Separating autotrophic and heterotrophic soil CO2 effluxes in afforested peatlands

Peatlands are a significant global carbon (C) store, which can be compromised by drainage and afforestation. Quantifying the rate of C loss from peat soils under forestry is challenging, as soil CO2 efflux includes both CO2 produced from heterotrophic peat decomposition and CO2 produced by tree roots and associated fungal networks (autotrophic 15 respiration). We experimentally terminated autotrophic belowground respiration in replicated forest plots by cutting through all living tree roots (“trenching”), and measured soil surface CO2 flux, litter input, litter decay rate and soil temperature and moisture over two years. Annual peat decomposition (heterotrophic CO2 flux) was 115 ± 16 g C m y, representing c. 40% of total soil respiration. Decomposition of needle litter is accelerated in the presence of an active rhizosphere, indicating a priming effects by labile C inputs from roots. This suggests that our estimates of peat mineralization in our trenched plots are 20 conservative, and underestimate overall rates of peat C loss. Considering also input of litter from trees, our results indicate that the soils in these 30 year-old drained and afforested peatlands are a net sink for C, since substantially more C enters the soil as organic matter, than is decomposed heterotrophically. However, the C balance for these soils should be taken over the lifespan of the trees, in order to determine if the soils under these drained and afforested peatlands are a sustained sink of C, or become a net source over longer periods of forestry. 25

in decomposition and greenhouse gas (GHG) fluxes (Ojanen et al., 2013). However, the effects of drainage and afforestation on temperate peatlands that were previously dominated by smaller-statured vegetation (e.g. blanket bogs, moorland, heathland) are more uncertain (Sloan et al., 2018). While changes in hydrology and soil redox potential are anticipated to accelerate soil C loss and alter the composition of GHG emissions, there is a very little quantitative data on how this shift in land-use changes 35 soil C dynamics and GHG emissions in temperate peatlands (Hermans et al., 2018). Since northern hemisphere peatlands are estimated to store a third of global terrestrial carbon (Gorham, 2010), changes linked to drainage and afforestation of temperate peatlands could have significant impacts on regional C dynamics and GHG balances.
There is a broad consensus that peatland drainage accelerates the loss of endogenous peat C stocks, but the impacts of drainage and afforestation on total soil C stocks and soil respiration are less certain (Hargreaves et al., 2003;Mayer et al., 40 2020;Simola et al., 2012;Vanguelova et al., 2019;Zerva and M. Mencuccini, 2005). This is because our conceptual and numerical models of peat C dynamics are based on studies where drainage is assumed to be the principal human intervention, and there is no effective functional shift in plant community composition linked to land-use change (e.g. bogs drained to improve the productivity of the existing plant community to increase the quantity of forage for livestock). Afforestation of previously open peatlands in the British Isles (e.g. Mazzola et al., 2020) and Fennoscandia (e.g. Tolvanen et al., 2020) differ 45 from other types of drained peatlands because this land-use change involves a wholesale shift in the functional composition of the plant community (i.e. replacement of short-statured vegetation with trees), leading to potential interactions or synergistic effects arising from changes to both soil hydrology and plant community composition.
The comparatively high rates of net primary production and larger stature of the trees on drained and afforested peatlands can represent a significant net ecosystem C sink, and consequently represent a large source of detrital material to soil (Yamulki et 50 al., 2013). Thus, it is possible that this input of C from more highly productive trees could partially or wholly offset some of the losses of C derived from mineralization of the original peat. Moreover, changes to soil organic matter (SOM) chemistry due to a shift towards inputs of more recalcitrant and nutrient-poor coniferous litter could further enhance soil C storage in afforested peatlands in the British Isles and Fennoscandia. This is because the trees planted in these regions (e.g. Sitka spruce) produce tissues that are of poorer quality than the organic matter (OM) generated by the short-statured vegetation that they 55 have replaced (i.e. larger proportion of woody debris generated with higher C:N ratios and greater proportion of recalcitrant compounds like lignin) (Hermans et al., 2018). This may affect soil C stocks by inhibiting or slowing overall rates of decay due to the tree litter's chemical recalcitrance (Liski et al., 2002), but has not been studied in tree plantations on deep peat so far.
In order to close these knowledge gaps, we need to determine the C balance of drained and afforested peatland soils, 60 tracking C inputs from tissue turnover (e.g. litterfall), as well as C losses from SOM decay. However, separating the direct and indirect effects of trees on peat mineralization is methodologically challenging (Subke et al., 2006). Soil CO2 efflux includes both CO2 released from peat mineralization (heterotrophic CO2 flux) and CO2 produced from root-rhizosphere (i.e. autotrophic) respiration; experimental manipulation of autotrophic C supply to the rhizosphere allows a separation of these two main component fluxes, but introduces a number of potential artefacts (see Subke et al., 2006  including method-related assumptions and artefacts). Root trenching, where roots are severed throughout the depth of the rooting zone to create areas of forest floor with no recent C input from trees to roots or associated mycorrhizas, has been used across many forest sites, and provides useful insights into relative contributions of autotrophic and heterotrophic soil CO2 efflux and the respective temporal dynamics. This disruption of root exudation and continuous root turnover could influence peat decay through processes such as microbial priming effects (Kuzyakov, 2006;Subke et al., 2006), where microbial activity 70 (linked with an active rhizosphere) in the soil is stimulated by the addition of easily accessible C from roots (Kuzyakov, 2006).
Further experiments are required to evaluate the effects of these other plant-facilitated processes on peat decay, to better constrain estimates of heterotrophic activity in soil.
In this study, we experimentally determined the C budget of a drained and afforested peat soil. Carbon inputs were tracked by quantifying litterfall, while C outputs were determined using a trenching technique to partition root-rhizosphere 75 and heterotrophic respiration. In addition, we conducted a root decomposition study in rhizosphere and living root-free (i.e. trenched) soil, to constrain artefacts associated with the trenching and in order to determine if rhizosphere-linked processes (e.g. priming effects) facilitated OM decay. We hypothesised that 1) the soils under these forest plantations are a net C source, 2) soil CO2 efflux is dominated by autotrophic respiration and 3) interactions between C supply to the rhizosphere by trees result in higher decomposition rates of the litter due to rhizosphere priming effects. 80

