Variation and Controlling Factors of Carbon Flux over a Humid Region Kiwifruit Orchard in Southwest China

Abstract

Investigating the carbon flux in orchard ecosystems is crucial for assessing agroecosystem productivity and optimizing management practices. We measured and estimated carbon fluxes (gross primary productivity, GPP; ecosystem respiration, R_e; and net ecosystem exchange, NEE) and environmental variables in a seven-year-old kiwifruit orchard over two years. Our results showed that diurnal carbon fluxes exhibited bell-shaped patterns, peaking between 12:30 and 15:30. Daily carbon fluxes exhibited a seasonal trend, characterized by an increase followed by a decrease. The average daily GPP, R_e, and NEE values were 6.77, 4.99, and −1.79 g C m⁻² d⁻¹ in 2018, and 5.88, 4.78, and −1.10 g C m⁻² d⁻¹ in 2019, respectively. The orchard sequestered −444.25 g C m⁻² in 2018 and −285.77 g C m⁻² in 2019, which accounted for 26.4% and 18.6% of GPP, respectively. Diurnal GPP and NEE were significantly influenced by photosynthetically active radiation (PAR), with direct path coefficients of 0.75 and 0.88 (p < 0.01), while air temperature (T_a) significantly affected GPP and NEE through PAR, with an indirect path coefficient of 1.12 for both. PAR had a similar effect on daily GPP and NEE, while both were indirectly influenced by soil temperature (T_s) at a 5 cm depth and vapor pressure deficit (VPD). R_e was primarily impacted by VPD, with a direct path coefficient of 0.64 (p < 0.01), while T_a and the concentration of air carbon dioxide (CCO₂) significantly affected GPP through VPD, with indirect path coefficients of 0.82 and −0.80. The leaf area index (LAI) and soil water content (SWC) at a 20 cm depth exhibited a significant correlation with carbon fluxes during the vigorous growing period.

1. Introduction

Agricultural land accounts for one-third of global land area but contributes about 31% of greenhouse emissions produced by anthropogenic activities [1]. The potential of a cropland ecosystem to act as a carbon sink or source depends on the balance between carbon captured by plants through photosynthesis and carbon lost through microbial and plant respiration. Orchard ecosystems, like other agricultural ecosystems, release carbon via respiration and fix carbon through photosynthesis. Compared to annual crops, orchards have a greater carbon sink potential over a 15–20-year period of carbon sequestration [2]. The annual carbon accumulation can be significantly enhanced by regulating orchard management strategies [3]. Therefore, understanding the variation patterns of orchard ecosystem carbon fluxes and their responses to environmental changes is crucial for assessing orchard carbon sink function and optimizing horticultural management [4,5,6,7].

The eddy covariance (EC) technique has been widely recognized as a reliable and extensively adopted method for measuring carbon fluxes in situ across various terrestrial ecosystems [8,9,10,11,12,13]. In agricultural systems, the EC technique has been employed to investigate how biotic and abiotic factors influence carbon exchange [14,15,16,17,18,19,20,21]. Yang et al. (2020) revealed that elevated temperatures not only increased daily net ecosystem exchange (NEE) but also accelerated apple tree sprouting [15]. In northeast China, NEE in an apple orchard showed negative correlations with net radiation, soil temperature (T_s), air temperature (T_a), and vapor pressure deficit (VPD) [22]. Similarly, in citrus orchards, diurnal NEE was strongly influenced by the interplay of photosynthetically active radiation (PAR) and T_a [23]. Gross primary productivity (GPP) and ecosystem respiration (R_e) were sensitive to T_a, with their responses further influenced by crop phenology and VPD [24]. Under nearly saturated light conditions, the impact of VPD on GPP was greater than that of soil water content (SWC) and T_a [25]. Water availability remained a critical determinant, as highlighted by Zanotelli et al. (2022), who reported that humidity significantly affected carbon fluxes, particularly during dry periods [26]. R_e is widely acknowledged as primarily temperature-driven in water-sufficient regions [27,28,29]. Orchard ecosystems, unlike unmanaged natural systems, are shaped by human interventions such as irrigation, pruning, and fertilization, which alter soil moisture, surface radiation dynamics, and local microclimatic conditions. Moreover, diverse underlying surface conditions suggest variations in the carbon balance [30], yet the complex interactions among these variables remain unclear.

