Continuous Measurements of Temporal and Vertical Variations in Atmospheric CO 2 and Its δ 13 C in and above a Subtropical Plantation

: Atmospheric CO 2 dynamics in forest ecosystems are dependent on interactions between photosynthesis, respiration, and turbulent mixing processes; however, the carbon isotopic composition of atmospheric CO 2 ( δ 13 C) is not well established due to limited measurement reports. In this study, a seven-inlet proﬁle system with a Picarro analyzer was developed to conduct continuous in situ measurements of CO 2 and its δ 13 C in and above a subtropical plantation from 2015 to 2017. Results showed that ecosystem CO 2 concentration was the lowest in the afternoon and reached its peak at dawn, which mirrored variations in its δ 13 C in and above the canopy. Inverse seasonal variations were apparent between CO 2 and its δ 13 C in and above the canopy, and δ 13 C was positive during the peak growing season and negative at other times. Diel and seasonal variations in ecosystem CO 2 and its δ 13 C were mainly affected by the vapor pressure deﬁcit, followed by photosynthetic active radiation, temperature, and the enhanced vegetation index in and above the canopy; however, environmental and physiological factors had reverse or no effects near the forest ﬂoor. Nocturnal gradients of vertical variations in atmospheric CO 2 and its δ 13 C were greater than diurnal variations due to weak turbulent mixing under more stable atmospheric conditions overnight. These results implicate that photosynthesis and respiration dominated CO 2 dynamics above the canopy, while CO 2 recycling by photosynthesis and turbulent mixing changed CO 2 dynamics in the canopy.