Study site
The research took place in RSPB's Forsinard Flows National Nature Reserve in the north of Scotland (58° 22' N, 3° 53' W).
Four paired plots were established in the beginning of June 2014 in three separate forestry plantation blocks of identical age containing a mixture of Sitka Spruce (Picea sitchensis) and Lodgepole Pine (Pinus contorta). The plantations were drained 85 around 30 years old and very dense (about 5000 trees per ha), with no vascular plant understory, but sporadic patches of moss, predominantly feather moss, e.g. Hypnum jutlandicum, Hylocomium splendens and in some instances, Sphagnum fallax and S. capillifolium in furrows. The average diameter at breast height (DBH) for Sitka Spruce was 13.3 cm and for Lodgepole Pine 17.9 cm, with an average ratio per area of Sitka Spruce : Lodgepole Pine of 0.6. The average canopy cover was 76.3% (Smith and Hancock, 2016). Peat depths in these three forestry blocks varied between 30 and 260 cm, with depths at research plots 90 between 137 and 204 cm (Smith et al., 2014). The average annual precipitation in the research area between 1981-2010 was 970.5 mm with an average air temperature of 11.4°C measured at the Kinbrace weather station approximately 20 km from the plots (Location: 58º13'89''N, 3º55'1.

Experimental set up
Candidate locations for trenched and control areas in each plot were initially identified at random and soil surface respiration measured. Based on respiration results, two closely matched plots were selected, and randomly allocated a treatment (trenching or control). Paired plots were no more than 10 metres apart from each other.
The double ploughing of the peat at the time of afforestation created a regular micro topography with low-lying furrows (c.

