Northern typical forested swamp soil CO₂ efflux dynamics

1 Introduction

Since the Industrial Revolution, atmospheric concentrations of major greenhouse gases have increased dramatically, contributing to anthropogenic global warming (Cao et al., 1987). As one of the most critical greenhouse gases, CO₂ source-sink dynamics exert decisive influence on the direction and intensity of ecosystem responses to climate warming (Piao et al., 2009; Tao et al., 2024). Temperature serves as a key environmental variable regulating greenhouse gas emissions. Research indicates that elevated temperatures enhance microbial activity and accelerate organic matter decomposition rates, thereby promoting CO₂ emissions (Chen et al., 2021; López Lloreda et al., 2024). The relationship between carbon emissions and temperature rise under varying regional CO₂ concentration contexts presents complex patterns, emerging as a prominent frontier scientific issue (Filonchyk et al., 2024; Jarvis and Forster, 2024).

Globally, the correlation between CO₂ emissions and global temperature variations has been extensively studied and validated (Stips et al., 2016). Andrew Jarvis proposed that cumulative carbon emissions since the industrial era exhibit a linear relationship with temperature changes (Jarvis and Forster, 2024). Li Chen identified shared trends between global temperature and atmospheric greenhouse gas concentrations using common trend analysis. UK researchers established the first mathematical relationship between global warming and carbon emissions through mathematical modeling, highlighting a positive feedback loop between rising temperatures and CO₂ emissions (Goodwin et al., 2015). Research by the Chinese Academy of Sciences revealed regional heterogeneity in temperature responses to cumulative CO₂ emissions across different zones (Cheng et al., 2022). Current studies primarily focus on macro-regional scales, emphasizing agricultural soils and crop production systems. However, investigations on finer-scale forest landscapes, particularly soil carbon emissions in northeastern forested wetlands, remain inadequate (Han et al., 2019; Chen et al., 2016). Forest wetlands constitute a critical wetland ecosystem type and play a significant role in the global carbon cycle (Zhou et al., 2018; Wang and Liang, 2013). As important wetland ecosystems, forested swamps exert dual functions as both carbon sources and sinks, and their greenhouse-gas emissions constitute a critical component of regional carbon budgets. Meanwhile, these ecosystems possess considerable carbon-sequestration potential, playing an irreplaceable role in regulating terrestrial carbon cycling under climate change. Under accelerating climate warming, understanding temperature-dependent variations in greenhouse gas fluxes from forested swamps has become a key research priority.

Internationally, numerous field investigations have addressed temperature-driven soil CO₂ efflux from boreal forested swamps in North America and Europe (Davidson et al., 2024). Most of this have mainly focused on soil temperature and water-table depth as core controlling variables for soil respiration in thick-peat swamp ecosystems (Karim et al., 2026; Butlers et al., 2025). Although near-surface microenvironmental air temperature directly reflects above-ground thermal conditions and can transmit signals to topsoil, yet few studies have systematically quantified how antecedent air-temperature conditions over different time windows modulate soil CO₂ emission dynamics (Mamkin et al., 2023). In boreal thick-peat sites, the substantial thermal buffering of peat layers creates pronounced time lags between air-temperature fluctuation and soil-respiration responses (Ranniku et al., 2025). Whether such lag patterns apply to mineral-soil forested swamps under monsoon-dominated continental climates remains unclear. Furthermore, field-observed temperature-flux coupling relationships are strongly ecosystem-specific, and conclusions derived from European and North-American peat-dominated forest-swamps cannot be directly transferred to East-Asian monsoon forest-swamp systems with thin humus mineral soils. Hence, further field observations are required to clarify how in-situ microenvironmental air temperature (including multi-day antecedent temperature metrics) shapes soil CO₂ efflux across different plant-community types in northern monsoon-affected forested swamps.

This study compares the multidimensional effects of forest swamp types on soil CO₂ emissions and their microclimate feedbacks. Our objective is to gain deeper insights into how the microclimate of forested swamps promotes or inhibits soil CO₂ emissions. The research focuses on the forested swamps of Laojieji Forest Farm in the Zhangguangcai Range, a typical distribution area for northern forest swamps in China. The findings can provide a scientific basis for carbon management, carbon reduction potential assessment, and the development of targeted mitigation strategies for northern forested swamps.

The overall aim of this work is to examine five typical northern forested swamps at Laojieji Forest Farm of the Zhangguangcai Range. We aim to characterize multi-temporal-scale spatiotemporal variations in soil CO₂ efflux across different swamp types and elucidate the regulatory mechanisms of micro-environmental air temperature on soil CO₂ emissions, thereby providing scientific support for carbon assessment and refined ecological management of northern forested swamps in China. To achieve this goal, we address three specific objectives: (1) to reveal the dynamics of soil CO₂ efflux across five forested swamp types at annual, growing−/non-growing-season, and seasonal (spring, summer, autumn, winter) scales, and identify spatial differences in CO₂ emissions among swamp types; (2) to quantify and statistically test inter-community differences in soil physicochemical properties and micro-environmental temperature parameters among these swamp types; (3) to characterize the responses of soil CO₂ efflux to micro-environmental air-temperature indices (antecedent temperature metrics of different time windows) and identify key temperature-driving factors.