Introduction
Forest ecosystems fix approximately one-third of the current anthropogenic CO 2 emissions from the atmosphere [1,2]; hence, an accurate assessment of forest carbon sink is important to better understand the global carbon budget [3,4]. The stable carbon isotope composition of ecosystem CO 2 (δ 13 C) is a powerful tool for tracing carbon cycling and its exchange with the atmosphere [5]. CO 2 dynamics in forest ecosystems are the results of canopy photosynthesis, respiration of different components (leaf, stem, root, and soil microbes), and turbulent mixing processes [6,7]; however, patterns of δ 13 C are not well established due to limited measurements, particularly during the day [8]. Higher CO 2 concentration is always associated with more negative δ 13 C of forest air [8,9]. Isotope ratio infrared spectroscopy (IRIS) technology has been used to continuously observe ecosystem CO 2 and its δ 13 C [8][9][10][11], providing insight into the underlying mechanisms of δ 13 C dynamics in forest ecosystems [12,13].
Biogenic CO 2 has a significant imprint on diurnal variations in ecosystem δ 13 C [14]. Mass discrimination or isotope effects, which will cause changes in isotope abundance, result in isotopic fractionation during chemical, physical and biological processes [15]. Figure 1 shows that photosynthetic discrimination ( 13 ∆) against heavier 13 C leads to isotopic enrichment of forest canopy 13 CO 2 during the daytime [16,17]. Conversely, postphotosynthetic fractionation during plant respiration [14,18], metabolic fractionation from soil microbial respiration [19,20], and diffusion fractionation of soil efflux [21,22] cause isotopic depletion of the ecosystem CO 2 . The refixation of respired CO 2 affects the δ 13 C of source CO 2 during canopy photosynthesis [6]. Furthermore, the troposphere exchanges 12 CO 2 and 13 CO 2 with ecosystem air by turbulent mixing [23], which varies the relative contributions of ecosystem photosynthesis and component respiration during the daytime; limited mixing among these component respiration creates vertical stratification of ecosystem CO 2 and its δ 13 C under stable nighttime conditions [24]. However, forest CO 2 dynamics in a hilly ecosystem are generally influenced by complex terrain [25]; local flows and weather conditions will also influence on the short-term variations in δ 13 C of forest CO 2 . Biogenic CO2 has a significant imprint on diurnal variations in ecosystem δ 13 C [14]. Mass discrimination or isotope effects, which will cause changes in isotope abundance, result in isotopic fractionation during chemical, physical and biological processes [15]. Figure 1 shows that photosynthetic discrimination ( 13 Δ) against heavier 13 C leads to isotopic enrichment of forest canopy 13 CO2 during the daytime [16,17]. Conversely, post-photosynthetic fractionation during plant respiration [14,18], metabolic fractionation from soil microbial respiration [19,20], and diffusion fractionation of soil efflux [21,22] cause isotopic depletion of the ecosystem CO2. The refixation of respired CO2 affects the δ 13 C of source CO2 during canopy photosynthesis [6]. Furthermore, the troposphere exchanges 12 CO2 and 13 CO2 with ecosystem air by turbulent mixing [23], which varies the relative contributions of ecosystem photosynthesis and component respiration during the daytime; limited mixing among these component respiration creates vertical stratification of ecosystem CO2 and its δ 13 C under stable nighttime conditions [24]. However, forest CO2 dynamics in a hilly ecosystem are generally influenced by complex terrain [25]; local flows and weather conditions will also influence on the short-term variations in δ 13 C of forest CO2. The processes contributing to forest CO2 and its stable carbon isotope (δ 13 C), and the multi-inlet profile system used to sample CO2 and its δ 13 C in and above the canopy. 13 Δ represents photosynthetic carbon isotope discrimination, R represents the respiration of different components, and 13 D represents the carbon isotopic disequilibrium between ecosystem photosynthesis and respiration. SOM represents soil organic matter. The gray dashed lines represent the recycling of respired ecosystem CO2 by photosynthesis, black broken lines represent turbulent mixing, and black dotted lines represent the CO2 diffusion of soil respiration ignoring chemoautotrophic and anaplerotic fixation. Std1, Std2, and Std3 represent three standard gases of the profile system. Picarro G2201 represents the Picarro G2201-i CO2 δ 13 C analyzer (Picarro Inc., Sunnyvale, CA, USA).
The response of forest ecosystem CO2 and its δ 13 C to environmental stress and physiological activity may differ below and within the canopy. Ecosystem photosynthesis, respiration, and turbulent mixing are all associated with variations in canopy structure, temperature, moisture conditions, etc. [26,27]. There are empirical relationships between climate variation and the ecosystem response, which are negatively correlated with soil Figure 1. The processes contributing to forest CO 2 and its stable carbon isotope (δ 13 C), and the multi-inlet profile system used to sample CO 2 and its δ 13 C in and above the canopy. 13 ∆ represents photosynthetic carbon isotope discrimination, R represents the respiration of different components, and 13 D represents the carbon isotopic disequilibrium between ecosystem photosynthesis and respiration. SOM represents soil organic matter. The gray dashed lines represent the recycling of respired ecosystem CO 2 by photosynthesis, black broken lines represent turbulent mixing, and black dotted lines represent the CO 2 diffusion of soil respiration ignoring chemoautotrophic and anaplerotic fixation. Std1, Std2, and Std3 represent three standard gases of the profile system. Picarro G2201 represents the Picarro G2201-i CO 2 δ 13 C analyzer (Picarro Inc., Sunnyvale, CA, USA).
The response of forest ecosystem CO 2 and its δ 13 C to environmental stress and physiological activity may differ below and within the canopy. Ecosystem photosynthesis, respiration, and turbulent mixing are all associated with variations in canopy structure, temperature, moisture conditions, etc. [26,27]. There are empirical relationships between climate variation and the ecosystem response, which are negatively correlated with soil Forests 2021, 12, 584 3 of 14 moisture (SWC) and precipitation, and positively correlated with vapor pressure deficit (VPD), air temperature (Ta), and photosynthetic active radiation (PAR). 13 ∆ increases with reductions in water content in boreal forest mosses [28], and there is a time lag between carbon isotopic composition of ecosystem respiration (δ R ) and VPD [29][30][31][32]. However, the relative contributions of ecosystem photosynthesis and respiration and their isotopic signatures vary with phenological activity and environmental disturbance [25,26]; they respond differently to precipitation [33,34] and drought [35,36], suggesting that there may be fundamental differences in the environmental factors and physiological activity at the canopy spatial scale. Some studies reported that δ 13 C values of forest respiration were significantly correlated with VPD within the canopy [30][31][32]37], and soil moisture near the forest floor [25].
Subtropical forest ecosystems in the East Asian monsoon region have a high CO 2 uptake, similar to that of North American and European temperate forests [38]. We developed a multi-inlet 12 CO 2 and 13 CO 2 profile system combined with a Picarro analyzer to conduct in situ measurements of atmospheric CO 2 and its δ 13 C in and above the canopy in a subtropical plantation from 2015 to 2017. This study aims to (1) examine the temporal (diel and seasonal) and vertical variations of atmospheric CO 2 and its δ 13 C in and above the forest ecosystem, and (2) elucidate the effects of environmental and physiological factors and atmospheric conditions on temporal and vertical variations.