Trenching 110
A 40 cm deep and 30 cm wide trench was cut to just below the main rooting depth of the trees, cutting through all roots present.
The trench was double-lined using polypropylene gardening cloth, and re-filled with peat soil in between the two layers of cloth to prevent in-growth of roots ( Figure 2). The dimensions of each trench plot were about 3.5 x 1.5 meters and included

CO2 measurements 120
Three pairs of PVC collars of 10 cm height with a diameter of 20 cm where installed to a depth of three cm within the three microforms ( Figure 1) of both trenched and control plots. CO2 measurements were taken using custom-built dark dynamic closed chamber with a height of five cm and a diameter of 20 cm, which were placed on the permanent collars for three minutes. Elasticated rubber material placed over the joint of chambers and collars for the duration of the measurement provided an air-tight seal. The chamber was connected to an EGM 4 Infrared Gas Analyser (PP-Systems, Amesbury, MA, USA), 125 recording CO2 concentrations every 4-5 seconds. Fluxes were calculated from increases in CO2 concentration within the chamber over 3 minutes. Measurements were carried out ten times between August 2014 and July 2016.

Litter
Six litter traps (0.07 m 2 each) were located close to each plot, and litter (falling needles and twigs) collected at each sampling visit. Dry weight of all litter was recorded as area-based averages for each plot separately. 130 https://doi.org/10.5194/bg-2021-126 Preprint. Discussion started: 12 May 2021 c Author(s) 2021. CC BY 4.0 License.
Litter was allowed to fall onto the soil surface within collars for the duration of the experiment. To be able to distinguish peat respiration from litter respiration, surface litter was removed manually from one (always the same) of the two paired collars in each microform before measuring respiration. The litter present in the "collar with litter" was weighed after a respiration measurement and then placed back in the collar. Litter adjacent to the collar was also collected and weighed in the field, then taken back to the lab, dried and weighed again to establish the wet to dry mass ratio of litter and calculate litter dry mass within 135 each collar.

Roots
Root biomass was determined from soil cores (0-20 cm deep and 6.5 cm diameter) taken from each microform in all plots, at the start (June 2014) and end (July 2016) of the experiment. Roots from each core were carefully separated and sorted into three root diameter classes: smaller than 2 mm, 2 to 5 mm, and greater than 5 mm. All roots and the root-free soil were dried 140 at 50°C for 7 days, and weighed to establish percentage roots per gram of soil. Root depth was found to be 25 cm when digging trenches.
To estimate root decomposition directly, roots were taken from soil collected in each plot, dried (50°C for 7 days) and separated in the same size classes as described previously. Between 0.36 and 0.69 g of dried root material (separately for each size class) were placed in polyester mesh bags (10 x 10 cm; mesh size of 0.5 mm) for field incubations. Bags were soaked 145 in water for 2 days prior to field placement, to mimic field conditions. Four replicate bags of each size class where buried at 5-10 cm depth in all three microforms in all plots four weeks after trenching. To account for any weight loss that may have occurred prior to field incubation, five bags of each size class where taken into the field and not buried, but taken back to the lab; the proportional mass loss of litter in these bags was used to correct the initial root mass of all other bags.
One bag per root class per microform was collected from all sites in November 2014, March 2015, July 2015 (except root class 150 >5 mm, since there was not enough material for four bags) and July 2016. After retrieval, bags were dried for seven days at 50°C, and root dry mass recorded.
Root decay was fitted to an exponential decay function: With Mt the remaining amount of root biomass after collection from the field, M0 the initial root biomass, t time and k the 155 decay constant. Data fits were performed separately for root size and microform.