2 Study area

The study area is situated in the Laojieji Forest Farm on Zhangguangcai Ridge, Heilongjiang Province, China (127.93°–127.98°E, 44.85°–44.88°N). This region experiences a humid continental monsoon climate characterized by long, cold, dry winters and warm, humid, rainy summers. Mean annual temperature ranges from 2.3 to 3.7 °C, with an average annual freezing period of 160–229 days. Annual precipitation varies between 440 and 640 mm, concentrated primarily from June to September. Seasonal frost depth reaches 1.8–2.5 m (Zhou and Tang, 2023). Dominant vegetation includes mixed coniferous-broadleaf forests, birch forests, deciduous broadleaf forests, broadleaf-Korean pine forests (Pinus koraiensis), Mongolian oak (Quercus mongolica) forests, and shrub/herb communities. Representative tree species comprise Manchurian walnut (Juglans mandshurica), Manchurian ash (Fraxinus mandshurica), Manchurian maple (Acer mandshuricum), painted maple (Acer mono), large-fruited elm (Ulmus macrocarpa), and Japanese elm (Ulmus davidiana var. japonica). Predominant soil types are dark brown soils, luvisols, and black soils (Su and Wang, 2017).

3 Methods

3.1 Plot establishment

This study established five forested swamp types: silver-birch swamp (B), herbaceous swamp (C), shrub swamp (G), larch swamp (L), and alder swamp (M). Three permanent sample plots (20 m × 30 m) were established for each swamp type, totaling 15 plots. Within each forest swamp type, three static dark chambers were installed as replicates, resulting in a total of 15 chambers for soil CO₂ gas sampling and flux monitoring (Zhou et al., 2016). The location of the study area and the distribution of forested swamp types plots are shown in the Figure 1.

3.2 Field measurements and analysis

3.2.1 CO₂ flux measurement and analysis

Gas samples were collected using the static dark chamber-gas chromatography method. The chamber consisted of a stainless steel top box (50 cm × 50 cm × 50 cm, L × W × H) and base (50 cm × 50 cm × 10 cm, L × W × H). The exterior of the top box was covered with foam insulation to minimize temperature fluctuations (Gu et al., 2010). The top box was equipped with two 1-cm diameter ports: one for gas sampling and another for insertion of a digital thermometer probe. A 10-cm diameter fan was installed inside the top box and operated during sampling to prevent gas concentration gradients. Prior to sampling, the base was inserted 10 cm into the soil at each sampling point and remained fixed at the site until completion of annual sampling to minimize soil disturbance. During sampling, water seals were filled in the peripheral grooves of the base before installing the top box. Vegetation and roots within the base were removed before each chamber deployment to exclude autotrophic respiration from plants.

We sampled gases every 10 days from 8 May 2023 until 8 May 2024. A 10 day interval (i.e., three samplings per month), is routinely used in year round static dark chamber GC monitoring for forested swamps in the mountainous areas of Northeast China (Ma et al., 2017). Cold temperate swamp studies emply 7–10 day intervals, which can capture seasonal patterns and inter community differences in soil CO₂ emissions for annual scale field surveys (Chang et al., 2020). Soil CO₂ efflux stays low and changes slowly during the long frozen non growing season in this high latitude seasonally frozen swamp site, where rainfall triggered emission pulses rarely occur. Accordingly, based on published regional studies, local ecosystem characteristics, our research objectives and field practical constraints, this study adopted a 10-day sampling interval.

To minimize diurnal variation bias, all gas sampling campaigns over the whole monitoring period were carried out between 08:00 and 11:00 local time. For each sampling campaign, quadruplicate gas samples were collected from each static chamber at 0, 10, 20, and 30 min after chamber closure. Samples were obtained using 60-mL polyethylene medical syringes connected via three-way valves to stainless steel needles inserted through the chamber’s sampling port. Collected samples were transferred to 100-mL gas-tight bags for transport to the laboratory, where concentrations were analyzed by gas chromatography (GC, Agilent 7890A). Soil CO₂ efflux values were subsequently calculated according to the following Equation 1.

F = dc dt ⋅ M V 0 ⋅ P P 0 ⋅ T 0 T ⋅ H (1)

Where:

F = gas flux (mg·m⁻²·h⁻¹);

dc/dt = linear rate of change in gas concentration inside the chamber (μL·L⁻¹·min⁻¹);

M = molar mass of target gas (g·mol⁻¹);

V₀ = molar volume of ideal gas under standard conditions (L·mol⁻¹);

P₀ = standard atmospheric pressure (kPa);

T₀ = absolute temperature under standard conditions (K);

P = actual atmospheric pressure at sampling site (kPa);

T = absolute air temperature inside the chamber during sampling (K);

H = height of the static chamber above the soil surface (m).Positive F values represent gas emission from soil to atmosphere, while negative values indicate gas uptake by soil. Unit-derivation note: The term dc/dt describes concentration change over time. Combining molar mass, standard molar volume, pressure–temperature correction and chamber height converts concentration-change rates into mass-based flux normalized by soil-surface area and sampling time, yielding the final flux unit of mg·m⁻²·h⁻¹.

