How do harvesting methods applied in continuous-cover forestry and rotation forest management impact soil carbon storage and degradability in boreal Scots pine forests?

Forest management affects soil carbon (C) storage through forest composition, microclimate and litter inputs. How two major forest management systems, continuous-cover forestry (CCF) and clear-cut-based rotation forest management (RFM), differ in their impact on soil C in boreal forests is still poorly understood, however. We compared their effects on soil organic carbon (SOC) storage and quality in boreal Scots pine (Pinus sylvestris L.) dominated forests in eastern Finland. We tested the hypotheses that (1) colder microclimates and continuous litter inputs will lead to higher SOC stocks in CCF plots than in clear-cuts and (2) the more labile litter in clearcuts with varying ground vegetation will enhance SOC decomposition rates. We sampled uncut mature forests, clear-cuts, retention-cuts and gap-cuts, in which we analysed SOC concentrations and calculated the stocks. We measured stand characteristics such as diameter-at-breast height, basal area, dominant tree height, and understorey species coverage of the various treatments and modelled the aboveand belowground litter inputs based on these parameters. We used laboratory incubation and sequential fractionation of SOC to assess its degradability under standardized conditions. To estimate the decomposition rate in the various environments we incubated cellulose bags in situ. We assessed the impact of microclimate on SOC decomposition, using data from soil-temperature and soil-moisture field measurements. We quantified the microbial biomass C pool, using chloroform fumigation extraction to gain insight on the impact of forest management practice on soil microbes. The SOC concentrations and SOC stocks did not differ significantly between the treatments, despite the presence of a warmer microclimate and lower litter inputs in the clear-cut plots. However, we found differences in the quality of the SOC. Soils in clear-cut sites showed lower proportions of labile SOC compounds than did the other treatments. As hypothesized, the decomposition rates were elevated in clear-cuts, but were equally as high within the canopy gaps on gap-cut stands. Our work highlights that forest management affects the quality, degradability, long-term accumulation and storage of SOC. We conclude that the accumulation of labile compounds in uncut forests and retention-cuts, combined with the decreased decomposition rates, indicate a higher potential for future C accumulation in the soil than in clear-cuts.


