Environmental Effects on Carbon Isotope Discrimination from Assimilation to Respiration in a Coniferous and Broad-Leaved Mixed Forest of Northeast China

Abstract

Carbon (C) isotope discrimination during photosynthetic CO₂ assimilation has been extensively studied, but the whole process of fractionation from leaf to soil has been less well investigated. In the present study, we investigated the δ¹³C signature along the C transfer pathway from air to soil in a coniferous and broad-leaved mixed forest in northeast China and examined the relationship between δ¹³C of respiratory fluxes (leaf, trunk, soil, and the entire ecosystem) and environmental factors over a full growing season. This study found that the δ¹³C signal of CO₂ from canopy air was strongly imprinted in the organic and respiratory pools throughout C transfer due to the effects of discrimination and isotopic mixing on C assimilation, allocation, and respiration processes. A significant difference in isotopic patterns was found between conifer and broadleaf species in terms of seasonal variations in leaf organic matter. This study also found that δ¹³C in trunk respiration, compared with that in leaf and soil respiration, was more sensitive to seasonal variations of environmental factors, especially soil temperature and soil moisture. Variation in the δ¹³C of ecosystem respiration was correlated with air temperature with no time lag and correlated with soil temperature and vapor pressure deficit with a lag time of 10 days, but this correlation was relatively weak, indicating a delayed linkage between above- and belowground processes. The isotopic linkage might be confounded by variations in atmospheric aerodynamic and soil diffusion conditions. These results will help with understanding species differences in isotopic patterns and promoting the incorporation of more influencing factors related to isotopic variation into process-based ecosystem models.

1. Introduction

Stable carbon (C) isotope measurement is a useful tool for disentangling C cycling processes in forest ecosystems, as a unique C isotopic signature is imprinted along the C transfer pathway from CO₂ in the atmosphere to C in different ecosystem pools, which is affected by isotopic discrimination (Δ). During photosynthesis, C₃ plants discriminate heavily against ¹³C through photosynthetic discrimination; the resulting C isotopic signature (δ¹³C) in the photosynthate is more negative compared to the δ¹³C of atmospheric CO₂ (δ¹³C_air) and is linearly dependent on the ratio of the intercellular and atmospheric CO₂ concentration (c_i/c_a) [1,2]. After photosynthesis, plants rapidly release photo-assimilated C into the soil, and the post-photosynthetic discrimination effects become involved in a number of biochemical processes alongside mixing and exchange with other C storages, which leads to different δ¹³C levels among woody plant parts and soil [3,4]. Subsequently, respiration-associated discrimination and the consumption of substrates with different δ¹³C values further modify the δ¹³C of respiratory fluxes, resulting in apparent differences between the δ¹³C of respired CO₂ (δ¹³C_R) and the respiration substrates [5,6,7,8]. The accurate measurement of δ¹³C in different transfer pathways is important to explore the complexity of the C cycle under the changing climate.

Moreover, the discrimination processes encode biochemical and physiological information into δ¹³C, and this information is transferred through different stages of the C cycle. Thus, thoroughly investigating δ¹³C in each C pool and respiratory flux can help evaluate the coupling between each of the processes in the pathways of C cycling. For example, strong evidence of the linkage between plant photosynthesis and soil respiration was found by Kuzyakov and Gavrichkova [9], indicating that the δ¹³C of soil-respired CO₂ may also be related to photosynthetic discrimination via root respiration or exudation. Ecosystem respiration combines the CO₂ emitted from leaves, trunks, and soil within the ecosystem, and it also integrates their C isotopic signals. Thus, the δ¹³C in ecosystem respiration (δ¹³C_eco) can vary substantially with changes in the relative contribution of the respiration of each component.

Discrimination also encodes information about the dynamics of environmental drivers into δ¹³C. Seasonal variation in environmental factors, such as vapor pressure deficit (VPD) and soil moisture, influence stomatal conductance, which then alters photosynthetic discrimination and further leads to seasonal dynamics in the δ¹³C of CO₂ respired from leaves [7]. Due to the linkages between the photosynthetic metabolites and the substrates of different respiratory fluxes, several studies also found that changes in environmental drivers may be responsible for seasonal dynamics in the δ¹³C_R of trunks, roots, and soil [10,11,12]. Accordingly, the δ¹³C of ecosystem respiration, which integrates all respiratory components of the ecosystem, was also found to have significant short-term and seasonal variations [13,14,15], which have been directly or implicitly linked to environmental factors with a time lag. This indicates either a fast linkage between assimilates and ecosystem respiration or a linkage delayed by the time in which the assimilates are transported to the sites of respiration [10,16].

However, a significant response of δ¹³C_eco to environmental factors does not always exist. For example, McDowell et al. [13] found only a weak relationship between canopy conductance and δ¹³C_eco over a 2-week period, which they attributed to noncanopy controls, such as belowground respiration on δ¹³C_eco. Similarly, the results of long-term measurements with high time resolution showed that δ¹³C_eco was not tightly coupled with environmental factors in some of the study periods [17]. These studies collectively indicate that the response of δ¹³C_eco to environmental changes is highly variable and that the influence of complicated environmental conditions, combined with other factors, may weaken the linkage between assimilates and ecosystem respiration. Thus, there is a practical need to explore what factors control δ¹³C_eco and to what extent δ¹³C_eco reveals the links between C cycling processes.

Photosynthetic discrimination can be calculated at the leaf level by determining the difference between the δ¹³C of leaf organic matter and δ¹³C_air [1], or it can be measured online using branch chambers [17]. Alternatively, photosynthetic discrimination can be estimated using the comprehensive Farquhar’s model at a canopy scale by combining δ¹³C and CO₂ flux measurements [18,19]. On the other hand, respiration-associated discrimination can be evaluated by chamber measurements [20,21] using the Keeling plot approach [22,23]. This approach is based on the theory of the two-component gas-mixing model, where δ¹³C_R emerges as the y-intercept of the mixing relationship between δ¹³C and the inverse of the CO₂ concentration (CO₂) [24,25]. The Keeling plot approach can also determine the δ¹³C_eco using nighttime CO₂ and δ¹³C across the vertical profile of the forest [14,17]. Recently, optical laser spectroscopy techniques have allowed in situ continuous monitoring of CO₂ isotopes within open forest canopies or closed gas exchange chambers, thus tracing the high-resolved dynamics in δ¹³C_air, δ¹³C_R, and δ¹³C_eco over more weather conditions [12,17]. However, few studies have simultaneously measured δ¹³C_air, δ¹³C_R, and δ¹³C_eco, which represents a limitation of our knowledge on discrimination in the process of C exchange.

