Beyond the Growing Season: Variability of ¹³C-CO₂ Fluxes in Temperate Forests and Peatlands

Abstract

Winter processes are increasingly recognised as important components of ecosystem carbon cycling, yet ¹³C-CO₂ fluxes from temperate forests and peatlands remain poorly quantified. This study quantified cold-season ¹³C-CO₂ fluxes in a Scots pine forest and a temperate fen in western Poland using manual closed chambers coupled with a Picarro G2201-i isotope analyser. Measurements were conducted during the cold half of the year and related to soil temperature, air temperature and, at the forest site, soil moisture. Median ¹³C-CO₂ fluxes were about twice as high in the forest (607 µg·m⁻²·h⁻¹) as in the fen (290 µg·m⁻²·h⁻¹), indicating stronger winter respiratory activity in the mineral soil than in the water-saturated peat. In the forest, ¹³C-CO₂ fluxes showed a weak, non-significant tendency to increase with temperature, whereas in the fen they were significantly negatively correlated with soil temperature and tended to peak near 0 °C, pointing to an important role of zero-curtain and freeze–thaw conditions. These plot-scale measurements provide rare constraints on winter ¹³C-CO₂ losses from temperate forest–peatland mosaics and highlight the need to represent cold-season isotopic fluxes in carbon–climate assessments. From a biogeochemical perspective, the findings emphasize that ¹³C losses during the cold season can occur as transient, high-intensity ‘hot moments’. Such episodic fluxes should therefore be explicitly incorporated into winter carbon accounting and isotopically enabled carbon–climate feedback assessments to improve the fidelity of annual net ecosystem exchange projections.

1. Introduction

Global climate warming is reshaping the seasonality of carbon cycling in mid- and high-latitude ecosystems. While most carbon flux studies have focused on the growing season, there is increasing evidence that the non-growing season—including late autumn, winter and early spring—can contribute substantially to annual CO₂ losses from soils and ecosystems [1,2,3,4]. In many boreal and temperate systems, cold-season CO₂ emissions already account for 10%–40% of annual soil respiration and, in some cases, may offset a large fraction of growing-season carbon uptake [1,3,5]. As warming is often strongest outside the growing season, particularly in winter months, these fluxes are expected to increase and may shift some forest and peatland ecosystems from long-term carbon sinks towards net annual sources [2,6,7,8]. The role of non-growing season emissions is therefore crucial for understanding the global carbon balance.

Temperate forests and peatlands are key terrestrial carbon reservoirs. Globally, forest soils store more carbon than the atmosphere, much of it in slowly cycling organic matter fractions and deep soil horizons [9,10]. Peatlands, although they occupy only approximately 3% of the global land surface, store roughly 30% of the world’s soil organic carbon, making them among the most carbon-dense ecosystems on Earth [2,6,8]. In temperate regions, forest and peatland landscapes often co-occur under similar climatic conditions but differ strongly in hydrology, soil oxygen availability and microbial communities, which can lead to contrasting controls on winter CO₂ production and emissions [7,11,12]. Drained or partially drained peatlands may be especially vulnerable to enhanced winter decomposition under warming, because previously anoxic soil layers become aerated and remain unfrozen for more extended periods [7,8]. Winter warming has been shown to modify CO₂ flux dynamics substantially, stimulating heterotrophic respiration even at sub-zero temperatures [13,14]. Empirical evidence indicates that microorganisms in forest and peat soils retain the ability to hydrolyse complex biopolymers, such as cellulose, at temperatures of −5 °C and lower [14,15]. This activity is strongly moderated by snow cover, which acts as an insulating layer and decouples the soil from extremely low air temperatures [2,13,16]. Snow insulation can maintain a so-called zero-curtain layer, in which partially frozen soil remains biologically active for many weeks, emitting significant amounts of carbon [4].

A key mechanism controlling the variability of winter fluxes is the occurrence of freeze–thaw cycles, which can trigger rapid pulses of gas emissions [5,13]. These processes have both physical and biological components: physically, the release of previously trapped gases from soil pores, and biologically, the sudden availability of nutrients and substrates from microbial cells damaged or lysed by frost [5,13]. Studies using stable carbon isotopes (¹³C) allow differentiation of the sources of these emissions and tracking of the decomposition dynamics of fresh organic matter versus older soil carbon during such critical transition periods [14,16,17].

Over the past decade, an expanding body of research has demonstrated that non-growing-season carbon fluxes are neither negligible nor temporally uniform. Rather, they exhibit pronounced sensitivity to variations in soil temperature, snow cover, freeze–thaw dynamics, and hydrological conditions [1,3,18,19]. Despite these advances, most syntheses and ecosystem-scale modelling studies continue to rely predominantly on eddy-covariance measurements from boreal and Arctic environments, while high-resolution chamber observations in temperate forests and peatlands remain comparatively limited [2,3]. Furthermore, measurements of the isotopic composition of respired CO₂ during the cold season are rare, even though ¹³C signatures offer critical insights into substrate origin and turnover, the relative contributions of autotrophic and heterotrophic respiration, and the influence of recent photosynthetic inputs on ecosystem-level CO₂ efflux [10,20,21,22,23].