Soil moisture and temperature
Between June 2014 and July 2016, soil moisture and soil temperature at 5 and 20 cm soil depth were recorded at 30-minute intervals in all three microforms in a nearby plot, using 12-bit smart temperature sensors, S-TMB-M002 (Onset Computer In addition to this, soil temperature (10 cm thermistor) and moisture (HH2 moisture meter, Delta-T Devices, Cambridge) were measured at 5 cm depth next to each collar during sampling. Air temperature was also measured one meter above the ground during sampling. 165

Statistical analyses and flux calculations
Data were analysed using R (RStudio Team, 2016). All CO2 data was log transformed to meet the criteria of normality, and a linear mixed-effect model (LMM) using the nlme package in R (Pinheiro et al., 2017) was used to predict CO2 fluxes based on environmental drivers. All numerical predictors were standardized to one standard deviation prior to statistical analyses, to allow relative effect sizes of predictors to be compared directly (Nakagawa and Schielzeth, 2010). Model selection was done 170 based on information theory (Burnham and Anderson, 2002). First a model of maximum complexity was built, with interactions between soil moisture, soil temperature, trench treatment and microform plus interactions between trench treatment, microform and litter treatment, with plot as a random effect. Then, all possible combinations of this model were identified using the 'dredge' function in the MuMIn package (Barton, 2017). Goodness of model fit was assessed with the small-sample size corrected Akaike's Information Criterion (AICc), which is calculated using the number of parameters and 175 either the maximum likelihood estimate for the model or the residual sum of squares. "Likelihood" here is a measure of the extent to which a sample provides support for particular values of a parameter in a parametric model. AICc values of different models can be compared and the model with the lowest AICc is selected as the 'best approximating model', hereafter called 'top model' (Burnham and Anderson, 2002). Any of the models with a delta AICc of less than 2 are considered to be as good as the best model (Richards, 2005). 'Dredge' also gives a weight to the models it produces, ranging between 0 and 1; with for 180 example a weight of 0.7 meaning that there is a 70% chance that that model is the best approximating model of the models considered. If the weight of the best model is low, it is not possible to say that that model really is the best model, meaning other models also have to be considered. In this study, the model with the best AICc and highest weight was used. Significance (p-values) for the mixed effect model were calculated using the package lmerTest (Kuznetsova et al., 2016).
Annual fluxes were calculated using the predict function over the mixed effects model from library lme4 in R (Bates 185 et al., 2015). The predictions were made over half-hourly measurements of soil moisture and soil temperature at 5 cm soil depth in all three microforms just outside the plots.
From these predictions, partitioned fluxes were calculated from the collars without litter as: 190 Where Fa is autotrophic CO2 flux, Fcontrol is the CO2 flux from the control plots, Ftrench is the CO2 flux from trenched plots, Fdead roots is the CO2 flux coming from the dead roots in the trenched plots created by the trenching technique and ԑ the associated error terms. The annual flux from litter was calculated from the difference in modelled annual fluxes between collars with and without litter. 195

Temporal trends in soil CO2 fluxes
Trenching initially led to an increase in soil respiration, followed by a significant reduction in soil CO2 flux. Soil respiration fluxes from both control and trenched plots showed a clear annual cycle, with highest fluxes in summer. After the initial 200 perturbation fluxes from trenched plots are significantly lower than fluxes from control plots (p<0.001) and this difference is greater in the summer (Figure 3). Soil CO2 fluxes were best explained with a combination of soil moisture, soil temperature, trenching treatment, microform and litter treatment, with an interaction between soil moisture and soil temperature, including 'plot' as a random effect.  (1.03 ± 0.05 µmol m -2 s -1 ) were significantly higher than fluxes from collars without litter (0.91 ± 0.05 µmol m -2 s -1 , p=0.008). 225

Role of environmental drivers in modulating CO2 flux
Observed soil CO2 efflux values correlated positively with higher soil temperatures, whilst soil moisture showed an inconsistent correlation with flux values; a significant (p=0.008) interaction between soil temperature and soil moisture means that at high temperatures CO2 flux decreases with increasing soil moisture, but at low temperatures flux increases when soil moisture increases (Figure 4). 230 There was no difference in soil temperature between trenched and control plots, but soil moisture was significantly higher in trenched plots than in control plots (p<0.001). To account for the artificially elevated soil moisture in trenched plots,

