Photosynthetic rate dominates the seasonal variation of tree intrinsic water-use efficiency in the humid East Asian Monsoon region

Abstract

Intrinsic water-use efficiency (iWUE) in trees, defined as the ratio of photosynthetic rate (A) to stomatal conductance (g_s), is a key indicator characterizing the carbon–water balance in trees. Previous studies have suggested that seasonal variation in iWUE is primarily controlled by g_s in arid regions; however, the main driving factors in humid regions remain unclear. Therefore, this study utilized 8 years of high-resolution tree-ring δ¹³C and δ¹⁸O data from Pinus massoniana (Lamb.) (coniferous) and Sassafras tzumu (Hemsl.) (broadleaf) in the humid East Asian monsoon region to reconstruct the seasonal dynamics of iWUE. We estimated iWUE based on δ¹³C and derived the leaf water δ¹⁸O enrichment (Δ¹⁸O_lw) from δ¹⁸O to represent g_s, thereby distinguishing the relative contributions of A and g_s to variations in iWUE. Both species exhibited synchronous seasonal iWUE patterns: decreasing from spring to summer before autumn recovery. Dual-isotope analysis revealed that iWUE seasonal variations were primarily driven by fluctuations in A, contrasting with g_s-dominated mechanisms in arid regions. Summer iWUE declines resulted from reduced A, constrained by relatively low CO₂ and high temperatures. This study reveals an A-dominated regulatory mechanism of iWUE in humid regions, providing a theoretical basis for more accurate predictions of forest carbon–water coupling under varying moisture conditions.

Introduction

Since the Industrial Revolution, rising atmospheric carbon dioxide concentrations (C_a) have enhanced tree photosynthesis and intrinsic water-use efficiency (iWUE), thereby promoting plant growth, vegetation biomass and soil organic matter (Walker et al. 2021). This process facilitates the fixation of atmospheric carbon dioxide into terrestrial ecosystems, strengthening the role of trees as a carbon sink. Furthermore, if the terrestrial carbon sink continues to intensify, it could reduce the rate of C_a increase, thus contributing to the mitigation of global climate warming (Mo et al. 2023). As a key physiological parameter, iWUE represents the ratio of the net photosynthetic rate (A) to stomatal conductance (g_s) (Farquhar et al. 1982). It reflects the balance between carbon assimilation and transpirational water loss in plants and is fundamental to vegetation–climate interactions and terrestrial water cycles at the global scale (Guerrieri et al. 2019, Li et al. 2023).

Tree-ring stable carbon isotope (δ¹³C) records enable the reconstruction of iWUE over long-term scales, revealing its dynamics from interannual to centennial variations (Huang et al. 2017, Reed et al. 2018, Adams et al. 2020, Treml et al. 2022, Rubio-Cuadrado et al. 2024, Wang et al. 2024). Globally, tree-ring δ¹³C chronologies have consistently shown an increasing trend in iWUE over recent decades (Mathias and Thomas 2021). To disentangle the relative contributions of A and g_s to iWUE trends, a dual-isotope approach combining δ¹³C and stable oxygen isotope ratios (δ¹⁸O) (Farquhar and Lloyd 1993, Roden et al. 2000, Scheidegger et al. 2000, Farquhar et al. 2007) has been widely employed (Guerrieri et al. 2019, Mathias and Thomas 2021, Lin et al. 2022, Rubio-Cuadrado et al. 2024, Wang et al. 2024). Dual-isotope studies indicate that the rise in iWUE at the global scale is primarily attributed to enhanced A driven by increasing C_a (Guerrieri et al. 2019, Mathias and Thomas 2021). Concurrently, local environmental factors—including temperature (T), precipitation (P), relative humidity (RH) and vapor pressure deficit (VPD)—also influence iWUE by modulating the balance between A and g_s (McAdam and Brodribb 2015, Dusenge et al. 2019, Grossiord et al. 2020). Specifically, increases in iWUE in humid regions are mainly driven by enhanced A, whereas in arid regions they rely more on reduced g_s (Guerrieri et al. 2019). Furthermore, tree species with different functional traits (e.g., leaf and wood functional types) exhibit distinct iWUE regulatory mechanisms (Mathias and Thomas 2021, Wang et al. 2024). However, recent comprehensive studies have highlighted significant complexities and methodological uncertainties in inferring g_s from tree-ring δ¹⁸O (Lin et al. 2022, Carter et al. 2025). These uncertainties stem from interspecific differences in the Péclet effect (Barbour et al. 2021), the potential seasonal variability of the oxygen exchange proportion (p_ex) during cellulose synthesis (Leppä et al. 2026) and the confounding background noise from environmental fluctuations, such as atmospheric water vapor δ¹⁸O (δ¹⁸O_v), temperature and VPD. Therefore, when utilizing the dual-isotope approach, explicitly evaluating these physiological and physical uncertainties is crucial for ensuring the robustness of δ¹⁸O-derived g_s interpretations.

However, most existing studies have been conducted at an interannual scale, limiting our understanding of the mechanisms regulating seasonal iWUE dynamics and the response differences among different functional tree types (e.g., conifers vs. broadleaf trees). While interannual studies provide valuable long-term trends, they often mask the physiological responses of trees to rapid environmental fluctuations within a single growing season (Miyahara and Locosselli 2024). Consequently, the emergence of intra-annual stable isotope analysis has become a cornerstone for achieving higher temporal resolution in dendrochronology, enabling researchers to capture tree responses to seasonal climate variations (Wilson and Grinsted 1975, Ogle and McCormac 1994, Vaganov et al. 2009, Schubert and Timmermann 2015, Andreu-Hayles et al. 2022, Kagawa and Battipaglia 2022, Xu et al. 2022, Miyahara and Locosselli 2024, Xu et al. 2024). Recent studies have successfully extended tree-ring δ¹³C applications to seasonal-scale iWUE estimation, with cross-validation through gas exchange measurements and eddy covariance techniques further confirming its reliability (Tang et al. 2023). Investigating the seasonal variation of iWUE is crucial for understanding carbon sequestration, transpiration and related physio-ecological strategies of trees in terrestrial ecosystems, providing important insights into tree growth and carbon–water cycling (Martínez-Sancho et al. 2022, Tang et al. 2023, Portarena et al. 2024).

