Ecosystem carbon dynamics differ between tundra shrub types in the western Canadian Arctic

Shrub expansion at high latitudes has been implicated in driving vegetation ‘greening’ trends and may partially offset CO2 emissions from warming soils. However, we do not yet know how Arctic shrub expansion will impact ecosystem carbon (C) cycling and this limits our ability to forecast changes in net C storage and resulting climate feedbacks. Here we quantify the allocation of photosynthate between different ecosystem components for two common deciduous Arctic shrubs, both of which are increasing in abundance in the study region; green alder (Alnus viridis (Chaix) DC.) and dwarf birch (Betula glandulosa Michx., B.). Using 13C isotopic labelling, we show that carbon use efficiency (i.e. the fraction of gross photosynthesis remaining after subtracting respiration) in peak growing season is similar between the two shrubs (56 ± 12% for A. viridis, 59 ± 6% for B. glandulosa), but that biomass production efficiency (plant C uptake allocated to biomass production, per unit gross photosynthesis) is 56 ± 14% for A. viridis, versus 31 ± 2% for B. glandulosa. A significantly greater proportion of recent photosynthate is allocated to woody biomass in A. viridis dominated plots (27 ± 5%), compared to plots dominated by B. glandulosa (4 ± 1%). Allocation of C to belowground pools also differs significantly; after 2.5 weeks we recovered 28 ± 6% of recent photosynthate in root-free soil under B. glandulosa, but under A. viridis we were unable to detect recent photosynthate in the soil. We provide the first evidence that the impact of shrub expansion on Arctic C cycling will be species dependant. Where Betula dominates, ~1/3 of recently photosynthesised C will be rapidly allocated belowground to soil and microbial pools. Where Alnus dominates, more recently fixed C will be allocated to woody biomass. We conclude that models driven by remotely-sensed aboveground canopy characteristics alone (i.e. greenness) will be unable to accurately represent the impact of vegetation change on Arctic C storage.