Partitioned fluxes
Soil CO2 efflux was partitioned into component fluxes for all measuring dates from August 2014 onwards to remove disturbance related artefacts observed immediately after trenching. Flux simulations based on the soil model details indicate 240 significantly lower autotrophic fluxes than heterotrophic fluxes (p=0.01, Figure 5). Across all microforms, heterotrophic fluxes represented 61% and autotrophic fluxes represented 39% of the total fluxes. From these predictions, annual sums for autotrophic and heterotrophic fluxes have been calculated, giving an average peat decomposition flux of 184 ± 21 g C m -2 y -1 .
Total soil respiration without needle litter is 301.3 ± 34.2 g C m -2 y -1 , with needle decomposition adding 41.2 ± 53.5 g C m -2 y -1 to annual fluxes, giving a total soil respiration including needle litter of 343 ± 35 g C m -2 y -1 .

Impact of litter and roots
Litter fall was 719 ± 71.3 grams of litter per m -2 y -1 , with no detectable difference between trenched and control plots. Assuming a C fraction of biomass of 50% (Mathews, 1993), this represents an input to the soil of 359 g C m -2 y -1 via litter fall.
CO2 flux from surface litter is calculated from the difference in the modelled annual fluxes between the collars with 255 and without litter, with the fluxes from collars with litter significantly higher than the CO2 flux from the collars without litter (p = 0.008, Table 1). C emitted by litter in the control plots appears to be higher than in trenched plots. Further, the average amount of litter in the collars (per m -2 ) of the trenched plots is higher than in the collars of the control plots, resulting in a 1.7 to 3.6 times higher CO2 flux per gram of litter from the control plots than from the trenched plots (Table 3). 260 Table 3 Mean amount of C emitted as CO2 by surface needle litter in 'litter collars', for both years (standard error in brackets).

Microform
Litter CO2  Root biomass per m 2 with an assumed rooting depth of 25 cm in August 2014 was 761 ± 324 g, 603 ± 110 g and 715 ± 257 g for < 2 mm roots, 2-5 mm roots and > 5 mm roots, respectively. For both the control and the trench plots, roots smaller than 2 mm declined in total biomass from the start of the experiment to the end of the experiment, but there was no significant 265 difference between the control and trenched plots at the start and end of the experiment. Despite an apparent trend towards higher root biomass in control plots in July 2016, these differences were not statistically significant. There was also no significant differences between the beginning and end of the experiment for root classes 2-5 mm and >5 mm ( Figure 6).

Root decomposition
Root mass in decomposition bags showed a consistent decline over the duration of the experiment (Figure 7). The calculated 275 decay constant (k) showed systematic differences by microform, with highest decay rates tending to occur for incubations at "original surface" (Table 4).

C flux from dead roots
The amount of C emitted from the decaying roots is calculated using the exponential decay function (1), with the biomass of 285 roots per m 2 to a depth of 20 cm in the trenched plots at the beginning of the experiment as M0. It is assumed that all biomass lost is emitted as CO2 and that 50% of roots is C (Mathews, 1993), as conservative assumptions, meaning that estimates are maximum possible CO2 flux from decaying roots in trenched plots. To correct for soil CO2 efflux generated in trenched plots as an artefact of creating dead root biomass, annual estimated CO2 emissions were corrected by subtracting root-decay based estimates from trenched plot CO2 emissions (Table 5). 290 https://doi.org/10.5194/bg-2021-126 Preprint. Discussion started: 12 May 2021 c Author(s) 2021. CC BY 4.0 License. Table 5 Corrected heterotrophic (peat only; Fh) and autotrophic (Fa) fluxes (standard error in brackets) in g C m -2 y -1 for dead root decay in trenched plots for both first (August 2014 -August 2015) and second year (August 2015 -August 2016) of the study. 295 Year 1 Year 2 (21) Heterotrophic fluxes represents approximately 24% and autotrophic fluxes 76% of the total soil fluxes in the original surface, 37% and 63% respectively in the plough throw, 51% and 49% respectively, in the furrow and 38% and 62% respectively averaged over all microforms.