Site Description
Field measurements were conducted at the ChinaFLUX Qianyanzhou site (26 •  The soil type is mainly acid red soil, and the bulk density of the surface soil (0-40 cm) is 1.57 g cm −3 . CO 2 emissions from soil inorganic carbon were not considered.

Profile System In Situ Measurements
The mixing ratios of 12 CO 2 and 13 CO 2 were measured using a multi-inlet profile system from 14 January 2015 ( Figure 1). The multi-inlet profile system was comprised of a Picarro G2201-i CO 2 δ 13 C analyzer (Picarro Inc., Sunnyvale, CA, USA) using wavelengthscanned cavity ring-down spectroscopy, seven air sampling inlets, and three standard gases (Std1, Std2, and Std3). The distribution of sampling inlets at 1.6, 7.6, 11.6, 15.6, 23.6, 31.6, and 39.6 m formed a vertical atmospheric profile below, within, and above the canopy.
In a sampling sequence, the three standard gases were each scanned for five minutes; then, the seven ambient air inlets were selected in turn for three minutes each and scanned 14 times. The standard gases and sampling air were pumped continuously at a flow rate of 0.03 L min −1 into the analyzer and signals were recorded at 0.3 Hz with standard temperature and pressure. Allan variance analysis showed that the G2201-i analyzer had the best precision for CO 2 (0.01 ppm) and δ 13 C (0.01‰) at 7600 s [35]. Some data during the summer of 2017 were missing due to extreme thunderstorm events. A more detailed instrument configuration can be found in the literature [11].

Calibration of Forest CO 2 and Its δ 13 C
A three-point linear calibration scheme was applied to correct the measured ecosystem atmosphere 12 CO 2 and 13 CO 2 and standard gases from 2015 to 2017 [11]. The corrected Forests 2021, 12, 584 4 of 14 atmospheric 12 CO 2 and 13 CO 2 were then used to calculate atmospheric CO 2 and its δ 13 C as follows: and where CO 2 is the total mixing ratio (µmol mol −1 ), f is the fraction of all CO 2 isotopomers (0.00474), δ 13 C represents the delta notation of isotopic mixing ratios (‰), and R VPDB is the standard molar ratio of 0.0111797. The differences (mean ± SD) between the calibrated and true values of Std1, Std2, and Std3 ranged from −1.63 ± 0.13 to 3.09 ± 0.28 µmol mol −1 for CO 2 and from −0.027 ± 0.075 to 0.017 ± 0.048‰ for δ 13 C ( Table 1). The strong performance of the multi-inlet CO 2 and its δ 13 C profile sampling system ensured the long-term stability and accuracy of observations. Table 1. Validation data for the multi−inlet CO 2 and its δ 13 C vertical profile sampling system. Year

Meteorological Measurements and Atmospheric Conditions
Meteorological measurements were completed with an eddy covariance (EC) flux system which was mounted on a tower at 39.6 m. The EC system was comprised of an open-path CO 2 /H 2 O analyzer (Model Li-7500, Licor Inc., Lincoln, NE, USA) and a 3D sonic anemometer (Model CSAT3, Campbell Scientific Inc., Logan, UT, USA). Meteorological measurements from 2015 to 2017 provided the photosynthetic active radiation (PAR), air temperature (Ta), humidity (RH), wind velocity (WS) and direction, precipitation, soil temperature (Ts), and soil moisture (SWC). Meteorological data were averaged at half hourly intervals. More detailed data processing and quality control are provided in [39].
We used two estimates to describe atmospheric stability: the friction wind velocity (u*) and scaled Obukhov length atmospheric stability (ξ, = (z − d)/L). The u* was calculated at a height of 39.6 m using mean horizontal and vertical wind speeds. ξ was determined using the displacement height (d), and the Obukov length (L) was calculated.