Introduction
Shrub expansion is occurring in many tundra ecosystems and has been linked to widespread patterns of vegetation 'greening' across the Arctic (Sturm et al 2001, Myers-Smith et al 2011, Forbes et al 2010, Jia et al 2009. Greater photosynthetic CO 2 uptake by more productive Arctic vegetation has the potential to compensate for expected carbon (C) losses from warming permafrost soils over the next decades (Schuur et al 2015); recent model-based estimates suggest that increased plant growth at high latitudes will offset permafrost emissions until about mid-century (Schaefer et al 2011, Koven et al 2011. Field-based observations, however, suggest a different pattern, with more highly productive forest and shrub vegetation associated with less total ecosystem C storage than neighbouring tundra (Hartley et al 2012, Parker et al 2015, Wilmking et al 2006. If these field observations reflect a widespread phenomenon, 'Arctic greening' may not offset permafrost C release to the extent currently predicted, and we risk underestimating future permafrost C cycleclimate feedbacks (Abbot et al 2016). The key issue is that enhanced photosynthetic activity in vegetation may also be accompanied by changes in the way C is allocated throughout the plant-soil system, which in turn can modify rates of C turnover, and thus the trajectory of total stocks. Understanding patterns of plant C allocation in tundra shrubs is therefore a prerequisite for understanding the impact of greening on Arctic C cycle-climate feedbacks.
Carbon partitioning in plants is critical in determining ecosystem C sink strength in a number of ways (Bradford and Crowther 2013). Firstly, the proportion of gross photosynthesis (gross primary productivity, GPP) used to produce new organic matter within the plant, versus that immediately re-respired, determines the total input of fixed C to the ecosystem per unit time. This total flux of fixed C is how we define net primary productivity (NPP) here (Roxburgh et al 2005), and thus the ratio of NPP:GPP is equivalent to ecosystem carbon-use-efficiency (CUE). Not all NPP is used for plant growth; this term includes C allocated to non-plant biomass pools such as volatile organic compounds (VOCs), root-exudates and root symbionts. In this study we assume that VOC emissions contribute only a small fraction of NPP (VOC fluxes were <0.01% of gross photosynthesis in Arctic birch forest at a similar latitude (Faubert et al 2012). Recent work suggests that CUE across biomes may be tightly-constrained at around 0.5 as a result of the close biochemical interdependence between respiration and photosynthesis, but that plant growth per unit GPP (the biomass-production-efficiency, BPE) varies substantially-largely in response to site fertility (Campioli et al 2015, Vicca et al 2012. Predicting plant growth from GPP within terrestrial ecosystem models has been identified as a major source of uncertainty in modelling the changing C balance of northern permafrost regions, partly because the difference between BPE and CUE is often ignored (Xia et al 2017).
Once C is fixed from the atmosphere, plant C allocation between biomass pools and the expected lifetime of C in those pools  determines the size of plant C stocks over time. Rates of C turnover can differ by orders of magnitude between plant tissues; the mean residence time (MRT) of C in wood might be decades or centuries, while the MRT of C in leaves or fine roots may only be days or weeks. The greater the amount of carbon, per unit of increasing plant productivity, that is allocated to long-lived biomass pools, the greater the expected long-term carbon sequestration potential of Arctic vegetation (Hobbie et al 1998).
The influence of plant C allocation on ecosystem carbon sink strength is not limited to productivity and patterns of biomass allocation. It is now wellestablished that belowground C allocation to non-plant pools (root exudates and root-associated symbionts) can directly moderate or 'prime' rates of soil organic matter (SOM) turnover in the rhizosphere (the soil immediately surrounding and influenced by roots) and thus impact soil C stocks (Kuzyakov et al 2000, Walker et al 2016. In Arctic systems, rhizosphere priming effects (RPEs; an increase in the rate of SOM decomposition in the presence of plants compared with the rate of decomposition in plant-free soil) associated with ectomycorrhizal (ECM) fungi have been implicated as an over-riding factor explaining smaller total terrestrial carbon stocks under deciduous shrub and tree vegetation compared to ericaceous heath tundra (Parker et al 2015, Hartley et al 2012. Plant partitioning of C to non-plant pools belowground, and thus the significance of RPEs, may therefore be critical in determining tundra C balance as shrubs expand. In many regions, tundra shrub expansion involves patch expansion, infilling and encroachment of tall deciduous shrubs into sedge or ericaceous dwarf shrub-dominated plant communities (Lantz et al 2013, Fraser et al 2014. Tall deciduous shrub vegetation (> 0.4 m) in the Arctic is dominated by three genera; Salix, Betula and Alnus (Myers-Smith et al 2011), all of which share a woody growth form and are able to form ECM associations. Alnus species are also able to gain N from specialized root nodules through symbiosis with N-fixing bacteria of the genus Frankia Ruess 2009b, Horton et al 2013). At our study site, Alnus viridis (Chaix) DC. and Betula glandulosa Michx. dominate the tall deciduous shrub vegetation, with Salix confined to riparian areas.
In this study we address the question: do ecosystem C dynamics differ between the common deciduous shrub types currently expanding in the NW Canadian Arctic? Specifically: (1) does ecosystem CUE and BPE differ between vegetation dominated by A. viridis versus vegetation dominated by B. glandulosa? (2) How does the allocation of recent photosynthate to aboveground and belowground biomass, as well as non-plant pools belowground, differ between these shrub types? Our expectation was that CUE would be similar between shrub types, but that BPE would be greater for Alnus because N inputs from symbiotic fixation would increase overall plant N availability, and thus reduce the need for non-plant-biomass C allocation (e.g. to ECMs) belowground. This study is the first to quantify and compare patterns of C allocation common deciduous shrub species expanding into the tundra biome and thus provides crucial insightand data-to support predictions of terrestrial Arctic C balance.

Materials and methods
We use 13 C isotopic pulse-chase labelling methods to quantify ecosystem CUE, BPE and C allocation patterns. Vegetation in 0.64 m 2 experimental plots was exposed to isotopically enriched 13 CO 2 for approx. 2 hrs, and the fate of the photosynthesised 13 C label was then monitored over a period of 18 days in plant tissues, respiration fluxes and root free soil. We also assess plant N availability by comparing total leaf N pools and understorey leaf 15 N values.