3.2.2 Soil physicochemical property measurements

The soil physicochemical indicators monitored in this study included soil temperature (Tsoil), soil moisture content (MCsoil), soil bulk density (BDsoil), soil organic carbon content (SOC), total soil nitrogen content (TSN), and soil carbon-to-nitrogen ratio (C/N). A portable digital thermometer (JM624, Tianjin Jinming Co., Ltd.) was used to monitor the temperature at a soil depth of 0–20 cm. In the center of each standard plot, soil profiles were established in a triangular pattern, with each profile consisting of two soil layers (0–10 cm and 10–20 cm). Collected soil samples were brought back to the laboratory for further physicochemical analysis. Soil bulk density was determined using the cutting ring method (100 cm³). Soil gravimetric water content was measured by the oven-drying method (soil samples were dried at 105 °C for 48 h); soil organic carbon was analyzed using a C/N analyzer (Multi N/C 3100, Analytik Jena AG, Germany); total soil nitrogen was determined using a Kjeldahl nitrogen analyzer (KjeltecTM 8,400, Foss Tecator AB, Sweden). Soil temperature measurements and sample collection were conducted synchronously with gas sampling.

3.2.3 Forest microenvironmental temperature measurement

Simultaneously with CO₂ gas sampling, microclimate monitoring instruments (COS-03-XUSB temperature loggers) were installed 0.5 m above ground within standard plots of each forested swamp type. These instruments recorded forest swamps microenvironmental temperatures at 2-h intervals.

We measured four different temperature metrics including maximum, minimum, mean, and accumulated temperature. Temperature metrics were calculated for the one day prior, three days prior, and five days prior to the monthly gas sampling dates (the 10th, 20th, and 30th of every month). The one day window includes the day directly preceding the sampling day (9th, 19th, 29th); the three day window includes the three consecutive days preceding the sampling day (7th 9th, 17th 19th, 27th 29th); the five day window includes the five consecutive days preceding the sampling day (5th 9th, 15th 19th, 25th 29th). Eleven temperature variables were produced altogether. The symbol notations are explained below:

• T_1a, T_3a, T_5a: mean temperature for the 1-day, 3-day, and 5-day antecedent windows, respectively;
• T_1i, T_3i, T_5i: minimum temperature for the 1-day, 3-day, and 5-day antecedent windows, respectively;
• T_1x, T_3x, T_5x: maximum temperature for the 1-day, 3-day, and 5-day antecedent windows, respectively;
• T_3j, T_5j: accumulated temperature for the 3-day and 5-day antecedent windows, respectively.

3.3 Data analyses

To address our first objective (1) regarding the spatiotemporal differentiation characteristics of soil CO₂ emissions in forested swamps, we analyzed the intra-annual dynamic trends of soil CO₂ emissions for the five forest swamp types based on field CO₂ gas sampling and laboratory flux measurements conducted from 2023 to 2024. One-way ANOVA followed by Tukey’s HSD post-hoc test was applied to compare group-level mean soil CO₂ efflux among five swamp types at annual, growing-season, non-growing-season and seasonal scales.

In order to investigate our second objective (2) regarding the variation in microclimate indicator variables between forest types, we applied ANOVA and Tukey’s HSD tests for pairwise comparison to evaluate the differences in soil physicochemical properties among forest types at the same year. Ultimately it verified that Tsoil, MCsoil, and BDsoil play a significant role in describing the differences among forest types. Additionally, in order to analyze the meteorological background conditions of the forest swamps, we calculated and analyzed daily mean temperature and daily precipitation according to air temperature and precipitation data obtained from meteorological stations in the study area (corresponding to the CO₂ gas sampling year). We then performed linear regression to analyze trends in those temperature variables during the spring freeze–thaw period. In order to explore the role of forest swamp microclimate in characterizing differences among forest types, we conducted ANOVA and Tukey’s HSD tests for pairwise comparison to identify variations of daily mean, minimum, maximum, and accumulated temperatures varying with forest type.

To determine the effects of environmental indicators on CO₂ emissions (3), the correlation between CO₂ efflux and the physicochemical properties of forest swamp soil as well as environmental temperature was analyzed via the Pearson correlation coefficient. Building on the established roles of Tsoil, MCsoil, BDsoil, and environmental temperature (T) in characterizing differences among forest types, and considering the identified relationships between Tsoil, MCsoil, and environmental temperature (T), we employed scatter plot fitting to analyze the trend of forest swamp soil CO₂ efflux in response to environmental temperature—specifically, the mean (T_1a, T_3a, T_5a), minimum (T_1i, T_3i, T_5i), maximum (T_1x, T_3x, T_5x), and accumulated temperatures (T_3j, T_5j)—which serve to characterize the distinctions among forest swamp types. Finally, to parse the feedback effects of soil CO₂ efflux on forest swamp type, we constructed quantified coupling relationships between soil CO₂ efflux and environmental temperature at the microclimate scale (T_1a, T_3a, T_5a, T_1i, T_3i, T_5i, T_1x, T_3x, T_5x, T_3j, T_5j) using stepwise regression models. Prior to constructing multivariate coupling regression models, variance inflation factor (VIF) was used to diagnose multicollinearity among predictors. Severe multicollinearity was defined as VIF > 10. Backward stepwise regression was applied for iterative elimination of non-significant and collinear predictors until VIF values for all retained independent variables were below 10 to obtain stable final regression models. These analyses spanned five spatial scales (B, L, M, G, and C) and seven temporal scales: annual (12 months), growing season (May–September), non-growing season (October–April), spring (March–May), summer (June–August), autumn (September–November), and winter (December–February).

3.4 Statistical analysis

All statistical analyses were performed with SPSS 17.0, while further data processing and graphing were conducted in OriginPro 2021.