Vegetation and Aridity Indexes and Statistical Analyses
The enhanced vegetation index (EVI) is considered to be a physiological factor of forest carbon flux that is more responsive to leaf area and canopy biophysical characteristics [40]. We obtained 16-EVI data at a spatial resolution of 250 m from the MOD13Q1 and MYD13Q1 product subsets [41], merged the two subsets by time, and then linearly interpolated the data to obtain daily values from 2015 to 2017.
Budyko's aridity index (AI) was expressed as the ratio of monthly precipitation and potential evapotranspiration, where AI < 1 indicates periods of drought stress [42], and potential evapotranspiration is calculated in the same way as in [43].
Correlations were calculated using the Pearson method to analyze the relationships between environmental and physiological factors and atmospheric CO 2 and its δ 13 C in and above the forest ecosystem. Statistical significance was fixed to p-value < 0.001.    Figure 3 shows the monthly mean diel variation of half hourly ecosystem CO2 and its δ 13 C below (1.6 and 7.6 m), within (11.6 and 15.6 m) and above (23.6, 31.6, and 39.6 m) the canopy (2015-2017); the original time series of atmospheric CO2 and its δ 13 C at each individual height are shown in Figures S1 and S2, respectively. Atmospheric CO2 decreases and its δ 13 C becomes more positive with canopy height; diel CO2 mirrors variations in its δ 13 C at different canopy heights (Figure 3). The diel pattern of CO2 is lowest in the afternoon (14:00-15:00) and peaks at dawn (5:00-6:00). Conversely, peak δ 13 C occurred in the afternoon, and was the lowest in the early morning.

Temporal and Vertical Variations of Ecosystem CO2 and Its δ 13 C
The diel patterns of CO2 and its δ 13 C were more apparent near the forest floor (1.6 m) due to the contribution of soil-and understory-respired CO2 (Figure 3). Differences in mean diel ranges were 0.92 to 290.8 μmol mol −1 for CO2, and 0.0006 to 9.3‰ for δ 13 C; the largest differences at 1.6 m were 76.25 ± 51.64 μmol mol −1 for mean diel CO2 and −3.03 ± 2.06‰