Natural-abundance δ¹³C measurements of soil-respired CO₂ in forest and peatland soils have revealed strong temporal variability, reflecting shifts in substrate use, diffusion processes and meteorological forcing [20,22,24]. Experimental work also indicates that snow cover and litter dynamics can modify both the magnitude of soil CO₂ efflux and its isotopic signature during winter and freeze–thaw transitions [20,23,25]. At the ecosystem scale, recent studies have shown that extreme precipitation events and hydrological anomalies can dampen or obscure the isotopic signal of recent photosynthates in ecosystem respiration, underscoring the complex coupling between above- and below-ground processes under a changing climate [3,10,26]. Nevertheless, there is still a lack of integrated understanding of how winter CO₂ and ¹³C-CO₂ fluxes vary across contrasting temperate ecosystems such as forests and peatlands, and how these fluxes respond to microclimatic variability during the cold half-year.

The non-growing season in temperate regions is also changing rapidly. Warmer winters, thinner or intermittent snowpacks, more frequent rain-on-snow and freeze–thaw events, and shifts in the timing and duration of soil freeze–thaw and zero-curtain periods have already been documented and are projected to intensify [3,18,19,27]. These changes are likely to alter soil thermal and moisture regimes, oxygen availability and substrate accessibility for decomposers, thereby modifying the magnitude and timing of winter CO₂ losses, as well as the partitioning of ¹³C between ecosystem carbon pools and the atmosphere [7,19,28]. However, winter processes are still poorly represented in most land-surface and Earth system models, which often assume minimal cold-season respiration or rely on simple temperature functions calibrated on growing-season data [2,3,29]. This modelling gap directly translates into uncertainty in projections of forest and peatland feedbacks to climate.

Here, these knowledge gaps were addressed by quantifying ¹³C-CO₂ fluxes during the cold half-year in a temperate Scots pine forest and a temperate fen in western Poland using manual closed chambers coupled with a cavity ring-down spectroscopic analyser (Picarro). The focus was on the period beyond the growing season, when photosynthetic activity is strongly reduced but soil and ecosystem respiration persist [27]. Specifically, this study (i) characterises the magnitude and variability of winter ¹³C-CO₂ fluxes in forest and peatland ecosystems; (ii) examines how these fluxes respond to microclimatic variability, with particular emphasis on soil and air temperature and, in the forest, soil moisture, including the occurrence of near-freezing and frozen conditions; and (iii) compares winter ¹³C-CO₂ fluxes between the forest and the peatland to infer ecosystem-level differences in carbon turnover during the cold half-year. By combining repeated plot-scale chamber measurements with in situ ¹³C-CO₂ mole fraction data, this work provides rare constraints on winter ¹³C losses from temperate forest–peatland mosaics and clarifies the role of near-freezing conditions in shaping peatland emissions. These results underscore the need to explicitly represent cold-season isotopic fluxes in assessments of ecosystem carbon balance and in isotopically enabled carbon–climate models.

2. Materials and Methods

2.1. Study Sites

The study was conducted in two temperate ecosystems located in Greater Poland Voivodeship, western Poland, approximately 20 km apart (Figure 1):

(1)

a forest stand in the Oborniki Forest District (52.7° N; 16.5° E), representing mesic mixed forest and mesic pine forest, and

(2)

the Rzecin transitional peatland, classified as a fen (52.4° N; 16.2° E).

Both sites experience a temperate climate with cold winters and warm summers and are exposed to similar regional meteorological conditions. Mean annual air temperature and precipitation are approximately 9.5 °C and 555 mm, respectively (own data from Rzecin station, 2018 to 2024). Soils at the forest site are classified as Dystric Brunic Sideralic Arenosols according to the World Reference Base 2014. The Rzecin peatland is a mesotrophic, minerotrophic fen formed by terrestrialisation, with a floating peat mat overlying peat deposits several metres thick and a water table that typically remains within about 0–30 cm below the peat surface during the growing season [11]. Vegetation at the forest site is dominated by Scots pine (Pinus sylvestris L.) with a closed canopy and continuous ground vegetation [30]. In contrast, the peatland is characterised by a mosaic of mosses dominated by Sphagnum teres (Schimp.) and vascular species typical of peatland ecosystems (e.g., Drosera rotundifolia L., Carex limosa L., Oxycoccus palustris Pers., etc.) [31]. The two ecosystems lie within the same regional climate but differ strongly in hydrology, soil aeration, and organic matter accumulation.

2.2. Chamber Measurements of ¹³C-CO₂ Fluxes

Soil and ecosystem ¹³C-CO₂ fluxes were measured during the cold period (October–March, 2018–2021 in the peatland and December 2023, January–March 2024 in the forest) using two manual dynamic closed-chamber configurations coupled to an isotope-ratio cavity ring-down spectrometer (Picarro G2201-I, Picarro Inc., Santa Clara, CA, USA). At the forest site, permanent chamber collars (25 cm diameter) were installed randomly before the start of measurements to minimise soil disturbance. In the peatland site, square collars measuring 75 × 75 cm were used to better accommodate the heterogeneous surface microtopography. All collars were inserted into the soil and peat and left in place throughout the study.