Introduction:
Boreal forests play a significant role in climate change mitigation due to their ability to bind and store carbon (C). Boreal forests constitute a sink for about 0.5 Pg C annually and account for 32% of the global forest C stock (Pan et al., 2011). Approximately 60% of the C stored in boreal forests is located in the soil, mainly as soil organic matter (SOM) (Pan et al., 2011). Characteristically, boreal forest soils feature accumulations of partly decomposed organic material occurring as an organic layer (Olayer) on top of the forest floor. Globally, SOM forms a larger C pool than the atmosphere and global vegetation combined (Lehmann and Kleber, 2015). Climate change affects soil C storage, especially in boreal forests at high latitudes, where rising temperatures may turn the soils into sources of C (Crowther et al., 2016;Gauthier et al., 2015). Hence, it is important to understand the processes driving C storage in boreal forests and to adopt forest management strategies that protect soil C storage. While it is relatively easy to quantify and compare the aboveground biomass and productivity of different management options, the high spatial variability and laborious measurement methods make it more difficult to assess the effects of different management practices on the soil organic carbon (SOC) stocks as the expected change is comparably small against a large and variable background (Schrumpf et al., 2011).
Forest management affects C storage in the soil and aboveground by altering stand structure, microclimate and species composition (Jandl et al., 2007). Two major forest management systems, rotation forest management (RFM) and continuous cover forestry (CCF) likely differ in their impact on soil C storage. The first (RFM) is characterized by repeating cycles of a growth period followed by a final timber harvest as clear-cutting. Before the clear-cut, the stand is tended with several thinnings. A regeneration phase follows the harvest, often including the introduction of artificial regeneration and/or mechanical site preparation. The use of RFM leads to even-aged forest stands predominantly consisting of single tree species (Gustafsson et al., 2020;Pukkala et al., 2012b). Contrastingly, CCF is defined by maintaining a permanent forest cover over substantial parts of a given stand and harvesting of selected single trees or tree groups (Gustafsson et al., 2020). In CCF, selective cuts are the means used for harvesting timber as well as for enabling secondary diameter growth of the remaining trees and opening up the stand to initiate natural regeneration. Cuts can be conducted as single-tree selection cuts, gap-cuts or retention-cuts (amongst others). In retention forestry, the canopy tree cover is continuously maintained in substantial parts of the original stand, with retention trees spread quite evenly or as groups to maintain structural and functional variation (Gustafsson et al., 2012). Due to the continuity of the original stand structures, we will hereafter also address retention forestry under the term CCF. However, clear-cuts with low levels of retention (up to 5%) as commonly applied in Finland, are insufficient means for maintaining the continuity of structural variation in harvested stands (Kuuluvainen et al., 2019) and are not included under the term CCF. Typically, CCF leads to uneven-aged stand structures (Kuuluvainen and Aakala, 2011) and promotes multiple-species stands (Köhl and Baldauf, 2012). In Fennoscandia RFM is currently the most commonly used forest management system (Koivula et al., 2014;Kuuluvainen et al., 2019;Lundmark et al., 2016;Pukkala et al., 2012a) and is also widely applied in boreal forests worldwide (Burton et al., 2010). In Finland, alternative management approaches were discouraged politically via the Declaration Against Uneven-Aged Management (1948 to increase timber revenues after the Second World War (Pukkala et al., 2012a).
The immediate effect of forest harvesting on the soil C pool is the increased input of C as logging residues and dead roots (Strukelj et al., 2015). Input of fresh organic matter typically leads to accelerated decomposition of older SOC, the so-called "priming effect". At first, the freshly added litter increases microbial activity. Once the substrate is exhausted, microbes resort to the decomposition of older SOC (Kuzyakov, 2010). However, it is difficult to assess the long-term effects of harvesting on the soil C storage, due to the high spatial variation and slow rate of C accumulation (Jandl et al., 2007). After clear-cutting, the soil C storage declines rapidly for several years, due to accelerated decomposition and increased leaching of dissolved organic C. Following the initial C pulse of logging slash, the litter input is decreased for some years. Young regrowing forests are a source of C until accelerating tree growth and SOC accumulation turns them again into C sinks (Lindroth et al., 2018). Several decades are required after clear-cutting until the amount of SOC has recovered to the pre-cutting level (Covington, 1981;James and Harrison, 2016). In podzols, the most common soil type in boreal forests, the recovery in the mineral soil can take>100 years (James and Harrison, 2016). Unintended C losses from the soil during forest operations can be avoided by minimizing disturbances in the soil structure, such as soil preparation.
The application of CCF may increase C sequestration in the soil, because the harvesting effects of selective cuts on microclimate, litter input and soil disturbance are moderated, compared with clear-cutting (Jandl et al., 2007;Lindroth et al., 2018;Mayer et al., 2020). Continuous C inputs from living roots sustain the belowground ecosystem and may replenish the SOM and C stocks (Prescott and Grayston, 2023). Models of SOC storage in unevenly aged forests have produced conflicting results, some of them leading to a higher (Kellomäki et al., 2019) and others to a lower C storage (Shanin et al., 2016), compared with in RFM. Post-harvest tree density and length of harvesting intervals were positively correlated with SOC stocks (Shanin et al., 2016). Some studies have led to the assumption that the effects of harvesting methods applied in CCF on C dynamics are similar to the effects of thinnings (Jandl et al., 2007;Lindroth et al., 2018). A major knowledge gap regarding the effects of CCF on soil C storage is the current shortage of empirical studies (Prescott and Grayston, 2023). The few available come to different results (Nilsen and Strand, 2013;Pötzelsberger and Hasenauer, 2015;Strukelj et al., 2015). The present study will shed light on the effects of logging methods applied in CCF on SOC storage and quality.
We compare soil C storage and degradability under two major forest management regimes -CCF and RFMin boreal Scots pine (Pinus sylvestris L.) stands in eastern Finland. We hypothesized that (1) CCF with more constant production of tree root litter and colder microclimate under the canopy results in higher soil C stocks than RFM based on clearcutting; and (2) the more labile litter of the grass-and herb-rich vegetation associated with clear-cuts enhances the decomposition rates at clear-cut sites in comparison to the less decomposable litter of the dwarf shrub dominated vegetation characteristic of closed boreal forest stands.

Study area
Our research plots are part of a long-term research project aimed at determining the effects of uneven-aged forest management on forest structures and biota on a landscape level (Koivula et al., 2014). The plots are located within a 4-km radius in Ruunaa, Lieksa Municipality, eastern Finland (63 • 23.25Ń, 30 • 32.18É) at the southern edge of the Middle Boreal Zone (Ahti et al., 1968) at 140-150 m above sea level (Karttapaikka -Maanmittauslaitos [WWW Document], 2023) (Fig. 1). The mean annual temperature lies between + 2 • C and + 3 • C and annual precipitation ranges between 650 mm and 700 mm (Vuositilastot -Ilmatieteen laitos [WWW Document], 2023). The forests are situated in dry sandy soils, representing medium-and low-fertility sites, and feature albic podzols on glacial till, according to the World Reference Base for Soil Resources (WRB) classification (IUSS Working Group, 2022). The humus form is a mor. To minimize the effect of the sites on the SOC, the plots were intentionally chosen to have similar soil types, soil texture, elevation and aspect. For more detailed information, see Supplementary Material Table S1 and section Methods S1.
The plots in this study were harvested in winter 2010/11 to establish uneven-aged structures. Prior to logging, the stands were conventionally managed mature, even-aged Scots pine-dominated forests with occasionally admixed birch (Betula L. spp.) and Norway spruce trees (Picea abies (L.) H. Karst.). The treatments included in this study were gap-cuts with 15-20 m cleared gap radii (about 20-30% of the stand area harvested in multiple gaps), retention-cuts (20% of trees [volume] left after harvest, dispersed and grouped), clear-cuts (with 5% retention) and uncut even-aged reference forests ( Fig. 2 (A)-(D)). Uncut forest has been previously managed and thinned. The RFM is represented by the two stages of rotation: clear-cut and closed even-aged forest (uncut treatment), CCF by gap-cuts and retention-cuts. On gap-cuts the sampling points were positioned in the canopy gaps. Clear-cut sites were subjected to soil scarification and left to regenerate naturally.