Temporal and Vertical
Variations of Ecosystem CO 2 and Its δ 13 C Figure 3 shows the monthly mean diel variation of half hourly ecosystem CO 2 and its δ 13 C below (1.6 and 7.6 m), within (11.6 and 15.6 m) and above (23.6, 31.6, and 39.6 m) the canopy (2015-2017); the original time series of atmospheric CO 2 and its δ 13 C at each individual height are shown in Figures S1 and S2, respectively. Atmospheric CO 2 decreases and its δ 13 C becomes more positive with canopy height; diel CO 2 mirrors variations in its δ 13 C at different canopy heights (Figure 3). The diel pattern of CO 2 is lowest in the afternoon (14:00-15:00) and peaks at dawn (5:00-6:00). Conversely, peak δ 13 C occurred in the afternoon, and was the lowest in the early morning.
for mean diel δ 13 C. Differences in nocturnal CO2 and its δ 13 C among different canopy heights were greater than daytime differences, especially during the peak growing season (from April to October), which was related to the overnight atmospheric conditions.   Figure 4 shows seasonal variations in daily atmospheric CO2 and its weighted mean δ 13 C below, within, and above the canopy (2015-2017). Seasonal patterns were more significant at increased canopy height. Variations in CO2 within and above the canopy followed a downward trend in July and August, when forest photosynthesis became more vigorous, and peaked in November and December. Seasonal variations in δ 13 C mirrored those of CO2 and were more positive during the peak growing season and negative at other times.
Seasonal patterns of CO2 and its δ 13 C were not apparent below the forest canopy (Figure 4). CO2 recycling from the soil and understory affected the biogenic contributors to forest atmosphere, and potentially changed the seasonal patterns of CO2 and its δ 13 C below the canopy. Overall, CO2 had a seasonal range of 389.6 to 518.7 μmol mol −1 and δ 13 C ranged from −12.2‰ to −4.8‰ between 2015 and 2017. Statistics indicated that differences in seasonal CO2 and its δ 13 C at different canopy heights were greater during the peak growing season compared to other periods in response to environmental stress and physiological activity. The diel patterns of CO 2 and its δ 13 C were more apparent near the forest floor (1.6 m) due to the contribution of soil-and understory-respired CO 2 (Figure 3). Differences in mean diel ranges were 0.92 to 290.8 µmol mol −1 for CO 2 , and 0.0006 to 9.3‰ for δ 13 C; the largest differences at 1.6 m were 76.25 ± 51.64 µmol mol −1 for mean diel CO 2 and −3.03 ± 2.06‰ for mean diel δ 13 C. Differences in nocturnal CO 2 and its δ 13 C among different canopy heights were greater than daytime differences, especially during the peak growing season (from April to October), which was related to the overnight atmospheric conditions. Figure 4 shows seasonal variations in daily atmospheric CO 2 and its weighted mean δ 13 C below, within, and above the canopy (2015-2017). Seasonal patterns were more significant at increased canopy height. Variations in CO 2 within and above the canopy followed a downward trend in July and August, when forest photosynthesis became more vigorous, and peaked in November and December. Seasonal variations in δ 13 C mirrored those of CO 2 and were more positive during the peak growing season and negative at other times.
Seasonal patterns of CO 2 and its δ 13 C were not apparent below the forest canopy ( Figure 4). CO 2 recycling from the soil and understory affected the biogenic contributors to forest atmosphere, and potentially changed the seasonal patterns of CO 2 and its δ 13 C below the canopy. Overall, CO 2 had a seasonal range of 389.6 to 518.7 µmol mol −1 and δ 13 C ranged from −12.2‰ to −4.8‰ between 2015 and 2017. Statistics indicated that differences in seasonal CO 2 and its δ 13 C at different canopy heights were greater during the peak growing season compared to other periods in response to environmental stress and physiological activity.

Effects of Atmospheric Conditions on Variations in Ecosystem CO2 and Its δ 13 C
We selected a rainy (DOY137-140 in 2015) and drought (DOY215-218 in 2015) period as typical study periods to analyze the impact of different atmospheric conditions during the daytime and overnight. Black areas in Figure 5 show the occurrence of stable conditions (u* < 0.4 m s −1 , ξ > 0.1) during the daytime (10:00-16:00) and night (22:00-4:00) for both the rainy and drought periods. Due to weak turbulence in the nocturnal flow, the stable atmospheric conditions overnight lasted longer than during the day for both the rainy and drought periods. However, nocturnal atmospheric conditions were more stable during the rainy period, which might be related to the local circulations in hilly terrain during the drought period.