For each flux measurement in the forest, an opaque, dome-shaped chamber (approximately 10 L) was placed gas-tight on the collar for 30 min. Air from the chamber headspace was circulated in a closed loop through tubing to the analyser, and the ¹³C-CO₂ mole fraction was recorded every 1 s. Raw time series were visually inspected, and measurements affected by possible leaks or instrument instability were discarded. Due to the different vegetation structure and surface conditions at the peatland site, the chamber system was configured differently. The opaque chamber, constructed from white PVC, measured 76 × 76 × 50 cm and had a volume of approximately 306 litres. Despite the differences in chamber shape and volume, both chambers were equipped with appropriate fans (peatland chamber: two 7 × 7 cm, 2 W, Sunon, Kaohsiung, Taiwan; forest chamber: 0.07 W, Zephyr 40 MM, SilentiumPC, Sokołów, Poland) to ensure adequate air mixing. To prevent pressure changes during measurements, both chambers were fitted with a pressure equalisation system. A thermohygrometer (HygroVue5, Campbell Sci., Logan, UT, USA) was placed inside the peatland chamber to monitor internal conditions. Because of the large chamber volume and the relatively low flow rate of the analyser, the peatland system was equipped with an additional pump (flow rate 10 L·min⁻¹) to increase airflow between the analyser and the chamber. The analyser, which drew air for analysis, was connected to this loop. This configuration reduced the time required to detect changes in ¹³C-CO₂ concentration to approximately 3 min. The ¹³C-CO₂ mole fraction was recorded every 1 s.

The ¹³C-CO₂ flux (F, µmol·m⁻²·s⁻¹) for each chamber closure was estimated from the linear increase in ¹³C-CO₂ concentration over time [32]:

$$F={\displaystyle \frac{1}{A}}\cdot {\displaystyle \frac{dC}{dt}}\cdot {\displaystyle \frac{V}{R\cdot T}}$$

(1)

where A is chamber area, V chamber volume, dC/dt the slope of ¹³C-CO₂ concentration versus time, R the gas constant and T the mean chamber air temperature during the closure. Fluxes were first calculated in µmol·m⁻²·s⁻¹ and subsequently converted to µg·m⁻²·h⁻¹ for statistical analysis and presentation.

Different chamber geometries were used because of ecosystem-specific constraints: the larger peatland chamber better captured heterogeneous moss–vascular plant microtopography and reduced edge effects on the floating mat, whereas smaller collars are standard for mineral forest soils. Before carrying out forest measurements, comparative tests using both chamber types were conducted, demonstrating that flux estimates were consistent and unambiguous across configurations.

2.3. Microclimatic Measurements

At each site, air temperature (Tair), soil temperature (Ts) and soil moisture were measured continuously. At the forest site, TMS-4 dataloggers (Tomst s.r.o, Prague, Czechia) were installed near the previously installed soil collars. Ts was measured at 5 cm depth and Tair was measured near the ground surface and at 15 cm height. At the peatland, Tair was measured using a HygroVue5 thermohygrometer (Campbell Sci., Logan, UT, USA), while Ts was measured using a T107 thermistor (Campbell Sci.; UK) installed 5 cm beneath the peat surface. Both sensors were connected to a CR1000 datalogger (Campbell Sci., UK). At the peatland, measurements were recorded every 30 min; in the forest, recordings were made every 15 min.

2.4. Statistical Analysis

All statistical analyses were performed in R (v4.4.1, R Core Team, Vienna, Austria) [33]. Because fluxes were derived from chamber CO₂ time series, an initial quality filter was applied based on the coefficient of determination (R²) of the regression used to estimate each ¹³C-CO₂ flux. Measurements with poorly constrained regressions (R² < 0.80) were excluded from further analysis. This threshold represents a compromise between retaining a sufficient number of winter observations and ensuring robust flux estimates under low-flux, high-noise conditions typical of the cold season. As a sensitivity check, all key analyses were repeated using a more conservative cut-off (R² ≥ 0.90), which reduced the sample size but did not qualitatively change the results or the main conclusions.

For each ecosystem, descriptive statistics were calculated for ¹³C-CO₂ fluxes and environmental variables (Ts at 5 cm, Tair 15 cm above the surface and soil moisture). The number of observations, mean, standard deviation, median, interquartile range (25th–75th percentile) and range (minimum-maximum) are reported to characterise the overall magnitude and variability of winter ¹³C-CO₂ fluxes in the forest and peatland sites.

Differences in ¹³C-CO₂ fluxes between the forest and peatland were evaluated using the non-parametric Mann–Whitney U test (Wilcoxon rank-sum test), which does not assume normality and is suitable for unequal sample sizes. In addition to p-values, median fluxes for each ecosystem and the difference in medians with its 95% confidence interval are reported to emphasise the magnitude of ecosystem differences.

Relationships between ¹³C-CO₂ fluxes and environmental variables were examined separately for the forest and peatland. Ts and Tair were treated as primary drivers. Spearman rank correlations were used to quantify monotonic associations between ¹³C-CO₂ fluxes and Ts, and between ¹³C-CO₂ fluxes and Tair. Correlation coefficients (ρ) and associated p-values are reported and interpreted in an exploratory sense rather than as definitive evidence.

To explore the role of freezing conditions at the peatland site, peatland measurements were classified into two categories based on peat temperature: “frozen” (Ts ≤ 0 °C) and “unfrozen” (Ts > 0 °C). ¹³C-CO₂ fluxes were then compared across these categories using the Mann–Whitney U test, and flux distributions (medians, interquartile ranges) were summarised for each thermal state. This analysis was designed to assess whether near-freezing and frozen conditions are associated with systematically different winter ¹³C-CO₂ fluxes. At the forest site, a Spearman rank correlation between forest ¹³C-CO₂ fluxes and daily mean soil moisture was also computed.

Scatterplots were used throughout to visualise relationships between ¹³C-CO₂ fluxes and environmental drivers, with trend lines added to aid graphical interpretation. These lines are intended solely for exploratory visualisation and are not used for formal inference. Given the limited sample size, emphasis is placed on consistent patterns across methods (descriptive statistics, non-parametric tests, correlations and graphical exploration) rather than on individual p-values.

