Effects of Snow Cover on Carbon Dioxide Emissions and Their δ¹³C Values of Temperate Forest Soils with and without Litter

Abstract

The presence of litter and winter snow cover can affect the decomposition of organic matter in forest soils and changes in δ¹³C values of soil-respired carbon dioxide (CO₂). However, limited information is available on the responses of CO₂ emissions from forest soils and their δ¹³C values to snow cover and litter addition over the year. We experimentally manipulated snow cover to study the impacts of light and heavy artificial snow cover on soil heterotrophic respiration and its δ¹³C values, using undisturbed large soil columns collected from two typical temperate forests in Northeastern China. Based on the average temperatures of surface forest soils in four seasons of the year in this study region, the simulations of autumn freeze–thaw, winter freeze, spring freeze–thaw, and the growing season were sequentially carried out under laboratory conditions. A set of novel analysis systems, including automated chamber equipment and laser spectroscopy analysis with high-frequency measurements for CO₂ concentrations and the ¹³C/¹²C isotopic ratios in CO₂, was used to study the effects of artificial snow cover and the presence of litter on soil heterotrophic respiration and its δ¹³C values. During the autumn freeze–thaw simulation, there were larger CO₂ emissions and less negative δ¹³C values of soil-respired CO₂ upon heavy snow cover than upon light snow cover, indicating that the presence of increased snow cover prior to winter freeze can increase the decomposition of organic C in subsurface soils under temperate forests. The δ¹³C values of soil-respired CO₂ in all treatments were, on average, less negative as the simulated spring freeze–thaw proceeded, which was contrary to the variations of the δ¹³C during the autumn freeze–thaw simulation. Soil heterotrophic respiration and its δ¹³C values during the spring freeze–thaw simulation were, on average smaller upon heavy snow cover than upon light snow cover, which differed from those during the autumn freeze–thaw and growing season simulations, respectively. Taken together, the results highlight that the effects of snow cover on soil heterotrophic respiration and its δ¹³C values under temperate forests may vary with different seasons of the year and the presence of litter.

1. Introduction

Permanent and seasonal snow cover regions occupy about 60% of the global land surface [1], which accounts for more than 70% of global soil organic carbon (C) stocks [2,3]. Snow cover directly affects the heat exchange between topsoils and the atmosphere, reducing the impacts of soil temperature fluctuations. Thicker surface snow cover (>30 cm) may even prevent soil frost under severe cold conditions, supplying relatively stable living environments for underground microbial activity [4]. However, in the context of global climate change, the area of land snow cover in the Northern Hemisphere in spring is currently decreasing year by year and is predicted to reduce by 25% until the end of this century [5]. It has been confirmed that snow cover in the Northern Hemisphere reduced by 7% in March and 11% in April 2010 compared to that in 1970 and that this reduction is accelerating [6]. Seasonal snow cover patterns can have complex and diverse responses to global climate change, probably affecting above- and belowground processes in forest ecosystems via the redistribution of resources such as light, heat, water, and nutrients [7,8]. As a vital process of soil C cycling, soil respiration can account for more than 60% of forest ecosystem respiration [9,10], and a small change in soil respiration due to climate changes can even give rise to important changes in forest ecosystem respiration. The interaction between snowpack accumulation and increased air temperature affects soil freeze–thaw cycles, which can lead to changes in soil properties and soil CO₂ emissions [11].

In deciduous forests in the cold temperate zone, plant litterfall occurs as a pulse in late autumn and is considered one of the largest annual inputs of dissolved organic matter (DOM) into the soil [12,13], which can, in turn, affect soil CO₂ emissions and δ¹³C values of soil-respired CO₂ [14,15,16,17]. Furthermore, the variations of winter snowpack and following freeze–thaw variations in non-growing seasons in the cold climate zone can substantially influence the decomposition of litter and the dynamics of DOM released into the soil [11,18,19,20,21,22,23,24]. More importantly, there has been a trend of decreasing winter snowpack in cold temperate and boreal zones, especially in East Asia, under the background of global warming [25,26,27]. These phenomena can, in turn, affect the changes in soil respiration and its δ¹³C values in cold temperate forests over the year [15,16,24,28,29]. In recent decades, snow manipulation experiments have been widely carried out under different terrestrial ecosystems around the world, and there are some contrary results regarding the effects of snowpack on soil CO₂ emissions in forest ecosystems [11,28,30,31] due to different soil moisture levels as well as temperature and the duration of soil freeze–thaw events after snow manipulation [28,32]. For example, natural variations in snow cover did not affect annual soil CO₂ emission under a temperate forest in Austria [30], whereas soil CO₂ emission was reduced by snow removal under a temperate forest in northeastern China [28]. Based on a one-year winter snow manipulation experiment under a mid-temperate forest in northeastern China, 50% removal of snowpack did not affect soil respiration and microbial biomass C, whereas a 50% increase of snowpack temporarily enhanced the ratio of fungi to bacteria [31]. A recent meta-analysis showed that snow addition significantly increased soil respiration by 15.5%, whereas snow removal had no effect [33]. The deep continuous snow cover was reported to be closely associated with relatively high soil heterotrophic respiration in winter [34]. Unfortunately, to date, there have been few studies describing the impacts of snow cover and litter on soil heterotrophic respiration and its δ¹³C values during non-growing and growing seasons over the year [15,16].

Based on the δ¹³C measurements of soil respiration in trenched and non-trenched plots, Moyes et al. showed that seasonal δ¹³C variability of approximately 4‰ enrichment in summer versus spring and fall soil respiration was ascribed to the heterotrophic CO₂ source [35]. Under a mature temperate forest in northeastern China, the δ¹³C values of soil-respired CO₂ in the trenched plots were, on average, less negative (by 0.6‰) than those in the non-trenched plots in the growing season [36]. Probably, heterotrophic respiration from decomposing soil organic matter (SOM) and litter would affect the changes in δ¹³C values of soil-respired CO₂, with relatively large contributions of litter decomposition and surface soil layers to the δ¹³C in non-growing seasons [15,16]. Furthermore, aboveground litter inputs have an important impact on soil respiration and soil C cycling in various forest ecosystems over the year, mainly via their physical, chemical, and biological functions on soil properties [17]. However, to date, few studies have specifically documented how snow cover and the presence of litter affect soil CO₂ emissions and their δ¹³C values in temperate forests over the year [14,15,16,36,37], which is indeed associated with evaluating more precisely the responses of forest soil C cycling to winter climate changes.