Soil sampling and stand inventory
We sampled the 16 study sites (Table S1) in August 2020 in four replicate plots of each harvesting treatment. In each plot, we took 20 soil samples along two transects ( Fig. 1(D)). We sampled the O-layer with a stainless-steel soil auger with a sampling cylinder of 6-cm inner diameter. We sampled the mineral soil at 0-10-cm depths, using a smaller soil corer (diameter 1.9 cm). We pooled the samples for laboratory analyses to form one composite sample per each replicate plot and soil layer. In one of the four replicate plots, we analysed the 20 samples individually, i.e., without pooling, to examine the spatial variation within the plots. In each replicate plot we took four volumetric samples of the O-layer and mineral soil layer to determine the bulk density (BD) (<2 mm for mineral and < 4 mm for organic soil samples) and to collect living roots with a diameter > 2 mm (Poeplau et al., 2017). We took the samples, using a soil corer with a diameter of 6 cm. We estimated the stoniness, using Viroś (Viro, 1952) rod-testing method of 12 points in each replicate plot to a depth of 15 cm in the mineral soil ( Fig. 1 (D)). We measured the stand basal area (BA), the dominant tree height (DTH) and the diameterat-breast height (DBH) in each replicate plot. We also counted the number of saplings and seedlings smaller than 3 m within a rectangle of 6 × 12.5 m in each plot.
We measured the soil pH values, using a PHM210 standard pH meter from Radiometer Analytical (SA) (Villeurbanne, France) in a soil-water suspension and we determined the particle size distribution by laser diffraction.
We installed a temperature and soil-moisture logger in the ground at the centre of each plot. The loggers measured air temperature, soil surface temperature and soil temperature at a soil depth of about 8 cm. They recorded soil moisture down to a depth of approximately 14 cm (Wild et al., 2019). For detailed methods description, see Supplementary Material section Methods S1.

Soil carbon pools
We measured the C and nitrogen (N) concentrations (%) from the airdried mineral and organic soil samples by dry combustion, using a LECO CN828 analyser (LECO Corporation, St. Joseph, MI, USA). We calculated the C stocks (g C m − 2 ), considering the C concentration, the bulk density of the fine earth fraction, soil-layer thickness and stoniness. We quantified the C pool of the coarse tree roots that were picked from the samples. We estimated the tree fine-root (diameter ≤ 2 mm) biomass in kg ha − 1 , using a model by Lehtonen et al. (2016b). The results are presented in Table S2 in g C m -2 . In the understorey vegetation, we estimated the coverage of five plant functional groups (dwarf shrubs, grasses, herbs, bryophytes and lichens) during the field sampling and calculated the above-and below-ground biomass (in kg ha − 1 ) with mixed linear models with cover percentage as independent variable (Lehtonen et al., 2016a). The data can be found in the Supplementary Material (Table S3). The C pool was estimated as 50% of the dry biomass. For a more detailed method description see Supplementary Material (Methods S1).

Carbon input through litter
We modelled the annual total aboveground litterfall in g m − 2 using a multiple-linear-regression model from Starr et al. (2005). The model applied includes latitude and stand basal area as variables. We estimated the annual belowground litter production for coarse roots and fine roots based on turnover rates reported in Lehtonen et al. (2004) and Ding et al. (2021) respectively. For the understorey, we estimated the annual aboveground and belowground litter production based on turnover rates from Lehtonen et al. (2016a) (Supplementary Material, Methods S2).