3. Results

3.1. Environmental Conditions During Measurement Periods

Chamber measurements of ¹³C-CO₂ fluxes were conducted during the cold half-year. After applying the quality filter (R² ≥ 0.80), the final dataset comprised 40 flux estimates. Soil temperature at 5 cm depth (Ts) during flux measurements ranged from 2.3 to 6.3 °C in the forest (median 3.7 °C) and from −0.37 to 7.54 °C in the peatland (median 1.0 °C), indicating more frequent near-freezing and slightly sub-zero conditions in the peatland. Near-surface air temperature showed a similar pattern, with generally lower values at the peatland site and several episodes close to or below 0 °C. Average meteorological conditions for the study areas on flux-measurement days are summarised in

Table 1. Mean values and standard deviations of meteorological conditions during the cold half-year on days with flux measurements at the peatland and forest sites.

Site	Season	Soil/Peat Temperature (°C)	Air Temperature (°C) *	Soil Moisture (%)	Soil Water Content (m3·m−3)
Peatland	01.10.2018–31.03.2019	3.70 ± 3.54	3.79 ± 5.72 (4.78)	-	-
	01.10.2019–31.03.2020	4.47 ± 3.22	4.49 ± 5.28 (5.81)	-	-
	01.10.2020–31.03.2021	3.53 ± 4.16	2.89 ± 6.05 (3.87)	-	-
Forest	01.10.2023–31.03.2024	7.23 ± 3.37	5.0 ± 6.16 (5.51)	8.0 ± 3.0	0.24 ± 0.08
	01.10.2024–31.03.2025	5.86 ± 2.91	3.68 ± 5.27 (4.32)	7.0 ± 3.0	0.21 ± 0.11

* numbers in parentheses represent mean Tair at the Poznań Ławica meteorological station, which is part of the Polish national meteorological network.

.

During field measurements, snow cover was present at the forest site in January 2024, whereas the peatland study area was snow-free during the peatland campaigns.

At the forest site, continuous soil-moisture monitoring over two cold seasons (2023/2024 and 2024/2025) showed daily mean volumetric water content (SWC) between 0.03 and 0.46 m³·m⁻³ (median 0.25 m³·m⁻³, interquartile range 0.18–0.29 m³·m⁻³). On the campaign days when forest ¹³C-CO₂ fluxes were measured, daily mean SWC ranged from 0.19 to 0.27 m³·m⁻³ (median 0.24 m³·m⁻³), i.e., within the central part of the seasonal distribution. This indicates that forest flux measurements were not biased towards unusually dry or wet conditions. In contrast, peat at 5 cm depth in the fen remained close to full water saturation throughout the cold season, consistent with a high and relatively stable water table.

Because the peatland chamber collars were installed on a floating peat mat that remained water-saturated throughout the year, soil moisture was effectively constant and was not analysed as a separate driver.

3.2. Magnitude and Variability of Winter ¹³C-CO₂ Fluxes

Both ecosystems acted as sources of ¹³C-CO₂ during the cold half-year. Fluxes derived from high-quality chamber closures (R² ≥ 0.80) showed substantial variability, reflecting differences among plots, dates and local microclimatic conditions.

In the forest, ¹³C-CO₂ fluxes ranged from 171 to 731 (µg·m⁻²·h⁻¹), with a mean of 537 µg·m⁻²·h⁻¹ and a median of 607 µg·m⁻²·h⁻¹ (interquartile range 418–693). In the peatland, fluxes ranged from 122 to 621 µg·m⁻²·h⁻¹, with a mean of 301 µg·m⁻²·h⁻¹ and a median of 290 µg·m⁻²·h⁻¹ (interquartile range 201–359). Although some peatland measurements overlapped with the lower part of the forest distribution, the overall flux level was clearly reduced relative to the forest site (Figure 2). Median winter ¹³C-CO₂ fluxes were approximately two times higher in the forest than in the peatland, with a larger spread of values in the forest and a more compact distribution in the peatland.

In addition, the mean of total CO₂ flux in the forest was approximately 2 µmol·m⁻²·s⁻¹, compared with 0.38 µmol·m⁻²·s⁻¹ in the peatland. For clarity of presentation, total CO₂ fluxes are reported in µmol m⁻² s⁻¹, whereas ¹³C-CO₂ fluxes are expressed in µg·m⁻²·h⁻¹ to reflect their smaller magnitudes and improve the readability of the figures and tables. In the forest, the lowest mean total flux during winter was observed under snow-covered conditions in January 2024 (mean 1.3 µmol·m⁻²·s⁻¹), whereas measurements one month later indicated the highest winter total flux (mean 2.8 µmol·m⁻²·s⁻¹).

3.3. Differences Between Forest and Peatland

The forest–peatland contrast in winter ¹³C-CO₂ fluxes was statistically distinct. A Mann–Whitney U test indicated significantly higher fluxes in the forest than in the peatland (U = 281, p ≈ 0.0009). Median fluxes were 606.9 µg·m⁻²·h⁻¹ in the forest and 289.8 µg·m⁻²·h⁻¹ in the peatland, with a bootstrap-estimated difference in medians of approximately 320 µg·m⁻²·h⁻¹ (95% confidence interval ≈ 117–443 µg·m⁻²·h⁻¹). Under the observed winter conditions, forest soils therefore released substantially more ¹³C-CO₂ than the peatland surface in terms of both central tendency and overall distribution, even though ranges overlapped at intermediate flux levels (Figure 2). This forest–peatland contrast in ¹³C-CO₂ fluxes provides a baseline for parallel analyses of total CO₂ fluxes using the same non-parametric framework.