Kiwifruit is a valuable cash crop widely cultivated worldwide [31,32], and its planted area has been increasing in China over the past decades due to its high commercial value [33]. The kiwifruit cultivation area in southwest China has exceeded 60,000 ha, accounting for one-third of the total national planting area. The intensification of climate change, such as rising CCO₂ and T_a, will alter the transport of matter and energy at the land–atmosphere interface. How the climate variables affect orchard ecosystem carbon flux variations remains unclear, though it is important for understanding the regional carbon flux dynamics and orchard production management. In this study, we measured carbon fluxes and meteorological variables in two consecutive growing seasons using an EC system in a kiwifruit orchard located in a seasonal-dry region in southwest China. The objectives were to (1) characterize diurnal and daily variations in carbon fluxes and quantify carbon exchange during the growing season, and (2) identify the key environmental variables influencing diurnal and daily carbon flux dynamics and investigate how their relationships were mediated by the leaf area index (LAI).

2. Materials and Methods

2.1. Site Description

The study area was a 7-year-old golden kiwifruit orchard (Actinidia chinensis deliciosa, cv. Jin Yan) within the 100,000-acre kiwifruit growing base in the National Modern Agricultural Industrial Park (30°19 N, 103°25 E, ~537 m above sea level), Pujiang County, Sichuan Province, China (Figure 1). The region is characterized by a subtropical monsoon humid climate, with an average annual temperature of 16.3 °C, an annual average precipitation of 1280 mm occurring mainly from May to September, and prevailing southeast winds. The experimental site was situated on a hilly terrain, and the soil was silty loamy with a bulk density of 1.35 g·cm⁻³, comprising 76.80% silt, 12.29% sand, and 10.91% clay. The soil saturated water content is 47% and the field soil capacity is 34%.

The kiwifruit was planted at a density of roughly 450 trees ha⁻¹, with a spacing of 5.0 × 4.5 m. The canopy height of kiwifruit trees varied from 1.7 to 2.0 m, and the diameter at breast height ranged from 8 to 10 cm. The primary growth period extended from March to November, with the fruits harvested in late October. The yield of the kiwifruit orchard was 8667–12,920 kg·ha⁻¹. Pruning was conducted in late April or early May for spring, and after harvest for winter. The orchard was equipped with a sprinkler irrigation system, with a flow rate of 70 L·h⁻¹, which was activated for four hours whenever the volumetric soil moisture fell below 25%. The irrigation was mainly applied in March–May. During the kiwifruit growing seasons of 2018 and 2019, the study area received 1443 mm and 1247 mm of precipitation, respectively, with evapotranspiration estimated at 500 mm. Combined with adequate irrigation, this ensured the trees experienced no water deficit, and flooding did not occur because of the hilly terrain.

2.2. Eddy Covariance Measurements and Data Processing

A closed-path eddy covariance (EC) system was installed in the east-central section of the orchard, with the sensors positioned at 8.0 m above the land surface. The system consisted of a fast-response three-dimensional sonic anemometer (CSAT3A, Campbell Scientific Inc., Logan, UT, USA), a closed path CO₂/H₂O gas analyzer (EC155, Campbell Scientific Inc., USA) capable of measuring the concentration of CO₂ (CCO₂) and concentration of water vapor (CH₂O), a temperature and humidity sensor (HMP45C), a data logger (CR6 Campbell Scientific Inc., UT, USA), an instrument calibration control system, and a power supply system. To alleviate aerodynamic interference, the three-dimensional sonic anemometer was mounted on a 3 m long arm, oriented perpendicular to the prevailing wind direction at the flux station.

Data measured by the eddy covariance technique were processed using the Datalogger Support Software 4.4 (Campbell Scientific Inc., USA), following standard procedures to convert the origin 10 Hz data into 30 minute flux data. The processing methods include the following: (i) using quadratic rotation to eliminate the vertical (w) and lateral (v) velocity components over the 30 minute interval; (ii) using covariance maximization to correct the time delays caused by the distance between the sensor and the analyzer; (iii) correction for high-frequency losses due to the spacing between the ultrasonic instrument and the water vapor and CO₂ probes, and low-frequency losses due to inadequate averaging time; and (iv) WPL density correction (adjustment for air density fluctuations) for water vapor and CO₂ fluxes [34].