The Changbai mountain region of northeastern China belongs to a typical seasonal snow-covered area, with continuous surface snow cover lasting from late November to early April each year. Due to its unique geographical location and climatic conditions, the region is currently becoming sensitive to global climate change, which leads to changes in the aboveground snowpack and the quantity and quality of aboveground litter in forest ecosystems [38,39,40]. This can ultimately affect C cycling in forest soils and C fluxes between soils and the atmosphere in the region. The mature temperate mixed broadleaf and Korean pine forest (BKPF) belongs to the main zonal forest ecosystem in northeastern China, and its aboveground litter composition, root distribution, SOM, and microbial characteristics are quite different from those of nearby secondary white birch forest (WBF) soils [19,38,41,42]. Compared with the BKPF, the stand structure of WBF is simple, and the soil environment under the WBF stand is becoming more sensitive to solar radiation and air temperature due to the relatively low canopy [43], which can result in different snow cover and changes in soil temperature and moisture conditions under the two forest stands in the non-growing season. Probably, the above-mentioned factors can remarkably lead to different responses of soil respiration and its δ¹³C values to changes in snow cover and litter decomposition between the two forests over the year.

We propose that winter snow cover can affect soil microbial respiration and its δ¹³C during different seasons of the year, which would vary with forest stands and the presence of litter. To reduce the effects of other environmental factors as much as possible, laboratory-controlled incubation has been considered an effective method to study the interaction effects of snow cover and the presence of litter on soil CO₂ emissions and their δ¹³C values. Based on the long-term laboratory-controlled incubation of undisturbed large soil columns collected from BKPF and WBF stands, the objectives of this study were to study the effects of snow cover and the presence of litter on soil heterotrophic respiration and its δ¹³C values of typical temperate forests in the Changbai mountain region in northeastern China. The results would provide a nice basis for understanding the responses of soil C cycling in temperate forests to global winter climate change.

2. Materials and Methods

2.1. Site Description and Collection of Soil Columns

The study area is in the temperate forest stands near the National Research Station of Changbai Mountain Forest Ecosystems (42°24′ N and128°6′ E) (Figure S1), with flat terrain and an average altitude of 738 m. The Changbai mountain region is located on the northeastern edge of the global monsoon climate and is influenced by the continental climate, which is characterized by long cold winters and short summers. A basic message regarding the thickness of ground snowpack, daily average air temperatures in winter, and the status of soil freezing was given by Xu et al. [29]. Within the climate region, the mature temperate BKPF belongs to the main zonal forest ecosystem in northeastern China, and its aboveground litter composition, SOM, and soil microbial characteristics are quite different from those of nearby secondary WBF [19,29]. Dominant trees of the BKPF stand include Tuan linden (Tilia amurensis Rupr.), Mongolian oak (Quercus mongolica Fisch. ex Turcz), Manchurian ash (Fraxinus mandshurica Rupr.) and Korean pine (Pinus koraiensis Sieb. et Zucc.), mostly >200 years old, and their stand densities are 117.1, 37.1, 27.2 and 98.9 trees ha⁻¹, respectively [44]. The WBF stand is dominated by white birch (Betula platyphylla Suk.) and mountain poplar (Populus davidiana Dode), mostly >70 years old, and its stand density is 1402 stem ha⁻¹; both white birch and mountain polar account for approximately 51.5% and 23.1% basal area of tree species, respectively [43]. Due to the differences in vegetation cover and tree species phototaxis, both air temperature and understory PAR transmittances under WBF stand in the non-growing season are higher than those under BKPF stand [43], resulting in the relative rapid melting of surface snowpack each year. The temperate forest soil belongs to an Andosol, containing a 3–5 cm depth of organic layer and approximately 10 cm depth of A horizon. Main soil properties, such as pH, texture, SOM, and bulk density, as well as the properties of aboveground litter, were described by Xu et al. [45,46] and Wu et al. [19,42].

We established 12 2 m × 2 m small plots with the well-preserved aboveground litter layers under BKPF, and WBF stands each in October 2017, prior to the occurrence of the autumn freeze–thaw process, for collecting large soil columns at approximately 15 cm depth under the two forests. At the center of each small plot, a stainless-steel cylindrical grinding tool (20 cm in diameter) was firstly smashed in the soil at approximately 15 cm depth and then replaced by thickened UPVC tubes (20 cm in diameter and 20 cm in height), and the soil column was finally taken out as one undisturbed soil column with aboveground litter. All forest soil column samples were sealed by covers and transported to the laboratory immediately, and they were stored in the dark at 4 °C prior to laboratory incubation.

2.2. Setup of Incubation Experiments

Half of all undisturbed soil columns (n = 6) under BKPF and WBF stands each removed the litter layer, and the other half retained the litter layer (n = 6). Referring to the average monthly precipitation in winter in this study area from 2005 to 2015, two levels of snow cover were considered as light and heavy snow cover treatments, respectively (

Table 1. Amount and time of snow addition at the different stages of the incubation.