For example, Martínez-Sancho et al. (2022) found that during extreme summer drought, iWUE increased by 11%, but carbon fixation decreased by 67% due to constraints in cell enlargement and wall thickening, indicating that the increase in iWUE could not fully compensate for drought-induced growth inhibition. Most current research focuses on arid regions, where summer drought typically leads to reduced g_s, thereby elevating iWUE (Barbour et al. 2002, Battipaglia et al. 2014, Martínez-Sancho et al. 2022, Portarena et al. 2024). In contrast, the East Asian monsoon region—a critical global carbon sink—experiences synchronous warm and wet summers with minimal water limitations, potentially leading to a distinctly different iWUE regulatory mechanism compared with arid zones. Nevertheless, the seasonal drivers of iWUE in this region remain poorly understood.

Therefore, this study examined Pinus massoniana (Lamb.) (coniferous) and Sassafras tzumu (Hemsl.) (broadleaf) in the humid monsoon region of East Asia (sampling site: Lin’an, Zhejiang Province). By collecting tree-ring samples and establishing high-resolution intra-annual isotope sequences, this study aims to physiologically elucidate the seasonal variations in iWUE. First, tree-ring δ¹³C was used to reconstruct the seasonal dynamics of iWUE from 2007 to 2014; second, combined with tree-ring δ¹⁸O and the oxygen isotope composition of atmospheric precipitation, the degree of leaf water enrichment (Δ¹⁸O_lw) was estimated to quantify the contribution of g_s to changes in iWUE; finally, the relative roles of A and g_s in the seasonal regulation of iWUE were clarified. Specific research objectives include: (i) revealing the seasonal patterns of tree-ring δ¹³C, δ¹⁸O and iWUE in two functional tree species; (ii) determining the relative contributions of A and g_s to seasonal variations in iWUE; and (iii) identifying key climatic drivers influencing the seasonal changes in iWUE.

Materials and methods

Sampling sites and tree species

Tree-ring samples were collected from a mixed coniferous and broad-leaved forest at the Lin’an Atmospheric Background Station (30.3°N, 119.72°E, 138.6 m a.s.l.) in Hangzhou, Zhejiang Province, southeastern China, using 1-cm-diameter increment borers at a height of 1.3 m above the ground (Figure 1a). P. massoniana and S. tzumu are the dominant tree species in this forest. P. massoniana is a non-porous evergreen coniferous tree (Figure 1b), and S. tzumu is a ring-porous deciduous broad-leaved tree (Figure 1c).

Figure 1.

(a) Sampling points and meteorological station locations. Triangles represent the locations of sampling points, Lin’an Meteorological Station and Lin’an Atmospheric Background Station (LAN), while circles indicate the locations of oxygen isotope measurements in atmospheric precipitation; (b, c) Wood anatomical characteristics of P. massoniana and S. tzumu; (d) Monthly average temperature, precipitation and relative humidity at LAN Station from 2007 to 2014.

Climate data, atmospheric CO₂ concentration and precipitation δ¹⁸O

The Lin’an Meteorological Station (30.22°N, 119.70°E, 117.9 m a.s.l.) is ~8 km away from the sampling site (Figure 1a). From 2007 to 2014, the annual mean temperature in this region ranged from 15 to 17 °C, with an average annual precipitation of 1481 mm. The month with the highest precipitation was June, and the primary rainy season occurred between June and August, reflecting typical features of a subtropical monsoon climate (Figure 1d). The mean annual RH was 73.3%, with maximum values observed during summer and autumn.

Vapor pressure deficit (kPa) was calculated based on the monthly temperature and RH data using the formula proposed by Murray (1967):

(1)

where T is the monthly average temperature (°C), and RH is the relative humidity (%).

Monthly atmospheric CO₂ concentration data from 2007 to 2014 were acquired from the Lin’an Atmospheric Background Station. Studies indicate that the atmospheric CO₂ concentration at this station from 2011 to 2013 exhibits a significant linear relationship with its δ¹³C_CO2 value (slope = −0.037, R² = 0.8) (Xia et al. 2015, Xia 2016). Based on this, this study utilized the following formula to estimate the δ¹³C_CO2 from 2007 to 2014:

(2)

where δ¹³C_CO2 is the carbon isotope value of atmospheric carbon dioxide, CO₂ is the carbon dioxide concentration and 6.36 is a constant calculated by substituting the CO₂ and δ¹³C_CO2 for 2011.

The precipitation δ¹⁸O data for 2011 were retrieved from measurements in Anhui (PLX, 30.23°N, 117.53°E) (Duan et al. 2016), while the data for 2012–2014 were from Nanjing (NJ, 32.12°N, 118.95°E) (Li et al. 2020a). Due to the lack of long-term in situ measured data of source water and water vapor isotopes, we utilized the LMDZ4 model data (2007–2010, 2.5° × 3.75° resolution), which has been widely validated and used in tree-ring cellulose δ¹⁸O research (Hourdin et al. 2006, Shi et al. 2011, An et al. 2019, Huang et al. 2019). Specifically, we extracted both the simulated precipitation δ¹⁸O (δ¹⁸O_p) and the atmospheric water vapor δ¹⁸O (δ¹⁸O_v).

Stable isotope analysis

At the study site, we collected 40 tree cores from P. massoniana and 15 cores from S. tzumu. The samples were air-dried, mounted on wooden holders and progressively polished using graded sandpapers until tree-ring boundaries were clearly visible. Ring widths were measured followed by cross-dating to verify chronological accuracy. The period 2007–2014 was chosen for isotopic analysis due to the availability of monthly resolved C_a and δ¹³C_CO2 data (coverage: 2011–2013). High-resolution isotope analysis was conducted on two trees per species (P. massoniana: LA02A, LA18A; S. tzumu: LA36A, LA39A), which were selected from the cross-dated samples for their nearly parallel ring boundaries. Following separation from wooden holders via hot water soaking, all cores were meticulously cleaned of surface adhesive to eliminate contamination risks. Each annual ring was subjected to serial thin-sectioning using a rotary microtome: initially sectioned at consecutive 60 μm intervals, followed by pooling the sections into 20 parts per annual ring (i.e., each growth year comprises 20 equally divided samples). This high-resolution partitioning approach effectively deconstructs the annual growth process into 20 consecutive temporal units, enabling precise tracking of intra-annual isotopic dynamics during the growing season (refer to Figure 1b and c for schematic illustrations of the ring partitioning in P. massoniana and S. tzumu). All thin-section samples were subsequently processed for cellulose extraction and isotopic measurements.