3.4. Relationships Between ¹³C-CO₂ Fluxes and Temperature

Associations between ¹³C-CO₂ fluxes and temperature were ecosystem-specific (

Table 2. Spearman rank correlations (ρ, p) between ¹³C-CO₂ fluxes and soil temperature (Ts), air temperature (Tair) in the forest and peatland during the cold half-year.

	Forest		Peatland
	13C-CO2 vs. Ts	13C-CO2 vs. Tair	13C-CO2 vs. Ts	13C-CO2 vs. Tair
ρ	0.30	0.44	−0.43	−0.09
p	0.34	0.15	0.023	0.65

number of forest observations n = 12 and number of peatland observations n = 28.

). In the forest, ¹³C-CO₂ fluxes tended to increase with both soil and near-surface air temperature, but correlations were not statistically significant. These values indicate a tendency towards higher fluxes under warmer conditions. However, the limited number of observations and the relatively narrow winter temperature range at the forest site preclude strong inferences.

In the peatland, a different pattern emerged. A moderate, statistically significant negative correlation was observed between ¹³C-CO₂ fluxes and soil temperature at 5 cm depth, whereas the relationship with near-surface air temperature was weak and not significant (Table 2).

These results indicate that the highest ¹³C-CO₂ fluxes tended to occur at near-freezing or slightly sub-zero soil temperatures, rather than under the warmest winter conditions, whereas air temperature alone did not capture this behaviour. Scatterplots of ¹³C-CO₂ fluxes against Ts, with separate points for forest and peatland, showed a weak positive trend in the forest and a clear negative trend in the peatland, consistent with the correlation analysis (Figure 3).

3.5. Influence of Freezing Conditions in the Peatland

To further examine the role of freezing in the peatland, measurements were classified into “frozen” (Ts ≤ 0 °C) and “unfrozen” (Ts > 0 °C) categories. In total, four observations were classified as frozen and 24 as unfrozen. Median ¹³C-CO₂ fluxes under frozen conditions were approximately 385 µg·m⁻²·h⁻¹ (interquartile range ≈ 234–528), compared with approximately 290 µg·m⁻²·h⁻¹ (IQR ≈ 186–338) under unfrozen conditions. Thus, fluxes tended to be higher under frozen than under unfrozen soil temperatures, despite the expectation that very low temperatures might suppress microbial activity. However, this difference was not statistically significant (Mann–Whitney U test, U = 61, p ≈ 0.43), likely due to the small number of frozen observations.

The combination of a significant negative correlation between ¹³C-CO₂ fluxes and Ts and slightly elevated median fluxes under frozen conditions suggests that near-freezing and zero-curtain periods may favour enhanced winter ¹³C-CO₂ emissions in the peatland. These periods likely involve partially unfrozen pore water, changing gas-diffusion pathways and potential freeze–thaw pulses, which together can stimulate short-term CO₂ production and release despite low ambient temperatures.

3.6. Soil Moisture Content at the Forest Site

At the forest site, soil moisture during flux measurements was representative of typical winter conditions rather than of extremes. Daily mean SWC over the two cold seasons ranged from 0.03 to 0.46 m³·m⁻³, with a median of 0.25 m³·m⁻³ and an interquartile range of 0.18–0.29 m³·m⁻³. On the five days when forest ¹³C-CO₂ fluxes were measured, SWC values ranged from 0.19 to 0.27 m³·m⁻³ (median 0.24 m³·m⁻³), closely matching the seasonal median and lying within the interquartile range.

A Spearman correlation between forest ¹³C-CO₂ fluxes and daily mean SWC was weak and not significant (ρ ≈ 0.19, p ≈ 0.56, n = 12), indicating no clear monotonic relationship between winter ¹³C-CO₂ fluxes and soil moisture at the forest site under the conditions sampled. This likely reflects both the modest variability in soil moisture on the measurement days and the limited number of observations. The available data therefore suggest that, within the typical winter moisture range observed, temperature is a more important short-term driver of forest ¹³C-CO₂ fluxes than soil moisture.

In contrast, all peatland measurements were performed under water-saturated conditions on a floating peat mat, and no within-site gradient in soil moisture was available to test its effect on winter ¹³C-CO₂ fluxes.

4. Discussion

4.1. Winter CO₂ Emissions as a Key Component of Annual Carbon Balance

The measurements confirm that both temperate forests and peatlands remain active sources of ¹³C-CO₂ during the cold half-year. Median winter ¹³C-CO₂ fluxes were approximately twice as high in the forest as in the peatland, indicating that forest soils can sustain substantial respiratory activity even under low-temperature conditions. In this study, the presence of snow cover at the forest site was associated with only a slight, statistically non-significant reduction in soil CO₂ efflux compared to snow-free conditions. Previous work has shown that natural variations in snow cover can modify winter soil respiration but exert only a minor influence on annual soil CO₂ efflux [34], and that winter CO₂ emissions generally account for only a small fraction (on the order of a few percent) of the annual carbon budget [35]. In peatlands, recent work has likewise highlighted that non-growing-season carbon fluxes contribute substantially to annual CO₂ budgets and are projected to increase under future warming scenarios [2].