Coniferous and broad-leaved mixed forests are widely distributed in northeast China. These forests are well-protected old-growth virgin forests characterized by rich species diversity as well as high biomass. Previous studies have shown that coniferous and broad-leaved mixed forests are strong C sinks and have experienced climate warming over the past 50 years [26]. To our best knowledge, variations of δ¹³C in forest storage C pools and respiratory fluxes have mainly been studied in pure forests [17,27,28,29], while mixed forests with multiple tree species have rarely been covered by previous studies [12]. In the present study, we continuously investigated in situ seasonal variations in the δ¹³C_air and δ¹³C_R of leaves, trunks, and soil as well as δ¹³C_eco in a 200-year-old coniferous and broad-leaved mixed forest over a full growing season. We also explored the Δ variation of the dominant species in this mixed forest. Firstly, we aimed to identify species-specific differences in photosynthetic Δ and variations in the Δ subsequent photosynthesis in this mixed forest. Secondly, we assessed the environmental effects on seasonal variations in δ¹³C for CO₂ respired by leaves, trunks, and soil as well as the ecosystem. Finally, we tested whether δ¹³C_eco reveals links between aboveground and belowground processes.

2. Materials and Methods

2.1. Site Description

The study was carried out in a coniferous and broad-leaved mixed forest site (42° 24.149′ N, 128° 05.768′ E, 732 m elevation) in the Changbai mountain natural reserve, Jilin, northeast China. The site is pristine and undisturbed by anthropic activities and characterized by richness of species in both its overstorey and understorey. The density of the trees is about 560 trees ha⁻¹ (stem diameter > 8 cm). Dominant species include one evergreen needle species, Korean pine (Pinus koraiensis), and four deciduous broadleaf species, including Manchurian ash (Fraxinus mandshurica), Mono maple (Acer mono), Mongolian oak (Quercus mongolica), and Tuan linden (Tilia amurensis). P. koraiensis, F. mandshurica, and T. amurensis account for about 40%, 30%, and 20% of the total tree number, respectively, while Q. mongolica and A. mono account for the remaining 10%. The average age of the dominant species has been estimated to be over 200 years, and the mean canopy height is about 26 m. The growing season of the vegetation extends from May to September. The site’s topography is nearly flat. The soil is classified as dark brown forest soil with a depth of 60–100 cm. The climate is temperate, with a mean annual temperature of 3.6 °C and annual precipitation of 695.3 mm measured between 1982 and 2003; over 80% of the rainfall occurs within the growing season.

2.2. Eddy Covariance and Meteorological Measurements

A 62 m high meteorological tower is located at the study site. Temperature and relative humidity probes (HMP155A, Vaisala, Helsinki, Finland) and anemometers (WindSonic1, Campbell Scientific, Logan, UT, USA) were installed at seven heights (2.5, 8, 22, 26, 32, 50, and 60 m) on the tower. A pyranometer (CMP22, Kipp&Zonen, Delft, Netherlands) was installed at 40 m on the tower. Rainfall was measured at 62 m using a rain gauge (TE525MM, Texas Electronics, Dallas, TX, USA). Soil temperature and moisture were measured at a 5 cm depth using a thermocouple burial probe (105E-L, Campbell Scientific, Logan, UT, USA) and a water content reflectometer (CS616, Campbell Scientific, Logan, UT, USA), respectively. All of the meteorological data were recorded via a CR1000X data logger (Campbell Scientific, Logan, UT, USA) at 30 min intervals.

An eddy covariance system was installed on the tower at 40 m above the ground. The system consisted of a three-axis anemometer (WindMaster Pro, Gill Instruments, Lymington, UK) and an open-path CO₂/H₂O infrared gas analyzer (Li-7500DS, LI-COR, Lincoln, NE, USA). Raw data were collected by a SmartFlux 3 System (LI-COR, Lincoln, NE, USA).

2.3. Canopy CO₂ Concentration and Carbon Isotope Composition Profile

In this study, the canopy CO₂ and δ¹³C_air were measured using a wavelength-scanned cavity ring down spectroscopy (WS-CRDS) analyzer (G2201-i, Picarro Inc., Santa Clara, CA, USA). This instrument performed well in measuring the δ¹³C of CO₂ with a 5 min precision of 0.11‰ and a maximum drift of 0.25‰ day⁻¹. Data were collected at about 1 Hz and expressed on the VPDB (Vienna Pee Dee Belemnite) scale.

A profile control system (PRI-8500, PRI-ECO Inc., Beijing, China) was used as a peripheral of the WS-CRDS analyzer to automatically change ports in the δ¹³C profile measurements (Figure 1a). Air inlets were installed at the seven heights on the tower. Different levels of air were simultaneously pumped from the inlets by a vacuum pump (R410, Rocker Scientific Co., Ltd., Taiwan, China) through 1/4-inch Teflon tubes followed by flow control valves (ASC200-08, AirTAC, Taiwan, China). Each of the inlets was selected and then scanned for 5 min one after another from lower to higher levels through a 16-port rotation valve (Valco Instruments Co. Inc., Houston, TX, USA). The air was pumped into the WS-CRDS analyzer at a flow rate of 30 mL/min using a Picarro external pump. The first minute of each measurement was discarded to remove the data of the response time after the port was switched.