Degradability and quality of soil organic carbon
Forest management affects the stability of SOC through altered microclimate and through altered litter quality. To distinguish between these two options, we examined different aspects of decomposition separately.
To determine the impact of the environment on the decomposition rate of organic material, we applied cellulose bags in the field in situ and left them to decompose for 12 weeks throughout the vegetation period. We estimated decomposition as the mass loss (%) during this period of time.
We determined microbial biomass carbon (MBC) as a driver of C turnover with chloroform (CHCl 3 ) fumigation extraction (modified after Blagodatskaya et al., 2014;Vance et al., 1987) measured as the flush of C released after fumigating the soil with chloroform. We calculated MBC as the difference between total organic carbon (TOC) of the nonfumigated subsamples and TOC of fumigated samples and present the results in relation to total soil C in mg g − 1 . We did not apply correction factors (Leckie et al., 2004).
To determine the degradability of SOC ex situ, we carried out an incubation laboratory experiment. We incubated about 40 g of organic soil and 8 g of mineral soil in 120-ml incubation bottles for 28 days at 14 • C, resembling midsummer outdoor conditions in Finland. We calculated the cumulative CO 2 production rate in mg CO 2 -C per g SOC as an indicator for the degradability of SOC (Rousk and Frey, 2015). A more detailed method description can be found in the Supplementary Materials (Methods S2).
To characterize the SOC quality of the O-layer, we used sequential chemical fractionation, first by removing nonpolar extractives, such as fats, oils and waxes, with dichloromethane. After this step, we dissolved polar extractives (soluble carbohydrates, ketones, pectins, tannins) with hot water, followed by an extraction of acid-soluble substances (cellulose and hemicellulose) with sulphuric acid (H 2 SO 4 ) hydrolysis (Hilasvuori et al., 2013;Wieder and Starr, 1998). We determined the amounts of extractable substances gravimetrically as the weight loss of each sample after the extraction, filtering through glass sinters and drying overnight at 105 • C. The remaining insoluble residue is defined as Klason lignin and is considered chemically recalcitrant. We determined the ash content of the samples by loss-on-ignition at 550 • C for 2 h and subtracted from the insoluble fraction.

Statistical analyses
We ran all analyses using R Statistical Software version 4.2.0 (R Core Team, 2019). We applied the Shapiro-Wilk test to check whether the data were normally distributed, except for the data from the temperature and moisture measurements. For these, we applied the Kolmogorov-Smirnov test, due to the high number of measurements. We used Levene's test for homogeneity of variance to confirm the equality of variances of the data for the different treatments. In case of normally distributed values or equal variances, we compared the treatments with parametric one-way analysis of variance (ANOVA) with Fisher's F test, followed post-hoc by unpaired t-test with Bonferroni correction. As a non-parametric test for different variances or non-normally distributed values we used Kruskal-Wallis ANOVA followed by Mann-Whitney U post-hoc test with Bonferroni correction.
To explore the associations between variables, we conducted correlation analyses. We calculated the correlations with Pearson's correlation coefficient when both variables were normally distributed or with Spearman's correlation coefficient when they were not. We analysed the temperature data with cross-tabulation and Pearson's chi-squared test. We cross-tabulated the air-temperature and soiltemperature recorded with the treatments, distinguishing between highair-temperature recordings (>25 • C) and low-air-temperature (<5 • C) as well as high-soil-temperature (>18 • C) and low-soil-temperature recordings (<8 • C). Two temperature and moisture sensors broke during the measuring period; therefore, we could only integrate three data points for the gap and uncut treatments.
We performed principal component analysis (PCA) to illustrate the associations between the field and laboratory measurements, SOC storage and decomposability, variables related to stand and soil characteristics and the treatments. The PCA was conducted separately for Olayer and mineral soil, using the factoextra package in R, after scaling the dataset with the prcomp function (Kassambara and Mundt, 2020).

Soil microclimate
Summer 2021 was the hottest recorded in Finland since 1937, with a long heatwave lasting from June till August (Finnish Meteorological Institute, 2021). The stand microclimatei.e. soil moisture, soil temperature, and air temperature -was significantly affected by the forest management practice during this season (Fig. S1, Table S4). The RFM treatments (uncut forest and clear-cuts) represented the two opposing patterns, while the CCF treatments, gap-cuts and retention-cuts, were intermediate. The mean soil temperatures during June, July and August were significantly higher in clear-cuts than in the other treatments and significantly lower in uncut forests ( Fig. S1 (a)). Soil-temperature extremes (high and low temperatures) were recorded most frequently in clear-cut plots and least frequently in uncut forests (see Supplementary Material, Results S1). The soil was driest in retention-cuts throughout the vegetation season and highest in the canopy gaps (Table S4).

Carbon input as litter
The amount of root biomass was significantly lower at the clear-cut sites in contrast to the retention or uncut forests (Table S2). This difference between treatments was clearest for the fine roots; their highest amount was found in the uncut treatment (Table S2).
The treatments affected the C entering the soil as aboveground and belowground litter (Table 1). Similar to the microclimatic parameters, the two RFM treatments formed the extreme ends, while both CCF treatments were intermediate. The annual litter input in g C m − 2 was highest in the above-and belowground measurements in uncut forests. The lowest annual litter input was in the clear-cuts. The two CCF treatments, retention-cuts and gap-cuts, did not differ in their annual above-and belowground litter inputs.
The differences in ground vegetation composition did not result in significant differences in the quantity of annual above-or belowground understorey litter input in the various treatments (Table 1). However, the uncut forests and retention forests showed marginally higher aboveground litter input from bryophytes, caused by a more extensive moss cover, yet this did not affect the total amount of litter input.