Levels of Snow Cover	Simulation Stages
	Autumn Freeze–Thaw	Winter Freeze	Spring Freeze–Thaw
Light snow cover	35 g and 35 g on the first and 8th days of the incubation, respectively	160 g on the 15th day of the incubation	70 g on the 205th day of the incubation
Heavy snow cover	150 g and 150 g on the first and 8th days of the incubation, respectively	600 g on the 15th day of the incubation	300 g on the 205th day of the incubation

). Throughout the incubation, all soil columns with and without litter layers were covered four times with artificial snow made by the snowflake ice machine (FM50, Beijing Yijialin Technology Ltd., Beijing, China), and the scheme of artificial snow addition was shown in Table 1. The total amount of snow addition in the heavy snow cover treatment accounted for approximately 40% of the average annual snowfall to avoid water and nutrient leaching inside soil columns, which was twice the total amount of snow addition in the light snow cover treatment. Based on the average distribution of annual snowfall in autumn, winter, and spring in this study area, the total amount of snow cover in autumn and spring was assumed to be the same as in winter, and snow was added evenly twice in autumn. Based on the variations of soil temperature and air temperature in the field over the year (Figure S2), long-term incubation experiments were conducted by using cryogenic incubators (LRH-250CA, Shanghai Yiheng, Shanghai, China) to reasonably set up the changes in simulated temperature levels at four different seasons of the year (

Table 2. Temperature setting at the four stages of the incubation.

Simulation Stage	Autumn Freeze–Thaw		Winter Freeze	Spring Freeze–Thaw		Growing Season
Duration (day)	7	7	180	7	7	61
Daily freeze–thaw alternation time (hour)	12/12		24 (a)	12/12		24 (a)
Temperature setting (°C)	−5/10	−10/5	−10	−10/5	−5/10	15

^(a) 24 h each day at a constant temperature.

). During the whole incubation, the outer surface of all UPVC pipes containing soil columns was wrapped with insulation materials to simulate freeze–thaw fluctuations in the field as closely as possible.

2.3. Measurements of CO₂ Emissions from Soil Columns and Their δ¹³C Values

During the whole incubation, a CO₂ isotope spectrum analyzer (CCIA-38d-EP, Los Research Inc., Fremont, CA, USA) along with its supporting fully automated soil respiration analysis system (SF-3000, Beijing LICA United Technology Limited, Beijing, China) was nicely combined with cryogenic incubators (LRH-250CA, Shanghai Yiheng, Shanghai, China), to determine CO₂ emissions from large soil columns and the ¹³C/¹²C isotope ratios of soil-respired CO₂ in the chamber headspace (Figure S3). The sampling frequency was 1 Hz for the measurement of CO₂ concentration and its ¹³C/¹²C ratio in the headspace, and the output results of the stable isotopic ratios were based on Pee Dee Belemnite (PDB) as the standard, expressed by δ¹³C:

δ¹³C (‰) = (R_sample/R_standard − 1) × 100

(1)

where R_sample and R_standard represent the molar ratios of sample and standard ¹³C/¹²C, respectively.

Both CO₂ concentrations and the ¹³C/¹²C isotope ratios in CO₂ were determined using the δ¹³C-CO₂ automatic analysis system (Figure S3), with a precision of 0.05 μmol mol⁻¹ and 0.1‰ for a 5-min integration time, respectively. The δ¹³C values of CO₂ produced via microbial respiration in soil columns were calculated based on the Keeling plot approach [47,48], namely using linear regressions of the δ¹³C values of soil-respired CO₂ against the reciprocal of headspace CO₂ concentrations at second scales during the closure ranging from 90 to 200 s as mentioned afterward. The closure time as short as possible (e.g., within several minutes) can minimize the disturbance of the soil-atmosphere system and concentration dependence of δ¹³C measurement of headspace CO₂ with high precision [49,50,51]. The effect of water vapor on the δ¹³C measurement was nicely eliminated using a dehumidification device attached to the CO₂ isotope spectrum analyzer (CCIA-38d-EP, Los Research Inc., Fremont, CA, USA). Therefore, the δ¹³C-CO₂ automatic analysis system with high-frequency measurements can be nicely combined with laboratory incubation experiments to investigate the changes in soil CO₂ emissions and their δ¹³C values using large soil columns (diameter = 20 cm) (Figure S3). Prior to each assay, the CO₂ isotope analyzer was calibrated using two standard CO₂ mixture gases containing two different CO₂ concentrations (CO₂ concentrations and its δ¹³C values: 398.60 μL L⁻¹ and −22.31‰; 794.30 μL L⁻¹ and −22.34‰), 1% Argon (w/w), and balance gas of air. The standard CO₂ mixture gases were supplied by Air Products (Beijing Helium Pu North Branch Gas Industry Ltd., Beijing, China), and their δ¹³C and δ¹⁸O values were, respectively, measured with a mass spectrometer combined with a Trace Gas System (Isoprime-100, Elementar Ltd., Langenselbold, Germany), using reference materials NBS 19 and IAEA-CO-8 by the Environmental Stable Isotope Laboratory, Chinese Academy of Agricultural Sciences. During the whole incubation, CO₂ emissions from soil columns in all treatments were also determined using a portable greenhouse gas analyzer (915-0011, Los Research Inc., Fremont, CA, USA) nicely coupled by a smart respiratory chamber (SC-11, Beijing LICA United Technology Limited, Beijing, China) (Figure S3). Due to sharp fluctuations of soil CO₂ emissions upon the simulated freeze–thaw processes, the frequency of CO₂ emission determination was six times each day during the periods of simulated autumn and spring freeze–thaw, respectively (Table 2). Within the period of winter freeze simulation (Table 2), soil CO₂ emissions in all treatments were measured only once every 60 days. During the period of growing season simulation (Table 2), soil CO₂ emissions in all treatments were measured initially five times each day, and then measured once every 5–7 days as it became gradually stable. Upon the determination of CO₂ emissions, the duration of the closure varied from 90 to 200 s to ensure the measurement accuracy as much as possible, which was dependent on the size of soil microbial respiration at the different stages of the incubation. Throughout the incubation, the method of weighing and adding water was used at intervals to compensate for the evaporation loss of water in soil columns. During the non-freezing period of soil columns, soil volumetric water contents (v/v, %) at 5 cm depth inside soil columns in all treatments were determined at regular intervals using a portable MPkit (ICT, Sydney, NSW, Australia).