4 Analysis of results

4.1 Analysis of forest swamp CO₂ emissions

4.1.1 Annual flux and variation trends of CO₂ emissions across five swamp soil types

In addressing our first objective of evaluating and comparing temporal and spatial variation trends of soil CO₂ emissions, our analysis of field sampling and monitoring data revealed considerable and statistically significant CO₂ emission variation characteristics. The five swamp types exhibited distinct intra-annual CO₂ efflux dynamics, primarily differing in peak frequency and timing. Silver-birch swamps (B) demonstrated a unimodal pattern, while the other four types (larch/L, alder/M, shrub/G, herbaceous/C) showed bimodal patterns. All types displayed gradual increases starting in early spring, peaked between June–July, then experienced fluctuating declines before stabilizing at low levels from late autumn through winter (Figure 2a).

Annual mean CO₂ efflux differed significantly among the five swamp types, ranging from 122.79 to 235.18 mg·m⁻²·h⁻¹. Shrub swamps (G) showed the highest efflux (1.35–1.92 times that of C/L/B/M, p < 0.05), followed by herbaceous swamps (C) (1.02–1.24 times that of L/B, p > 0.05; 1.41 times that of M, p < 0.05). Larch (L) and birch (B) swamps exhibited intermediate values (0.191–0.382 times higher than M, p > 0.05), while alder swamps (M) had the lowest annual efflux. The overall ranking was G > C > L ≈ B > M (Figure 2b).

4.1.2 Seasonal variations in CO₂ fluxes across five swamp soil types

During the growing season, significant differences in soil CO₂ efflux were observed among the five swamp types. The growing-season mean fluxes across swamp types ranged from 234.21 to 469.99 mg·m⁻²·h⁻¹. The distribution generally maintained the same ranking as annual fluxes (G > C > L > B ≈ M). Unlike annual patterns, birch swamps (B) exhibited lower growing-season efflux, closely aligning with alder swamps (M). During the non-growing season, differences among swamps were non-significant (p > 0.05), with mean fluxes across swamp types ranging from 51.07 to 77.57 mg·m⁻²·h⁻¹. Contrary to annual and growing-season scales, the CO₂ efflux ranking was G > L > B > C > M (box chart in Figure 3a).

CO₂ efflux responses to seasonal variations differed among swamp types. Seasonal mean efflux ranged from 110.24–158.37 mg·m⁻²·h⁻¹ in spring, 242.74–569.70 mg·m⁻²·h⁻¹ in summer, 106.95–185.238.69 mg·m⁻²·h⁻¹ in autumn, and 3.45–22.60 mg·m⁻²·h⁻¹ in winter. Spring efflux followed the order G > L ≈ C ≈ B ≈ M, with shrub swamps (G) differing significantly from silver-birch swamps (B) and alder (M) (p < 0.05). Summer efflux ranked G > C > L ≈ B > M; the inter-type difference pattern of soil CO₂ efflux in summer was consistent with that observed at the annual scale. Inter-swamp differences in soil CO₂ efflux were not statistically significant in autumn. Winter efflux followed G > C > M > L ≈ B, only the differences between shrub swamps (G) and larch (L), silver-birch (B) swamps reached statistical significance (box chart in Figure 3b).

The difference in CO₂ efflux contribution rates during the growing season among swamp types was not significant. CO₂ efflux during the growing season accounted for over 87% of annual emissions across all swamp types, with herbaceous swamps (C) showing the highest contribution, followed by shrub (G), birch (B), larch (L), and alder (M) swamps. Tree-dominated swamps (B, L, M) showed lower contributions than shrub and herbaceous swamps, whereas non-growing season contributions inversely mirrored growing season trends (bar chart in Figure 3c).

Seasonal contributions to annual efflux were consistent across swamp types: summer dominated (>50% of annual emissions), followed by comparable contributions from spring and autumn, with winter accounting for <1%. Among the five types, herbaceous swamps (C) showed the highest summer contribution, followed by shrub (G), birch (B), larch (L), and alder (M) swamps. Overall, spring and summer collectively contributed 90% of annual CO₂ efflux, indicating most emissions occur during these two seasons (bar chart in Figure 3d).

Spring (March–May) is the seasonal soil freeze–thaw period in our study area, during which mean fluxes across swamp types ranged from 110.24–158.37 mg·m⁻²·h⁻¹. Spring-time fluxes represent combined signals of freeze–thaw pulses and spring warming. Spring soil CO₂ emissions make substantial contributions to annual total efflux (Figure 3d), indicating that freeze–thaw-driven emissions constitute a non-negligible component of carbon fluxes in northern forested swamps.

4.2 Soil environment

In addressing our second objective to quantify the impact of forest type on forest microenvironment parameters, we found that Tsoil, MCsoil, and BDsoil vary with forest type (Figure 4). Statistical analysis of the annual mean values and ANOVA of various soil physicochemical indicators across different forested swamp types quantified the influence of swamp type on soil properties. The results showed that among the six indicators (Tsoil, MCsoil, BDsoil, SOC, TSN, C/N), Tsoil, MCsoil, and BDsoil exhibited significant variation with forest type (p < 0.01), while the others (SOC, TSN, C/N) showed no significant differences. This indicates that Tsoil, MCsoil, and BDsoil play a dominant role in characterizing differences among forest swamp types.