Site description
The experimental site is located within Siksik Creek catchment (68 • 44'54.5" N, 133 • 29'41.7" W), a small subsidiary of Trail Valley Creek, ∼55 km NNE of Inuvik in Canada's NW Territories. The site is underlain by continuous permafrost with soils composed of 0.05-0.5 m organic-rich cryosols overlaying 1 m thick Quaternary Pleistocene till (Teare 1998). The topography of the catchment and surrounding area is characterized by gentle rolling hills, with a mean elevation of ∼ 80 m above sea level. Siksik Creek is at the southern limit of Arctic bioclimatic subzone E (Walker et al 2005), just beyond the northern limit of the spruce treeline. Mean air temperatures were 16.7 • C for October 2013-April 2014, and 7.1 • C for May-September 2014 .
Vegetation within Siksik Creek catchment is composed of heath tundra on hilltops, with ericaceous shrubs, sedges, prostrate Salix species and a welldeveloped lichen layer. Sedge and moss-dominated tundra is present in wetter areas. Patches of deciduous shrubs are a significant component of the landscape, consisting of green alder (Alnus viridis (Chaix) DC.), dwarf birch (Betula glandulosa Michx., B.) and Salix species. A. viridis and B. glandulosa are present in distinct patches on hillslopes or in riparian zones, while the taller Salix shrub species are confined to water tracks and riparian areas. We located our experimental plots in patches of alder-and birchdominated vegetation on a gently-sloping south-west facing hillslope ca. 200 m from the Siksik Creek stream channel. The understorey vegetation was dominated by ericaceous Rhododendron tomentosum Harmaja and Vaccinium vitis-idaea L., with forbs (e.g. Petasites frigidus (L.) Fr.) and graminoids (including Carex spp.) also present.

Pulse-labelling
In order to trace the fate of photosynthetic carbon, we pulse-labelled replicate plots with CO 2 enriched in 13 C (> 99 atom%). We labelled four replicate 0.8 × 0.8 m plots in a patch of A. viridis and a nearby patch of B. glandulosa within ca. 100 m of each other. Individual plots were pulse-labelled over a two-hour period, and labelling was carried out over three consecutive days from 26-28 July 2013 following a block design to avoid bias from unavoidable variations in light and temperature conditions during this period. Further methodological details and environmental conditions during labelling (table S1) available at stacks.iop.org/ERL/13/084014/mmedia are given in supplementary material (appendix S2).
Tracing the fate of assimilated 13 C after labelling We traced the fate of the assimilated 13 C label over time by sampling leaves (n = 7 time points), stems and roots of the dominant plant species (n = 2 time points). We quantified 15 N on leaves and % N by mass on leaf and stem samples at a single time point (4 days after labelling). Eighteen days after labelling we removed soil cores (0.05 m diameter) to quantify the total 13 C remaining in total root biomass and in rootfree SOM (fine roots were removed by hand from a small subsample of soil). We quantified the 13 C content (by manual gas sampling) of ecosystem respiration (n = 3 time points) and in belowground respiration (n = 9 time points), in subplots from which understorey aboveground biomass had been removed. Appendix S2 contains further details on isotopic sampling methods for plants, soils and respiration fluxes. Natural abundance 13 C isotope ratios were also quantified for all pools and fluxes from unlabelled Betula and Alnus plots (n = 3).

C in soil and biomass pools
The total mass of 13 C (mg 13 C m −2 ) originating from the applied label ('pulse-derived') in soil and biomass pools is calculated as: where 13 C x is the mass of pulse-derived 13 C (mg 13 C m −2 ) in C pool x, C x is the total mass of C contained in pool x (g C m −2 ) and x is the 13 C content of pool x in units of atom% excess above natural abundance. Total carbon pools sizes were quantified by destructive sampling (details in appendix S1). For leaf and stem pools, the total pulse-derived 13 C pool for the whole community was calculated as the sum of pulse-derived 13 C for leaf or stem pools of each species. Understorey species for which 13 C content was measured (P. frigidus, R. tomentosum, V. vitis-idaea plus graminoids) accounted for, on average, 94% of understorey leaf biomass-for species that were not sampled we assumed a 13 C atom% value equal to the average value for the other understorey species.