2.4. Statistical Analysis

We calculated CO₂ emission per unit area (m⁻²) merely concerning the surface area of soil columns. The δ¹³C values of soil-respired CO₂ in all treatments during the whole incubation were calculated according to the method of the Keeling plot [47,48]. Based on the two methods mentioned previously for measuring soil CO₂ emission (Figure S3), the results were almost the same and averaged at each sampling. Means and standard errors were calculated for selected variables at each sampling of the incubation based on three replicates. All variables were examined for homogeneity of variance using Levene’s test and normality using Shapiro–Wilk test and log-transformed as needed. The soil CO₂ emissions and δ¹³C values of soil-respired CO₂ in all treatments at each stage of the incubation were, on average, described with box plots using OriginPro 2021 (OriginLab Corporation, Northampton, MA, USA). Three-factor repeated analysis of variance (ANOVA) was used with forest type, snow cover, and the presence of litter to assess individual and interactive effects on cumulative soil CO₂ emissions at each stage of the incubation and throughout the incubation. Two-factor repeated ANOVA was used with snow cover and the presence of litter to assess individual and interactive effects on the average soil CO₂ emissions and average δ¹³C values of soil-respired CO₂ at each stage of the incubation. When ANOVA was significant, pairwise comparisons of means were examined using Duncan’s multiple-range test. The differences in soil volumetric water contents across all treatments, average soil CO₂ emissions, and average δ¹³C values of soil-respired CO₂ between the first and second week of the simulated autumn and spring freeze–thaw simulations (Table 2) were, respectively, assessed at a significant level of a = 0.05 using Student’s test. Least significant differences (LSDs) were calculated at a significant level of a = 0.05 to assess the differences between light and heavy snow cover-treated soil columns in soil CO₂ emissions and δ¹³C values in CO₂ at each sampling of the incubation. The relationships between soil CO₂ emissions, δ¹³C values of soil-respired CO₂, and soil moisture contents at 5 cm depth across soil columns were fitted with nonlinear and linear regressions. All statistical analyses were conducted using SPSS for Windows software (version 19.0) (IBM Corp., New York, NY, USA).

3. Results

3.1. Changes in Soil Moisture at the Different Stages of the Incubation

Prior to the incubation, volumetric water contents at 5 cm depth in all BKPF soil columns were, on average, 26.0 ± 1.0%, which was much larger than in all WBF soil columns (16.0 ± 1.0%). However, the spatial heterogeneity in initial volumetric water contents across all soil columns under each forest stand was negligible. Following the first addition of artificial snow, volumetric water contents of BKPF and WBF soil columns at 5 cm depth were, on average, increased by 29.0 ± 1.0% and 17.0 ± 1.0%, respectively (Figure 1), which were slightly larger than the background values as mentioned previously. Furthermore, following the second addition of artificial snow during the autumn freeze–thaw simulation, volumetric water contents of BKPF and WBF soil columns at 5 cm depth would increase substantially in the heavy snow cover treatment, especially in the absence of litter, and were mostly significantly larger than in the light snow cover treatment (p < 0.05) (Figure 1). Afterward, all soil columns treated with two different snow cover levels remained to be frozen until the second week of the spring freeze–thaw simulation. During the early spring freeze–thaw simulation, the task of the fourth snow addition was completed (Table 1 and Figure 2), and the following soil moisture did not change much. Therefore, soil volumetric water contents measured during the simulated growing season were averaged to represent the final volumetric water contents of soil columns in all treatments. During the simulated growing season, volumetric water contents of BKPF and WBF soil columns at 5 cm depth were, on average, 65.0 ± 0.5% and 56.0 ± 1.0% in the heavy snow cover treatment, and 40.0 ± 1.0% and 29.0 ± 1.0% in the light snow cover treatment, respectively (Figure 1).

3.2. Effect of Snow Cover and the Presence of Litter on CO₂ Emissions from Soil Columns

Prior to the incubation, CO₂ emissions from BKPF and WBF soil columns (n = 12, each forest type) at 15 °C were, on average, 1.48 ± 0.06 and 1.52 ± 0.09 μmol CO₂ m⁻² s⁻¹, respectively. However, the soil CO₂ emissions across all soil columns varied little between the two forest stands and within a single stand. The results suggested that the spatial variability of background conditions was negligible in the subsequent analysis of soil CO₂ emissions in all treatments. During the autumn freeze–thaw simulation, CO₂ emissions from BKPF and WBF soil columns in all treatments showed a zigzag fluctuation with the freeze–thaw alterations (Figure 2), and with the decrease of freeze and thaw temperatures, the emissions had a decreasing trend (p < 0.001) (Figure 2 and Figure 3). The increased snow cover could increase the soil CO₂ emissions, especially at the early stage of the autumn freeze–thaw simulation (Figure 2 and Figure 3). This phenomenon resulted in relatively larger cumulative CO₂ emissions from soil columns with and without litter during the simulated autumn freeze–thaw, compared with those from the light snow cover treatment (p < 0.05) (

Table 3. Cumulative CO₂ emissions from forest soil columns during the different stages of the incubation and the results of ANOVA.