Tree-ring α-cellulose was extracted according to a standardized protocol by Xu et al. (2011). Subsequently, 60–100 μg of α-cellulose samples were encapsulated in Sn foil for analyzing the carbon isotope ratios (¹³C/¹²C), while 120–200 μg of α-cellulose samples were encapsulated in silver foil for determining the oxygen isotope ratios (¹⁸O/¹⁶O). The carbon and oxygen isotope ratios of tree-ring α-cellulose were measured at the Institute of Geology and Geophysics, Chinese Academy of Sciences, using a coupled isotope ratio mass spectrometer (Delta V Advantage; Thermo Scientific, Germany) and a pyrolysis-based high-temperature conversion elemental analyzer (Flash 2000HT; Thermo Scientific, Germany). The sample stable isotope value were calibrated using a laboratory standard (Merck cellulose) after every set of eight samples. Analytical precision was ±0.23‰ for δ¹⁸O (n = 80) and ±0.05‰ for δ¹³C (n = 90) based on repeated measurements of Merck cellulose standards. δ¹³C is the ¹³C/¹²C ratio of the sample normalized to the international standard Vienna Pee Dee Belemnite, while δ¹⁸O is the ¹⁸O/¹⁶O ratio of the sample normalized to Vienna Standard Mean Ocean Water, both expressed in per mil (‰).

Assigning a date to each section of the tree ring

The P. massoniana growing season is from April to October (He et al. 2012). S. tzumu flowers in early spring and then grows leaves. The flowering period lasts from mid-February to late March, and the leaves fully develop by mid-to-late April to provide photosynthetic products (Chen et al. 2020). Research indicates that in some deciduous tree species, xylem development initiates before the onset of photosynthesis within the same growing season (Michelot et al. 2012, McCarroll et al. 2017), with non-structural carbohydrates stored during the previous growing season serving as the primary carbon source (Helle and Schleser 2004). Leaf senescence of S. tzumu begins in mid-to-late September and continues into early October, coinciding with the onset of dormancy (Jiang et al. 2016). Based on these observations, the growing season of S. tzumu is estimated to range from mid-February to mid-to-late September.

To account for inherent individual growth rhythm variations and potential offsets in manual slicing, we first synchronized the intra-annual δ¹³C and δ¹⁸O curves. By identifying and correcting shifts of one to four segments at the starting positions, we aligned physical segments from different trees to consistent physiological growth stages. Since both isotopes were measured from the same physical slices, a simultaneous shift calibration was applied. The synchronized data are presented in Figure 2, with the original raw measurements provided in Figure S1 available as Supplementary Data at Tree Physiology Online.

Figure 2.

Seasonal variations in synchronized tree-ring δ¹³C (a, c) and their mean seasonal patterns (b, d) for P. massoniana (a, b) and S. tzumu (c, d) during the period 2007–2014; seasonal variations in synchronized tree-ring δ¹⁸O (e, g) and their mean seasonal patterns (f, h) for P. massoniana (e, f) and S. tzumu (g, h) during the period 2007–2014.

Near the study area, the absence of systematic observations on the growth patterns of P. massoniana and S. tzumu using tree radial growth measurement instruments makes it challenging to precisely determine the exact formation periods of each annual ring slice. Although tree-ring growth typically follows an S-shaped function (Ogée et al. 2009, Song et al. 2014), this study focuses on seasonal-scale variation characteristics, with daily-scale errors being within an acceptable range. Based on the phenological traits of P. massoniana and S. tzumu, we adopted the methodological framework developed by Berkelhammer and Stott (2009). Using the established mathematical formula, we performed the date registration for the subsamples within each annual ring. It is noteworthy that, during the synchronization process to achieve inter-individual phase alignment, the number of effective intra-annual segments for the same species was expanded from the initial 20 to a range of 20–24. Despite variations in n_t across different years or individuals, the estimation of all dates strictly followed the same mathematical framework to ensure the consistency of the time-allocation criteria. For conciseness, Table S1 available as Supplementary Data at Tree Physiology Online only presents the calculated dates for the standard 20-segment scenario, serving as a representative benchmark for this dating methodology.

(3)

where g_i is the start date of growth, g_l is the length of the growing season, n_s is the relative position of each section among the total number of sections per tree ring and n_t is the total number of sections per annual ring. When calculating the corresponding dates for P. massoniana, g_i was set as 1 April, and g_l was 214 days (from 1 April to 31 October). For S. tzumu, g_i was 15 February, and g_l was 218 days (from 15 February to 20 September). Based on the calculated dates for each part, the isotopic mean values of tree-ring parts corresponding to each month were obtained.

Calculation of iWUE

We estimated the iWUE based on high-resolution tree-ring δ¹³C data using two models: the simplified model iWUE_sim (Farquhar et al. 1982), which assumes infinite mesophyll conductance (g_m), and the improved model iWUE_mes (Ma et al. 2021), which incorporates g_m. The discrimination of atmospheric CO₂ by plants (Δ¹³C) serves as a critical indicator of plant iWUE, and the formula is presented below:

(4)

where Δ¹³C is the stable carbon isotope discrimination; δ¹³C_a and δ¹³C_t are the monthly mean values of δ¹³C for atmospheric CO₂ and tree-ring cellulose, respectively.

(5)

where A is the photosynthetic rate of the plant; g_s is the stomatal conductance; C_a is the monthly mean of atmospheric CO₂ concentration; a (4.4‰) is the isotope discrimination during CO₂ diffusion through stomata; b (27‰) is the isotope discrimination value due to the carboxylation; 1.6 is the molar diffusivity ratio of CO₂ to H₂O.