Weighted average for soil CO2 flux in Flow Country forest plantations 300
In order to scale soil CO2 fluxes (excluding litter) from different microforms to the level of an entire forest stand, fluxes from individual microforms were weighted by their fractional area (Table 6). This results in a slight shift in proportion of heterotrophic and autotrophic CO2 flux sources to 40% and 60% respectively and a total area weighed soil CO2 flux of 289.4 ± 18.6 g C m -2 y -1 . 305 Table 6 Area-weighted heterotrophic (peat only) (Fh) and autotrophic (Fa) fluxes (standard error in brackets) in g C m -2 y -1 averaged over both years measured.

Microform
Fractional area

Discussion
Mass balance calculations indicate that the soils in these 30 year old drained and afforested peatlands are a net sink for C, as substantially more C enters the soil as organic matter, than is decomposed heterotrophically. The C balance of the soil under 310 these forest plantations is visualised in Figure 8, with the annual CO2 fluxes of the forest plantation based on the area-weighted fluxes. We found a C input of 359 g C m -2 y -1 via litter fall, similar to other Sitka Spruce forests of similar age to these forest plantations in the UK, which range from 273 to 573 g C m -2 y -1 (Morison et al. 2012). This is balanced by total soil efflux including litter-320 derived CO2 of 342.5 ± 34.7 g C m -2 y -1 , i.e. the amount of C entering the soil as surface litter alone falls within a similar range to C leaving as CO2. Missing in the C budget in Figure 8 are the input of C from root exudates that remain in the soil C pool (net rhizodeposition), potential losses through aquatic C export, and root growth and turnover. Gaffney, et al. (2020)  to the CO2 soil efflux. Root:shoot allocation in forest ecosystems is usually in the order of 1:3 (Laiho and Laine, 1997), so it is possible that belowground productivity could account for a significant C input only partially sampled by our approach (i.e. excluding large root stocks), adding to these plantations being a potential carbon sink.
The rate of peat decomposition in these drained and afforested peatlands is substantial, but overall soil CO2 efflux and ratio of heterotrophic/autotrophic respiration falls within boreal averages for forested ecosystems (Figure 9; Bond-330 Lamberty and Thomson, 2010). Average soil efflux corrected for microform area without litter (to determine the peat decomposition rate) over the two measurement years was 289 ± 19 g C m -2 y -1 of which 174 ± 10 g C m -2 y -1 is autotrophic and 115 ± 16 g C m -2 y -1 is heterotrophic.
Our total soil respiration (including litter) of 342.5 ± 34.7 g C m -2 y -1 is slightly higher than values of around 260 g C m -2 y -1 reported for a similar forest plantation in Ireland, a 39-year old drained Sitka spruce plantation on naturally treeless 335 blanket bog (Byrne and Farrell, 2005). Our peat oxidation rates of 115 ± 16 g C m -2 y -1 are also higher than found by Hargreaves et al. (2003), who found <100 g C m -2 y -1 in a 26 year old Sitka spruce plantation on drained deep (> 2 m) peat in Scotland.
However, they point out that their estimate has a big uncertainty, since it was calculated from the difference between total net C exchange and net tree gain, which both have a large uncertainty.
Comparing our results to a study in 18 to 44 year old Sitka spruce plantations on poorly drained Dystric Histosols in 340 Southern Ireland, our results of total soil respiration is much lower than their 972 g C m 2 yr -1 . This could be due the difference in the south of Finland over 4 years. They found that the ecosystem was a strong sink of CO2 in all studied years, with an 345 average NEE for the 4 years of -234 g C m -2 y -1 . By subtracting the carbon sink of the tree stand from NEE the authors show that the soil was a carbon sink of -60 g C m -2 y -1 . By modelling their forest soil respiration fluxes from chamber measurements they found the peat only respiration made up 53% of the total forest floor respiration flux, litter 22%, roots 16% and autotrophic respiration of above-ground vegetation 8%. Our results show a lower percentage of peat only respiration of 38% of the total soil respiration minus the litter flux. The higher percentage found by Minkkinen et al (2018)   19/04/2020. Our study is included in the red square. See Table 7 in Appendix for references to studies used to create this graph.