Forest Type (FT)	Litter Cover (L)	Levels of Snow Cover (S)	Simulation Stage
			Autumn Freeze–Thaw	Winter Freeze	Spring Freeze–Thaw	Growing Season	Total
			(g CO2-C m−2)
BKPF	Without	Light	4.84 ± 0.11	8.43 ± 0.24	22.15 ± 1.55	128.67 ± 3.77	164.08 ± 5.21
		Heavy	5.74 ± 0.26	10.44 ± 0.77	11.98 ± 0.49	139.46 ± 2.41	167.62 ± 2.30
	With	Light	6.51 ± 0.31	16.02 ± 0.47	25.08 ± 0.30	161.92 ± 2.08	209.53 ± 2.34
		Heavy	8.58 ± 0.79	20.42 ± 1.68	17.13 ± 1.19	197.17 ± 13.47	243.30 ± 16.82
WBF	Without	Light	5.71 ± 0.11	15.03 ± 0.73	24.01 ± 1.51	154.06 ± 3.69	198.81 ± 5.59
		Heavy	7.67 ± 0.14	19.73 ± 0.81	16.30 ± 0.47	182.48 ± 2.67	226.17 ± 3.69
	With	Light	6.55 ± 0.65	27.04 ± 1.65	23.99 ± 2.05	157.92 ± 9.07	215.51 ± 12.40
		Heavy	8.20 ± 0.54	30.57 ± 1.77	17.83 ± 0.39	203.71 ± 3.14	260.31 ± 5.76
Three−way factorial ANOVA with FT, L, and S as fixed factors (p-value)
FT			0.338	<0.001	0.396	0.064	0.025
L			0.032	<0.001	0.167	0.005	0.002
S			0.018	0.040	<0.001	0.004	0.033
FT × L			0.227	0.434	0.337	0.084	0.154
FT × S			0.801	0.786	0.531	0.442	0.469
L × S			0.733	0.856	0.576	0.259	0.326
FT × L × S			0.564	0.594	0.922	0.845	0.789

). The results of ANOVA showed that the presence of litter and increased snow cover could significantly increase soil CO₂ emissions, but there were no significant differences between the two forest stands and no interaction effects of snow cover and the presence of litter on the soil CO₂ emissions (Table 3).

During the simulated winter freeze period, CO₂ emissions from BKPF and WBF soil columns in all treatments remained at a low level, with a range from 0.03 to 0.17 μmol CO₂ m⁻² s⁻¹ (Figure 2 and Figure 3). Based on the results of ANOVA, cumulative soil CO₂ emissions during the winter freeze simulation were significantly affected by forest stands, the presence of litter, and snow cover (Table 3). During the spring freeze–thaw simulation, CO₂ emissions from BKPF and WBF soil columns in all treatments showed a zigzag fluctuation with the soil freeze–thaw alterations (Figure 2), and with the increase of freeze and thaw temperatures, the emissions increased substantially, especially under light snow cover conditions (p < 0.001) (Figure 2 and Figure 3). In contrast to that during the autumn freeze–thaw simulation, the increased snow cover significantly reduced CO₂ emissions from BKPF and WBF soil columns during the spring freeze–thaw simulation (Figure 2 and Figure 3). The decrease in the increased snow cover-induced cumulative CO₂ emissions appeared larger in BKPF soil columns than in WBF soil columns, but the difference was not significant (Table 3). During the late period of the spring freeze–thaw simulation, there were several flush emissions of CO₂ from soil columns treated with the light snow cover, regardless of forest stands and the presence of litter (Figure 2).

During the simulated growing season, soil columns treated with heavy snow cover completely melted, and cumulative soil CO₂ emissions were significantly larger than those upon light snow cover, especially in the presence of litter (p < 0.01) (Table 3). Throughout the incubation, there were relatively large increases in the increased snow cover-induced cumulative CO₂ emissions from WBF soil columns, and the emissions were significantly affected by forest stands, snow cover, and the presence of litter (Table 3). In the absence of litter, cumulative CO₂ emissions from WBF soil columns treated with two different snow cover levels became larger than those from BKPF soil columns at the different stages of the incubation and throughout the incubation, especially during the winter freeze and growing season simulations (p < 0.05). However, regardless of snow cover levels, there were no significant differences in the cumulative soil CO₂ emissions in the presence of litter between BKPF and WBF stands, except during the winter freeze simulation (Table 3). Taken together, the results indicated that the impacts of snow cover and the presence of litter on soil heterotrophic respiration in temperate forests would vary with forest types and different seasons of the year.

3.3. Effect of Snow Cover and the Presence of Litter on the δ¹³C Values of Soil-Respired CO₂

During the autumn freeze–thaw simulation, the δ¹³C values of soil-respired CO₂ fluctuated with the soil freeze–thaw alterations (Figure 4), and with the decrease of freeze and thaw temperatures, δ¹³C values of soil-respired CO₂ in all treatments decreased on average substantially (p < 0.001) (Figure 5). Furthermore, regardless of the presence of litter, the δ¹³C values of soil-respired CO₂ in the heavy snow cover treatment became, on average, less negative than those in the light snow cover treatment (p < 0.05), especially at the early stage of the autumn freeze–thaw simulation (p < 0.001) (Figure 5). However, the effect of the presence of litter on the average δ¹³C values of soil-respired CO₂ did not occur during the autumn freeze–thaw simulation (Figure 5). Together with the finding that the heavy snow cover treatment can significantly increase soil heterotrophic respiration (Figure 3), the results indicate that the increased snow cover may remarkably promote the decomposition of organic C in subsurface soils under temperate forest stands during the autumn freeze–thaw period under field conditions, regardless of the presence of litter.

In contrast to those during the autumn freeze–thaw simulation, δ¹³C values of soil-respired CO₂ in all treatments during the spring freeze–thaw simulation appeared to have a V-shaped fluctuating trend with the freeze–thaw alterations (Figure 4), and with the increase of freeze and thaw temperatures, there were on average less negative δ¹³C values of soil-respired CO₂ in all treatments (p < 0.001) (Figure 5). Furthermore, under heavy snow cover conditions, the δ¹³C values of the CO₂ released from BKPF and WBF soil columns were, on average more negative, especially at the early stage of the simulated spring freeze–thaw (p < 0.01) (Figure 5). This indicates that the presence of increased snow cover in winter may accelerate the decomposition of organic C in surface mineral soils and/or litter layers under temperate forests in the field during the spring freeze–thaw period. During the simulated growing season, δ¹³C values of soil-respired CO₂ in all treatments increased gradually, and then remained relatively less negative until the end of the incubation, compared with those at the other stages of the incubation (Figure 4). On average, the increased snow cover could result in less negative δ¹³C of CO₂ released from BKPF and WBF soil columns during the simulated growing season (p < 0.05), which was the same as shown during the simulated autumn freeze–thaw (Figure 5). Throughout the incubation, there was a shift of about 4.2‰ ¹³C-enrichment on average in the simulated growing season versus non-growing season soil respiration measured on all soil columns (Figure 5).