Effects of Atmospheric Conditions on Variations in Ecosystem CO 2 and Its δ 13 C
We selected a rainy (DOY137-140 in 2015) and drought (DOY215-218 in 2015) period as typical study periods to analyze the impact of different atmospheric conditions during the daytime and overnight. Black areas in Figure 5 show the occurrence of stable conditions (u* < 0.4 m s −1 , ξ > 0.1) during the daytime (10:00-16:00) and night (22:00-4:00) for both the rainy and drought periods. Due to weak turbulence in the nocturnal flow, the stable atmospheric conditions overnight lasted longer than during the day for both the rainy and drought periods. However, nocturnal atmospheric conditions were more stable during the rainy period, which might be related to the local circulations in hilly terrain during the drought period. Figure 6 shows vertical variations in atmospheric CO 2 , its δ 13 C, and wind speed (WS) for each day during the rainy and drought periods. Nocturnal gradients of vertical variations in forest ecosystem CO 2 and its δ 13 C were greater than diurnal variations, and daytime ecosystem CO 2 throughout the canopy was similar to that of CO 2 in the well-mixed atmosphere above the canopy. Vertical variations in ecosystem CO 2 and its δ 13 C during the rainy period were more significant than those in the drought period both during the day and overnight. Moreover, forest ecosystem CO 2 increased and its δ 13 C became more negative as the WS decreased below the canopy because high CO 2 from soil autotrophic and heterotrophic respiration was diffused and concentrated near the forest floor.   Figure 6 shows vertical variations in atmospheric CO2, its δ 13 C, and wind speed (WS) for each day during the rainy and drought periods. Nocturnal gradients of vertical variations in forest ecosystem CO2 and its δ 13 C were greater than diurnal variations, and daytime ecosystem CO2 throughout the canopy was similar to that of CO2 in the well-mixed atmosphere above the canopy. Vertical variations in ecosystem CO2 and its δ 13 C during the rainy period were more significant than those in the drought period both during the day and overnight. Moreover, forest ecosystem CO2 increased and its δ 13 C became more negative as the WS decreased below the canopy because high CO2 from soil autotrophic and heterotrophic respiration was diffused and concentrated near the forest floor.  Ecosystem CO2 and Its δ 13 C Figure 7 shows the Pearson correlations of atmospheric CO2 and its δ 13 C with environmental (PAR, Ta, Ts, VPD, RH, and SWC) and biological factors (EVI) at diel (Figure   Figure 7 shows the Pearson correlations of atmospheric CO 2 and its δ 13 C with environmental (PAR, Ta, Ts, VPD, RH, and SWC) and biological factors (EVI) at diel (Figure 7a,b) and seasonal (Figure 7c,d) time scales (2015-2017). Atmospheric CO 2 within and above the canopy was significantly negatively correlated with photosynthetically active radiation (PAR), temperature (Ta and Ts), and EVI (p < 0.001) at the diel and seasonal scales, and was positively correlated with moisture conditions (VPD, SWC, and RH). Accordingly, the δ 13 C of atmospheric CO 2 had a significant inverse correlation with these factors. The effects of environmental and biological factors on atmospheric CO 2 and its δ 13 C varied with canopy height. At diel scales, PAR, VPD, and RH had more obvious effects on canopy (15.6 m) CO 2 and its δ 13 C compared to other layers; correlations of other factors at the seasonal scale became stronger with canopy height.

Diel Variations and Effects of δ 13 C of Ecosystem CO2
The diel cycle of ecosystem atmospheric CO2 and its δ 13 C essentially reflect shortterm variations in ecosystem photosynthesis, respiration, and their related carbon isotopic fractionation and environmental factors. Observed patterns in δ 13 C of ecosystem CO2 showed distinct diurnal variations in and above a subalpine forest [8]; values of δ 13 C reached its maximum in the afternoon and minimum in the morning with values ranging from −11.01 to −7.94‰ near the floor of the deciduous forest [10]. Diurnal variations in photosynthetic discrimination ( 13 Δ) in C3 plants generally range from 10 to 35‰ [9,44,45], and high 13 Δ values have been measured or modeled at dawn and dusk [44,46]. Ecosystem exchange studies have often assumed that there was no short-term variation in the carbon isotope of ecosystem respiration (δR) between day and night; hence, the intercept of the nightly Keeling plot [47] or the slope of the nightly Miller-Tans plot [48] reflected the daily There were reverse effects of temperature (Ta and Ts) and EVI on CO 2 and its δ 13 C near the forest floor (1.6 m), and correlations of PAR and SWC at this layer were low or insignificant (Figure 7). This indicates that the response of forest ecosystem CO 2 and its δ 13 C to environmental stress and physiological activity varies at canopy spatial scales. The overall effects of environmental and biological factors were VPD > RH > PAR > Ta > Ts >SWC for CO 2 and its δ 13 C at the diel scale, and VPD > PAR > Ta > EVI > Ts > SWC for CO 2 and VPD > Ta > PAR > Ts > EVI > SWC for δ 13 C at the seasonal scale.