The 30 min flux data were further screened based on the following criteria: (1) incomplete measurements or data measured within rain events were excluded; (2) data with quality flags ranging from 7 to 9 were also excluded due to low measured quality [35]; (3) flux data were deemed accurate only under high turbulent conditions, indicated by the friction velocity u* > 0.15 m s⁻¹ [36,37]. Based on these criteria for quality control, 27% of all data during 2018–2019 were rejected. For the data gaps less than 2 h, the data were filled with linear interpolation. For gaps exceeding two hours, daytime carbon flux data were interpolated using the following Michaelis–Menten equation [38]:

$$\mathrm{N}\mathrm{E}\mathrm{E}={\mathrm{R}}_{\mathrm{d}}-{\displaystyle \frac{\mathsf{\alpha }\cdot \mathrm{P}\mathrm{A}\mathrm{R}\cdot {\mathrm{P}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}{\mathsf{\alpha }\cdot {\mathrm{P}\mathrm{A}\mathrm{R}+\mathrm{m}\mathrm{a}\mathrm{x}}_{\mathrm{m}\mathrm{a}\mathrm{x}}}}$$

(1)

where NEE is the measured net ecosystem exchange (mg CO₂ m⁻² s⁻¹), R_d is the daytime ecosystem respiration (mg CO₂ m⁻² s⁻¹), PAR is photosynthetic active radiation (µmol phon m⁻² s⁻¹), α is the apparent quantum output (mg CO₂ µmol phon⁻¹), and P_max is maximum ecosystem photosynthetic rate.

The missing nighttime carbon flux data was filled using the equation given by [39]:

$${\mathrm{R}}_{\mathrm{e}}={\mathrm{R}}_{10}\cdot \exp \left[{\mathrm{E}}_0\left({\displaystyle \frac{1}{{\mathrm{T}}_{10}-{\mathrm{T}}_0}}\right)-\left({\displaystyle \frac{1}{{\mathrm{T}}_{\mathrm{a}}-{\mathrm{T}}_0}}\right)\right]$$

(2)

where R_e is ecosystem respiration, R₁₀ is ecosystem respiration at 10 °C (T₁₀), E₀ is a temperature-sensitive parameter, T_a is air temperature, and T₀ is a reference temperature (−46.02 °C).

The gross primary productivity (GPP) is calculated from the following:

$$\mathrm{G}\mathrm{P}\mathrm{P}={\mathrm{R}}_{\mathrm{e}}-\mathrm{N}\mathrm{E}\mathrm{E}$$

(3)

2.3. Measurements of Meteorological Factors and Leaf Area Index (LAI)

Meteorological and soil sensors (see

Table 1. Variables and sensor models observed by automatic weather stations.

Variables	Sensor Model	Installation Height (m)	Abbreviation	Unit
Air temperature	PTS-3	2.5	Ta	°C
Air relative humidity	PTS-3	2.5	RH	%
Precipitation	L3	0.6	P	mm
Photosynthetically active radiation	ZF-1	3.5	PAR	W·m−2
Soil temperature	PTWD-2A	−0.05, −0.10, −0.15, −0.20	Ts	°C
Soil moisture content	TDR-3	−0.20, −0.40, −0.60, −0.80	SWC	%

for details) were installed in the orchard to automatically measure the air temperature (T_a), relative humidity (RH), precipitation (P), photosynthetically active radiation (PAR), soil temperature (T_s) at the depth of 5/10/15/20 cm, and soil water content (SWC) at the depth of 20/40/60/80 cm. The leaf area index (LAI) was measured every 1–2 weeks using the LAI-2000 Canopy Analyzer (LAI-2000, Li-Cor, Inc., Lincoln, NE, USA). The measurements were taken at a location 50 cm below the canopy top along the branch-extending direction. Each tree was measured at 10 locations spaced horizontally by 30 cm, and eight trees were measured at each timing point. Daily LAI outside timing points in the measurement were estimated using spline interpolation [40]. To explore the mechanisms of the LAI in mediating environmental factors acting on the variation in carbon flux, we divided the growing season of the kiwifruit into a vigorous and non-vigorous growth period based on the growth rate of the LAI. In 2018, the vigorous growing period was from 21 April to 4 September, while in 2019, it was from 2 May to 19 August, accounting for 55% and 42% of the whole growing season.

2.4. Statistical Analyses

Path analysis is a multivariate statistical method used to explore direct and indirect effects of independent variables on a dependent variable [41,42]. SPSS software (Version 18.0, SPSS Inc., Chicago, IL, USA) was used to conduct a pathway analysis for investigating the correlation relationship between carbon fluxes and environmental variables at different time scales, as well as the interaction between these variables in their influences on carbon fluxes [43].

For a correlative system with one dependent variable y and numerous independent variables x_i (i = 1, 2, …, n), the multiple linear equation is expressed as follows:

$$\mathrm{y}={\mathrm{b}}_0+{\mathrm{b}}_1{\mathrm{x}}_1+{\mathrm{b}}_2{\mathrm{x}}_2+\cdots +{\mathrm{b}}_{\mathrm{n}}{\mathrm{x}}_{\mathrm{n}}$$

(4)

where x_i represents the ith explanatory variable (one of VPD, CCO₂, CH₂O, T_a, RH, PAR, T_s, SWC, and LAI in this study), and n (=9) is the total number of the explanatory variables.