C in respiration fluxes
We used the Keeling plot method to quantify the 13 C isotope ratio of belowground and ecosystem respiration ( figure S3). This method assumes that the y-intercept of a linear regression between chamber headspace 13 C and the inverse of CO 2 concentration provides an estimate of the isotopic composition of the respiration source (Pataki et al 2003). The pulse-derived 13 C respiration flux is then calculated as: where 13 C R is the respiratory 13 C flux (mg 13 C m −2 min −1 ), R is the total respiratory flux (mg C m −2 min −1 ) and resp is the 13 C content of the respiration source expressed as atom% excess above natural abundance. We calculated the 13 C 'return flux' in ecosystem and belowground respiration, based on the 13 C content of respired CO 2 and measured/modelled respiration flux rates (details on respiration flux modelling are given in the supplementary material appendix S4).
We then express the 13 C content of plant biomass, soil and respiration fluxes as a percentage of total 13 C assimilated over the labelling period ( 13 C uptake ; mg 13 C m −2 ), which we assume is equal to total 13 C leaf immediately after labelling (at t = 0). In reality, a small amount of 13 C label will move out of the leaves over the labelling period and the gross uptake of 13 C will be slightly underestimated. With a continuous two hour labelling period, there will be approximately one hour's worth of losses from leaves (either through respiration or translocation) on average over the labelling period. Based on the fitted values for 13 C retention in leaves over time (supplementary material appendix S3) we estimate 13 C loss from leaves during the labelling period to be ∼ 5% (figure S6). We quantify CUE over the study period using the total label recovered from plant and soil pools after 18 days, and BPE as the total label recovered in plant biomass. We aimed to capture the dynamics of metabolically active plant C-i.e. photosynthesised C which is used to support plant respiration or growth (Ostler et al 2016)-and previous labelling experiments suggest that respiratory return of 13 C becomes undetectable after approximately this period .

Statistical analysis
All analyses were carried out in R version 3.3.1 (R Core Team 2016).We use linear mixed effects models to compare root and SOM 13 C values. We use paired t-tests to assess differences in the total recovery of 13 C in respiration fluxes, soil and biomass pools, and ANOVA to compare understorey leaf N concentrations and 15 N values.

Carbon and nitrogen pools
Total aboveground biomass varied between 244-598 g C m −2 in the Alnus plots and 256-355 g C m −2 in the Betula plots. A. viridis stems contributed the largest single aboveground C pool, containing approximately double the amount of C found in B. glandulosa stems, but aboveground C in the leaves and stems of understorey species was similar between vegetation types ( figure 1(a) and (b)). The root C pool in surface soils (0-0.15 m depth) was also similar between the vegetation types (Alnus 480 ± 231 g m −2 , Betula 433 ± 203 g m −2 ), and there were no clear differences in root distributions with depth ( figure 1(c)). Total SOM-C within the organic horizon was 3640 ± 1400 g C m −2 for Alnus and 3080 ± 4091 g C m −2 for Betula ( figure  1(d)).

C dynamics in leaves and stems
The total amount of 13 C label assimilated into leaf tissue across all plots was 40-141 mg 13 C m −2 (figure S4). Initial enrichment was followed by a consistent decline in leaf 13 C ratios over time for all species (figure S5). At the canopy-scale there was an initial rapid decrease in the fraction of assimilated 13 C retained within the leaves, then after 3-4 days leaf 13 C contents declined more gradually. 13 C enrichment was lower in stems than in leaf material (figure S7). A. viridis stems tended to have higher levels of enrichment than B. glandulosa stems, and in three of the four plots A. viridis 13 C content increased slightly between t = 4 and t = 18-19 (we identified one data point as an outlier, see figure  S7); for B. glandulosa stem 13 C isotope ratios were similar between the two time-points (figure S7).