Two standard gases (Std1: 370.7 ppm for CO₂ and −20.8709‰ for δ¹³C; Std2: 799.0 ppm for CO₂ and −21.5685‰ for δ¹³C) were scanned for 15 min once a day to conduct two-point gain and offset calibrations for CO₂ measurements and single-point delta value offset calibrations for δ¹³C measurements [30]. The dependence of δ¹³C on CO₂ for G2201-i was 0.09‰ per 100 ppm in the range of 368.1–550.1 ppm, as tested by Pang et al. [31]. Thus, it can be presumed that the CO₂ dependence of δ¹³C is linear within the small CO₂ range (about 400–495 ppm for the air profile measurement and 400–600 ppm for the respiration measurements; see Section 2.4) in the present study. The first 5 min and last 1 min of each standard gas measurement were discarded to remove biased data introduced by port switching.

2.4. Carbon Isotope Composition of CO₂ Respired from the Leaves, Trunks, Soil, and Ecosystem

Five ports of the profile control system along with the respiration chambers were used to build a closed-circuit system under non-steady-state conditions [21,32] to conduct measurements of the δ¹³C in the leaf-, trunk-, and soil-respired CO₂ (δ¹³C_R) (Figure 1a). The outlet of each chamber was directly connected to the rotation valve in the profile control system and then connected to the inlet of the WS-CRDS analyzer. The outlet of the WS-CRDS analyzer, which is the outlet of the Picarro external pump, was divided into five sub-outlets by a 5-way air distribution manifold connected to the inlet of each chamber. All tubes were 1/8-inch Teflon tubes, thereby reducing the system volume and response time.

The leaf chambers were made from cylindrical PET (polyethylene terephthalate) jars (10 cm in diameter and 20 cm in height) (Figure 1b,c). The lid of each jar was fixed on the branch by letting the branch pass through the center of the lid. Mastic tape and silicone sealant were used to fill the space between the branch and the lid to make the leaf chamber airtight. The inlet and the outlet were on the sides and the bottoms of the jar bodies. No fans were installed inside the jars because of limited space.

The trunk chambers were made from cylindrical PP (polypropylene) airtight containers (18 cm in diameter and 9.5 cm in height) (Figure 1b,c). The bottoms of each container were removed, and the rest of the body was fixed onto the trunk using silicone sealant after gently sanding the bark. An inlet and an outlet were located on the lid of the container. A mini fan was installed inside the lid for air mixing.

The soil chamber was made from the same material and of the same size as the trunk chambers (Figure 1b,c). The body of the soil chamber was fixed on a soil collar (16 cm in diameter and 10 cm in height) using hot-melt adhesive. The soil collar was inserted into the soil to a depth of 5 cm. Here, only one soil chamber was used due to the limited number of sampling ports. The leaf chambers and the soil chamber were permanently covered by aluminum foil tape to isolate sunlight and maintain a steady temperature.

Two leaf chambers were installed on a branch of a P. koraiensis and an A. mono. The branches were mature and healthy with a diameter of about 1.5 cm and located at about 5 m in height on the sunny side of the trees. Two trunk chambers were installed on the sunny side of the trunks on a P. koraiensis and a F. mandshurica at breast height (1.3 m). The selected trees and the soil collar were about 5 m apart from each other, thereby sharing similar meteorological and environmental conditions. The chambers were installed one month before the first sampling. Each of the chambers included a silicon sealing ring on the inside of the lid to ensure airtightness. The airtightness of the respiration chamber was tested by connecting a CO₂/H₂O analyzer (LI-850, LI-COR, Lincoln, NE, USA) to a closed chamber (screwing the body back on the lid for leaf chambers or locking the lid with clamps on the body for trunk and soil chambers). If there was no obvious change in the measured values of CO₂ concentration in the chamber when blowing along the joints of the chamber, then the sealing performance was considered good.

The δ¹³C_R measurements were conducted from early May to late September of 2019. To begin measuring, the chamber was first closed after selecting an inlet and an outlet corresponding to the chamber by manually operating the rotation valve and the air distribution manifold. The measurement continued until the internal CO₂ concentration increased by about 200 ppm (about 20–40 min for the leaf and trunk chambers and about 5–10 min for the soil chamber), as suggested in [21], and then we switched to the next chamber. The chambers were closed only during measuring. We periodically closed the chambers in the following sequence: the P. koraiensis leaf chamber (L1), the A. mono leaf chamber (L2), the P. koraiensis trunk chamber (T1), the F. mandshurica trunk chamber (T2), and the soil chamber (S1). Four rounds of measurements were conducted from around 7:00 to 18:00 each day. The δ¹³C_R of each measurement was estimated based on the widely used Keeling plot approach [22,23]. The daily δ¹³C_R of the leaf, trunk, and soil was averaged from the four measurements each day.

The δ¹³C_eco was determined by the Keeling plot approach [22,23] using the nighttime (21:00–03:00) CO₂ and δ¹³C_air profile measurements. Considering that small a CO₂ range would cause uncertainties in fitting the Keeling plot [24], we set a restriction that only nights with a CO₂ range greater than 60 ppm were used to conduct the regression, and the ordinary least squares regression (OLS, model Ι) was used to obtain the intercept of regression, as suggested by Chen et al. [25].

2.5. Sampling and Carbon Isotope Analysis of the Ecosystem Compartments

Leaves of the five dominant species were collected in the mornings of sunny days once a month from May to September 2019 around the flux tower. Specifically, nine mature trees of each species were randomly selected on each sampling day; then, mature and healthy leaves were collected at three canopy heights (lower, middle, and upper). Since the trees are different in height and the canopy structure is different among species, the three canopy heights of each tree were determined relative to their canopy height range rather than using absolute heights. We collected bark and xylem at breast height and coarse and fine roots of the five species, as well as litters in the undecomposed layer (Oi) and decomposed layer (Oe + Oa) and the soil at four depths (0–5, 5–10, 10–20, and 20–40 cm) with nine replicates around the flux tower at the end of September 2019.

All samples were immediately transported to the lab and dried at 65 °C for 4 days to a constant weight and then ground with a ball mill (MM 400, Retsch, Haan, Germany). The C isotope compositions of nine sample replicates were determined separately using an elemental analyzer (Flash EA1112, Thermo Finnigan, Milan, Italy) coupled with a mass spectrometer (Finnigan MAT 253, Bremen, Germany). The overall precision of the δ¹³C measurement was < ±0.2‰. All δ¹³C values are reported on the VPDB (Vienna Pee Dee Belemnite) scale.