(6)

(7)

where C_a is the monthly mean of atmospheric CO₂ concentration; a (4.4‰) is the isotope discrimination during CO₂ diffusion through stomata; b′ (29‰) and f′ (11‰) are the discriminations due to carboxylation and photorespiration, respectively; a_m (1.8‰) is the discrimination associated with CO₂ dissolution and diffusion in the mesophyll. The g_s/g_m ratio is a constant value of 0.79, which was determined through measurements conducted across a broad spectrum of plant species (Ma et al. 2021); Γ^* is the CO₂ compensation point in the absence of mitochondrial respiration (Brooks and Farquhar 1985) and T is the average monthly temperature.

Calculation of the enrichment of leaf water ¹⁸O over source water (Δ¹⁸O_lw)

We calculated the enrichment of ¹⁸O in tree rings relative to source water (Δ¹⁸O_t) following the method of Cheesman and Cernusak (2017):

(8)

where δ¹⁸O_t is the monthly mean δ¹⁸O value of tree rings, while δ¹⁸O_p is the monthly mean δ¹⁸O value of precipitation during tree-ring formation. We assume that δ¹⁸O_p is indicative of the oxygen isotope composition of soil water, which serves as source water. Based on the method proposed by Guerrieri et al. (2019), we quantified Δ¹⁸O_lw using the following equation:

(9)

where p_x (~1) represents the proportion of oxygen exchanged with local xylem water during cellulose synthesis, and p_ex denotes the fraction of oxygen atoms participating in this exchange. Traditionally, p_ex is assumed to be a constant value of 0.4 (Guerrieri et al. 2019). However, to account for the potential seasonal variability of p_ex as suggested by recent modeling observations, we additionally conducted a sensitivity simulation. By comparing the standard constant p_ex scenario (0.4) with a linearly decreasing p_ex scenario—simulating a dynamic decline of p_ex from 0.5 to 0.2 across the growing season (Leppä et al. 2026)—we quantified the potential interference of this physiological uncertainty on the derived Δ¹⁸O_lw signal. εwc is the temperature-dependent discrimination related to the exchange of carbonyl oxygen atoms with water during cellulose synthesis, εwc = 0.0084 T² – 0.51 T + 33.172 (Sternberg and Ellsworth 2011), T is the average monthly temperature.

Statistical analysis

Pearson and partial correlation analyses were conducted to evaluate the linkage between intra-annual iWUE and Δ¹⁸O_lw, and associations of iWUE with key climatic variables (T, P, RH, VPD and C_a). Partial correlation analysis controlled for covarying effects among environmental factors. Furthermore, Pearson correlation analysis was employed to evaluate the relationships between the simulated Δ¹⁸O_lw under different p_ex scenarios and iWUE_mes, aiming to assess the potential impact of physiological uncertainties. The same correlation method was also applied to validate the consistency between the LMDZ4-simulated δ¹⁸O_v and precipitation δ¹⁸O_p.

To decouple the potential influence of mathematical coupling between iWUE and C_a—arising from C_a being an explicit multiplier in the iWUE calculation formula—a ‘null-hypothesis model’ was introduced to isolate true biological signals. We calculated a theoretical slope (k_null), representing the iWUE increase rate assuming a constant carbon isotope discrimination (Δ¹³C). This baseline was compared against the observed slope (k_obs), derived from the linear regression of iWUE against C_a. A k_obs significantly exceeding k_null indicates an active biological signal where physiological regulation enhances the iWUE response beyond the mathematical expectation of rising C_a.

To quantify the relative contributions of environmental factors to seasonal variations in iWUE, we applied multiple linear regression analysis.

First, all variables were standardized using Z-score normalization to eliminate scale effects and improve comparability across the model. The standardization formula is:

(10)

where x is the original variable value, μ is the mean and σ is the standard deviation.

The multiple linear regression analysis requires a linear relationship between the independent and dependent variables. By examining scatter plots and calculating correlation coefficients, we excluded two climatic variables, P and RH.

Subsequently, a multiple linear regression model was constructed:

(11)

where iWUE is the dependent variable, β₀ is the constant (intercept) term, T, VPD and C_a are the independent variables, β₁–β₃ are the partial regression coefficients for T, VPD and C_a, respectively, and ε is the residual term.

To check for multicollinearity among the independent variables and accurately quantify the independent explanatory contributions of each environmental factor to the variation in iWUE, we computed the variance inflation factor (VIF), which measures the extent of variance inflation in coefficient estimates due to multicollinearity. A VIF > 10 (or more strictly, >5) indicates severe multicollinearity (O’Brien 2007).

Results

Seasonal variations of tree-ring δ¹³C and δ¹⁸O

δ¹³C analysis of two specimens per species (P. massoniana: LA02A, LA18A; S. tzumu: LA36A, LA39A) confirmed synchronous intra-specific variation (Figure 2a and c). After applying the synchronization process to align physiological growth stages, statistical analysis revealed a significant positive correlation between the two P. massoniana individuals (n = 144, r = 0.43, p < 0.01), while an even stronger coordinated variation was observed between the two S. tzumu individuals (n = 149, r = 0.78, p < 0.01). Pinus massoniana exhibited an intra-annual pattern characterized by a shift from ¹³C depletion in the early growing season to pronounced enrichment in the mid-to-late growing season (Figure 2b), though this pattern was not consistently observed across all investigated years. In contrast, S. tzumu maintained strict interannual synchrony in δ¹³C variation, displaying a typical triphasic seasonal pattern: initial isotopic enrichment, mid-season depletion and a transient resurgence in the late season (Figure 2d). The seasonal amplitude of δ¹³C variation was similar between the two species, averaging 2.3‰ for P. massoniana and 2.4‰ for S. tzumu.

Also, both P. massoniana and S. tzumu exhibited significant intra-specific synchronicity in δ¹⁸O values following synchronization (Figure 2e and g). Statistical analysis revealed significant positive correlations between the two individuals of P. massoniana (n = 144, r = 0.85, p < 0.01) and similarly between the two individuals of S. tzumu (n = 149, r = 0.84, p < 0.01). Pinus massoniana exhibited strict temporal coherence in interannual δ¹⁸O seasonal patterns, characterized by a V-shaped trajectory: an initial depletion phase followed by progressive enrichment (Figure 2f). Sassafras tzumu displayed a three-phase δ¹⁸O pattern: early-growing-season enrichment, mid-season depletion and transient late-season resurgence (Figure 2h). The seasonal amplitude of δ¹⁸O variation was more species-specific compared with δ¹³C, with 9.7‰ for P. massoniana and 7.3‰ for S. tzumu.