The present plot-scale chamber measurements complement ecosystem-scale eddy-covariance studies that integrate fluxes across heterogeneous snow and vegetation patterns [36,37], as well as recent regional assessments of winter soil CO₂ fluxes based on snowpack diffusion-gradient methods [3]. By explicitly quantifying ¹³C-CO₂ fluxes at the plot scale, the results provide a first-order constraint on the magnitude of winter ¹³C losses from temperate forest and peatland soils, which is rarely reported explicitly in non-growing-season studies that focus on total CO₂.

4.2. Forest–Peatland Contrast in Winter Fluxes

The consistently higher winter ¹³C-CO₂ fluxes in the forest relative to the peatland likely reflect differences in substrate availability, soil physical environment and microbial functioning between mineral and organic soils. Temperate forest mineral soils typically have better aeration and higher gas diffusivity than waterlogged peat, and often contain substantial pools of relatively labile litter- and root-derived carbon, which can sustain appreciable soil respiration throughout the non-growing season [2]. In contrast, fen peat is characterised by high water content and a shallow water table, conditions known to restrict oxygen availability and slow aerobic decomposition, thereby reducing CO₂ production relative to better-drained mineral soils [38,39].

Additionally, the partial overlap between forest and peatland flux ranges indicates that peatland microsites can occasionally emit ¹³C-CO₂ at rates comparable to the lower part of the forest distribution. Similar spatial heterogeneity in cold-season CO₂ efflux has been reported for northern bogs and fens, where hummocks, sedge-dominated patches and zones influenced by fluctuating water tables show elevated winter fluxes, especially during thaw events or under thin snow cover [38,40].

From a carbon-budget perspective, the observed forest–peatland contrast suggests that, under present climatic conditions, winter CO₂ and ¹³C-CO₂ losses from temperate mineral soils can be proportionally larger than from temperate fens, even though peatlands store much more carbon per unit area and are typically small annual CO₂ sinks [8,41]. At the same time, non-growing-season CO₂ fluxes from northern peatlands have been shown to contribute on the order of 10%–30% of annual respiration in some sites and to increase under warmer and less stable winter conditions [8,41], suggesting that the relative difference between forest and peatland winter emissions may diminish under future climate scenarios. Higher CO₂ emission rates in forest ecosystems are driven by efficient aerobic respiration and a constant supply of labile carbon from extensive root systems and litter, which stimulates microbial metabolism [42]. In contrast, high water saturation in peatlands induces anaerobic conditions, which, together with lower soil temperatures and the presence of biochemically resistant organic substrates, significantly limit the rate of carbon mineralization [16,42,43]. Importantly, because the forest and peatland campaigns were conducted in different years, the magnitude of the forest–peatland contrast should be considered indicative, since interannual variability in snow cover, freeze–thaw frequency and winter temperatures can modulate cold-season fluxes.

4.3. Temperature and Freeze–Thaw Controls on Winter ¹³C–CO₂ Emissions

The ecosystem-specific relationships between ¹³C-CO₂ fluxes and soil temperature are broadly consistent with previous work on winter soil respiration. In the forest, ¹³C-CO₂ fluxes tended to increase with soil and air temperature, although correlations were not statistically significant. Similar weak but positive temperature responses have been reported in other temperate forests, where winter soil respiration is frequently controlled by small thermal changes around 0–5 °C and by the insulating effects of snow [2]. Experimental studies have further shown that apparent Q₁₀ values can be higher in winter than in the growing season, implying a high sensitivity of CO₂ efflux to temperature under cold conditions [2,44]. The present results are consistent with this picture but do not allow robust quantification of Q₁₀ due to the limited number of observations and narrow winter temperature range.

In the peatland, a significant negative correlation between ¹³C-CO₂ fluxes and soil temperature, alongside slightly higher median fluxes under frozen than under unfrozen conditions, points to a different control regime. Similar patterns have been observed in northern bogs and other wetland systems, where near-freezing zero-curtain periods and freeze–thaw transitions can enhance CO₂ production by maintaining liquid water films, mobilising substrates and inducing microbial pulses during thaw [2,38,45]. In such water-saturated systems, the transition between frozen and unfrozen states, rather than absolute temperature, often emerges as a key driver of winter CO₂ fluxes. In addition, Zhao et al. (2026) [46] demonstrated that in terrestrial ecosystems with mean annual temperatures below 3.75 °C (characteristic of, among others, the Tibetan Plateau and cold boreal zones), annual soil respiration (Rs) exhibits a U-shaped response, with negative responses to warming at lower temperature ranges. The primary mechanism responsible for this phenomenon is moisture-mediated microbial suppression: rising temperature accelerates the thawing of snow cover and permafrost, rapidly increasing soil moisture content and constraining aerobic microbial activity. Given the small frozen subset in this study, the observed peatland pattern should be regarded as hypothesis-generating. One likely explanation is a zero-curtain behavior in saturated peat, where temperatures remain near 0 °C, and physical constraints on gas transport (ice barriers, changing diffusivity and episodic release) and/or short-term microbial pulses during thawing can result in elevated fluxes without a monotonic temperature response. The weak and non-significant relationships between peatland ¹³C-CO₂ fluxes and near-surface air temperature are also consistent with previous studies showing a decoupling between air temperature and subsurface processes in snow-dominated environments, where snow depth, ice formation and water-table position strongly mediate soil temperatures and gas transport [2,38,40]. Taken together, these findings support the emerging view that winter peatland CO₂ dynamics are more directly linked to the subsurface thermal and hydrological regime than to air temperature alone.