Soil carbon content and soil organic carbon stocks
The C concentrations did not differ between the treatments (Table S7). The O-layer tended to be thicker in uncut forests than in the clear-cuts, with respective mean thickness being 6.0 (±0.2) and 5.4 (±0.4) cm. However, the mean BD of the O-layer was significantly higher in the clear-cuts than in the uncut sites. This resulted in equal SOC stocks of the O-layers in all treatments (Table S7, Fig. 3A). The soil C content, BD and C stock in the mineral soil showed no differences between the treatments (Table S7, Fig. 3A).

Soil microbial carbon
The retention-cuts and clear-cuts showed the highest values for microbial C (Fig. 4). In the gap-cuts and uncut forests, the MBC was significantly lower. The MBC values measured for the mineral soil were lower, even partly below the detection limit of the method resulting in some negative values. There were no significant differences between the treatments.

In-situ decomposition rate of cellulose
The mass loss of cellulose in standardized litterbags was higher in the clear-cuts and in gap-cuts than in the retention-cuts and the uncut forests (Fig. 5). The mass loss varied more in the clear-cuts and gap-cuts than in the two other treatments.

Laboratory incubation indicating degradability of soil organic carbon
Based on the lab incubation, the O-layer of uncut forests released a higher amount of CO 2 g -1 C in soil than soils from the clear-cut sites, with a cumulative mean respiration of 14.33 (±1.2) mg CO 2 -C g − 1 SOC and 8.59 (±0.73) mg CO2-C g − 1 SOC, respectively (Fig. 6A). Samples from gap-cuts and retention-cuts were intermediate and not statistically   Table 1 Annual above-and belowground litter input from trees and understorey in the four different treatments in g C m − 2 . The numerical values show the mean (±standard error), while different letters indicate significant differences between the treatments tested with ANOVA followed by unpaired t-test with Bonferroni correction (p < 0.05). The letters in parentheses depict a marginal difference with p < 0.1.  different. The degradability of SOC, measured under standardized conditions in the laboratory, correlated negatively with the in-situ decomposition rate of cellulose (Pearsonś correlation coefficient = -0.507, df = 14, p-value < 0.04). The release of CO 2 from the mineral-soil samples during the incubation did not differ significantly between the treatments (Fig. 6B).

Sequential chemical fractionation describing soil organic carbon quality
In agreement with the contrasting results of incubation, cellulose decomposition rate, and microbial biomass that point towards the SOC quality as the limiting factor under standardized conditions, but not in the field, we found the lowest amount of labile fractions in clear-cuts (Fig. 7). Here, the more labile fractions, i.e., polar, nonpolar, and acidhydrolysable, together formed 51.7% (±0.7%) of SOC, while on uncut plots they amounted to a mean of 61.2% (±1.2%). Again, CCF treatments were intermediate. In retention plots, we recorded 56.3% (±1.0%) and in the gaps 54.6% (±3.1%) of the labile fractions. The proportions of the nonpolar extracts, the non-extractable fraction and the ash content did not differ significantly between the treatments.

Principal component analysis
For both layers, the organic and the mineral soil, the first principal component (PC1) separated the two RFM treatments, clear-cut and uncut, while the two CCF treatments were intermediate (Fig. 8). In the O-layer (Fig. 8A), PC1 was associated (p < 0.05) with the stand characteristics and variables associated with the degradability of SOC. It linked increasing labile C fractions, respiration, C concentration, and N concentration with increasing tree litter, DBH, BA, and DTH and with decreasing decomposition, soil temperature and understorey litter. The second principal component (PC2) was associated with internal variation within treatments, but also separated the two CCF treatments. PC2 correlated with pH, and percentage of silt and clay, and with C stock and the C/N ratio. The MBC and soil moisture that were negatively correlated, contributed to the third dimension, which explained an additional 13.7% of the variation in the O-layer data.
The PCA result for the mineral soil (Fig. 8B) resembled the results for the organic soil (Fig. 8A). Similarly, PC1 was formed by increasing C concentration and C stock with increasing tree litter, DBH, BA, and DTH, and with decreasing decomposition, soil temperature and amount of understorey litter. PC2 was significantly associated with percentage of silt and clay, pH, C concentration, decomposition and C stock. The correlation coefficients of the various variables with the principal components can be found in table S8 in the Supplementary Materials.
The SOC concentration and storage in the mineral soil were correlating, associated with both principal components and correlating with the amount of silt and clay (Fig. 8B). The C stocks in the O-layer, however, correlated negatively with the amount of silt and clay and were associated only with PC2. The decomposition correlated strongly with the soil temperature.