Diel Variations and Effects of δ 13 C of Ecosystem CO 2
The diel cycle of ecosystem atmospheric CO 2 and its δ 13 C essentially reflect shortterm variations in ecosystem photosynthesis, respiration, and their related carbon isotopic fractionation and environmental factors. Observed patterns in δ 13 C of ecosystem CO 2 showed distinct diurnal variations in and above a subalpine forest [8]; values of δ 13 C reached its maximum in the afternoon and minimum in the morning with values ranging from −11.01 to −7.94‰ near the floor of the deciduous forest [10]. Diurnal variations in photosynthetic discrimination ( 13 ∆) in C 3 plants generally range from 10 to 35‰ [9,44,45], and high 13 ∆ values have been measured or modeled at dawn and dusk [44,46]. Ecosystem exchange studies have often assumed that there was no short-term variation in the carbon isotope of ecosystem respiration (δ R ) between day and night; hence, the intercept of the nightly Keeling plot [47] or the slope of the nightly Miller-Tans plot [48] reflected the daily δ R values. However, considering changes in the substrate of autotrophic respiration at the diel scale and the relative contribution of components to ecosystem respiration, some studies have found that there was a more positive trend in the evening and a gradual decrease overnight [30,49], a general pattern trending to negative [50] or positive [51], or no short-term variation in δ R [52]. Short-term and diurnal variations ranged from 3 to 10‰ in forest ecosystem respiration [14,53].
There were deviations in the phases of the diel cycle between ecosystem CO 2 and its δ 13 C because of the carbon isotopic disequilibrium ( 13 D = δ A − δ R ) between ecosystem photosynthesis (δ A ) and δ R . The 13 D value determined the weights of ecosystem CO 2 resulting from isotopic enrichment by 13 ∆, and isotopic depletion by post-photosynthetic fractionation in plants and metabolic fractionation from soil microbes. Observational and modeling studies have frequently produced estimates of 13 D > 0 at the ecosystem scale [44,45,54], suggesting that isotopic depletion during ecosystem respiration was stronger than the enrichment by 13 ∆. Thus, ecosystem δ 13 C trending to positive would peak earlier than CO 2 in the afternoon. Conversely, the maximum diel variation in ecosystem CO 2 occurred at dawn (Figure 3), and the minimum of δ 13 C occurred earlier than the CO 2 peaks because the 13 ∆ values were high in the initial stages of photosynthesis. In summary, the peak and trough of ecosystem δ 13 C, in theory, occurred earlier than that of CO 2 in the diel cycle.

Seasonal Variations and Effects of δ 13 C of Ecosystem CO 2
Seasonal variations in ecosystem atmospheric CO 2 and its δ 13 C generally reflected seasonal fluctuations of background CO 2 , ecosystem photosynthesis, and respiration. Since the Industrial Revolution, fossil fuel burning and land use changes have decreased δ 13 C values in the troposphere CO 2 , and the increase in 13 ∆ slows down this declining trend [55,56]. The background values of δ 13 C of ecosystem CO 2 at two typical atmospheric stations in China (Waliguan and Shangdianzi) were more positive in summer than in spring [57]. Meanwhile, there were seasonal variations in ecosystem photosynthesis and respiration. The 13 ∆ values were obtained by gas exchange measurements and Farquhar's model [16]. Wingate et al. [9] reported that broad seasonal changes in 13 ∆ were reflected in the carbon isotopic composition of the stem, soil, and ecosystem in a maritime pine stand (Pinus pinaster Ait.); however, they became decoupled from soil respiration during rainy periods. Similar to seasonal variations in background CO 2 [31,58], the seasonal variations in δ 13 C of forest ecosystem CO 2 were more than 7‰ [30,59].
Forest understory provided important contributions to ecosystem respiration [13]; CO 2 recycling from the soil and understory changed atmospheric CO 2 and its δ 13 C in the canopy. The mechanisms of seasonal 13 ∆ are still unclear due to a lack of field observations; however, Choi et al. [60] reported that 13 ∆ was increased by irrigation due to increased stomatal and/or mesophyll conductance in a loblolly pine stand (Pinus taeda L.), Wingate et al. [9] showed that rain events caused 13 ∆ to increase above 20‰ in a maritime pine stand (Pinus pinaster Ait.), and another study further indicated that precipitation controlled the latitude distribution of 13 ∆ across the world [61]. In this study, seasonal variations in ecosystem atmospheric CO 2 and its δ 13 C were affected by VPD, PAR, Ta, and EVI, and there were reverse effects for Ta and EVI near the forest floor (Figure 7). Schaeffer et al. [25] also found stronger correlations between δ R within the canopy and VPD, and δ R near the ground and SWC in a subalpine coniferous forest. The effects of SWC on ecosystem CO 2 and its δ 13 C were not significant in this study, as forest ecosystems may use the VPD to regulate the effects of moisture conditions on vegetation photosynthesis, or the groundwater buffering of plant uptake [43] to mitigate drought stress.