C enrichment in roots and soil
The 13 C isotope values of bulk fine root material from Alnus plots were enriched by ca. 1.7 ‰ compared to natural abundance, though this was not statistically significant ( figure S8(a), table S2). The 13 C isotope values of fine roots from the Betula plots however increased by ∼2.7 ‰, significant at = 0.1 ( figure S8(b), table S2). We could detect no incorporation of 13 C in root-free SOM in the Alnus plots, with the overall effect of the labelling on SOM 13 C being < 0.2 ‰ ( figure S8(c)). In contrast, the 13 C of SOM in the labelled Betula plots consistently increased by ca. 1.0 ‰ (table S2, figure S8(d)). We found only two Alnus root nodules in labelled soil cores, but levels of 13 C enrichment in these nodules were high at ca. 10 ‰ above natural abundance values (figure S8(e)).

C in respiratory fluxes
Belowground respiration rates (excluding understorey aboveground biomass) were 3.5 mol CO 2 m 2 s −1 on average in the Betula plots and 3.0 mol CO 2 m −2 s −1 in the Alnus plots; there were no clear difference in belowground respiration rates between vegetation types (figure S9). However, the percentage of assimilated 13 C recovered in belowground respiration was higher for the Betula plots (accounting for 26% of total label uptake) than for the Alnus plots (13% of total label uptake; t = 2.89, df = 3, p = 0.06; figure 3(a), figure S10). Measured ecosystem respiration rates varied between 1.2-4.8 mol m −2 s −1 for Alnus and 0.6-8.5 mol m −2 s −1 for Betula. 13 C ecosystem respiration fluxes however were consistently greater for Alnus than for Betula (figure S12). Over 18 days following labelling, the total ecosystem respiratory return of 13 C was significantly higher for the Alnus plots (39% of initial label uptake) than for the Betula plots (23% of initial label uptake; figure 3(b); paired ttest, t = 3.35, d.f. = 3, p = 0.04).

Partitioning of recent photosynthate in biomass after 18 days
Assimilated 13 C recovered from plant biomass pools (indicating BPE) over 18-19 days was 56 ± 14% for Alnus and 31 ± 2% for Betula, although this was not a statistically significant difference (two sample t-test, t = 1.7, d.f. = 3, p = 0.22). Both vegetation types retained a similar proportion of assimilated label in leaf and root pools ( figure 4(a)). The two vegetation types differed significantly in the proportion of 13 C retained in stems, and in root-free SOM. In the Betula plots 28 ± 6% of the assimilated label was recovered from the root-free SOM component, accounting for a greater proportion of the total 13 C recovered than any other pool after  18 days ( figure 4(b)). In the Alnus plots there was no detectable excess 13 C in the root-free soil (figure 4(a)) and instead stem biomass accounted for the largest fraction of recovered 13 C (27 ± 5%). Total recovery of assimilated 13 C in all plant and soil pools, or the CUE over 18 days, did not differ between vegetation types, 56 ± 12% for Alnus and 59 ± 6% for Betula (two sample t-test, t = 0.20, d.f. = 3, p= 0.85). The unrecovered component of assimilated label (i.e. assimilated 13 C not detected in soil, biomass or respiration) was on average 8% for Alnus and 16% for Betula ( figure 4(b)).