Seasonal variation of iWUE and Δ¹⁸O_lw

The iWUE_sim model overestimated δ¹³C-derived iWUE in P. massoniana by 50.31 μmol CO₂ mol⁻¹ H₂O and in S. tzumu by 38.13 μmol CO₂ mol⁻¹ H₂O compared with iWUE_mes measurements. Although neglecting g_m limitations in the iWUE_sim approach resulted in systematic overestimation of absolute iWUE values, both models captured seasonal water-use efficiency patterns consistently (Figure S2 available as Supplementary Data at Tree Physiology Online). The iWUE_mes framework, which explicitly incorporates g_m constraints, demonstrated superior estimation accuracy. We therefore employed iWUE_mes-derived values for subsequent physiological analyses.

Both P. massoniana and S. tzumu showed significant inter-individual correlations in iWUE (Figure 3a), with correlation coefficients of r = 0.73 (n = 53, p < 0.001) and r = 0.95 (n = 59, p < 0.001), respectively. The two species exhibited synchronized seasonal iWUE patterns during their shared growing seasons, characterized by a decline during the early growth phase, a minimum value in summer and a rebound in autumn (Figure 3b). The mean seasonal amplitudes of iWUE variations were 12.95 μmol CO₂ mol⁻¹ H₂O for P. massoniana and 18.11 μmol CO₂ mol⁻¹ H₂O for S. tzumu. Notably, while P. massoniana displayed higher iWUE values overall, its seasonal fluctuation amplitude was smaller compared with S. tzumu (Figure 3b).

Figure 3.

Seasonal variations of iWUE of P. massoniana and S. tzumu from 2007 to 2014 (a) and their mean seasonal patterns (b); seasonal variations of Δ¹⁸O_lw of P. massoniana and S. tzumu from 2007 to 2014 (c) and their mean seasonal patterns (d).

Both P. massoniana and S. tzumu exhibited significant inter-individual correlations in Δ¹⁸O_lw (Figure 3c), with correlation coefficients of r = 0.87 (n = 52, p < 0.001) and r = 0.95 (n = 43, p < 0.001), respectively. The two species displayed synchronized seasonal Δ¹⁸O_lw patterns during their shared growing seasons, characterized by an initial rise during the early growth phase, peak values in summer and a subsequent decline in autumn (Figure 3d). The mean seasonal amplitudes of Δ¹⁸O_lw variations were 12‰ for P. massoniana and 13.25‰ for S. tzumu. Notably, while P. massoniana showed higher Δ¹⁸O_lw values overall, its seasonal fluctuation amplitude was smaller compared with S. tzumu (Figure 3d).

Relationships between iWUE, Δ¹⁸O_lw and climatic factors

To discern whether seasonal variations in iWUE were primarily driven by A or g_s, we examined the relationship between iWUE and Δ¹⁸O_lw. From 2007 to 2014, a significant negative correlation was observed between iWUE and Δ¹⁸O_lw in P. massoniana from April to October (r = −0.32, n = 52, p < 0.01) and in S. tzumu from April to September (r = −0.38, n = 43, p < 0.01).

We further investigated the relationships between iWUE and climate factors. Both P. massoniana and S. tzumu exhibited similar response patterns of iWUE to climatic factors including T, P, RH and VPD. Pearson correlation analysis (Figure 4; Table S2 available as Supplementary Data at Tree Physiology Online) revealed significantly negative correlations between iWUE and T for both P. massoniana and S. tzumu. Significantly negative correlations were also found between iWUE and VPD in P. massoniana and S. tzumu. iWUE showed weak correlations with P and RH in both species.

Figure 4.

Relationships between iWUE and monthly mean temperature (a, f), precipitation (b, g), relative humidity (c, h), vapor pressure deficit (d, i) and atmospheric CO₂ concentration (e, j) in P. massoniana and S. tzumu. The distinct biphasic response of S. tzumu iWUE to temperature is highlighted in (f), where the regression lines represent the piecewise linear regressions below and above the 20 °C threshold, with the corresponding slopes indicated by the adjacent numbers.

To eliminate potential confounding effects from other climatic variables, partial correlation analyses were conducted. The results indicated that significant partial correlations between iWUE and T were maintained in P. massoniana (Table 1) and S. tzumu (Table 2) after controlling for other climatic factors. Notably, no significant correlation was observed between iWUE and VPD when T was held constant (Tables 1 and 2).

Table 1.

Partial correlation coefficients between iWUE and climatic factors for P. massoniana (n = 56) from 2007 to 2014

	Control variable: T	Control variable: P	Control variable: RH	Control variable: VPD	Control variable: C  a
T		−0.522***	−0.574***	−0.441***	−0.422***
P	−0.207		0.342*	−0.398**	−0.339*
RH	0.097	0.095		−0.397**	0.054
VPD	−0.085	−0.459***	−0.545***		−0.348**
C a	0.587***	0.681***	0.679***	0.657***

Significance levels: ^***p < 0.001, ^**p < 0.01, ^*p < 0.05.

Table 2.

Partial correlation coefficients between iWUE and climatic factors for S. tzumu (n = 64) from 2007 to 2014

	Control variable: T	Control variable: P	Control variable: RH	Control variable: VPD	Control variable: C  a
T		−0.743***	−0.761***	−0.660***	−0.662***
P	−0.066		−0.137	−0.278*	−0.174
RH	−0.283*	−0.189		−0.664***	−0.398***
VPD	−0.151	−0.531***	−0.746***		−0.341**
C a	0.718***	0.789***	0.817***	0.748***

Significance levels: ^***p < 0.001, ^**p < 0.01, ^*p < 0.05.