Vertical Variations and Effects of δ 13 C of Ecosystem CO 2
Vertical variations in ecosystem CO 2 and its δ 13 C formed the "canopy effect" [24], in which ecosystem CO 2 increases, while its δ 13 C is progressively isotopically depleted towards the forest floor [8,29]. Heterogeneity was apparent in the distribution of ecosystem components and the allocation of resources and environmental factors at the canopy spatial scale [62,63]. Reasons for the vertical profile of δ 13 C of ecosystem CO 2 are as follows: (1) turbulent mixing between canopy CO 2 and isotopically enriched CO 2 from the troposphere; (2) PAR becomes progressively weaker near the floor, causing an increase in the CO 2 concentration ratio in the intercellular space to canopy air (C i /C a ); thus, 13 ∆ values are increased [17,64] and δ A is depleted; and (3) the isotopic dilution effect of CO 2 below the canopy due to soil respiration.
Atmospheric conditions impacted turbulent mixing between the isotopically enriched background CO 2 and depleted biogenic CO 2 . The forest atmosphere under stable nighttime conditions was stratified with poor mixing within and below the canopy, forming a distinct profile structure of forest ecosystem CO 2 and its δ 13 C ( Figure 6). The atmospheric boundary layer brought high momentum down to the canopy top under unstable conditions, which impacted the vertical profiles of turbulence moments and integral length scales within and above the canopy [65]. In this study, the subtropical high controlled the drought period; consequently, more sunny days and drainage flow with the development of temperature inversions influenced the transport of CO 2 in complex terrain [25]. Furthermore, forest CO 2 dynamics are dependent on interactions between photosynthesis, respiration, and turbulent mixing processes (Figure 1). It is likely that photosynthesis and respiration dominated CO 2 dynamics above the canopy, while CO 2 recycling by photosynthesis and turbulent mixing changed CO 2 dynamics in the canopy. Further investigations are needed to quantify the effects of photosynthesis, respiration, and turbulent mixing processes to the forest CO 2 dynamics.

Conclusions
In this study, we developed a multi-inlet profile system with a Picarro analyzer to continuously measure CO 2 and its δ 13 C below, within, and above the canopy in a subtropical plantation from 2015 to 2017. Diel and seasonal patterns of ecosystem CO 2 mirrored variations in δ 13 C at seven different canopy heights. Temporal variations in ecosystem CO 2 and its δ 13 C were affected by VPD > PAR > T > EVI; however, environmental and physiological factors displayed reverse or no effects near the forest floor. Nocturnal gradients of vertical variations in ecosystem CO 2 and its δ 13 C were greater than diurnal variations due to weak turbulent mixing under stable conditions, and vertical gradients during the rainy period were more significant compared to those during the drought period, possibly due to drainage flow during the drought period.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/f12050584/s1, Figure S1: The original time series of ecosystem CO 2 at seven heights in and above a subtropical plantation between 2015 and 2017, Figure S2: The original time series of δ 13 C of ecosystem CO 2 at seven heights in and above a subtropical plantation between 2015 and 2017.
Author Contributions: C.C. and X.W. contributed to fieldwork; C.C. analyzed all data and prepared the manuscript. C.C., X.W., J.W., and Q.G. reviewed manuscript. All authors have read and agreed to the published version of the manuscript.