4.3 Meteorological environments in forest swamp

4.3.1 Meteorological environment in the study area

The daily mean air temperature and daily precipitation used for characterization of regional background conditions was acquired from the Shangzhi Municipal Meteorological Station. Located about 42 km away from the experimental plots; the meteorological station has an elevation of 186 m while elevations of study plots range from 340 m and 380 m, thus creating an elevation difference of 154–194 m. The station and the study area both lie in the northern Zhangguangcai Range and dominated by identical climatic systems. With similar seasonal freeze thaw regimes and intra annual precipitation patterns, observations from this meteorological station represent the regional background meteorological conditions throughout the study area.

The study area exhibited distinct seasonal temperature patterns during 2023–2024: a significant springtime increase (R²: 0.8114), peaking in summer, declining through autumn, and reaching winter minima. The annual mean temperature was 4.7 °C. Precipitation mirrored this seasonal trend, with summer maxima, lower spring/autumn levels, and minimal winter precipitation (Figure 5).

4.3.2 Temperature in forested swamp microenvironments

As shown in Figure 6a, monthly temperatures (month-scale daily-mean values derived from micro-environmental airtemperature records -at 2-h intervals) across the five forest swamp microenvironments ranged from −15.4 °C to 22.5 °C. All swamp types showed consistent annual thermal regimes: gradual warming beginning in early March, peaking in July, followed by cooling to January minima. During the growing season (May–September), daily mean temperatures exceeded 0 °C across all swamp types, while the non-growing season (October–April) featured subzero temperatures from November through March. The annual variation trend of air temperature in the microenvironment of forest swamps is consistent with that in the study area.

In addressing our second objective to quantify the impact of forest type on forest microclimate parameters, we found that daily mean, minimum, maximum, and accumulated temperatures in swamp microenvironments ranged from −18.8 °C to 24.4 °C, −29.0 °C to 19.4 °C, −9.0 °C to 39.0 °C, and −100.9 °C to 71.5 °C, respectively (these metrics were calculated from micro-environmental air temperature measured at 2-h intervals). Maximum and accumulated temperatures vary with forest type. No significant differences (p > 0.05) occurred in daily mean or minimum temperatures among the five swamp types. However, birch and larch swamps displayed significantly higher daily maximum and accumulated temperatures compared to other types (p < 0.05) (Figure 6b).

4.4 Environment response of forest swamp soil CO₂ emissions

4.4.1 Correlation analysis

Environmental temperature (T: calculated as the average of T_1a, T_3a and T_5a across micro-environments of the five swamp types) in the forest swamps showed a highly significant positive correlation (correlation coefficient = 0.92, p < 0.001) with Tsoil, followed by a weaker negative correlation with MCsoil, and no significant correlation with other soil physicochemical indicators. It can be inferred that Tsoil and MCsoil are the core factors associated with environmental temperature in forest swamps (Figure 7a).

Based on the above analysis, environmental temperature (T) can characterize the differences among forest swamp types. Meanwhile, soil CO₂ efflux in forest swamps maintained a highly significant positive correlation (p < 0.001) with both environmental temperature (T) and its core associated factor Tsoil, and a significant negative correlation with MCsoil (p < 0.05) (Figure 7b). This indicates a close synergistic relationship between soil CO₂ efflux and environmental temperature (T), which characterizes the differences among forest swamp types. Therefore, this study analyzes the feedback effects of soil CO₂ efflux on forest swamp type, Tsoil, MCsoil, and other indicators by constructing a quantitative coupling relationship between soil CO₂ efflux and environmental temperature (T) at the microclimate scale.

4.4.2 Variation in CO₂ emissions with temperature

Our results indicate that increases in the values of microclimate parameters in forest swamps will promote soil CO₂ emissions. Overall, soil CO₂ efflux across the five forest swamp types consistently demonstrated positive correlations with temperature variables (T_1a, T_3a, T_5a, T_1i, T_3i, T_5i, T_1x, T_3x, T_5x, T_3j, T_5j), with R² values ranging from 0.6397 to 0.7628. Mean and accumulated temperatures showed stronger associations (R² > 0.7). Temperature ranges for significant efflux were: -20.5–24.4 °C (mean temperatures: T_1a/T_3a/T_5a), −37.2–19.5 °C (minimum temperatures: T_1i/T_3i/T_5i), −9.1–40.9 °C (maximum temperatures: T_1x/T_3x/T_5x), and −163.3–116.6 °C (accumulated temperatures: T_3j/T_5j). Critical thresholds were identified at 14.1 °C (T_5x, maximum) and −12.3 °C (T_1i, minimum). Among all examined antecedent temperature metrics, the derived critical thresholds spanned a wide range, with the maximum threshold of 14.1 °C observed for T_5x (maximum antecedent temperature) and the minimum threshold of −12.3 °C for T_1i (minimum antecedent temperature). When below the minimum threshold, soil CO₂ efflux remained at a low level (9.9–42.2 mg·m⁻²·h⁻¹) with weak temperature sensitivity. When exceeding the maximum threshold, efflux showed a marked increase (24.6–812.8 mg·m⁻²·h⁻¹), accompanied by elevated variability (Figure 8; Table 1).

Table 1. Critical temperature thresholds and ranges for soil CO₂ emissions in forested swamps.