A matrix equation based on Equation (5) can be established as follows:

$$\left[\begin{array}{cccc}1& {\mathrm{r}}_{\mathrm{x}1\mathrm{x}2}& \dots & {\mathrm{r}}_{\mathrm{x}1\mathrm{xn}}\\ {\mathrm{r}}_{\mathrm{x}2\mathrm{x}1}& 1& \dots & {\mathrm{r}}_{\mathrm{x}2\mathrm{xn}}\\ ⋮& ⋮& \ddots & ⋮\\ {\mathrm{r}}_{\mathrm{xnx}1}& {\mathrm{r}}_{\mathrm{xnx}2}& \dots & 1\end{array}\right]\cdot \left[\begin{array}{c}{\mathrm{P}}_{\mathrm{x}1\mathrm{y}}\\ {\mathrm{P}}_{\mathrm{x}2\mathrm{y}}\\ ⋮\\ {\mathrm{P}}_{\mathrm{xny}}\end{array}\right]=\left[\begin{array}{c}{\mathrm{r}}_{\mathrm{x}1\mathrm{y}}\\ {\mathrm{r}}_{\mathrm{x}2\mathrm{y}}\\ ⋮\\ {\mathrm{r}}_{\mathrm{xny}}\end{array}\right]$$

(5)

where r_xixj is the correlation coefficient of variable x_i and x_j, and P_xiy is the direct path coefficient, expressing the direct effect of variable x_i to y.

$${\mathrm{P}}_{\mathrm{x}\mathrm{i}\mathrm{y}}={\displaystyle \frac{{\mathsf{\sigma }}_{\mathrm{x}\mathrm{i}}}{{\mathsf{\sigma }}_{\mathrm{y}}}}, (i=1,2,\dots ,n)$$

(6)

where σ_xi and σ_y are the standard deviations of x_i and y, and r_xixj. P_xjy is the indirect path coefficient, expressing the indirect effect of variable x_i through variable x_j to y.

3. Results

3.1. Variations in Environmental Factors and LAI

The variations in the monthly average abiotic factors and LAI during the kiwifruit growing season of 2018–2019 are shown in Figure 2. T_a, T_s, PAR, CH₂O, RH, VPD, and LAI increased firstly and peaked in July or August, followed by a gradually declined trend. The average T_a was 21.5 °C and 20.9 °C during the growing seasons of 2018 and 2019, and T_s closely followed the variations of T_a, with a maximum difference of less than 4 ℃. The monthly average PAR steadily increased from March to August, with its maximum being 80.50 and 87.66 W m⁻² d⁻¹ in 2018 and 2019, respectively. CCO₂ and CH₂O showed an opposite trend during the growth season. SWC and RH were relatively lower in the kiwifruit initial growing stages but increased sharply during the month of April–May. The averaged SWC and RH were 34.81% and 86.91% during the two growing seasons. The average monthly VPD in 2019 was higher than that in 2018, with the maximum of the former and latter being 2.08 and 1.69 kPa, respectively. As time elapsed, the monthly average LAI slightly decreased in late April or early May due to spring pruning, followed by a steady increase; it then decreased again in the late growing stage due to leaf senescence and shedding.

3.2. Variations of Kiwifruit Orchard Carbon Fluxes

3.2.1. Diurnal Variations of GPP, NEE, and R_e

Consistent bell-shaped profiles were observed in the average diurnal variations in GPP, R_e, and NEE across different months in 2018 and 2019 (Figure 3a), with carbon fluxes typically remaining low in the early morning and night, nearing zero, and peaking around midday. The kiwifruit ecosystem transitioned from being a carbon source to a carbon sink during the time of 7:30–9:30 and returned to being a carbon source at 17:30–19:00. The diurnal GPP and NEE consistently peaked at approximately 12:30–14:30, while R_e peaked at 13:30–15:30, signifying a distinct diurnal pattern. The mean maximum diurnal GPP was 1.36 and 1.20 mg C m⁻² s⁻¹ in the 2018 and 2019 growing seasons, respectively. The peak of mean diurnal NEE were −0.99 and −0.80 mg C m⁻² s⁻¹ in 2018 and 2019, respectively. The diurnal amplitude variability of R_e was minimal among carbon fluxes, reaching its maximum in August, with values of 0.47 mg C m⁻² s⁻¹ in 2018 and 0.43 mg C m⁻² s⁻¹ in 2019, respectively.