With the development of litter decomposition, the effect of litter on the δ¹³C values of soil-respired CO₂ would vary with forest stands and different stages of the incubation. The average δ¹³C values of soil-respired CO₂ in WBF soil columns were significantly affected by the presence of litter during the winter freeze simulation and at the early stage of the spring freeze–thaw simulation, respectively (Figure 5b). However, the δ¹³C values of the CO₂ emitted from BKPF soil columns were, on average, significantly affected by the presence of litter at the early stage of the spring freeze–thaw simulation and during the simulated growing season, respectively (Figure 5a).

3.4. Relationships among Soil CO₂ Emissions, δ¹³C Values of Soil-Respired CO₂, and Soil Moisture

Regardless of snow cover and the presence of litter, there was a nice nonlinear relationship between soil CO₂ emissions and δ¹³C values of soil-respired CO₂ throughout the incubation (Figure 6). Under experimental conditions, the δ¹³C values of soil-respired CO₂ had a relatively large variability when the soil CO₂ emissions were less than 1.0 mmol CO₂ m⁻² s⁻¹ (Figure 6). The addition of different snow covers would result in a large range of volumetric water contents in BKPF and WBF soil columns at 5 cm depth during the different stages of the incubation (Figure 1). During the whole incubation, CO₂ emissions from soil columns and their corresponding δ¹³C values increased with the soil moisture within the range from 14.2% to 65.2% (v/v), regardless of forest types and the presence of litter (Figure 7). Under experimental conditions, soil moisture at 5 cm depth could explain the 12% and 33% variability of the δ¹³C values of soil-respired CO₂ in the absence and presence of litter, respectively (Figure 7b).

4. Discussion

4.1. Effects of Snow Cover on Soil CO₂ Emissions Vary with the Presence of Litter and Forest Stands

As the simulated autumn freeze–thaw proceeded, CO₂ emission rates of BKPF and WBF soil columns decreased on average substantially due to a decrease in freeze-and-thaw temperatures (Figure 3), and cumulative soil CO₂ emissions significantly varied with levels of snow cover and the presence of litter (Table 3). During the autumn freeze–thaw simulation, heavy snow cover led to an increase in volumetric soil water contents due to the initial melting of artificial snow (Figure 1) and the incomplete freezing status of soil columns prior to the winter freeze. This would increase CO₂ emissions from soil columns upon the addition of heavy snow cover, especially during the early autumn freeze–thaw simulation (Figure 2 and Figure 3), hence enhancing cumulative soil CO₂ emissions (Table 3). Additionally, changes in freeze and thaw temperatures and following microbial activity, as well as the supply of labile dissolved organic C (DOC) (e.g., carbohydrates) initially released into the soil upon melting of artificial snow, could promote soil CO₂ emissions during the early autumn freeze–thaw simulation (Figure 3). The soil labile C pools released initially upon artificial snow cover can be partly supported by the results reported by Zhu et al. [52], who documented that the release of only O-alkyl C (e.g., carbohydrates) from fir and birch litter was increased by the presence of snow cover.

Based on the differences between CO₂ emissions from soil columns without and with litter, litter-derived CO₂ emissions following two levels of snow cover were coarsely calculated. During the autumn freeze–thaw simulation, the litter-derived CO₂ emissions from BKPF soil columns became, on average greater than those from WBF soil columns, especially upon heavy snow cover (Table 3). In this study area, the BKPF stand lies in a climax community of forest succession, with a greater SOM content and lower bulk density than the adjacent secondary WBF stand [19]. Furthermore, the thickness of litter layers under the BKPF stand is normally greater than that under the WBF stand due to the relatively slow degradable needles of Pinus koraiensis trees and different litter leaf accumulations of tree species [38,40]. The different properties of litter layers under the two forest stands and their impacts on the soils upon initial melting of artificial snow would explain the relatively great litter-induced CO₂ emissions from BKPF soil columns during the autumn freeze–thaw simulation (Table 3). The relatively high soil C availability (e.g., the microbial biomass C divided by soil organic C) under WBF stand [19] would explain larger cumulative CO₂ emissions from WBF soil columns without litter and with two levels of snow cover during the different stages of the incubation and throughout the incubation, compared with those from BKPF soil columns (Table 3). However, except during the winter freeze simulation, no significant differences in the cumulative CO₂ emissions were observed in the two forest soil columns with litter. In the presence of litter, the relatively great cumulative CO₂ emissions from WBF soil columns during the winter freeze simulation (Table 3) were partially attributed to the relatively great DOC concentrations and microbial degradable C pools (e.g., glucose C) in the water extracts of organic layers under WBF stand in autumn than under BKPF stand [42].

On the contrary to that, during the autumn freeze–thaw simulation, the reduced snow cover during the spring freeze–thaw simulation could increase soil CO₂ emission (Figure 2). This can result from more freeze–thaw cycles and greater soil temperature fluctuations under relatively low soil moisture conditions [53]. Following long-term winter freeze simulation, frequent soil freeze–thaw cycles could increase the contents of DOC released into the soil at thaw [22,24,29], especially at the late stage of the spring freeze–thaw simulation (Figure 2). The melting of artificial spring snow and following increased soil moisture would promote the activity of soil microorganisms and an associated flush of nutrients [22,54,55,56,57], which can, in turn, increase the soil CO₂ emissions during the spring freeze–thaw simulation, especially when the soil water content is below optimum [24,58]. This can explain several substantial flush CO₂ emissions from soil columns treated with light snow cover during the late spring freeze–thaw simulation, regardless of the presence of litter and forest stands (Figure 2). As the artificial snow cover increased, the complete melting of soil columns was detarded with the simulated spring freeze–thaw fluctuations. These phenomena resulted in relatively small cumulative CO₂ emissions from heavy snow cover-treated soil columns during the spring freeze–thaw simulation, compared with those from light snow cover-treated soil columns (Table 3). Since the incubation entered the simulated growing season, soil columns treated with heavy snow cover showed relatively larger CO₂ emissions compared with soil columns treated with light snow cover, especially in the presence of litter (Figure 2 and Figure 3, Table 3). Furthermore, the increased snow cover-induced cumulative CO₂ emissions from WBF soil columns with and without litter were larger than those from BKPF soil columns during the simulated growing season (Table 3). Taken together, the results indicate that the impacts of snow cover on CO₂ emissions from forest soils can vary with forest stands and the presence of aboveground litter as well as different seasons of the year.