2.6. Calculation of Photosynthetic Carbon Isotope Discrimination

C isotope discrimination of the leaf (Δ_leaf) was calculated as

$${∆}_{leaf}=\frac{{\delta }_{air}-{\delta }_{leaf}}{{\delta }_{leaf}/1000+1}$$

(1)

where δ_air is the averaged C isotope composition of the canopy CO₂ (‰) measured at 26 m within the canopy during the sampling day, and δ_leaf is the averaged C isotope composition of the leaves (‰) for all dominant species and canopy heights.

Canopy-scale photosynthetic discrimination (Δ_canopy) was calculated using Farquhar’s classical model (Δ_classical) [1] and compared with Δ_leaf to provide robust evidence for variations in C isotope discrimination over time. Δ_classical describes the fractionation in CO₂ diffusion, carboxylation fractionation, and respiratory fractionation as

$${∆}_{classical}=\overline{a}+\left(\mathrm{b}-\overline{a}\right)\frac{c_c}{c_a}-f\frac{{\mathsf{\Gamma }}^{*}}{c_a}-e\frac{R_d}{k{c}_a}$$

(2)

where $\overline{a}$ is the overall diffusional fractionation for CO₂ as calculated from the diffusional fractionation factors (fractionation across the boundary layer (2.9‰), fractionation across the stoma (4.4‰), and fractionation across the mesophyll, including dissolution (1.1‰ at 25 °C) and diffusion (0.7‰) in water) and the conductance factors of CO₂ across the foliar boundary layer, stoma, and mesophyll during photosynthesis based on the big-leaf model according to Wehr and Saleska [33]; b is the Rubisco fractionation in C₃ plants and assumed to be 27.5‰ [34]; f is the respiratory fractionation for photorespiration (11‰) [33]; e is the respiratory fractionation for daytime dark respiration (−5‰) [33]; Γ^* is the CO₂ compensation point (μmol mol⁻¹), which was calculated from the leaf temperature according to Brooks and Farquhar [35]; R_d is the daytime respiration rate (μmol m⁻² s⁻¹), which is assumed to be 3 μmol m⁻² s⁻¹ [36]; k is the carboxylation efficiency (μmol m⁻² s⁻¹), which is assumed to be 0.1 μmol m⁻² s⁻¹ [36]; c_c is the CO₂ concentration in the chloroplast (ppm); and c_a is the CO₂ concentration in the canopy air (ppm) measured at 26 m above the ground (see Section 2.3). The c_c/c_a was obtained by the numerical solution of the equation derived from the theory of isoflux-based isotopic flux partitioning (IFP) [19], which is expressed as

$$\frac{c_c}{c_a}=\frac{-({\delta }_a+b-{\delta }_{NR}-2\overline{a}+\frac{f{\mathsf{\Gamma }}^{*}}{c_a}+\frac{e{R}_d}{k{c}_a})\pm \sqrt{{({\delta }_a+b-{\delta }_{NR}-2\overline{a}+\frac{f{\mathsf{\Gamma }}^{*}}{c_a}+\frac{e{R}_d}{k{c}_a})}^2-4(\overline{a}-b)({\delta }_{NR}+\overline{a}-{\delta }_a-\frac{f{\mathsf{\Gamma }}^{*}}{c_a}-\frac{e{R}_d}{k{c}_a}-\frac{isoflux-{\delta }_{NR}NEE}{\overline{g}{c}_a})}}{2(\overline{a}-b)}$$

(3)

where δ_a is the δ¹³C of canopy CO₂ measured at 26 m above the ground (‰) (see Section 2.3); δ_NR is δ¹³C_eco (see Section 2.4) but was substituted by the weekly smoothed value to capture the seasonal variation; isoflux is the C isotopic flux calculated from the eddy covariance measurement (see Section 2.2), the canopy CO₂ δ¹³C profile measurement (see Section 2.3), and the intercept of the daytime (7:00–17:00) Keeling plot according to [18,37]; and NEE is the net ecosystem exchange flux (μmol CO₂ m⁻² s⁻¹). Note that Δ_canopy was calculated using data from the daytime and expressed as the daily average.

2.7. Statistical Analyses

Student’s t-test was used to assess the differences in mean δ¹³C (air, organic matter, and respiration) among sampling positions, C pools, or species. A one-way analysis of variance (ANOVA) was performed to assess the significance of the seasonal variations of δ¹³C_air, leaf δ¹³C, Δ_leaf, δ¹³C_R, and δ¹³C_eco. To evaluate how the δ¹³C_R of the leaf, trunk, and soil respond to environmental factors, we conducted correlation analyses and linear regression analyses using the δ¹³C_R of each measurement and the environmental data of the corresponding timespans. Correlation analyses were also conducted to assess the environmental effects on δ¹³C_eco and the potential lagged responses of δ¹³C_eco to environmental factors using daily δ¹³C_eco and daily average meteorological data. The environmental factors include air temperature, vapor pressure deficit (VPD), global radiation, soil temperature, and soil moisture. These factors were taken into consideration because they are expected to influence C isotope discriminations in above- and belowground processes. The lagged correlation was tested with shifted time periods from zero to 10 days.

3. Results

3.1. Seasonal Variation of δ¹³C_air

In the year 2019, the mean annual temperature was 5.4 °C and annual rainfall was 627 mm, making 2019 warmer and drier than the long-term average (see Section 2.1). During the growing season of 2019, the meteorological conditions showed significant seasonal fluctuations (Figure 2). The vapor pressure deficit (VPD) was higher in May and September, and air temperature was also higher during these periods, which resulted in remarkable soil water deficits. Soil temperature and air temperature reached their highest points at the end of July. In July and August, precipitation and soil moisture were the highest, especially at the end of July and in the middle of August, when the accumulated rainfall accounted for 15.7% and 19.6% of the annual total, respectively.