4.4.3 Coupling relationships between temperature and CO₂ emissions

Multivariate linear regression models were constructed using antecedent 1–5 d micro-environmental air-temperature indices. After VIF diagnosis and backward stepwise regression, the maximum VIF of retained predictors in all final models was below 10, confirming effective control of multicollinearity (Table 2).

Table 2. Stepwise regression models between soil CO₂ emissions and air temperature in forested swamps: silver-birch swamp (B), herbaceous swamp (C), shrub swamp (G), larch swamp (L), and alder swamp (M). Time scales included: annual, growing season (Gs), non-growing season (Ngs), spring, summer, autumn, and winter.3.2.3 for the explanation of temperature factor abbreviations.

At the annual scale, CO₂ efflux of all five swamp types was well explained by antecedent air temperature (R²: 0.746–0.868). Only T_1i was retained for birch (B), Larix (L) and alder swamps (M), while both T_1x and T_1i were selected for shrub (G) and herbaceous swamps (C), implying that 1-day antecedent air temperature was the key driver of annual flux variation. Growing-season models exhibited clear community differences: significant models were obtained for birch (B), shrub (G) and herbaceous swamps (C), driven mainly by T_5x, T_5j and T_1i; the alder swamp (M) model had low explanatory power (R²: 0.361); and no significant predictor was retained for Larix swamp (L). All five swamps yielded significant non-growing-season models (R²: 0.627–0.856), predominantly selecting 1–3 d mean and maximum temperatures, indicating strong explanatory power of short-term temperature fluctuations during freeze–thaw alternation.

Spring and autumn models were significant: spring was jointly driven by T_5a, T_1x and T_1i, and autumn by T_1i and T_5j. The winter model had low explanatory power (R²: 0.181) with only T_1i retained. No significant predictor was retained in summer.

5 Discussion

5.1 Spatial variability of soil CO₂ emissions

In this study, annual mean CO₂ fluxes from shrub swamps and herbaceous swamps were higher than those from tree-dominated swamps (silver-birch, larch, and alder swamps). Significant differences were observed between shrub swamps and other swamp types, as well as between herbaceous swamps and alder swamps (p < 0.05). Soil CO₂ efflux mainly derives from root autotrophic respiration and microbial heterotrophic decomposition of soil organic matter, rather than being a direct proxy for above-ground photosynthetic carbon-sink strength. The inter-community flux discrepancies found here can be mainly explained by differences in root characteristics, litter substrate supply and soil physicochemical properties among swamp types. Even though tree-dominated swamps generally have higher leaf area index and longer canopy duration, which give them stronger carbon sequestration potential at the ecosystem scale, their below-ground carbon allocation patterns constrain soil respiration intensity (Li et al., 2017; Yang et al., 2024; Zhou et al., 2012). Tree-dominated swamp communities allocate more assimilated carbon to long-lived woody tissues; root turnover rates are relatively slow, and labile organic substrate inputs to topsoil are limited, thereby restricting substrate availability for soil microbial decomposition (Zhu, 2015; Wang et al., 2016). By comparison, shrub and herbaceous swamps are characterized faster fine-root turnover and larger quantities of easily-decomposable above- and below-ground litter inputs. Such continuous labile substrate supply drives higher rates of root and microbial respiration, consequently generating higher soil CO₂ efflux (Song et al., 2006; Smith et al., 2018). In addition, the differences in soil microclimate and hydrological conditions among swamp types can also modify decomposition processes and enlarge inter-type differences in soil CO₂ emissions.

Comparisons with published studies on boreal forested swamps in Fennoscandia (Finland and Sweden) show obvious regional divergence in CO₂ emission patterns. Most European boreal forest swamp sites grow on thick peat substrates, many of them having been drained for forestry (Minkkinen et al., 2018). Their soil CO₂ fluxes being co regulated by water table depth and temperature where water table depth typically acts as the main constraint on flux variability. Our five forest swamp types develop on dark brown forest soils with thin humus layer rather than thick peat. Although at comparable latitude, our site is under a continental monsoon climate with distinct freeze thaw regime. Rapid spring warming triggers an extra emission peak during spring thaw, producing the bimodal spring-autumn flux pattern for most communities except silver-birch swamp, which is rarely reported for European drained boreal peat-forests.

Studies of European drained boreal peat-forests commonly emphasize the limiting effect of water-table depth on soil respiration, whereas temperature is of secondary importance (Tong et al., 2024). In our monsoon-influenced study area, strong seasonal air-temperature fluctuations make micro-environmental air temperature the predominant driver of inter-community CO₂-emission differences. Flux divergence among swamp types arises largely from microclimatic differentiation rather than gradients in peat thickness or water-table position. Furthermore, European boreal forest swamps are overwhelmingly conifer dominated; however, we have covered a diverse array of silver birch, larch, alder, shrub and herbaceous swamps. Our result showing that shrub swamps show the highest CO₂ efflux has rarely been reported in European counterparts, demonstrating the unique modulation effect of community composition on carbon emissions in Northeast Chinese forested swamps. Such discrepancies suggest that mechanisms observed from European boreal peat forest systems cannot be extrapolated to monsoon affected forested swamps in East Asia. We provide valuable full year observational data for forest swamp carbon cycling under the East Asian monsoon.

One way ANOVA combined with Tukey’s HSD test was used to compare group level mean flux differences between swamp types at different temporal scales. It does not completely remove possible bias due to temporal autocorrelation of repeated measurements within each static chamber. Linear mixed effects models are suggested for future research to further resolve differences in temperature response patterns.