Relationships between iWUE and C_a

Pearson and partial correlation analyses revealed that iWUE for both species was significantly and positively correlated with C_a (Tables 1 and 2; Table S1 available as Supplementary Data at Tree Physiology Online). However, to move beyond the potential artifacts of mathematical background forcing and isolate genuine physiological signals (as detailed in Statistical analysis), we compared the observed sensitivity (k_obs) of iWUE to C_a against the theoretical null-hypothesis slope (k_null).

The results (Figure S3 available as Supplementary Data at Tree Physiology Online) showed that k_obs substantially exceeded the corresponding k_null for both species (0.41 vs 0.21 for P. massoniana; 0.67 vs 0.23 for S. tzumu). This divergence confirms that the observed iWUE increase is not merely a product of the mathematical structure of the calculation formula, but is driven by active biological signals. Specifically, the k_obs for S. tzumu was nearly three times its theoretical value, indicating a significantly more robust active physiological regulation capability in this broadleaf species compared with the needle-leaved P. massoniana.

Relative contributions of environmental factors to the seasonal variation in iWUE

To quantify the relative contributions of environmental factors to the seasonal variation in iWUE, we performed multiple linear regression analyses. In the multiple linear regression model, a larger partial regression coefficient (Beta value) for an environmental factor indicates a stronger independent explanatory power of that factor regarding the variation in iWUE. Since multiple linear regression requires a linear relationship between independent and dependent variables, we excluded P and RH due to their weak correlations with iWUE, retaining only C_a, T and VPD. Given the correlation between VPD and T (r = 0.76, n = 72, p < 0.01), we assessed potential multicollinearity among the predictors by calculating the VIF, which measures the degree to which the variance of a coefficient estimator is increased due to multicollinearity.

The multiple linear regression model for P. massoniana was formulated as (Table S3 available as Supplementary Data at Tree Physiology Online): iWUE = 0.537 × C_a – 0.115 × VPD – 0.27 × T. For S. tzumu, the regression model was (Table S4 available as Supplementary Data at Tree Physiology Online): iWUE = 0.551 × C_a – 0.593 × T + 0.162 × VPD. Multiple linear regression models revealed that C_a, T and VPD collectively accounted for a substantial portion of the seasonal variation in iWUE, with a markedly higher explanatory power for S. tzumu (R² = 0.803) than for P. massoniana (R² = 0.565; Tables S3 and S4 available as Supplementary Data at Tree Physiology Online). A consistent driver pattern emerged across both species: C_a and T served as the primary predictors, exerting significant positive and negative effects on iWUE, respectively (Figure 5). Notably, while rising T significantly inhibited iWUE, VPD showed no significant independent influence in either model once other factors were accounted for. All models were statistically robust with no substantial multicollinearity (all VIF < 5).

Figure 5.

Multiple linear regression analysis of intrinsic water-use efficiency (iWUE) in response to climatic drivers in (a) P. massoniana and (b) S. tzumu. The numerical values represent standardized coefficients (Beta) for atmospheric CO₂ concentration (C_a), temperature (T) and vapor pressure deficit (VPD), with larger absolute values indicating greater contribution to the variation in iWUE. Significance levels are indicated as: ^**p < 0.01.

Discussion

Seasonal dynamics and interspecific differences in iWUE between P. massoniana and S. tzumu

During the shared growing season of P. massoniana and S. tzumu, although the iWUE differed in absolute values between the two species, they exhibited similar seasonal trends: a gradual decrease from spring to summer, followed by an increase in autumn (Figure 3a and b). These findings are consistent with previous studies reporting that although absolute iWUE values vary across tree species, their seasonal patterns tend to be similar (Klein et al. 2013, Morais et al. 2021, Portarena et al. 2024). Moreover, the two species showed comparable responses to climatic variables (Figure 4), suggesting that functional type differences did not alter the consistency in the direction of their responses to environmental drivers.

In terms of interspecific variation, S. tzumu exhibited greater seasonal amplitude in iWUE (Figure 3a and b), indicating higher plasticity in water-use regulation. In contrast, P. massoniana demonstrated a more conservative water-use strategy, with a narrower range of iWUE variation (Figure 3a and b). Regarding absolute differences, P. massoniana consistently exhibited higher iWUE. We attribute this to the generally lower hydraulic conductivity of conifers (non-porous wood) compared with broadleaf trees (ring-porous wood). This anatomical characteristic may enhance their ability to resist embolism under drought conditions (Choat et al. 2012, Carnicer et al. 2013, Guerrieri et al. 2019).

Photosynthesis-dominated mechanisms control seasonal iWUE dynamics in a humid monsoon region

Inferring g_s from tree-ring δ¹⁸O requires accounting for source water influences (Scheidegger et al. 2000). This issue was effectively addressed by calculating leaf water δ¹⁸O enrichment (Δ¹⁸O_lw) using well-established protocols (Barbour et al. 2004, Xu et al. 2018, Guerrieri et al. 2019, Mathias and Thomas 2021).

It should be noted that S. tzumu flowers from February to March, before leaf development (Chen et al. 2020). This aligns with tree physiological studies, indicating that deciduous broad-leaved species rely on stored non-structural carbohydrates in the early growth stage (Barbour et al. 2002, Helle and Schleser 2004, Kagawa et al. 2006, Fichtler et al. 2010, Simard et al. 2013). Therefore, Δ¹⁸O_lw was not calculated for S. tzumu during these months. We found a significant negative correlation between iWUE and Δ¹⁸O_lw (Figure 3), indicating that iWUE increases with rising g_s. This contradicts the classical inverse iWUE–g_s theory (Bögelein et al. 2012), confirming that g_s does not explain iWUE dynamics and establishing A as the dominant regulatory factor. This mechanism differs from that observed in summer-dry regions, where trees enhance iWUE through reduced g_s (Barbour et al. 2002, Battipaglia et al. 2014, Martínez-Sancho et al. 2022, Portarena et al. 2024). Our conclusion is further supported by Ren et al. (2024), who found that variations in bulk-leaf δ¹³C of Robinia pseudoacacia in a subtropical humid monsoon climate in China were controlled by photosynthetic capacity (as indicated by leaf nitrogen content), rather than by g_s regulation.