We found a significant seasonal variation in δ¹³C_air during the growing season (p < 0.01, ANOVA; Figure 3a). The average δ¹³C_air of the growing season was −9.25‰. At the beginning of the growing season, δ¹³C_air showed a slight decrease and then reached a maximum of about −8.5‰ in early July and then started to decrease until it reached a minimum of below −10‰ in the middle of August. In September, δ¹³C_air first showed an increase and then started to decrease in the middle of September. Moreover, δ¹³C_air showed remarkable day-to-day variation. The largest difference in δ¹³C_air between two consecutive days was up to about 2‰. Short-term rapid depletions of ¹³C in the forest occasionally occurred during the growing season, such as the relative depletion of ¹³C at the beginning of September. The difference of δ¹³C_air between the above-canopy (−9.13‰) and within-canopy (−9.18‰) was small but statistically significant (p < 0.01, t-test). δ¹³C_air within the canopy was significantly more positive (p < 0.001, t-test) than the δ¹³C_air below the canopy (−9.44‰), and δ¹³C_air below the canopy was always the lowest throughout the growing season.

3.2. Carbon Isotopic Compositions of Leaves and Photosynthetic Carbon Isotope Discrimination

Considering the complexity of the mixed forest, we determined the leaf δ¹³C for all five dominant species and along the vertical position of their canopies. In terms of temporal variation, leaf δ¹³C in broadleaved species showed significant seasonal variations (p < 0.001, ANOVA) except for A. mono (p = 0.516, ANOVA), with a gradual decrease from May to August by a magnitude of 2‰, followed by an increase after August (Figure 3b). In contrast to broadleaves, the δ¹³C in the needles of P. koraiensis did not vary significantly over time (p = 0.275, ANOVA) and showed a slight decrease throughout the growing season, except for a relatively small increase in July (Figure 3b). δ¹³C in the needles of P. koraiensis (−29.79‰) was also the lowest among the species (Figure 3b). Since broadleaved trees are the main components of the mixed forest in this study, variation in the mean leaf δ¹³C values of the five dominant species followed the same seasonal patterns as the four broadleaved species. The temporal variation of leaf δ¹³C corresponded to that of δ¹³C_air, except in May, when δ¹³C_air was relatively low, while leaf δ¹³C was the highest compared to the rest of the growing season (Figure 3a,b), indicating that leaf δ¹³C depends not only on the C isotopic signal carried by CO₂ in the air.

C isotope discrimination at both the leaf and canopy level is shown in Figure 3c. Δ_leaf was calculated directly from the leaf δ¹³C and δ¹³C_air of the corresponding sampling days (Equation (1)). Species-specific differences in Δ_leaf were found over the growing season. Except for A. mono (p = 0.084, ANOVA), Δ_leaf varied significantly (p < 0.05 for P. koraiensis and p < 0.001 for the other species, ANOVA) over time. On average, Δ_leaf showed an increasing trend from May (18.97‰) to July (20.70‰), while the Δ_leaf of both species showed a decrease in August compared to July and September (Figure 3c). This variation pattern of leaf level discrimination was strongly supported by the Δ_canopy estimated by the classical Farquhar’s model (Equation (2)), although Δ_canopy was generally lower than Δ_leaf (Figure 3c). Δ_canopy also showed a general increasing trend over time, with the lowest value observed in May. Since there are two periods of missing measurements, we can only deduce that the Δ_canopy reached its maximum in early to middle July, followed by relatively lower values in the middle of August, according to the increasing trend in June and the decreasing trend at the beginning of August (Figure 3c).

3.3. Carbon Isotopic Compositions of Ecosystem Organic Pools and Respired CO₂

Seasonal variation in the δ¹³C_R of different ecosystem pools (leaf, trunk, and soil), as well as δ¹³C_eco, is shown in Figure 4. Over the entire growing season, the δ¹³C_R of leaves and trunks varied within a range of 3.2‰ and 4.5‰, respectively. A statistically significant general increase of δ¹³C_R was found in the needles of P. koraiensis (p < 0.001, ANOVA), while no such phenomenon was found in the leaves of A. mono (p = 0.422, ANOVA; Figure 4a). The trunk δ¹³C_R of both P. koraiensis and F. mandshurica varied significantly (p < 0.001, ANOVA) throughout the growing season. Trunk δ¹³C_R showed a decreasing trend from May to August and a slight increase in September (Figure 4b). This variation trend was similar to the general seasonal pattern of leaf δ¹³C (Figure 3a and Figure 4b). Soil δ¹³C_R did not vary significantly over time (p = 0.084, ANOVA) and was also relatively stable (with a range of 2.3‰) compared to leaf and trunk δ¹³C_R (Figure 4c). Furthermore, a decrease of δ¹³C_R was found from May to June in leaf, trunk, and soil (Figure 4a–c).

δ¹³C_R exhibited a slight depletion from leaves (−26.33‰) to trunks (−25.67‰) and soil (−25.48‰), as did δ¹³C in the corresponding organic matter (leaf: −28.68‰, trunk: −26.55‰, soil: −25.72‰; Figure 5). A distinct spatial pattern of leaf δ¹³C was found throughout the growing season (Figure 5). Similar to δ¹³C_air, leaf δ¹³C values increased significantly with an increase of height (−29.16‰, −28.73‰, and −28.16‰ from the lower to middle and upper canopy; p < 0.001, ANOVA). Leaf δ¹³C_R was generally more enriched compared to the δ¹³C in leaf organic matter but closer to the δ¹³C in the upper leaf (−28.16‰), which indicates that a substantial quantity of photosynthate from the upper canopy was used as respiratory substrate. Trunk δ¹³C_R was also more enriched than the δ¹³C in trunk organic matter but showed a larger range of variation (Figure 5). Soil δ¹³C_R was more enriched than the δ¹³C of organic matter on the surface (0–5 cm) soil (−26.59‰), in the roots (−28.02‰), and especially in litters (−28.40‰). However, the soil δ¹³C_R was similar (p = 0.527, t-test) to the δ¹³C of soil in deeper layers (5–40 cm) (−25.43‰; Figure 5).