Based on the differential temperature responses of carbon emissions among the five types of forested swamps investigated in this study, differentiated wetland management strategies are proposed for forested wetland administration. Targeted regulation measures should be formulated for high-emission shrub swamps and herbaceous swamps, while priority conservation should be implemented for Alnus sibirica swamps and larch swamps with stronger carbon sequestration capacity.

5.2 Temporal dynamics of CO₂ emissions

In this study, CO₂ efflux during the growing season was significantly higher compared to the non-growing season, where all swamp types contributed >87% of annual emissions in the growing season, which dominate wetland CO₂ efflux. This agrees with seasonal variations in plant growth cycles caused by enhanced photosynthesis and respiration during the growing season (Song et al., 2021), whereas, suppressed efflux occurs because of lower temperatures in the non-growing season (Yin et al., 2022; Zhou L. et al., 2021). These results agree with those of Mansfield et al. (2020), Liu et al. (2024), and Zhou W. et al. (2021). Seasonal patterns of CO₂ flux were similar across swamp types: Summer exhibited the highest emissions (accounting for the largest proportion of annual fluxes); spring and autumn had comparable contributions but winter made minimal contributions (≤1% annually) (Smith et al., 2018). This parallels what is seen in southern Canadian riparian forest swamps (Wang and Liang, 2013; Zhou and Tang, 2023; Su and Wang, 2017; Gu et al., 2010; Dhandapani et al., 2019; Chang et al., 2020), temperature drives soil CO₂ efflux dynamics (Yin et al., 2022; Repo et al., 2021). Higher summer temperatures enhance microbial biomass and activity, increase soil enzyme activity and improve enzyme-substrate affinity (Su and Wang, 2017; Zhou et al., 2016) - all leading to an increase in CO₂ emissions. Such seasonal dynamics strongly affect wetland carbon cycling especially under climate change when shifting temperature regimes can further modulate carbon storage-release dynamics (Song et al., 2021).

In this study, larch, alder, shrub, and herbaceous swamps exhibited bimodal seasonal dynamics in CO₂ efflux, with distinct peaks occurring in spring and autumn. In contrast, birch swamps displayed a unimodal pattern—differing from the unimodal efflux reported in mid-northern U. S. riparian forest swamps. The bimodal trend aligns with seasonal variations in soil temperature, moisture, and precipitation events (Repo et al., 2021; He et al., 2017; Liu et al., 2019). As shown in Figure 2, the study area exhibited rapid springtime warming (early March to mid-June) at 0.2107 °C d⁻¹ (R² = 0.7524), followed by peak precipitation from late summer to autumn (exceeding other seasons). Spring temperature increases likely activated soil microbial and root activity, enhancing CO₂ production, while autumn precipitation events supplied critical soil moisture, further stimulating microbial activity and organic matter decomposition and increasing CO₂ efflux (He et al., 2017). Nevertheless, silver-birch swamp lacked this freeze–thaw-induced secondary spring peak, which can be explained by multiple intrinsic mechanisms. First, the phenology and physiological activation of fine-root systems in silver-birch differ from other communities; fine-root activity rises moderately in early spring, resulting to limited contribution from root autotrophic respiration (Liu et al., 2019; Kilpeläinen et al., 2022). Second, relatively high soil moisture in silver-birch swamp sites during spring restricts soil aeration, attenuating the freeze–thaw priming effect on microbial mineralization and reducing spring CO₂ pulses (Fierer and Schimel, 2002). Moreover, labile carbon substrates derived from silver-birch broad-leaf litter are partly consumed in early periods, lowering substrate availability during spring thaw (Deng et al., 2009; Wei and Man, 2021). Combined effects of above factors remote the secondary spring peak, and fluxes reach a sole maximum only under optimal summer hydrothermal conditions, forming a unimodal pattern. It should be noted that our discrete 10-day sampling interval may miss very short-lived freeze–thaw pulse events, imposing certain limitations on capturing faint spring emission pulses.

In this study, spring encompasses the soil-thawing freeze–thaw phase, and spring-time CO₂ fluxes (164.61–307.67 mg·m⁻²·h⁻¹) together with seasonal contribution ratios represent the integrated effect of freeze–thaw pulses. Swamp types differed in their responses to freeze–thaw processes: larch, alder, shrub and herbaceous swamps exhibited clear secondary spring emission peaks pointing to freeze–thaw priming. By contrast, silver-birch swamp showed no obvious freeze–thaw-driven spring secondary peak due to combined constraints from root phenology, soil aeration and substrate availability. It should be noted that our discrete 10-day sampling may miss short-term high-magnitude pulses occurring right at thaw onset, which might somewhat underestimate the true contribution of freeze–thaw events. Higher-frequency field monitoring would facilitate more accurate quantification of carbon loss during freeze–thaw periods.

Transient negative soil CO₂ fluxes, representing instantaneous net soil CO₂ uptake, have been occasionally reported in boreal and temperate forest-swamp ecosystems in previous studies. Such episodic carbon-sink signals generally emerge under specific conditions such as strong freeze–thaw cycles, post-heavy-rainfall saturation and low-temperature anoxic micro-environments. Possible underlying mechanisms include microbial chemosynthetic fixation and CO₂ dissolution within soil pore water. It should be noted that these negative-flux events are mostly short-lived and do not shift the annual carbon-source status of most forest-swamp soils (Wilson et al., 2017; Su et al., 2026).