4.4. Interpretation of ¹³C-CO₂ Fluxes and Isotopic Processes

Most previous work on carbon isotopes in forest and peatland soils has focused on δ¹³C of respired CO₂, rather than on ¹³C-CO₂ fluxes themselves. Natural-abundance δ¹³C measurements of soil-respired CO₂ have been widely used to infer the contributions of autotrophic versus heterotrophic respiration, the coupling between recent photosynthesis and root respiration, and the depth origin of CO₂ in different ecosystems [20,21,22,24,47]. These studies consistently show that soil respiration is highly dynamic, responding to changes in recent photosynthate supply, soil moisture, snow and litter conditions on timescales from days to seasons. Forest management and associated soil compaction can also alter CO₂ dynamics and soil aeration and should be considered in the case of managed forests [48]. At broader spatial scales, remote-sensing studies have documented pronounced changes in the extent and configuration of land in different regions [49], underscoring that land-use and land-cover changes can further modify the structure and functioning of terrestrial carbon pools.

By explicitly quantifying ¹³C-CO₂ fluxes, the present study adds complementary information on the magnitude of ¹³C export from soils during the cold half-year. This is relevant for isotopic mass balances in ecosystem and atmospheric models, which are increasingly incorporating ¹³C to constrain carbon turnover and source contributions [10,50]. Differences in ¹³C-CO₂ fluxes between the forest and peatland likely reflect not only contrasting total CO₂ fluxes but also differences in the isotopic composition of respired CO₂ arising from distinct vegetation types, organic matter pools and hydrological regimes. For example, peat formed from Sphagnum and other wetland plants is typically more ¹³C-depleted and decomposes under more reducing conditions than tree-litter-dominated organic matter in mineral forest soils, which can imprint different δ¹³C signatures on respired CO₂ [2,44]. Analysed data quantify absolute ¹³C-CO₂ fluxes and report mean total CO₂ fluxes for context, but measurements of the δ¹³C of soil-respired CO₂ or the isotopic composition of potential source pools are not provided. Consequently, inferences about isotopic discrimination or source partitioning are limited and are discussed as potential mechanisms rather than direct evidence.

Process-based models and conceptual studies indicate that post-photosynthetic fractionation, microbial processing, diffusion and dissolution can all generate differences between the behaviour of ¹²C and ¹³C in soils and in emitted CO₂ [8,51]. Although δ¹³C of soil-respired CO₂ was not measured here, combining future winter ¹³C-CO₂ flux measurements with isotopic signatures of CO₂ and source pools would help to separate the roles of flux magnitude and isotopic composition in driving ecosystem-level differences in ¹³C loss.

4.5. Role of Soil Moisture, Snow and Microclimate

At the forest site, soil moisture during flux campaigns was within the median and interquartile range of the seasonal winter distribution and showed no clear monotonic relationship with ¹³C-CO₂ fluxes. The relatively narrow moisture range observed during winter likely limited the ability to detect potential non-linear or threshold responses of soil respiration to moisture. Similar results have been reported in temperate forests, where winter soil respiration is often more strongly controlled by soil temperature and snow insulation than by soil moisture, provided that soils are neither extremely dry nor deeply frozen [2,6,52]. Under such conditions, snow cover can maintain relatively warm and stable soil temperatures near 0 °C, supporting continuous microbial activity despite low air temperatures and masking moisture effects that may become apparent only across broader hydrological gradients or during extreme events.

In the peatland, soil profiles were essentially water-saturated throughout all flux measurements, reflecting the presence of a floating peat mat and a persistently high water table. Consequently, no within-site moisture gradient was available to assess moisture controls on CO₂ fluxes statistically. Nevertheless, numerous studies [53,54,55] have demonstrated that water-table depth and soil saturation exert strong controls on peatland CO₂ emissions by regulating oxygen availability, redox conditions, and gas diffusivity, with deeper water tables and warmer soils generally favouring higher CO₂ efflux in drained or partially drained sites [38,39]. In intact or restored peatlands, however, interactions between water-table dynamics, ice formation and snow cover determine whether winter conditions suppress or enhance CO₂ emissions [8,56,57], highlighting the importance of coupled moisture–temperature processes rather than independent controls.

Overall, the results of the study are consistent with earlier studies showing that the effects of snow on cold-season carbon exchange are highly complex and context-dependent. Natural variations in snow cover can alter soil temperatures and winter soil respiration, yet may have a limited impact on annual soil CO₂ efflux in some peatlands and cold-climate ecosystems [34,35,40,51]. In contrast, experimental snow removal or reduction has been shown to decrease winter CO₂ fluxes by exposing soils to colder conditions and increasing frost penetration [2]. In peatlands, anticipated changes in snow and ice regimes, combined with shifts in water-table position, are expected to modulate both winter CO₂ fluxes and the timing and duration of zero-curtain events [2,38]. The observed negative dependence of peatland ¹³C-CO₂ fluxes and soil temperature, along with the tendency for elevated fluxes near 0 °C, is therefore consistent with mechanistic expectations related to freeze–thaw dynamics, gas–release processes and transient microbial activity under near-isothermal winter soil conditions.