Nights with a CO₂ range lower than 60 ppm were excluded from δ¹³C_eco calculations because a small CO₂ range could cause uncertainties in fitting the Keeling plot. After applying this exclusion, all of the values for R² in the regression were greater than 0.85 and yielded a sufficient quantity of reliable δ¹³C_eco data (Figure 4d). The variation in δ¹³C_eco was significant over the growing season (p < 0.001, ANOVA), with an average of −25.79‰. δ¹³C_eco steadily increased as the growing season progressed (Figure 4d), and the δ¹³C_eco values were within the range of the δ¹³C_R in the leaves, trunks, and soil (Figure 5). Some short-term δ¹³C_eco variations were observed, such as two gradual decreases in May and July, which may be attributed to the gradual ¹³C depletion of CO₂ in the air and respired from C pools during these periods (Figure 3; Figure 4).

3.4. Relationship between δ¹³C_R and Environmental Factors

The correlation analyses showed that the δ¹³C_R in the trunks of both P. koraiensis and F. mandshurica had a significant negative correlation with air temperature, soil temperature, and soil moisture (p < 0.01; Figure 6 and

Table 1. Pearson correlation coefficients for the δ¹³C_R from the C pools with environmental factors in the growing season. The δ¹³C_R of each measurement and the corresponding half-hour meteorological data were used to conduct this analysis.

Pearson’s Correlation Coefficient (r)	δ13CR (‰)
	Pk Leaf	Am Leaf	Pk Trunk	Fm Trunk	Soil
Air temperature (°C)	−0.035	0.009	−0.363 **	−0.443 **	0.234
VPD (kPa)	0.110	0.086	−0.146	−0.120	0.101
Global radiation (W m−2)	0.166	0.242	0.244	0.156	0.312 *
Soil temperature (°C)	−0.199	−0.101	−0.462 **	−0.513 **	0.092
Soil moisture (m3 m−3)	−0.000	−0.056	−0.484 **	−0.469 **	0.153

Note: * p < 0.05, ** p < 0.01.

). Soil δ¹³C_R only showed a significant positive correlation with global radiation (r = 0.312, p < 0.05; Figure 6 and Table 1). No significant correlation was found between leaf δ¹³C_R and any environmental factors (Figure 6 and Table 1).

We conducted a correlation analysis between δ¹³C_eco and environmental factors with a time window from zero to 10 days. It was found that δ¹³C_eco was negatively correlated with air temperature (p < 0.05;

Table 2. Pearson correlation coefficients for the δ¹³C_eco with environmental factors in the growing season. δ¹³C_eco and daily average meteorological data were used to conduct this analysis. The lag time in days and the Pearson’s correlation coefficient (r) are presented, and only the correlation that provided the best r value is presented.

	Lag	r
Air temperature (°C)	0	−0.256 *
VPD (kPa)	10	−0.280 *
Global radiation (W m−2)	0	−0.217
Soil temperature (°C)	10	0.085
Soil moisture (m3 m−3)	10	0.295 **

Note: * p < 0.05, ** p < 0.01.

) measured on the same day as δ¹³C_eco. And δ¹³C_eco was negatively correlated with VPD (p < 0.05; Table 2) and positively correlated with soil moisture (p < 0.01; Table 2) with a lag time of 10 days. Further, δ¹³C_eco was negatively correlated with global radiation and positively correlated with soil temperature, but this correlation was not statistically significant (Table 2).

4. Discussion

4.1. Variations in Carbon Isotope Discrimination from Assimilation to Respiration

Photosynthetic C isotope discrimination and δ¹³C_air both influenced leaf δ¹³C. In this study, the influence of photosynthetic discrimination on leaf δ¹³C was greater in the growing season most of the time. For example, from May to July, although δ¹³C_air showed a moderate increase, both Δ_leaf and Δ_canopy showed a strong increase, which directly led to a continuous decrease in leaf δ¹³C during this period (Figure 3). This increase in Δ_leaf and Δ_canopy may have resulted from an increase in stomatal conductance under the gradually warmer and wetter weather conditions from May to July (Figure 2). However, the influence of photosynthetic discrimination may have become weak for determining the leaf δ¹³C in August, at which point we also found that leaf δ¹³C reached its minimum, and Δ_leaf and Δ_canopy were not the highest (Figure 3). On the other hand, δ¹³C_air influenced leaf δ¹³C more profoundly in terms of height. Since ¹³C depleted CO₂ relative to the atmospheric CO₂ respired by the soil and understorey and due to the reduction in mixing between the air inside the forest and the atmosphere with a decrease in height, canopy δ¹³C_air was usually lower, as it was closer to the ground [38,39], which resulted in a similar spatial pattern between leaf δ¹³C and δ¹³C_air (Figure 5).

The seasonal pattern of photosynthetic C isotope discrimination was found to be species-specific. Compared with deciduous broadleaf species, the coniferous species (P. koraiensis) was found to have the highest Δ_leaf in the growing season (Figure 3), which is consistent with a previous study conducted at the same site as our study [40]. Low photosynthetic activity and higher stomatal conductance could result in a higher c_i/c_a and, consequently, higher Δ_leaf [41]. Therefore, P. koraiensis in this temperate mixed forest may have a lower photosynthetic rate and higher stomatal conductance compared to deciduous species. Moreover, δ¹³C in the needles of P. koraiensis showed smooth depletion during the growing season, resulting in less varied Δ_leaf over time compared to the deciduous broadleaf species (Figure 3). This difference is most likely due to the large proportion of slow-turnover compounds (e.g., starch and pinitol) in needles compared to broadleaves. These compounds generally have more stable seasonal δ¹³C variation than other carbohydrates [11,42,43] and can strongly dampen the overall δ¹³C signals of needle bulk organic matter [43,44].