3.2.2. Daily Variations of GPP, NEE, and R_e

The seasonal variations in carbon fluxes during the growing season of 2018 and 2019 are displayed in Figure 3b,c. Daily GPP and R_e increased gradually, peaking in July or August, and then decreased at the end of November. The value of daily R_e was compared to that of GPP at the beginning and end of the growing seasons. The trend in NEE was opposite to the GPP and R_e, and the value of NEE was negative almost throughout the growing seasons, but was negligible or slightly positive occasionally on cloudy or rainy days. The monthly average daily NEE peaked at −4.61 g C m⁻² d⁻¹ in June 2018 and −2.96 g C m⁻² d⁻¹ in August 2019, and the maximum monthly mean GPP was 10.25 g C m⁻² d⁻¹ in 2018 and 9.46 g C m⁻² d⁻¹ in 2019, respectively (

Table 2. Monthly mean daily, annual mean, and total gross primary productivity (GPP, g C m⁻² d⁻¹), ecosystem respiration (R_e, g C m⁻² d⁻¹), and net ecosystem exchange (NEE, g C m⁻² d⁻¹) in the kiwifruit orchard during the growing seasons in 2018 and 2019.

Year	Carbon Fluxes	Month									Mean	Total
		Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov
2018	GPP	2.75	5.51	9.41	10.25	9.85	8.18	5.13	3.67	2.66	6.77	1684.66
	Re	3.45	4.58	5.29	5.65	6.22	7.31	4.88	3.21	2.52	4.99	1240.41
	NEE	0.70	−0.93	−4.11	−4.61	−3.63	−0.87	−0.25	−0.46	−0.14	−1.79	−444.25
2019	GPP	2.82	4.98	6.95	6.74	8.47	9.46	5.65	3.89	2.24	5.88	1526.27
	Re	3.21	4.87	4.58	6.07	6.02	6.50	4.61	3.67	2.66	4.78	1242.09
	NEE	0.39	−0.11	−2.4	−0.67	−2.48	−2.96	−1.04	−0.22	0.42	−1.10	−285.77

). As indicated by the positive NEE in Table 2, the kiwifruit orchard ecosystem acted as a carbon source in March and November, and as a carbon sink in most of the days throughout the growing seasons. The total phenological GPP was 1684.66 g C m⁻² and 1526.27 g C m⁻² in 2018 and 2019, respectively, with NEE accounting for 26.4% and 18.6% of the total GPP in each year.

3.3. Driving Factors of Carbon Fluxes

3.3.1. Correlations Among the GPP, NEE, and R_e

Both diurnal and daily GPP, R_e, and NEE exhibited significant positive correlations (p < 0.01) among themselves (Figure 4). Diurnal GPP showed the strongest correlation with NEE, followed by R_e, with a Pearson correlation coefficient of 0.98 and 0.80 (p < 0.01), respectively. The correlation between diurnal NEE and R_e was relatively weaker, with a Pearson correlation coefficient of 0.66 (p < 0.01, Figure 4a,b). Daily GPP showed the strongest correlation with NEE and R_e, with a Pearson correlation coefficient of 0.89 and 0.80 (p < 0.01), and the correlation between NEE and R_e was relatively lower, with a value of 0.44 (p < 0.01, Figure 4c,d), respectively. During the vigorous growth period, the correlation between GPP and R_e, R_e, and NEE was relatively weak, with the correlation coefficient being 0.62 (p < 0.01) and 0.13 (p < 0.1), while the value was 0.80 and 0.33 (p < 0.01) at the non-vigorous growth period (Figure 5).

3.3.2. Driving Factors of Diurnal GPP, NEE, and R_e

All variables considered in the correlation analysis, except for CCO₂ and RH, exhibited significantly positive impacts on monthly average diurnal GPP, NEE, and R_e during the two growing seasons (Figure 4a,b). Among them, PAR and VPD were identified as the predominant drivers of both diurnal GPP and NEE, with a Pearson correlation coefficient of approximately 0.90 and 0.84 (p < 0.01). RH had significantly negative effects on GPP, R_e, and NEE, with correlation coefficients of −0.75, −0.60, and −0.74 (p < 0.01) during the two growing seasons, respectively. CCO₂ also exerted the negative influences on GPP, R_e, and NEE, with correlation coefficients of −0.68, −0.69, and −0.61 (p < 0.01). Although T_a was significantly correlated with GPP and NEE, it acted as the primary determinant of diurnal R_e, with a correlation coefficient of 0.94 (p < 0.01). T_s and CH₂O had little influence on GPP and NEE, yet their effects on R_e were remarkable (Figure 4).