Discussion
Our empirical study compared the effects of CCF and RFM on soil C storage and soil C degradability in boreal Scots pine forests, using a 10year-long field experiment. We included two different forms of cutting applied in CCF: retention-cuts and gap-cuts, and two extreme stages of RFM: uncut forests and clear-cut sites. In all parameters measured, the two RFM stages appeared to represent the extreme ends, while the two CCF cuts fell between.
In contrast to our first hypothesis that CCF plots have higher soil C  stocks than clear-cuts, due to lower temperatures and continuous litter input, we did not find higher C stocks at continuously managed sites. However, in agreement with the second hypothesis we found lower in-situ decomposition rates at sites where forest cover caused a cooler microclimate and higher tree litter input, i.e. in retention-cuts and uncut plots.
Our results indicate that a period of 10 years is not long enough to show measurable changes in the SOC stock, although we could still detect changes in the processes behind the accumulation and decomposition of SOC.
The higher in-situ decomposition rate of cellulose in clear-cuts and in gap-cuts with warmer microclimate indicates more rapid consumption of labile C, supported by the lower proportion of labile compounds found at the clear-cut sites. Thus, the decomposition rates were more rapid at clear-cut sites, as expected, but we found no support for the second part of our second hypothesis, suggesting that there would be more labile litter from grass-dominated vegetation at clear-cut sites.
There were also differences between the two CCF treatments in the microclimate, MBC pool and the cellulose decomposition rates that we will discuss in further detail below.

Do the logging methods applied in continuous-cover forestry enhance the soil organic carbon storage compared with rotation forestry?
Typically, the soil in uncut forests is better sheltered by the insolation and shade of the canopy and shows smaller diurnal and seasonal variation in temperature (Kulmala et al., 2014(Kulmala et al., , 2009Kumpu et al., 2018;Palviainen et al., 2004). Accordingly, our clear-cut plots experienced the largest amplitude in soil temperature and were on average warmer throughout the vegetation period. Soil moisture was higher in the canopy gaps and clear-cuts during June and July; during August and September moisture was higher in the gap-cuts, compared with all the other treatments ( Figure S1). These findings are both in line with previous studies (Kumpu et al., 2018;Lindo and Visser, 2003).
The annual litter input of trees was drastically smaller in clear-cut plots both above and below ground supporting our hypothesis on more continuous litter inputs in CCF. The inputs of above-and belowground litter were also lower in gap cuts and retention-cuts than in uncut forests; however, this was not reflected in the SOC storage. Even though we found no mature trees at the clear-cut sites, which strongly affected the aboveground tree litter input and the modelled fine-root litter, we did not find significantly less coarse roots in the soil samples. This was surprising, since we expected that the coarse root biomass was proportional to the biomass of living trees (Mokany et al., 2006). The relatively high amount of coarse roots remaining in clear-cut plots can be explained by the slow decay of these roots after harvest. Therefore, the annual amounts of coarse root litter was comparable to the annual amounts in the other treatments, with a tendency towards fewer coarse roots at the clear-cut sites and more coarse roots in retention plots. We can assume that the turnover rate will increase drastically at some point on clear-cuts as the roots are no longer connected to living trees. The models we used to estimate above-and belowground litter input are derived from datasets from even-aged stands and hold thus a higher uncertainty when predicting litter input in more heterogeneous forests managed with CCF. We anticipated that a different species composition in the understorey would lead to a different litter quality. However, there were no statistically significant differences in the composition of the understorey litter, except for marginally higher amounts of bryophyte litter in the retention plots and uncut plots.
In contrast to our first hypothesis, the colder microclimate and continuous litter inputs observed in the gap-cut, retention-cut and uncut treatments (compared with clear-cut), did not lead to higher SOC stocks. Unlike in many studies that show a decline in SOC storage following clear-cutting, especially in the O-layer (reviewed in Mayer et al., 2020) and a slow (75-100 years, or more in podzols) recovery of the C-stock to the pre-cutting level (James and Harrison, 2016), the SOC storage in our study had not declined 10 years after logging in any of the logging  N), the C stock, microbial biomass C (MBC), respiration during lab incubation (Respiration), mass loss of litterbags (Decomposition), percentage of the labile C fractions (Labile fraction), soil temperature (Temperature) and soil moisture (Moisture). All the variables were averaged for each plot, except soil temperature, which was expressed as temperature sum. The ellipses around the treatments depict a confidence interval of 95%.
treatments compared with the uncut plots. However, in line with our results, Senez-Gagnon et al. (2018) could not find links between the time since clear-cutting and soil C content or C stock in a boreal chronosequence study. This may have resulted from harvesting residues and tree coarse roots remaining in the soil after clear-cutting that have not yet been fully decomposed. In the O-layer we observed slightly lower average C contents in the clear-cuts than in the other treatments; however, the average C stock in clear-cuts was slightly higher than the Cstock in the other treatments (Table S7). The diverging trends between C stocks and C content may have resulted from the higher BD of the Olayer in the clear-cut plots, which was considered when calculating the C stocks. The BD in clear-cuts usually increases through soil compaction during harvesting (Lindo and Visser, 2003;McNabb and Startsev, 2022). Furthermore, we found fewer roots in the clear-cut plots, which is probably also associated with higher BD of the soil. A lower BD facilitates root penetration, while high rooting intensity in return lowers BD through the gradual increase in SOC (Binkley and Fisher, 2013).
We could not analyse in this study effects that may have been present in the deeper mineral-soil layers. Our sampling depth was limited to 10 cm, due to high stoniness in the deeper soil and limited resources. Even though the C concentration in the deeper soil layers is usually much lower than in the upper layers, they also constitute a significant C storage due to the large volume that these layers occupy (James and Harrison, 2016).