Our study also found apparent post-photosynthetic C isotope discrimination among the C pools from leaf to trunk and root (Figure 5). More importantly, our results show that the leaf (leaves of P. koraiensis and A. mono) and trunk (trunks of P. koraiensis and F. mandshurica) respired CO₂ that was ¹³C-enriched in comparison to the leaf and trunk organic matter of the domain species (Figure 5). Due to the long duration of a single measurement (see Section 2.4), it was unfeasible to establish more chambers to cover more species and at different heights or to study more replicates. Despite this weakness in the δ¹³C_R measurements, the results obtained by our study are more abundant than those in many previous studies [20,45,46] and indicate the effects of post-photosynthetic C isotope discrimination in the process of dark respiration. In the process of dark respiration, ¹³C-enriched C-3 and C-4 positions in glucose are preferably used to produce pyruvate during glycolysis; then, pyruvate is decarboxylated by pyruvate dehydrogenase (PDH) to release ¹³C-enriched CO₂. A greater contribution of CO₂ decarboxylated in the PDH will result in more positive δ¹³C_R [47].

Soil δ¹³C_R was more positive than δ¹³C in the potential respiratory substrates, including roots, litter, and soil on the surface; however, it was similar to δ¹³C in the deeper soil (Figure 5). Soil respiration is mainly performed by autotrophic components (roots and rhizosphere microbes) and heterotrophic components (the decomposition of litter and soil organic matter). Among the autotrophic components, root respiration contributes significantly to soil respiration with a generally more negative δ¹³C_R signal compared to δ¹³C in root organic matter, as reviewed in [4,46,48]. Along with the relatively negative root δ¹³C_R, if we further assume that litter δ¹³C_R was mildly more positive than δ¹³C in litter organic matter during the process of decomposition, as shown in previous studies [49,50], more positive δ¹³C_R signals would still be needed from other belowground processes to produce the exact soil δ¹³C_R values in our forest, considering that litter respiration generally contributes less to soil respiration. Many studies have shown that δ¹³C in microbial biomass is more positive (1–2‰) relative to that in total soil C [4,51,52], and that microbial CO₂ is also more ¹³C-enriched compared to soil organic matter [53]. Furthermore, the δ¹³C in soil CO₂ seems to behave similarly to the typical patterns of soil δ¹³C isotopic enrichment with soil depth. For example, Goffin et al. [54] found that soil air δ¹³C in the deeper layer (10–20 cm) was more positive than that on the surface (0–10 cm) in a Scots pine stand. Using incubated soil, Formánek and Ambus [55] also found more positive δ¹³C in the CO₂ respired from the deeper layer (15–38 cm) than that from the surface (3–10 cm). Thus, the soil δ¹³C_R values observed in this study may have resulted from mixing between the δ¹³C signal from autotrophic respiration and a relatively large portion of the δ¹³C signal from microbe decomposition of soil organic matter, especially soil organic matter in the deeper layers, which indicates that the soil CO₂ flux in our forest was mainly from heterotrophic components.

4.2. Seasonal Variations in the δ¹³C_R and its Linkage to Environmental Factors

The lack of replication of in situ online δ¹³C_R measurements using non-automated systems is a common problem in previous studies [27,56], especially for studies measuring the δ¹³C_R of different C pools simultaneously [12,17], including the present study. Nonetheless, the continuous δ¹³C_R measurements in the present study accurately captured the seasonal variation of δ¹³C_R in the target components. No consistent pattern was found in the δ¹³C_R between the needles of P. koraiensis and the leaves of A. mono throughout the growing season (Figure 4). One possible reason is that the δ¹³C in the leaf respiratory substrates of P. koraiensis and A. mono experienced relatively weak seasonal variations (Figure 3), because leaf δ¹³C_R found to be linked with the δ¹³C of leaf metabolites [5,57]. By contrast, both of the δ¹³C_R values in the trunks of P. koraiensis and F. mandshurica showed a decreasing trend over the growing season (Figure 4). This result is consistent with a previous study of Quercus petraea in a temperate forest [8]; they ascribed that to the use of ¹³C-enriched respiratory substrates (i.e., starch) in trunks at the beginning of the growth period. The seasonal changes in leaf δ¹³C_R differed from the δ¹³C of leaf organic matter (Figure 3 and Figure 4), which may indicate changes in the proportion of different respiratory substrates used in leaf respiration during the growing season [42]. However, trunk δ¹³C_R shared a similar seasonal variation pattern to the δ¹³C in leaf organic matter (Figure 3 and Figure 4). This may be because a large amount of assimilated photosynthate was transported through the phloem to the trunk, and these were subsequently used as substrates for trunk respiration over the entire growing season.

Furthermore, due to the weak seasonal variation of leaf δ¹³C_R in P. koraiensis and A. mono, we found no significant correlation between leaf δ¹³C_R and the environmental variables (Table 1). It is still unclear how the leaf δ¹³C_R of other dominant species in our forest are linked to environmental variables due to our limited chamber measurements. However, as the δ¹³C in the leaf organic matter of other dominant species showed larger seasonal variations (Figure 3), it is expected that the leaf δ¹³C_R of other dominant species could have more significant seasonal variations and could be more sensitive to environmental conditions. The trunk δ¹³C_R in our forest was found to be negatively correlated with air temperature (Table 1), which is consistent with the results of Maunoury et al. [8] on sessile oak in France. Interestingly, our study found that the δ¹³C in trunk respiration was more sensitive to environmental changes at a seasonal scale than that in other flux components, regardless of species. Comparably, the successful utilization of δ¹³C in tree rings to reconstruct climate demonstrates that δ¹³C in the structural carbohydrates in wood is closely related to environmental changes [58]. This suggests that the specific compounds responsible for tree ring δ¹³C may also be responsible for the δ¹³C in trunk respiration, resulting in the sensitivity of trunk δ¹³C_R to environmental factors.