The path analysis showed that the variance explained by the environmental factors on monthly average diurnal GPP, R_e, and NEE were 89%, 94%, and 85% in 2018, and 90%, 94%, and 86% in 2019, respectively (Figure 6a–f). Diurnal GPP and NEE were positively affected directly by PAR, with direct path coefficients of 0.75 and 0.88 (p < 0.01), respectively. The direct effect on diurnal R_e primarily stemmed from the positive influence of VPD and CH₂O, with direct path coefficients of 0.64 and 0.40 (p < 0.01). T_a exerted an indirect positive influence on GPP and NEE mainly through PAR, whereas RH had an indirect negative impact through PAR. Furthermore, T_a also exerted an indirectly positive effect on R_e, with an indirect path coefficient of 0.83, and mediated through VPD. The effect of CCO₂ on R_e was primarily indirect and negative, with VPD accounting for more than half of its influence.

3.3.3. Driving Factors of Daily GPP, NEE, and R_e

The primary driving factors of daily carbon fluxes were nearly comparable to those of average diurnal carbon fluxes (Figure 4c,d). The results indicated that PAR was the determinant of daily carbon fluxes, with Pearson correlation coefficients of 0.79 for GPP, 0.72 for R_e and 0.64 (p < 0.01) for NEE, respectively. Temperature terms (T_a and T_s) and VPD were also closely positively related to GPP and R_e, and the Pearson correlation coefficients between those factors and GPP and R_e were higher than that of NEE. CCO₂ and RH were the two factors negatively correlated to GPP, R_e, and NEE, and the effect of SWC on carbon fluxes was almost negligible. In addition to environmental factors, the LAI exerted a considerable influence on GPP, evidenced by a correlation coefficient of 0.59, 0.58, and 0.43 (p < 0.01) for GPP, R_e, and NEE in 2018, while the value was relatively lower in 2019. The correlation analysis showed that the environmental factors during the vigorous growing period are generally less prominent than daily carbon flux variations compared to that of non-vigorous periods (Figure 5). However, the LAI had a more positive impact on GPP and R_e during the vigorous growing period, with a correlation coefficient of 0.27 and 0.43 (p < 0.01), respectively.

The path analysis showed that the variance explained by the environmental factors on daily GPP, R_e, and NEE was 74%, 89%, and 53% in 2018, and 78%, 86%, and 57% in 2019 (Figure 6g–l). The effect of PAR on daily GPP and NEE was observed, primarily through its direct influence, with direct path coefficients of 0.56 and 0.78 (p < 0.01), respectively. T_s exerted a significant indirect effect on carbon fluxes through T_a and LAI, with indirect path coefficients of 0.91 for GPP, 0.54 for R_e, and 0.92 for NEE. T_a played a significant role in the daily variation in carbon fluxes, exhibiting substantial variability. In particular, its cumulative positive indirect influence exceeded its negative direct influence in 2019, which resulted in a net total positive impact. Biophysical factors predominantly affected R_e indirectly, whereas VPD played a significant role in direct influence, with a direct path coefficient of 0.42 (p < 0.01). As illustrated in the path diagrams in Figure 7, the direct effects on GPP and NEE during the vigorous growth period originated from PAR, T_a, T_s, SWC, and LAI, while the indirect effects, mediated by T_a, were mainly attributed to VPD and CH₂O. Regarding R_e during the vigorous growth period, the direct effects arose from T_a, CCO₂, and SWC, with indirect effects being derived from most factors influencing T_a, especially for T_s and CH₂O. Compared to the vigorous growth period, direct effects during the non-vigorous growth period were primarily from PAR and VPD, while RH exerted a main indirect influence on carbon fluxes through PAR. The explained variance during the vigorous growth period is significantly higher than that of the non-vigorous growth period, and the effect path of the LAI, SWC, and T_s is more obvious.

4. Discussion

Quantifying carbon fluxes and understanding their variation characteristics are of great interest for estimating the carbon fixation potential of kiwifruit ecosystems and comprehending the carbon cycle. The diurnal GPP, NEE, and R_e followed a unimodal pattern, with their peaks occurring at 12:30–15:30, coinciding with the daily maxima of solar radiation and temperature. The carbon sequestration capacity of kiwifruit ecosystems was lower during the early and late growth stages and peaked during the vigorous growth period due to the canopy development and favorable meteorological conditions conducive to growth. The peak monthly average daily NEE was −4.61 g C m⁻² d⁻¹ and −2.96 g C m⁻² d⁻¹ in summer (June/August), which suggests that suitable radiation and hydrothermal conditions are beneficial for carbon capture (Table 2 and Figure 2). The seasonal variations patterns were similar to the previous studies conducted in a multi-ecosystem (Table 3).