Evaluation of uncertainties in inferring stomatal conductance from tree-ring δ¹⁸O

The dual-isotope approach provides a classic theoretical framework for disentangling the forest carbon–water balance (Scheidegger et al. 2000). However, several recent comprehensive studies have highlighted that the linkage between tree-ring δ¹⁸O and g_s is indirect and weak, and its inferential process is highly susceptible to interference from various physical and physiological variables (Lin et al. 2022, Carter et al. 2025).

The validity of the dual-isotope model relies on the assumed universality of the Péclet effect and the stability of the effective path length (L) (Lin et al. 2022). Addressing this limitation, the two tree species investigated in this study still largely satisfy the prerequisites for applying the inverse Δ¹⁸O_lw–g_s relationship under their specific habitat conditions. On one hand, the Péclet effect in gymnosperms (P. massoniana) is typically minimal or negligible (Barbour et al. 2021). Although the absence of this effect diminishes the sensitivity of Δ¹⁸O_lw to g_s, it does not alter their inverse relationship. On the other hand, for broad-leaved angiosperms exhibiting the Péclet effect (S. tzumu) (Barbour et al. 2021), the signal fidelity is contingent upon the stability of L. The abundant precipitation and highly humid monsoon climate in our study region maintain relatively moderate leaf transpiration rates, thereby strongly restricting the drastic fluctuations in L that typically accompany extreme transpiration (Song et al. 2013).

Dual-isotope models typically assume p_ex to be a constant (0.4) (Guerrieri et al. 2019, Mathias and Thomas 2021); however, recent research indicates that p_ex decreases as the growing season progresses (Leppä et al. 2026). To quantify the potential confounding effect of this physiological uncertainty on the δ¹⁸O signal, we performed a sensitivity simulation. By comparing a constant p_ex scenario (0.4) with a linearly decreasing p_ex scenario (simulating p_ex dropping from 0.5 to 0.2 across the growing season), the results showed that Δ¹⁸O_lw and iWUE_mes consistently maintained a negative correlation for both P. massoniana (r = −0.30, p < 0.05) and S. tzumu (r = −0.24, p > 0.05) (Figure S4 available as Supplementary Data at Tree Physiology Online). While the correlation for S. tzumu under the dynamic p_ex scenario is not statistically significant, the seasonal Δ¹⁸O_lw trajectories from both constant and dynamic p_ex scenarios remain highly consistent and synchronized (Figure S4c available as Supplementary Data at Tree Physiology Online). This demonstrates that the potential dynamic variability of p_ex does not fundamentally alter the core seasonal covariation trend of the isotopic signals in these tree species.

Additionally, δ¹⁸O_v may introduce additional variability to Δ¹⁸O_lw, which confounds the g_s signal (Gat et al. 2001, Munksgaard et al. 2020, Lin et al. 2022). To evaluate the degree of interference from this environmental noise, we incorporated monthly δ¹⁸O_v and δ¹⁸O_p data simulated by the LMDZ4 model (2007–2010), alongside empirical measurements (2011–2014, including data derived from Li et al. 2020a), for auxiliary verification. The model’s capacity to capture the seasonal dynamics of local precipitation isotopes has been validated by comparing the LMDZ4-simulated data with measured δ¹⁸O_p near the sampling site, revealing highly similar seasonal fluctuation characteristics (Figure S5 available as Supplementary Data at Tree Physiology Online). The analysis reveals that in the humid East Asian monsoon region, δ¹⁸O_v is highly and positively correlated with precipitation δ¹⁸O_p, which serves as the reference source water (r = 0.58, p < 0.01), indicating that the two maintain an isotopic equilibrium state on a seasonal scale (Figure S5 available as Supplementary Data at Tree Physiology Online). Given that Δ¹⁸O_lw in this study is the enrichment value calculated relative to δ¹⁸O_p (Eqs (8) and (9)), this synchronous evolution of water vapor and precipitation isotopes computationally offsets a large portion of the physical interference caused by background δ¹⁸O_v fluctuations. Although the residual effects of δ¹⁸O_v cannot be completely excluded, this relative metric preserves the underlying stomatal physiological signal more effectively than the raw tree-ring δ¹⁸O.

Finally, it must be acknowledged that there are inherent methodological limitations in attempting to completely decouple the pure g_s signal from strong monsoon climate noise. In our study, tree-ring δ¹⁸O was significantly correlated with temperature and RH (Table S5 available as Supplementary Data at Tree Physiology Online), indicating unavoidable interference from the background noise of environmental fluctuations (Lin et al. 2022, Carter et al. 2025). Nevertheless, Carter et al. (2025) also noted that despite the methodological uncertainties of solely relying on δ¹⁸O to infer g_s, the conclusion drawn by Mathias and Thomas (2021) and Guerrieri et al. (2019) using such methods—that ‘g_s has not significantly changed at most sites’—may still hold true in real-world ecology. Based on plant physiological principles, it is reasonable to deduce that when water is not a limiting factor, trees will avoid reducing g_s to maintain the intercellular CO₂ supply, thereby ensuring photosynthetic assimilation. This physiological mechanism aligns well with our empirical inference: in the humid East Asian monsoon region, the seasonal dynamics of iWUE are primarily driven by seasonal variations in A, with g_s playing only a secondary role.

CO₂ and temperature jointly determine seasonal iWUE patterns through photosynthetic regulation

On interannual scales, numerous studies have demonstrated that the CO₂ fertilization effect is the primary driver of long-term iWUE increases (Guerrieri et al. 2019, Mathias and Thomas 2021, Treml et al. 2022, Mathias et al. 2023). At the seasonal scale, this study also found a highly significant positive correlation between atmospheric CO₂ concentration (C_a) and iWUE (Figure 4; Table S2 available as Supplementary Data at Tree Physiology Online). To rule out the ‘mathematical background forcing’ effect caused by C_a serving as an explicit multiplier in the calculation formula, we used a null-hypothesis model to confirm that this seasonal increase is fundamentally driven by active physiological regulation, rather than being a mere algebraic artifact (Figure S3 available as Supplementary Data at Tree Physiology Online).