Soil δ¹³C_R showed a relatively stable variation over the growing season (Figure 4). Variability in soil δ¹³C_R has been linked to environmental factors such as air temperature, soil moisture, VPD, and photosynthetically active radiation [6,13,59,60]. Our results found that soil δ¹³C_R had no significant correlation with environmental factors except for global radiation (Figure 6 and Table 1). This result is partly supported by Bowling et al. [56], who measured soil δ¹³C_R in a US subalpine forest using chambers. The authors’ results showed that environmental forcing does not induce temporal variation in soil δ¹³C_R. The lack of a linkage between soil δ¹³C_R and environmental factors at our site could be due to the lack of in soil δ¹³C_R variation during the growing season. The large range in potential soil respiratory substrates [3] and the correspondingly large range of δ¹³C values could also have dampened the isotopic linkage between soil δ¹³C_R and current environmental conditions. Moreover, a recent study suggested that variations in soil δ¹³C_R are mainly derived from diffusive fractionation rather than biological causes [32], which provides another possible explanation for the weak relationship between soil δ¹³C_R and the environmental factors we found.

4.3. δ¹³C_eco Reveals Short-Term Links between Aboveground and Belowground Processes

The δ¹³C_eco calculated by the Keeling plot method using nighttime δ¹³C and CO₂ profile data represents the integrated δ¹³C of the CO₂ produced by all above- and belowground respiring components in the forest. The results showed that the average δ¹³C_eco in the growing season is between the average δ¹³C_R of the trunk and the soil (Figure 5), which is consistent with the results reported by Wingate et al. [17].

From a short-term perspective, aboveground vegetation could determine the biogeochemical processes belowground through, for example, newly fixed photoassimilates transport to roots, which then contribute to soil respiration. This linkage between assimilates and respiration could be revealed by relating the δ¹³C of ecosystem respiration to environmental factors but with a time lag from hours to days [9,10,16,61]. For example, Fessenden and Ehleringer [59] found that soil moisture measured at a 40 cm depth was negatively correlated with δ¹³C_eco on the day of the measurements in an old-growth coniferous forest. Scartazza et al. [62] also found a negative relationship between soil moisture measured at 70–88 cm depth and δ¹³C_eco measured on the same day in a beech forest, whereas Mortazavi et al. [6] found that δ¹³C_eco was negatively correlated to soil moisture for 7 days before δ¹³C_eco measurements. The negative correlation between soil moisture and δ¹³C_eco was also found to be significant in a Douglas-fir forest but not in a ponderosa pine forest [63]. However, our results suggest that the correlation between soil moisture and δ¹³C_eco is positive but weak and that a lag of 10 days provides the best correlation (Table 2).

The VPD was found to be positively related to δ¹³C_eco after a lag of 3 days [63], 3–4 days [6], 4–5 days [14,61], or 5–10 days [16]. Again, our results show an opposite correlation pattern between VPD and δ¹³C_eco compared to previous studies, and a lag of 10 days gave the best (but still weak) correlation (Table 2). In previous studies, Bowling et al. [64] and Schaeffer et al. [65] found no correlation between δ¹³C_eco and soil moisture or VPD. We also found that δ¹³C_eco has no significant correlation with the other environmental factors we measured, regardless of the time lag (Table 2). The weak impact of air temperature, VPD, and soil moisture on δ¹³C_eco in our study reflects the weak link between assimilation and respiration at a seasonal scale. Additionally, stronger correlations were reported by Schaeffer et al. [65] after they separately calculated the δ¹³C_eco using data within or below the canopy by selectively removing periods when the air within and below the canopy was well mixed. These findings suggest that the environmental controls for δ¹³C_eco may also be influenced by other factors, such as the canopy’s aerodynamic properties.

Although we found different seasonal isotopic patterns in different respiratory C pools and different environmental factors can contribute differently to the δ¹³C_R of those pools over the entire growing season, some strong correlations were still observed between the δ¹³C_R of the leaf, trunk, and soil and the δ¹³C_eco. The most obvious occurred in May, when the δ¹³C from respiratory C pools as well as the ecosystem showed decreasing trends (Figure 4). These distinct decreases may have resulted from unique extreme weather events in May, when the ecosystem received a large amount of global radiation but less rainfall, which led to an overall increase in VPD, a decrease in soil moisture, and a large temperature difference between the air and the soil (Figure 2). The concurrent decrease in δ¹³C_R and δ¹³C_eco suggests that although the linkage between assimilates and respiration was weak at a seasonal scale in our study, recently fixed assimilates could be allocated to the location of respiration in a short period of time, indicating a quick link between different respiration pools and assimilation at the beginning of the growing season in the coniferous and broad-leaved mixed forest.

5. Conclusions

Overall, the observed difference in seasonal variation in photosynthetic discrimination between the deciduous broadleaved and evergreen needle species in our forest indicated a remarkable difference in C assimilation between them. As expected, post-photosynthetic discrimination explained the substantial isotopic differences among C pools and their respiratory fluxes during the process of photosynthate allocation and respiration. Compared to leaf and soil δ¹³C_R, trunk δ¹³C_R was more sensitive to environmental changes and related to the seasonal patterns of leaf δ¹³C, which suggests that trunk δ¹³C_R has great potential as a way to record discrimination-encoded biochemical and physiological information related to C cycling. In addition, the linkages between environmental factors and δ¹³C_eco were not constant, which complicates the use of δ¹³C_eco to determine the linkage between assimilation and respiration. We conclude that the effects of environmental changes on C isotope discrimination can be distracted through a number of processes during C transfer.

We suggest that the combined effect of multiple environmental factors and the potential time lags should be addressed carefully when evaluating the long-term temporal changes of δ¹³C_R and δ¹³C_eco. Apart from the biological impacts initiated by controlling the photosynthetic rate and stomatal openness, the physical parameters, such as aerodynamic conditions (e.g., atmospheric stability) and the conditions of the soil–atmosphere diffusion systems may also regulate δ¹³C_eco, although they seem to act as distracters, impairing the relationship between environmental factors and δ¹³C_eco. Furthermore, soil δ¹³C_R and δ¹³C_eco integrate the δ¹³C signals of multiple tree species with different C assimilation and transfer patterns in mixed forests. In these respects, the present study expands our knowledge of C isotope discrimination in mixed forests with multiple tree species and effectively promotes a comprehensive consideration of the influencing factors related to isotopic variation within process-based ecosystem models.