The total NEE measured at our study site in 2018 and 2019 was −444.25 and −285.77 g C m⁻² yr⁻¹, with an efficiency ratio (defined as the ratio of CO₂ sequestration to CO₂ emissions) of 1.36 and 1.23, respectively, indicating that the kiwifruit orchard functioned as a carbon sink at a moderate carbon sequestration efficiency during the growing seasons. The efficiency ratio of a 10-year-old kiwifruit orchard ecosystem in Italy was 1.15 and the NEE in 2012 was −493.30 g C m⁻² yr⁻¹, which was consistent with ours, despite the two research projects being conducted under different climatic conditions, soil characteristics, and management practices [44]. As shown in Table 3, the comparison analysis showed that the carbon absorption capability of kiwifruit ecosystems was relatively lower compared to other perennial fruit tree ecosystems, such as apple, pear, vineyard, citrus, and peach ecosystems [15,45,46,47,48]. This may be attributed to differences in the photosynthetic capacity among various fruit tree species and the shedding of kiwifruit leaves during the non-growing season. For other agricultural ecosystems, the annual NEE of winter wheat and agroforest ecosystems on the Loess Plateau in southern China were −437.35 and −545.42 g C m⁻² yr⁻¹, respectively, which were slightly higher than our findings [49]. Additionally, based on a yield of 1000 g m⁻² and a carbon content of 9% in the fruit dry matter, we estimated the carbon stored in the fruit to be about 90 g C m⁻². It can be inferred that the carbon stored in the soil and trees accounted for 69–80%. The lack of soil carbon monitoring limits our understanding of the below-ground carbon storage. Future studies should include soil organic carbon, woody biomass, and continuous carbon flux measurements to better characterize lifecycle dynamics and source-sink transitions. Optimizing agricultural practices could further enhance the orchard’s carbon sink potential.

Table 3. Comparison of annual net ecosystem exchange (NEE, g C m⁻² yr⁻¹) under different ecosystems.

Type of Ecosystem	Year	Annual NEE	Latitude	Longitude	Researcher
Kiwifruit	2018	−444.25	30°19′00″ N	103°25′00″ E	Our study
	2019	−285.77
Apple	2016	−698.00	35°08′52″ N	108°18′00″ E	[15]
	2017	−554.00
Pear	2012	−600.00	37°47′44″ N	114°55′57″ E	[45]
Vineyard	2008	−820.00	37°51′00″ N	102°51′00″ E	[46]
	2009	−824.00
	2010	−961.00
Citrus	2009	−385.50	39°27′15″ N	0°33′32″ W	[47]
Peach	2011	−1052.00	40°10′25″ N	116°07′53″ E	[48]
Kiwifruit	2003	−320.00	44°20′39″ N	11°59′02″ E	[44]
	2012	−493.30
Winter wheat	2006	−437.35	35°14′00″ N	107°41′00″ E	[49]
Agroforest	2006	−545.42
Grassland	2005	54.80	30°51′00″ N	91°05′00″ E	[50]
	2005	−51.70	37°39′55″ N	101°19′52″ E
	2005	139.90	43°32′00″ N	116°40′00″ E

Table 3. Comparison of annual net ecosystem exchange (NEE, g C m⁻² yr⁻¹) under different ecosystems.

Type of Ecosystem	Year	Annual NEE	Latitude	Longitude	Researcher
Kiwifruit	2018	−444.25	30°19′00″ N	103°25′00″ E	Our study
	2019	−285.77
Apple	2016	−698.00	35°08′52″ N	108°18′00″ E	[15]
	2017	−554.00
Pear	2012	−600.00	37°47′44″ N	114°55′57″ E	[45]
Vineyard	2008	−820.00	37°51′00″ N	102°51′00″ E	[46]
	2009	−824.00
	2010	−961.00
Citrus	2009	−385.50	39°27′15″ N	0°33′32″ W	[47]
Peach	2011	−1052.00	40°10′25″ N	116°07′53″ E	[48]
Kiwifruit	2003	−320.00	44°20′39″ N	11°59′02″ E	[44]
	2012	−493.30
Winter wheat	2006	−437.35	35°14′00″ N	107°41′00″ E	[49]
Agroforest	2006	−545.42
Grassland	2005	54.80	30°51′00″ N	91°05′00″ E	[50]
	2005	−51.70	37°39′55″ N	101°19′52″ E
	2005	139.90	43°32′00″ N	116°40′00″ E

Our study found that although the total R_e was comparable between the two growing seasons, GPP and NEE in 2019 were relatively lower than those of the 2018 growing season. On the one hand, this may be attributed to the lower SWC and higher VPD in the summer of 2019 (Figure 2), which limited photosynthesis due to reduced water availability. GPP would be reduced significantly due to a lack of precipitation and seasonal high temperatures [44]. On the other hand, the significant decrease in the LAI caused by spring pruning at the end of April 2019 likely resulted in overall lower carbon flux values. It can be speculated that farmland ecosystems’ carbon flux is strongly disturbed by both environmental and human management interventions [51,52].