Discussion
Patterns of plant C allocation between ecosystem components will determine the impact on Arctic C Figure 4. The fate of assimilated 13 C in biomass and soils after 18 days expressed as a percentage of the total 13 C assimilated (a) 13 C retention in plant and soil pools plotted as individual data points-closed symbols are Alnus plots, open symbols are Betula plots, with different shaped symbols corresponding to four experimental blocks; the grey line indicates the median value. (b) 13 C retention in plant and soil pools, and 13 C recovery in ecosystem respiration as an average of total 13 C uptake. SOM = root-free soil organic matter (which includes fungal hyphae).
balance if tundra vegetation becomes more productive. Our expectation was that the presence of an N-fixing shrub would lead to greater biomass production per unit photosynthesis (BPE) compared to an ECM-shrub dominated community. While our data do suggest that BPE in Betula vegetation may have been lower than in Alnus, (31% vs. 56%) we did not measure a statistically significant difference. As we expected, CUE over the study period did not differ between Alnus and Betula vegetation and was slightly greater than 50%, consistent with previous work (van Oijen et al 2010).
There were, however, clear differences in the fate of C once fixed as NPP. Around twice the amount of 13 C label (as a proportion of the total photosynthesised) was allocated aboveground in Alnus, mainly as a result of allocation to stem biomass (the amount of label in leaves after 18 days was similar between vegetation types). A. viridis stems comprised the single largest aboveground biomass pool and contained about double the total amount of C in B. glandulosa stems, so it is perhaps unsurprising to find a greater absolute sink for recent photosynthate in stems in Alnus vegetation. However, the allocation of photosynthate to stems was greater than would be expected on the basis of differences in stem biomass, being four to five times as high in Alnus compared to Betula. This is reflected in the higher isotopic enrichment in A. viridis stem tissue (> 5 × 10 −3 atom% excess) compared to B. glandulosa (< 5 × 10 −3 atom% excess, figure S7) and suggests that Alnus has a higher photosynthate requirement per unit stem mass. The N concentration of A. viridis stem tissue was approximately double that of B. glandulosa, and significantly higher than any of the other woody species in these communities, which also suggests higher metabolic demand (Reich et al 2008). Higher allocation of photosynthate to stem biomass in Alnus plots, therefore likely reflects more rapid growth in A. viridis, a greater allocation of photosynthate to stem storage pools (e.g. as non-structural carbohydrate), or a combination of both. High rates of allocation to rapidly respiring stem tissue also explains why total 13 C in ecosystem respiration fluxes was higher for Alnus than for Betula, even though 13 C in the belowground respiration flux was lower in the Alnus plots.
A further striking difference between shrub types in C allocation patterns was the substantial fraction of recent photosynthate recovered from root-free soil under Betula. The root-free soil pool we sampled includes fungal hyphae, microbial biomass and SOM, and incorporates many possible forms of non-plant biomass C. Whilst we did not measure enrichment in ECM hyphae directly, it is reasonable to assume that the extensive mycorrhizal networks which form under Betula (Deslippe and Simard 2011) facilitate the transfer of C from roots into the rhizosphere, and that ECM enrichment contributes to the 13 C signal we observed. The lack of detectable 13 C in Alnus soil suggests that in these communities the amount of recent photosynthate allocated to rhizosphere pools is very much smaller; so, while Alnus does associate with ECM (Horton et al 2013), in this context it seems likely that Alnus fungal symbionts do not receive the same C subsidy as the fungal community associated with Betula.
We did not recover all of the photosynthesized 13 C label from either vegetation type, probably because some label was transported outside of the plots. The logistical difficulties of working at a remote site restricted the size of the plots we could label, and if we had trenched around the plot margin we would have severely damaged the plants. It is likely then that at least some 13 C was transported outside of the plot (or deeper within the soil) via roots, hyphae, or diffusion of dissolved organic carbon or CO 2 (methane production in these mid-hill slope positions is low ) and the importance of these pathways may have differed between plots. The 'unrecovered' component was greater in Betula, which might be expected given that the 13 C was distributed more widely through the bulk soil, probably via hyphal networks; B. nana for example has been shown to transfer around 10% of photosynthate between individual plants via hyphae (Deslippe and Simard 2011).