In the context of urban ecosystems, studies by Bezyk et al. (2021, 2023) [58,59] confirm that soil temperature and moisture are key factors controlling biogenic CO₂ fluxes even in areas of high anthropogenic pressure, providing a useful reference point for the processes observed in temperate forests. These authors showed that the δ¹³C isotopic signature of ecosystem respiration is clearly enriched outside the growing season (during the so-called heating period) compared to the plant-growth period, reflecting a shift from the dominance of photosynthesis to microbial metabolic processes and heterotrophic respiration. At the same time, their studies highlighted the role of moisture as a moderator of thermal sensitivity. The Q₁₀ coefficient for CO₂ emissions ranged between 1.68 and 1.79, with the highest respiration rates observed under moderate moisture conditions (20%–25%), while extreme water saturation inhibited gas diffusion and microbial activity. Those findings emphasise the universal nature of the hydrothermal control of emissions outside the growing season, indicating that the biogenic contribution to the total carbon budget—amounting to about 16.3% in urban conditions—is a dynamic process and highly responsive to winter warming.

4.6. Climatic Implications and Limitations

The finding that winter ¹³C-CO₂ fluxes are substantial in both ecosystems, and markedly higher in the forest than in the peatland, reinforces the importance of explicitly representing non-growing-season processes in carbon-cycle assessments. Across a range of ecosystems, winter and non-growing-season CO₂ efflux has been shown to contribute from a few per cent up to roughly 20%–30% of annual soil respiration or ecosystem respiration, with peatlands and other cold-climate systems often falling at the upper end of this range [2,38,44,60]. For northern peatlands, recent analyses indicate that non-growing-season carbon emissions could increase substantially under future warming, thereby weakening or even reversing their role as long-term carbon sinks [2,8].

Several limitations of the present study should be considered when interpreting these implications. First, the number of flux observations is modest, particularly for frozen peatland conditions, and measurements were restricted to selected days within the cold half-year. Moreover, forest and peatland flux campaigns were conducted in different years, so part of the observed forest–peatland contrast may reflect interannual variability in winter conditions in addition to inherent ecosystem differences. Second, chamber measurements integrate over small spatial footprints and short time intervals, and may not capture lateral and advective fluxes or the full heterogeneity seen by eddy covariance systems. Third, δ¹³C of soil-respired CO₂ and potential source pools were not determined, precluding direct isotopic partitioning of the contributing autotrophic and heterotrophic respiratory sources, or attribution of the observed ¹³C-CO₂ fluxes to specific substrate pools. Finally, key winter microclimate variables such as snow depth, frost depth, ice-layer formation and water-table dynamics were not quantified directly, although they can influence soil thermal regimes and gas-transport regimes, particularly in the peatland. Campaign-based chamber sampling may miss short-lived pulses during rapid thawing or other transient events, and plot-scale measurements represent a small footprint relative to ecosystem-scale heterogeneity. In particular, the frozen peatland sample size was very small (n = 4), so freeze–thaw/zero-curtain processes are necessarily tentative.

Despite these limitations, the study provides rare plot-scale estimates of winter ¹³C-CO₂ fluxes from a temperate forest and fen, documents clear ecosystem contrasts in flux magnitude and highlights the apparent importance of near-freezing conditions for winter peatland emissions. Future work that combines continuous CO₂ and ¹³C-CO₂ measurements with δ¹³C-based source partitioning, detailed microclimate observations and model integration would help to refine estimates of winter carbon losses and their sensitivity to ongoing climatic change.

5. Conclusions

• Both ecosystems remained active ¹³C-CO₂ sources during the cold half-year.

Winter chamber measurements demonstrated measurable ¹³C-CO₂ fluxes from both the temperate forest and the temperate fen, confirming that non-growing-season processes contribute non-negligibly to ecosystem carbon loss.

2.

Winter ¹³C-CO₂ fluxes were substantially higher in the studied forest than in the peatland.

Median ¹³C-CO₂ fluxes in the forest were roughly twice those in the peatland, indicating a stronger winter respiratory activity in mineral forest soils than in water-saturated peat, despite the much larger carbon stocks in peatlands. Whether similar contrasts hold across other temperate forest–peatland systems remains to be tested with coordinated measurements at a broader set of sites.

3.

Controls on winter ¹³C-CO₂ emissions differed between the forest and the peatland.

In the forest, ¹³C-CO₂ fluxes tended to increase with soil and air temperature, whereas in the peatland, a significant negative relationship with soil temperature and slightly higher fluxes under frozen conditions pointed to an important role of near-freezing, zero-curtain and freeze–thaw conditions for winter emissions. However, the frozen peatland subset was small (n = 4), so these freeze–thaw-related patterns should be considered exploratory.

4.

Soil moisture effects were ecosystem-specific and constrained by data availability.

At the forest site, flux measurements were made under typical winter soil-moisture conditions, and no clear monotonic relationship between ¹³C-CO₂ fluxes and soil moisture was detected. In the peatland, soils remained water-saturated throughout, preventing analysis of within-site moisture effects but highlighting the importance of persistent saturation for limiting aerobic decomposition.

5.

Plot-scale ¹³C-CO₂ flux measurements provide complementary constraints to ecosystem-scale approaches.

The chamber-based ¹³C-CO₂ fluxes complement eddy-covariance and snowpack studies by resolving small-scale heterogeneity and explicitly quantifying ¹³C losses in winter, thereby offering useful constraints for isotopically enabled ecosystem and atmospheric models.

6.

Future work should integrate isotopes, microclimate and modelling to refine winter carbon budgets.

Extending measurements to include δ¹³C of respired CO₂ and source pools, continuous winter flux observations, and detailed snow-ice-soil microclimate, combined with process-based modelling, would improve understanding of how temperate forests and peatlands respond to ongoing changes in winter climate and freeze–thaw regimes.