5.3 Temperature response of soil CO₂ fluxes

Emission-climate model responses exhibited spatial-scale differences across forest swamp types. Nevertheless, temperature drivers remained consistent due to insignificant thermal variation among types (larch swamp, silver-birch swamp, alder swamp, shrub swamp, herbaceous swamp).

Our multiple regression results indicated that 1-day antecedent air temperature (T_1i) was the dominant predictor of annual soil CO₂ efflux across most swamp types, pointing to rapid responsiveness of soil carbon emission to short-term near-surface air-temperature fluctuations. Many boreal peat-swamp studies commonly select in-situ soil temperature or multi-day lagged environmental variables to interpret CO₂ flux variation, emphasizing the hysteresis effect of heat conduction within thick peat profiles. Unlike thick-peat boreal systems, our study sites are characterized by thin humus mineral soils; heat transfer between atmosphere and topsoil is relatively fast, so that the 1-day antecedent air temperature to effectively capture temperature signals driving root and microbial activities. Similar short-time-lag responses of soil respiration to air temperature have also been reported in temperate forested wetland ecosystems, where surface mineral soils lack buffered thermal inertia from deep peat layers (Li et al., 2018).

Interestingly, no antecedent-temperature predictor was retained for summer-period models, while non-growing-season models exhibited high explanatory power driven by short-term air-temperature indices. This contrasting pattern suggests strong seasonal-dependent temperature regulation for forest-swamp CO₂ emissions under a continental monsoon climate. In most boreal peat-dominated wetlands, temperature is the strongest controller of soil respiration during the warm growing-season, whereas winter fluxes are often weak and poorly predicted by temperature alone (Byun et al., 2021). The disappearance of significant temperature predictors in summer in our study may be attributed to co-limitation from multiple confounding factors: high substrate availability, saturated microbial activity, and complex joint regulation by soil moisture and plant physiological processes, which reduce the one-to-one temperature-flux coupling relationship (LaCroix et al., 2019). Conversely, frequent freeze–thaw alternations during the non-growing-season make short-term air-temperature fluctuations the primary trigger for CO₂ bursts, resulting in robust performance of antecedent-temperature-based models. Such seasonal divergence of temperature-driven patterns further demonstrates that the temperature-flux relationship derived from boreal peatlands cannot be directly transferred to monsoon-influenced mineral-soil forested swamps in Northeast China.

In accordance with the temperature-dominated emission patterns, forest tending, vegetation reconstruction, hydrological regulation and other management practices should be rationally arranged. Wetland maintenance activities are suggested to avoid the peak carbon emission period in summer to reduce additional carbon emissions caused by human disturbances. Based on the feedback relationship between microclimate and soil carbon emissions, local microclimate can be regulated by adjusting stand structure and optimizing understory vegetation allocation, so as to control soil carbon emission intensity from environmental perspectives and realize refined ecological management of forested wetlands.

6 Conclusion

On the basis of on one-year continuous monitoring of soil CO₂ efflux and micro-meteorological factors across five typical forested-swamp types (silver-birch swamp, larch swamp, alder swamp, shrub swamp, and herbaceous swamp) in the Zhangguangcai Range of Northeast China, we draw the following main conclusions.

Across annual, growing-season and non-growing-season scales, shrub swamps consistently exhibited the highest soil CO₂ fluxes and alder swamps the lowest. Except for silver-birch swamp showing a unimodal pattern, the other four swamp types displayed bimodal intra-annual CO₂ emission dynamics, with emission peaks concentrated in summer. The growing-season contributed more than 85% of total annual soil CO₂ emissions for all swamp types.

Statistical tests on microenvironmental temperature sho-wed significant inter-swamptype differences in daily maximum temperature and accumulated temperature (P < 0.05), while daily mean temperature and daily minimum temperature had no significant in-ter-group differences. The ranges of micro-environmental antecedent-temperature metrics were as follows: mean antecedent temperatures (T_1a, T_3a, T_5a): −20.5–24.4 °C; minimum antecedent temperatures (T_1i, T_3i, T_5i): −37.2–19.5 °C; maximum antecedent temperatures (T_1x, T_3x, T_5x): −9.1–40.9 °C; accumulated antecedent temperatures (T_3j, T_5j): −163.3–116.6 °C. Scatter-plot fitting showed positive trends between all antecedent-temperature indices and soil CO₂ efflux. Mean and accumulated temperature variables produced stronger fitting performance with R² values greater than 0.7.

Soil CO₂ efflux from monsoon-influenced mineral-soil forested swamps exhibits scale-dependent thermal lag responses to antecedent micro-environmental air temperature. No universal fixed lag-time window exists for temperature-flux coupling. Seasonal thermal background determines the magnitude of coupling strength: freeze–thaw-controlled cold periods promote strong short-time-lag coupling between air temperature and soil CO₂ efflux, while warm summer conditions disrupt this temperature-flux relationship. In addition, plant-community identity further distinguishes thermal-lag signatures during the growing season, producing community-specific response patterns.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Guido L. B. Wiesenberg, University of Zurich, Switzerland

Reviewed by: Xinwei Liu, Chinese Academy of Sciences (CAS), China

Velisav Karaklić, University of Novi Sad, Serbia