The correlation analysis of diurnal and daily carbon fluxes indicated that PAR was the primary driver for GPP and NEE, but R_e was predominantly influenced by T_a [53,54]. Previous studies have shown that water availability controls carbon fluxes in arid regions, whereas in humid regions, radiation and temperature are the primary factors, and excessive moisture conditions may restrict the efficiency of photosynthesis [55,56,57]. Our study area is located in the low-altitude Sichuan Basin, where frequent cloudy and rainy conditions limit available radiation, making radiation crucial for GPP. Owing to its direct involvement in photosynthesis, PAR can directly enhance GPP and NEE [58]. R_e is primarily driven by a series of enzymatic reactions, leading to an increase with rising temperature [55,59]. Soil temperature regulated surface temperature through heat exchange and radiation dynamics, influencing enzymatic activities and thereby altering carbon exchange rates. The results revealed that the influence of T_s at the depth of 5 cm on diurnal R_e was not as strong as on daily R_e (Figure 4). There may be small diurnal variations in soil characteristics (T_s and SWC), which cannot be reflected in the statistical analysis, and the calculation formula for R_e includes T_a but not T_s. Higher temperature (T_a and T_s) not only stimulated subsurface respiration but also extended the growing season of deciduous forests, resulting in a greater increase in GPP compared to R_e [60]. A similar phenomenon was observed in our study conducted in 2018. Moreover, studies have demonstrated that within an optimal temperature range, a moderate increase in air temperature (T_a) can enhance the activity of photosynthetic enzymes, thereby improving the efficiency of PAR utilization in photosynthesis [61,62,63].

In our kiwifruit orchard ecosystem, VPD exerted a substantial promotive effect on GPP and R_e (Figure 4, Figure 5, Figure 6 and Figure 7). This was possibly attributed to sufficient soil moisture and deeper roots maintaining high leaf water potential and stomatal opening [64]. VPD directly affects GPP and R_e by regulating stomatal closure, transpiration, and photosynthesis [65]. To some extent, VPD served as an indicator of atmospheric drought in humid regions [66], and VPD had a greater effect on GPP compared to RH and SWC, which was consistent with previously published results [67,68,69].

Numerous studies have established that the LAI is a critical driver of GPP [18,70]. In our study, the direct path coefficients of the LAI for GPP and NEE during the vigorous growth period were 0.34 and 0.43 (p < 0.01), respectively (Figure 7), indicating that the LAI plays an important role in regulating carbon flux exchange. However, during the non-vigorous growth period, the impact of the LAI on carbon fluxes appeared to be concealed. This discrepancy is primarily attributable to the high frequency of cloudy days in the study area, which results in substantial fluctuations in daily radiation and consequently prevents our current statistical methods from absolutely reflecting the relationship between daily variations in photosynthesis and the LAI. The indirect effects of environmental drivers (e.g., T_s via LAI) were acknowledged but not fully quantified using mechanistic or process-based models, limiting the understanding of system interactions. Nevertheless, this study offered valuable insights into the carbon balance and climate responsiveness of orchard ecosystems, contributing to enhancing orchard management strategies to bolster carbon sequestration while promoting sustainable agricultural practices.

5. Conclusions

The study systematically investigated the dynamics of carbon fluxes (GPP, R_e, NEE) and their driving factors in a kiwifruit orchard located in the humid region of southwest China, uncovering key characteristics and regulatory mechanisms of orchard carbon exchange. The findings indicated that (1) diurnal and daily carbon fluxes exhibited a pattern of increase followed by a decrease, with peaks observed in summer (June/August); (2) the kiwifruit orchard acted as a carbon sink; (3) PAR and VPD were the primary drivers of GPP/NEE and R_e, with other factors (T_a, RH and CCO₂) indirectly influencing carbon fluxes through their effects on these primary drivers; and (4) during the vigorous growing period, the LAI, SWC, and T_s showed a more significant correlation with carbon fluxes than in the non-vigorous growing period. The absence of soil carbon monitoring limited the understanding of below-ground carbon storage and lifecycle dynamics, highlighting the need for future studies to include continuous soil and biomass carbon measurements while exploring optimized agricultural practices to enhance the orchard’s carbon sink potential.