Trenching is likely to underestimate heterotrophic respiration and rates of peat mineralization since this approach does not account for rhizosphere effects such as priming (Walker et al., 2016). Therefore, it is likely that the estimated rate of peat oxidation from trenching is conservative. The observed difference of 1.7 to 3.6 times higher CO2 flux per gram of litter 360 from the control plots than from and trenched plots (Table 3) indicates that heterotrophic processes are reduced under trenching.
In presence of an active rhizosphere (control plots), decomposition of needle litter and/or soil organic matter (SOM) appears to be faster than when the rhizosphere is not active (trenched plots). This priming of organic matter decomposition is likely to be the result of microbial activity in the soil stimulated by the addition of easily accessible C from roots (Kuzyakov, 2006). Further, several studies have shown that mycorrhizal fungi are involved in soil priming (Kuyper, 2017;Paterson et al., 2016). 365 Therefore, in the control plots a slightly larger proportion of the total CO2 flux could be heterotrophic decomposition than the fluxes from the trenched plots suggest, which means there could be an underestimation of heterotrophic flux in our results, in line with results from literature (Subke et al., 2004(Subke et al., , 2011. To calculate the root biomass at the start of the experiment, one soil core per microform was taken and since trees were standing close to each other (1.5 to 2 m apart) this was assumed to be representative for the whole microform. It was not 370 possible to distinguish between living and dead roots in the soil cores, so living root biomass might have been overestimated.
The dead root emission correction made a big difference to the ratio of heterotrophic to autotrophic flux, going from 61% and 39% respectively over all microforms to 38% and 62% respectively (without area-weighting of fluxes), so a decrease in heterotrophic flux of 23%. This is in line with the corrections used in other studies; a comparison of corrections used in trenching studies indicates a range from 2% to 24%, with an average of 12% (Subke et al., 2006). This big difference in the 375 fraction heterotrophic : autotrophic flux suggests that even two years after trenching, decaying root biomass make significant contributions to the CO2 flux. That soils under a 30-year old conifer plantation on deep peat are a C sink is an interesting finding. However, in order to determine if these soils are a long-term sink or source of carbon, the balance between soil C input from roots and litter and C loss via oxidation should be quantified over the lifespan of the plantation. The average peat depth in these forest plots is 126 380 (±16) cm, with 0.47 kg C m -2 per centimetre depth (Cannell et al., 1993), which represents ca. 59.3 (±7.3) kg C m -2 stored in the peat under these plantations. The largest carbon losses most likely occurred in the initial planting phase of afforestation (Simola et al., 2012;Vanguelova et al., 2019;Zerva and M. Mencuccini, 2005), but fluxes have not been measured through this phase and cannot be quantified as new policy prevents planting on deep peat (Forestry Commission, 2016). In other parts of the world, peatland drainage is still actively happening and studies from these sites show very high rates of peat oxidation 385 during the first 5-10 years of conversion (e.g. McCalmont et al., 2021).

Code and data availability
The R code and datasets analysed in this paper are not publicly available. Requests to access the datasets should be directed to renee.kerkvliet-hermans@iucn.org.uk. 390

Author contribution
Funding acquisition and initial conceptualization of the whole PhD project was done by J-A Subke, R. Andersen, Y. Teh and N. Cowie. Further in-depth conceptualization of this particular work was done by R. Hermans and J-A Subke, with support from R. Andersen, Y Teh and N. Cowie. Investigation, methodology, project administration, data curation, formal analysis, software, visualization and writing was done by R. Hermans with supervision and guidance of J-A Subke, R. Andersen,Y. 395 Teh and N. Cowie.

Competing interests
The authors declare that they have no conflict of interest.

Acknowledgements
We are grateful for the PhD studentship that enabled this work, jointly funded by the University of Stirling and the Royal 400  Table 7 List of articles used in Figure 9 505

Authors
Year Title