4.2. Changes in the δ¹³C Values of Soil-Respired CO₂ and Influencing Mechanisms

The δ¹³C values of soil-respired CO₂ in all treatments became, on average, less negative during the first week than during the second week of the autumn freeze–thaw simulation (Figure 5). This was probably ascribed to an initial supply of DOC (e.g., carbohydrates) containing less depleted ¹³C in the soils upon the melting of artificial snow relative to the δ¹³C of soil organic C [59]. At the start of the simulated autumn freeze–thaw, there was an increase in labile C pools (e.g., carbohydrates) released into the soil [19,21], especially under high soil moisture conditions as caused by the initial melting of heavy artificial snow. The selective degradation of carbohydrates by microorganisms in our microcosm incubation was thus most likely to explain relatively less negative δ¹³C of the CO₂ respired initially during the autumn freeze–thaw simulation [60] (Figure 5). While the autumn freeze–thaw simulation proceeded, labile, isotopically heavier C compounds (e.g., carbohydrates) in soils could be preferentially and rapidly decomposed by microorganisms, and then the relatively refractory organic matter (e.g., lignin) with relatively low δ¹³C was likely to be selectively decomposed by microorganisms surviving in soils under freeze–thaw conditions [61]. Based on the measurements of δ¹³C values and specific ultraviolet absorbance at 254 nm (SUVA₂₅₄) of DOC collected from the intact organic horizons across the latitudinal transect over the year, the δ¹³C values of DOC would decrease on average with the increase in the SUVA₂₅₄ values [59]. Kalbitz et al. [62] also reported that greater SUVA₂₅₄ values of soil-dissolved organic matter in forest floors were associated with more negative δ¹³C values. The index of SUVA₂₅₄ values normally reflects the aromaticity and bioavailability of DOC in soil extracts and leachates [63], and the SUVA₂₅₄ values can increase following the simulated freeze–thaw fluctuations mainly due to microbial decomposition of labile DOC pools [19,21]. These could be responsible for the relatively lower δ¹³C values of soil-respired CO₂ in all treatments at the late stage of the autumn freeze–thaw simulation. Based on relatively larger CO₂ emissions and less negative δ¹³C of soil-respired CO₂ upon heavy snow cover than upon light snow cover (Figure 4 and Figure 5) and the fractionation variations of δ¹³C in SOM occur at different soil layers [37], it can be thus reasonably concluded that the presence of increased snow cover prior to winter freeze can increase the decomposition of organic C in subsurface soils under temperate forests in late autumn annually, which deserves further study in the future under field conditions to explore the magnitude of the soil C flush emission and the critical intrinsic influencing mechanisms involved.

Our results indicate a relatively great range of δ¹³C values of soil-respired CO₂ across all forest soil columns when soil CO₂ emissions are less than 1.0 mmol CO₂ m⁻² s⁻¹ (Figure 6). Furthermore, during the simulated autumn and spring freeze–thaw as well as winter freeze stages, δ¹³C values of soil-respired CO₂ at each sampling were mostly characterized by relatively large standard errors (Figure 4). The large uncertainty of δ¹³C values of soil-respired CO₂ during the non-growing season simulation corresponded with the results reported by Bowling et al. [64], who reported that the δ¹³C values of nocturnal whole-forest respiration in non-growing season ranged from −15‰ to −40‰ and varied more greatly than in growing season in a subalpine forest of Niwot Ridge, USA. This large uncertainty in the δ¹³C-CO₂ values would result from a small range in CO₂ concentrations and δ¹³C values in non-growing seasons upon using the Keeling plot approach [65] and the effects of diffusive transport-induced fractionation under non-steady state conditions caused by the freeze–thaw simulations [16,51].

The δ¹³C values of soil-respired CO₂ in all treatments became, on average, less negative as the simulated spring freeze–thaw proceeded (p < 0.001) (Figure 5), which was contrary to the variations of the δ¹³C values during the autumn freeze–thaw simulation. Furthermore, there were, on average, relatively more negative δ¹³C of soil-respired CO₂ upon heavy snow cover than upon light snow cover during the spring freeze–thaw simulation, which differed from the relatively less negative δ¹³C of soil-respired CO₂ upon heavy snow cover during the autumn freeze–thaw and growing season simulations, respectively (Figure 5). These phenomena could result in a shift of about 4.2‰ ¹³C-enrichment on average in the simulated growing season versus non-growing season soil respiration measured on all soil columns during the whole incubation (Figure 5). This shift is well supported by the results reported by Moyes et al. [35], who documented that a seasonal pattern of about 4‰ ¹³C-enrichment in summer versus spring and fall soil respiration under deciduous trees can be attributed to the seasonal variability of soil heterotrophic respiration.