How do the logging methods used in continuous-cover forestry affect the degradability and quality of soil organic carbon?
During the chemical fractionation of SOC, we observed smaller proportions of labile fractions in the O-layer at the clear-cut sites than in the other treatments. We did not find the differences in composition of the understorey litter we anticipated in our second hypothesis. This indicates that litter quality did not differ between treatments at this point, even though we did not directly measure it. This may indicate that the short-term changes in litter quality after clear-cutting do not influence the overall quality or decomposability of O-layer SOC, which is a product of longer-term balance between accumulation and decomposition of litter and transformation of SOC by microbes. Instead, the lower amounts of labile SOC compounds at the clear-cut sites could reflect the higher in-situ decomposition rate of labile SOC (as indicated by higher rates of cellulose decomposition), due to the higher levels of moisture and temperature that define the decomposition rate Liski et al., 2005;Prescott et al., 2000;Trofymow et al., 2002). Thus, the amount of labile SOC fractions may have been lower at the clear-cut sites because they were decomposed more efficiently. On the other hand, the cooler conditions in uncut forests may have limited the decomposition rates and allowed labile C to accumulate. Gap-cuts and retention-cuts show similar, but less strong effects than the uncut treatment. Furthermore there are no longer inputs of labile C from living roots through rhizodeposition in clear-cut sites, whereas they continue in uncut or partially cut stands (Table S2, Prescott and Grayston, 2023).
The in-situ decomposition rate of cellulose bags reflects the microbial activity in the field as influenced by ambient conditions. The decomposition rate of the cellulose bags was higher in the gaps and clear-cuts than in the covered treatments. Our results reflect those of other studies that observed higher soil respiration in gaps than under closed canopy (Kumpu et al., 2018;Roth, 2019). The mass loss through decomposition correlated negatively with the proportion of polar fractions of the SOC from the O-layer, i.e., the proportion of polar fractions was lower in gaps and clear-cuts and higher in uncut and retention plots (Fig. 7). The polar fraction contains tannins (among other substances) that are derived from decomposing roots and root exudates. Tannins decelerate microbial decomposition and stabilize the SOC obtained from root and leaf litter and fungal necromass in the soil (Adamczyk et al., 2019), which may illustrate why the decomposition of cellulose bags is slower in uncut and retention plots and indicate a potential for higher SOC stabilization in these treatments. The higher mass loss in clear-cuts and on gap-cuts can furthermore be explained with more favourable temperature and moisture conditions in these treatments during the field incubation period. The PCA results showed that soil temperature, soil moisture and the decomposition rate of cellulose were positively correlated at our study sites (Fig. 8). The retention-cuts and clear-cuts probably experienced desiccation during June and July (Fig. S1 B). Even though decomposition is usually limited by temperature in boreal forests, during summer 2020, soil moisture was the limiting factor, since volumetric moisture contents below about 25% (depending on the soil textural composition) probably limit microbial activity (Schjønning et al., 2003). Thus, the low moisture content of June and July, in addition to the cooler microclimate, may have contributed to the lower in-situ cellulose decomposition rates in the retention-cut and uncut control treatments.
We found more MBC on clear-cuts and retention-cuts in the O-layer. Microbial biomass should not only be regarded as a C pool, but also as an active driver of C and N turnover (Kuzyakov, 2010). However, the decomposition rate is not directly influenced by the amount of microbial biomass, but rather by the activity of these microbes. We can see this in our results, since the decomposition rate is higher in clear-cut plots and lower in retention plots, unrelated to the amount of microbial biomass. The PCA showed a negative correlation between MBC and soil moisture, i.e., lower moisture is associated with higher MBC (section 3.5). In contrast to our results, previous studies found that clear-cutting reduces microbial biomass, whereas partial cutting and thinning did not affect microbial biomass (Holden and Treseder, 2013). The MBC in clear-cuts and retention-cuts may have remained elevated 10 years post-harvest due to the greater input of harvesting slash in these treatments. However, we were not able to quantify the harvesting residues retrospectively. We could not see clear differences in the mineral soils, and the results were partly below the detection limit of the chloroform fumigation extraction method. This was due to the sandy, and poor mineral soils in dry pine forests that harbour only small amounts of microbial biomass.
The cumulative CO 2 evolution during the lab incubation furnished insight about the SOC degradability and can be used as a measure of labile C (Karhu et al., 2010). Interestingly, the soils in retention-cuts and uncut forests showed lower cellulose decomposition rates and higher degradability of SOC, which indicates that the labile C decomposition is limited by the cooler microclimate, as hypothesised, leading to higher accumulation of labile C (indicating SOC quality) in the retention-cuts and uncut forests. This labile C could be easily decomposed under optimal conditions in the laboratory, as shown by the higher CO 2 production from these soils in our incubation experiment. The degradability of SOC correlated negatively with the in-situ decomposition rate of cellulose, suggesting that the quality of SOC limits decomposition in the laboratory, while the environmental conditions become the limiting factors in the field. The CO 2 -production from the laboratory incubation correlated with the proportions of the chemical fractions, supporting the conclusion of more labile C accumulated under retention-cuts and uncut forests. Labile and recalcitrant C both contribute to the formation of a stable C pool (Adamczyk et al., 2020), hence, this could indicate a higher potential to accumulate soil C in the long-term. Soil microbes especially rely on labile SOC in their metabolism and transform it into more resistant compounds. This process is called the entombing effect of the microbial C pump and contributes strongly to the stabilization of C in the soil (Liang et al., 2017). Labile compounds are more sensitive to changes in the management in general, and also respond more quickly to changes in the stand microclimate, due to of their fast decomposition rates (Leifeld and Fuhrer, 2005). Thus we can see management effects first in the labile fractions, but these changes may indicate how the management will affect the more stable compounds in the longer term (Awale et al., 2017;Bongiorno et al., 2019). Hence, we cannot yet see a statistically significant increase in total soil C stocks in the plots with more labile compounds, but we can expect them in the future.