Isotopic labelling experiments such as this involve multiple potential sources of error and bias as a result of the many different kinds of measurement required to quantify the fate of 13 C across multiple fluxes and pools. We mention here some the most important limitations that should be considered. Firstly, our estimate of total gross 13 C uptake at t = 0 is likely an underestimate, because leaves will have lost carbon through respiration during the labelling period. We estimate this bias to be small (less than 5%) but if gross 13 C uptake is underestimated by 5%, our estimates of CUE/BPE will be overestimated by ∼ 3 percentage points (i.e. for Betula CUE would be 53%, rather than 56%). A further potential bias in our results arises because of the impact of vegetation removal within respiration collars on our estimates of belowground 13 C respiration fluxes. The removal of understorey vegetation likely reduces the 13 C return flux that would be measured were the vegetation intact. However, because understorey vegetation composition is similar between vegetation types, the impact of this is likely to be similar for both vegetation types. Vegetation removal is very unlikely to have significantly influenced measurements of ecosystem 13 C respiration fluxes, because the area affected (that inside the respiration collars) was < 2.5% of the total area of the labelled plot and the dominant shrub vegetation remained completely intact. The quantification of ecosystem 13 C fluxes was challenging, because we were only able to measure fluxes at three time points, and we rely on the model of Shaver et al (2013) to quantify the total return flux. This uncertainty does not impact our estimates of CUE/BPE however because these were based on tissue and soil data.
It was beyond the scope of this study to quantify N 2 fixation rates, but several lines of evidence suggest that the A. viridis plants at this site derive a substantial component of their N budget from fixation. An observed leaf 15 N value of −1.6‰ is typical of N 2 fixing species which derive N only from fixation (Boddey et al 2000), and is much closer to zero (i.e. to atmospheric values) than any other species in these communities. The presence of root nodules and high 13 C enrichment in those nodules after labelling also indicate that photosynthate was being allocated to support N 2 fixation. We only recovered two nodules in our soil cores, so our estimates are uncertain, but on the basis of the mass of the nodules collected (average mass of 4 mg dry weight), a rough estimate of total nodule biomass per plot in the top 0.15 m of soil would be ∼2.5 mg m −2 (this is lower than in boreal regions where nodule biomass has been recorded up to 44 g m −2 (Mitchell and Ruess 2009a)). The allocation of C to root nodules in our plots would then account for roughly 0.3% of recent photosynthesis.
We did not find clear evidence that the presence of N-fixing A. viridis plants had an impact on N availability at the community level; understorey species composition and abundance were very similar to the Betula understorey, and there were no differences in under-storey leaf N concentrations or 15 N values between vegetation types. The total amount of leaf N per unit ground area was also very similar between vegetation types. This suggests that the predominant mode of N acquisition (i.e. fixation vs. root/ectomycorrhizal uptake), rather than N availability per se, drives the differences in C allocation between these shrub types.
The continuing deepening of the Active layer and associated increase in nutrient availability (Schuur et al 2015) is projected to increase the expansion of deciduous shrubs in the circumpolar Arctic region. Recent modelling runs indicate a significant increase in the proportion of primary productivity associated with deciduous shrub species in tundra ecosystems over the 21st century (Mekonnen et al 2018). Our findings show that estimates of NPP associated with 'shrub' expansion are simplistic, and that it is critical to incorporate cascading impacts of productivity changes on belowground dynamics of C and N for specific shrub genera to predict ecosystem responses and feedbacks with the climate system.
In conclusion, we did not find evidence for a difference in CUE between the two vegetation types-thus, as shrubs encroach on tundra ecosystems, we would expect similar total inputs of fixed C to the ecosystem for every unit of additional GPP, regardless of shrub species. However, we did find clear differences in the fate of C photosynthesised as NPP. We show that not only does belowground allocation differ markedly between these shrub species, but also the distribution of that C throughout the soil profile. In Alnus dominated vegetation, recent photosynthate allocated belowground largely remains incorporated in plant root tissue, whereas in Betula it is distributed widely throughout the bulk soil. The amount and distribution of labile plant C inputs belowground therefore imply that the potential for priming of SOM (Hartley et al 2012) under Betula is much greater. Organic soils tended to be thinner in our Betula plots, suggesting faster turnover of SOM pools, and this is consistent with previous studies (Parker et al 2015, Hartley et al 2012. Our study implies that as Arctic vegetation changes, these common shrub species will differ significantly in their influence on ecosystem C dynamics, even if increases in aboveground productivity (=greening) are similar. laboratory analysis and data analysis. All authors contributed to writing the manuscript.