and changes in belowground allocation of assimilates in response to altered environmental conditions is therefore likely to impact on this flux component significantly.
The findings of the litter bag experiment indicate a promoting role of an active rhizosphere in decomposition processes in the litter layer. The significant difference in mass loss rate indicated in Fig. 2 confirms our third hypothesis, and substantiates results from the literature. The rhizosphere priming effect is well documented (Kuzyakov 2002; Fontaine et al. 2007), and Subke et al. (2004) showed similar acceleration in litter decomposition using isotopically labelled litter to measure flux partitioning in conjunction with a girdling experiment. Our results are the first to report changes in decomposition in response to autotrophic C supply based on actual litter mass loss. The dynamics of litter decomposition in Fig. 2 follows a classic sharp decline as early stages of litter mass loss are characterized by the decomposition of soluble compounds and non-lignified carbohydrates (including cellulose) (Berg 2000). The divergence in litter mass decline between girdled and control treatments occurs after this initial phase, when mass loss occurs as a result of the decomposition of more recalcitrant lignified components of litter biomass. The period of different mass loss change observed in the litter bags fell in the winter months, when air and soil temperatures were rarely above 5 °C. This further supports the finding that the rhizosphere remains active throughout winter (see Fig. 1a,b), making it likely that litter decomposition is in part being primed by an intact rhizosphere during that period. An alternative explanation of this trend is that plant nitrogen (N) uptake by roots is low during winter, so that there is more N available for decomposers to colonize litter in this period. Whilst the N dynamics during the different seasons were not part of our study, these dynamics are likely to be relevant for a better understanding of the link between rhizosphere activity and litter decomposition (Kuzyakov 2002), and should be considered in future experiments.
The small (and statistically insignificant) difference in soil moisture content observed throughout the experiment could create confounding conditions to explain the observed difference in litter decomposition rate, as the faster turnover was observed in control plots, which had the slightly higher soil moisture content (Figs 1c and 2). However, we encountered no dry conditions under which soil moisture could have been limiting for decomposition processes due to drought. Limitation of decomposition because of high soil moisture contents is similarly unlikely, as the surface layer where litter bags were incubated was never waterlogged, so that oxygen diffusion is not likely to have been significantly affected. Such conditions would in any case have led to the reversal of the observed pattern, such that litter decomposition in the ‘dryer’ girdled plots should have been accelerated compared with rates in control plots.
Given the sizeable contribution of RM to total RA documented in Fig. 1b, it is probable that the promoting effect of litter decomposition by the rhizosphere is strongly mediated by mycorrhizas. EM fungi form dense hyphal mats in the litter layer, and several species are known to produce ectoenzymes capable of decomposing organic matter (Read et al. 2003; Read & Perez-Moreno 2003; Talbot, Allison & Treseder 2008). Early work on plant–soil interactions in decomposition has suggested a retarding role on litter decomposition for mycorrhizas (Gadgil & Gadgil 1971). However, since the trenching approach in this earlier study also eliminated root water uptake, there is a confounding influence of SWC on litter decomposition that hampers firm conclusions regarding a possible retarding role of mycorrhizas on decomposition. A spatial separation of mycorrhizal and saprotrophic fungi in the soil, with saprotrophic species dominant in superficial layers and mycorrhizal fungi dominant in deeper layers has been reported previously (Lindahl et al. 2007). This contradicts our hypothesis that litter decomposition is primed by the C supply via EM fungi. There are however significant differences between the forest studied by Lindahl et al. (2007) and our site, as the forest floor vegetation in their study included shrubs and mosses. Needle and moss litter therefore extended relatively deeply into the organic surface layers, whilst our much shallower organic surface horizon was characterized by relatively dense layers of T. heterophylla needles at increasing stages of decay. The fact that we frequently observed fine roots growing through litter bags in control plots (but never in girdled plots) illustrates that fine root growth was concentrated in the surface layers, and associated EM fungi are therefore also likely to be found at shallow depth at our site. Whilst the evidence of EM involvement in heterotrophic decomposition in our study is circumstantial, firmer evidence is critical in order to obtain a better understanding of soil C processes, and warrants closer investigation in future experiments. Ecosystem models aiming to predict the impact of altered plant productivity under changed environmental conditions require information about where this component of soil C allocation impacts on decomposition.
The 14C results of soil respired CO2 give a clear indication that the reduction in flux observed between surface collars in the two girdling treatments is indeed caused by the elimination of C with a contemporary 14C signature (105%modern). This is significant, as it corroborates the findings of the flux comparison methods (both between girdled and control treatments and between collar types; see Fig. 3) that show little or no remaining flux contributions from root reserves following the initial transitory period. The good agreement in flux partitioning also with the deep collar method indicates that the collar insertion to 27 cm has been effective in terminating plant C transfer into the 1-μm mesh collars. An incomplete separation with a remainder of rhizospheric C entering these collars from depth would have resulted in an over-estimation of the RH/RS ratio (Subke, Inglima & Cotrufo 2006), but Fig. 3 indicates a slightly lower estimate (which does not however differ significantly from the estimates calculated by the other two methods).
A discrepancy between girdling estimates and 14C estimates for the partitioning could potentially have indicated a difference in the amount of SOM being decomposed in control plots compared with girdling plots. Whilst the size of the error bar is prohibitive of statistically robust evidence of priming by this technique, the indications are that the flux reduction following girdling can be explained well by the elimination of recent assimilates, with no additional source of older C in control plots.
Our results indicate a tight coupling between the aboveground assimilation by trees and belowground processes. However, the role of EM fungi as recipients of a significant amount of C allocated belowground by plants and their role in decomposition of soil organic matter is as yet only poorly documented. In order to parameterize models capable of capturing the response of soil C dynamics to changes in plant productivity and consequently altered belowground allocation may require an explicit partitioning of plant C allocation to root respiration per se and to other rhizospheric organisms. To date, only a small number of studies have attempted to partition C allocation into these fractions, and more research is needed to enable a better understanding of the conditions and constraints for the range of observed allocation patterns. The incorporation of such a more generalized allocation scheme into ecosystem models is likely to be a prerequisite to modelling of long-term C storage in the context of the rhizosphere priming response under altered environmental conditions.
Another significant conclusion of our work is that autotrophic soil activity in coniferous temperate forests persists throughout winter. Our data show that autotrophic flux components of RS, as well as the enhancement of litter decomposition by rhizosphere activity, occur in the coldest part of the year at least at our study site. Whilst activity is clearly reduced at low temperatures, measurements of autotrophic physiological activity that consider measurements during the growing season only are likely to be biased.
We would like to thank the Forestry Commission for England and Wales, and particularly Nick Short for the permission to conduct this girdling experiment. H. W. Vallack, J. E. Stockdale, C. S. Moody, R. D. Holden, L. Patterson and J. Stafford are acknowledged for their help during tree girdling, as is Sylvia Toet for the loan of gas analyser and chamber systems. This research was supported by the UK Natural Environment Research Council (grant ref. NE/E004512/1). Two anonymous referees are acknowledged for helpful comments on an earlier version of the manuscript.