Multiple linear regression analysis indicated that C_a was an environmental factor that significantly contributed to the variations in iWUE for both tree species (Figure 5a and b; Tables S3 and S4 available as Supplementary Data at Tree Physiology Online). Controlled experimental studies have confirmed that elevated CO₂, as the primary substrate for photosynthesis, directly enhances A while suppressing g_s, collectively increasing iWUE (Bert et al. 1997, Ainsworth and Rogers 2007). In contrast, humidity-related variables (P, RH, VPD) had weak effects on iWUE (Figure 4; Table S2 available as Supplementary Data at Tree Physiology Online), indicating minimal stomatal limitation under humid conditions and further underscoring the dominant role of A. As the main substrate for photosynthesis, CO₂ directly regulates photosynthetic activity, establishing C_a as the dominant factor controlling seasonal iWUE through A.

This study found that, on a seasonal scale, the iWUE of both tree species was generally significantly negatively correlated with temperature (Figure 4; Table S2 available as Supplementary Data at Tree Physiology Online). This is consistent with the observations of Bing et al. (2022) and Grossnickle et al. (2005), but differs from previous studies reporting a positive correlation between high temperature and iWUE (Seibt et al. 2008, Battipaglia et al. 2014). Multiple linear regression analysis further indicated that temperature is one of the key factors driving this seasonal variation (Figure 5a and b; Tables S3 and S4 available as Supplementary Data at Tree Physiology Online).

Notably, unlike the monotonic temperature response of P. massoniana (Figure 4e), the response of S. tzumu iWUE to temperature is not a simple monotonic decrease, but rather exhibits a biphasic characteristic (Figure 4f, as indicated by the red fitted line). From February to April, when temperatures are below 20 °C, the change in iWUE is relatively gradual, and its relationship with temperature is relatively weak (with a slope of only −0.31, Figure 4f). During this period, moisture conditions are relatively abundant, and a moderate temperature increase may promote the A while simultaneously driving a corresponding increase in g_s to support carbon assimilation (Li et al. 2020b). The coordinated changes in A and g_s likely maintain iWUE at a relatively stable level.

According to the temperature response theory of photosynthesis, the leaf photosynthetic rate typically follows a hump-shaped curve, increasing with temperature below the optimum temperature (T_opt) and decreasing once T_opt is exceeded. For subtropical tree species, T_opt is generally considered to be within the range of 25–35 °C (Li et al. 2020b). In this study, when the temperature exceeded 20 °C, the S. tzumu iWUE showed a distinct downward trend (reaching a slope of −1.51, Figure 4f), suggesting that 20 °C may approach or exceed the optimal temperature threshold for photosynthesis in this species on a seasonal scale. This decline likely reflects the inhibition of photosynthesis by high temperatures and its ‘decoupling’ from stomatal behavior. On one hand, high temperatures may lead to the downregulation of the maximum Rubisco carboxylation rate (V_cmax) and significantly increase the photorespiratory CO₂ compensation point (Γ*) (Brooks and Farquhar 1985, Bernacchi et al. 2001, Walker et al. 2017, Dusenge et al. 2019), thereby exerting an inhibitory effect on the A. On the other hand, under continuous heat stress, trees may tend to maintain relatively high g_s to sustain transpirational cooling and avoid thermal damage, rather than closing stomata synchronously with the decrease in A, thus leading to the decoupling of photosynthesis and stomatal behavior (Wang et al. 2026). Taken together, the observed decline in iWUE during the high-temperature season in this study may imply that photosynthesis is limited under high-temperature conditions while accompanied by high transpirational water loss; this speculation is consistent with the conclusions of Bing et al. (2022).

A key limitation of this study is the fact that intra-annual tree-ring dating remains approximate. Future research should combine continuous micro-coring sampling, high-resolution radial tree growth monitoring networks and simultaneous ecophysiological measurements to accurately determine cambial activity periods. Such approaches would significantly advance our understanding of the coupling between iWUE dynamics and tree carbon–water balance.

Conclusions

This study reveals that in the humid East Asian monsoon region, the seasonal variation in iWUE is predominantly governed by photosynthetic rate (A), establishing a ‘A-dominated’ regulatory mechanism distinct from the ‘stomata-dominated’ pattern observed in arid regions. Although P. massoniana and S. tzumu differ in their specific response strategies, the consistent seasonal iWUE pattern and its relationship with environmental factors jointly indicate that atmospheric CO₂ concentration and temperature are the key climatic drivers of iWUE variation. Humidity-related variables exhibit weak influences, confirming minimal stomatal limitation under humid conditions. The summer decline in iWUE is primarily attributable to the combined constraints of supra-optimal high temperatures (leading to photosynthetic limitation and stomatal decoupling) and limited CO₂ availability.

In humid forest ecosystems, carbon acquisition rather than water conservation is likely the key limiting factor for tree growth. Under future climate warming, even in moisture-sufficient regions, rising temperatures may directly suppress photosynthesis, thereby restricting tree growth and carbon accumulation and thus weakening forest carbon sink capacity. However, we acknowledge that current isotope-based physiological modeling entails inherent uncertainties, particularly regarding the background noise generated by environmental fluctuations. Therefore, it is essential for future forest carbon sink models to fully incorporate the seasonal dynamics of CO₂ supply and its interaction with temperature, and to couple high-resolution isotopic analysis with in situ continuous monitoring to constrain these model uncertainties, so as to avoid overestimating the carbon sequestration potential of forests in climate change projections.

Author contributions

C.X. was responsible for the design of this study. X.H., J.R. and S.L. conducted fieldwork. X.H. performed tree-ring cellulose δ¹³C and δ¹⁸O analysis. X.H., Y.L., C.X., K.T., R.H., Q.Z., and W.A. contributed to the interpretation of the results. X.H. and C.X. led the writing, with many contributions from the other co-authors.

Conflict of Interest

None declared.

Funding

This study was supported by the National Natural Science Foundation of China (Grant Numbers: 42425706, 42488201, 42407603), International Partnership program of the Chinese Academy of Sciences (Grant Number: 170GJHZ2023074GC), and National Key Research and Development Program of China (Grant Number: 2024YFF0807902).

Data availability

The dataset is available from the World Data Center for Geophysics, Beijing (https://doi.org/10.12197/2025GA041).