Do gap-cuts and retention-cuts differ regarding the soil organic carbon storage?
We found similar C content and SOC storage, similar amounts of litter inputs and similar shares of labile and recalcitrant fractions in gapcuts and retention-cuts. These two logging methods differed regarding the in-situ decomposition rate and the amounts of MBC, which were, respectively, higher and lower in gap-cuts. Furthermore, the microclimate differed between the two treatments, with moister soils in gap-cuts that experienced low temperature extremes more often, even though the mean soil temperature of the two treatments did not differ. Assuming that the decelerated in-situ decomposition rate will eventually lead to an increased long-term C storage, the potential for future C stabilization in the soil will be higher in the retention-cuts than in gap-cuts.

Outlook
The effects of CCF on C dynamics are often assumed to resemble the effects of a thinning (Jandl et al., 2007;Lindroth et al., 2018). However, gaps-cuts or retention-cuts open up the canopy more and thus affect microclimate, single-tree growth characteristics and SOC storage and quality differently than thinning from below. Furthermore, CCF may change the tree species composition, allowing more broadleaved species in the stands (Pukkala et al., 2012b). Different litter quality, different use of rooting space and different mycelia associated with the tree species will affect the soil C pool. Therefore, there is a need to compare CCF and RFM directly, regarding their C storage, and such direct comparisons are scarce. Our results show higher proportions of labile C in retention-cuts and uncut forests 10 years after the management conversion. However, a longer-term study of several rotations would be required to empirically determine the effects of CCF on the formation and accumulation of more recalcitrant SOC. In addition to longer-term field studies, we emphasize the need for modelling studies to investigate this effect of C transfer from labile to more stable pools between management practices. Forest growth models should account for the different structures in CCF and RFM and their effects on single-tree characteristics, such as live-crown ratio. More open stand structures lead to a higher live-crown ratio in boreal spruce forests (Bianchi et al., 2020;Kumpu et al., 2020), which would also increase root live biomass (Shinozaki et al., 1964b,a) of belowground litter input and thus impact the soil C storage. Since the canopy in the retention-cuts is sparse, they may affect the live-crown ratio differently than gap-cuts.

Conclusion
Ten years after the initial application of gap-cutting and retentioncutting to converge initially even-aged forests toward CCF, the SOC storage did not differ between treatments. Accumulation of SOC is defined by the balance of C input from above-and belowground litter and output rates i.e. decomposition. We found lower annual litter inputs in clear-cuts and a lower in-situ decomposition rate in retention-cuts and uncut reference forests. Thus, there is a potential for a build-up of greater SOC stocks in the future, since higher litter inputs from trees in continuously managed forests will continue, assuming that the decomposition rates will remain low. We already saw increased accumulation of labile SOC fractions in the uncut forests and retention-cuts, which could indicate a continuing increase in total SOC stock over time. The differences between retention-cuts and gap-cuts were small. Regarding the potential build-up of a stable SOC pool, retention cutting seems a more promising practice for boreal pine forests at less fertile sites.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
The data that support the findings of this study are openly available in Dryad publishing platform at https://doi.org/10.5061/dryad. pc866t1v2.