As the autumn and spring freeze–thaw simulation proceeded, the δ¹³C values of soil-respired CO₂ in all treatments became, on average, more negative at low temperatures (−10/5 °C) than at high temperatures (−5/10 °C), regardless of snow cover and the presence of litter (Figure 5), which indicates temperature-dependent changes in the δ¹³C values of soil-respired CO₂ under experimental conditions. This would result from the microbial-mediated decomposition of different C compounds available in the soils at the different stages of the incubation, which is likely related to the structure of the soil microbial community, as proposed by Andrews et al. [66]. Under frequent freeze–thaw disturbances caused by changes in snow cover, soil fungi, which are normally considered K-strategic microorganisms, can have a better performance than bacteria in substrate utilization in mid-latitude systems [4,67,68], and fungal biomass typically has lower δ¹³C values than bacterial biomass [69,70]. Normally, K-strategic microorganisms (mainly fungi) can degrade recalcitrant organic matter, whereas r-strategic microorganisms (mainly bacteria) depend on easily degradable organic matter and sufficient nutrients [71]. Furthermore, organic carbon compounds with different biodegradability contain different C isotopic compositions, with a decrease in the δ¹³C values from labile to recalcitrant organic matters [61]. At the early stage of soil freeze–thaw in autumn, r-strategic microorganisms can preferentially utilize labile C pools (e.g., carbohydrates) with more enrichment in ¹³C, resulting in relatively less negative δ¹³C of soil-respired CO₂ during the first week than during the second week of the autumn freeze–thaw simulation (Figure 5). As the autumn freeze–thaw proceeded or after a long-term severe winter freeze, soil microbial biomass decreased, and the structure of the microbial community would have a shift towards fungi-dominated microorganisms [29,67,68], due to less sensitivity of fungi than bacteria following soil freeze–thaw or freeze processes, especially under cold winter conditions [68,72,73]. Therefore, microorganisms surviving after a severe winter freeze can release extracellular enzymes to degrade recalcitrant organic C compounds [21,74]. Probably, a shift from more labile to more recalcitrant respiratory substances, or changes in soil microbial biomass and microbial community after long-term severe winter freeze can lead to more negative δ¹³C of the CO₂ respired initially during the spring freeze–thaw simulation (Figure 5). The temperature-dependent changes in the δ¹³C values of soil-respired CO₂ in this study differed obviously from the results reported in previous studies without freeze–thaw disturbances [66,75]. During the 98-day laboratory incubation using forest soils, Gauthier et al. [75] observed a linear increase in δ¹³C values of soil-respired CO₂ with decreasing temperature from 28 °C to 8 °C. Andrews et al. [66] also observed a larger increase in the δ¹³C values of soil-respired CO₂ at 4 °C than at 22 °C, which would be related to changes in the soil microbial community composition. Taken together, the results indicate that changes in the amount and quality of organic C compounds released into the soil upon freeze–thaw disturbances [19,21,76], soil aggregate distribution [76], and microbial properties such as the structure of the microbial community and microbial activity as well as soil enzyme activity [21,23,24,57,68,74,77] can most likely influence the temperature-dependent changes in the δ¹³C values of soil-respired CO₂.

The δ¹³C values of soil microbial respiration can reflect the decomposition of SOM and plant substrates and have been reported to vary greatly with environmental conditions [78,79]. Our results show that the δ¹³C values of soil-respired CO₂ under experimental conditions can increase with the increase in volumetric soil water contents at 5 cm depth within a range from 14.2% to 65.2% (v/v) (Figure 7b). This is different from the results reported by Barbour et al. [79], who observed a negative relationship between the δ¹³C values of soil-respired CO₂ and volumetric soil water content at a pasture site in New Zealand during nearly 10-day measurement. In a temperate desert steppe, the δ¹³C values of soil-respired CO₂ significantly decreased with increasing soil moisture within the range from 2.0% to 28.0% (v/v) (determination coefficient of regression, R² = 0.24) [80]. However, Phillips et al. reported that there were no changes in the δ¹³C values of heterotrophic soil-respired CO₂ with a large soil moisture range [81]. The differences would result from different site-specific experimental conditions and other soil properties. Furthermore, 12% and 33% variations in the δ¹³C values of soil-respired CO₂ in this study were, respectively, explained by volumetric soil water contents, in the absence and presence of litter (Figure 7b). This result indicates that in addition to soil moisture, other soil properties such as different organic C sources, microbial biomass, and microbial community may affect the δ¹³C of soil heterotrophic respiration under experimental conditions, which deserves further study in the future under field conditions.

5. Conclusions and Future Perspectives

The set of novel analysis systems was well used to study the changes in heterotrophic respiration of undisturbed large soil columns and their corresponding δ¹³C values under laboratory conditions. Based on laboratory simulation incubation experiments of four different seasons over the year, the effects of artificial snow cover on forest soil heterotrophic respiration and its δ¹³C values could vary with different seasons of the year, the presence of litter, and forest types. During the autumn freeze–thaw simulation, there were relatively larger CO₂ emissions and less negative δ¹³C values of soil-respired CO₂ upon heavy snow cover than upon light snow cover, indicating that the presence of increased snow cover prior to winter freeze may increase the decomposition of organic C in subsurface soils under temperate forests. The δ¹³C values of soil-respired CO₂ in all treatments became, on average, less negative as the simulated spring freeze–thaw proceeded, which was contrary to the variations of the δ¹³C values observed during the autumn freeze–thaw simulation. The δ¹³C values of soil-respired CO₂ during the simulated spring freeze–thaw were, on average, more negative upon heavy snow cover than upon light snow cover, which was contrary to the variations of the δ¹³C values observed during the simulated autumn freeze–thaw and growing season periods, respectively. Regardless of snow cover and the presence of litter, a shift of about 4.2‰ ¹³C-enrichment on average in the simulated growing season versus non-growing season soil heterotrophic respiration was observed on all large soil columns, and it would be attributed to changes in soil moisture, amount and quality of organic C compounds released into the soil, and soil microbial properties under experimental conditions. Normally, seasonal changes in δ¹³C values of soil heterotrophic and autotrophic respiration in terrestrial ecosystems can, to some extent, reflect the responses of SOM decomposition to environmental conditions [16,35,78,79]. The results of this study highlight the impacts of winter snow cover and the presence of litter on soil respiration and its δ¹³C values under temperate forests over the year, which deserves further study in the future under field conditions to explore the critical intrinsic influencing mechanisms, by moderately considering the functions of soil physicochemical and microbial properties as well as priming effects caused by fine root biomass in regulating the δ¹³C of soil respiration and soil C dynamics [11,16,36,61,76,77,82].